REMOVAL OF CRWTOSPORLDrun/iPAR C/ZRM BY GRANULAR MEDIA FILTRATION Monica Beata Emelko A thesis presented to the University of Waterloo in fulfillrnent of the thesis requirement for the degree of Doctor of Philosophy in Civil Engineering Waterloo, Ontario, Canada, 200 1 O Monica Beata Emelko 200 1 National Library I*l ,,fo Bibliothèque nationale du Canada Acquisitions and Bibliographie Services Acquisitions et services bibliographiques 395 Wellington Street Ottawa ON K IA OF14 395, rue Wellington Oaawa O N K IA ON4 Canada Canada Y w N e vom ~ ( B r e n t O Our üie Noae dfderxo The author has granted a nonexclusive licence allowing the National Library of Canada to reproduce, loan, distrî'bue or sell copies of this thesis in microform, paper or electronic formats. The author retains ownership of the copyright in this thesis. Neither the thesis nor substanttial extracts fiom it may be printed or otherwise reproduced without the author's pemiission. Lyauteur a accordé une licence non exclusive permettant à la Bibliothèque nationale du Canada de reproduire, prêter, distribuer ou vendre des copies de cette thèse sous la fome de microfiche/fihn, de reproduction sur papier ou sur format électronique. L'auteur conserve la propriété du drpit d'auteur qui protège cette thèse. Ni la thèse ni des extraits substantiels de celle-ci ne doivent être imprimés ou autrement reproduits sans son autorisation. The University o f Waterloo requires the signatures of al1 persons using or photocopying this thesis. Please sign below, and give address and date. ABSTRACT Increasingly stnngent regulations for drinking water qualïty have placed increased emphasis on a multi-barrier approach for providinp protection fiom waterborne pathogens. Widely found in surface waters, the pathogen Cryptospon'diurn pawum is particularly resistant to chemical disinfectants commonly used in drinking water treatment, underscorhg the need for multiple treatment strategies for inactivation or removal of C. p a m m fiom drinking water. When operation is optimized, granu1a.r media filtration systems are particularly effective barriers against C. parvum passage into potable water; however, Iess is known about the pathogen removal capacity of these systems at the beginning and end of the filter cycle (ie., filter t+ipening and breakthrough) or when particle removal processes are challenged (e-g-,coagulation upsets, hydraulic changes, etc.). The research presented in this thesis examined the passage of C.parvum and potential surrogates for C. parvurn through granular media filters during periods of optimal and non-optimal filter operation. A thorough review of the relevant filtration and C.parvzrrn Iiterature emphasized the difficulty in accurately enurnerating C. pawum fiom water samples. A relatively simple analytical method for concentrathg and enurnerating C.parmm during filtration studies was implemented and optirnized. Then, to address the uncertainty or reliability of C.pawlrm concentration and removal data, a new quantitative tool that incorporated several sources of error (representative sampling, random analytical error, and non-constant analytical recovery) was developed. The statistical model assumed a Poisson distribution for the true sample counts, a binomial distribution for modeling the recovered fraction of oocysts, and a Beta distribution for descnbing the uncertainty of oocyst recovery. A numerical technique (Gibbs sampler) was then applied to the statistical model to determine confidence intervals for C.parvzrm concentration and removal data. This method of descnbing the uncertainty associated with C.p a m m data (confidence intervals calculated via the Gibbs sampler) was used throughout this thesis research because it aIIowed for cornparison between different data sets with, in some cases, dif3erent analytical recoverïes. Bench-scale experïments were performed to determine if viable and chemicallyinactivated C.parvurn oocysts were similarly removed by granular media filters at a variety of operating conditions. These experiments were cntical because of the potential health rkks associated with the experimental use and release of viable oocysts. Since uncoagulated, chemically-inactivated oocysts have demonstrated slightly different surface charge properties (described by zeta potential) than viable oocysts, it had been speculated that the different oocysts might also be removed differently by granular media filters. Dual- (anthracite/sand) and tn-rnedia {anthracite/sand/garnet) investigations demonstrated similar removals of viable and chemically-inactivated C.pamrm oocysts during optimized operation, filter rïpening, and coagulation failure. While C. parvum removals were moderately lower (by -0.5 to 1-log) during ripening than during stable operation, the C. parvum removal capacity of both dual- and tri-media filters was severely and significantly compromised during coagulation failure when it decreased by >3-log relative to stable operation. C.pawum removals were not statistically different in dual- and tri-media filters, though increased replication may yield statistically significant differences in the marginally higher removals achieved by tri-media filtration. Pilot-sale experiments represented the majonty of the experimental efforts and focused on investigating design and operational strategies for maximizing C. p a m m removal by filtration. Multiple research platforms permitted investigation of different types of raw waters, water temperatures, coagulation regimes, and filter designs. Formalin-inactivated C.parvzim oocysts were seeded at al1 of the experimental locations. In addition to turbidity and particle concentration evaluations, polystyrene microspheres were evaluated as potential surrogates for C.parvum because they were sirnilar to oocysts in size and easy to ident* and enurnerate. The pilot-scale experiments demonstrated that excellent removals (>5-log) of C.parvlrrn couid be achieved during optimal operating conditions, even at temperatures as low as 1°C and during sprïng runoff conditions. These removals deteriorated substantially (by 3- to 4-Iog) during end-of-run and early breakrthrough filtration, even at filter effluent turbidities below 0.1 NTU. This result suggested that filter operation during breakthrough, as measured by turbidity or perhaps even particles, should be avoided. Coagulation failure and sub-optimal coagulation conditions (reductions in coagulant dose) also resulted in deteriorated C.p a m m removals. Relatively rapid changes in hydraulic loading demonstrated varied eEects on C. pamrm removal by filtration, though in most cases only little to no deterioration in filter effluent C.p a m m concentrations occurred. Turbidity monitoring proved more usefùl than particle counting in gauging the effects of hydraulic steps on C. panwm passage through filters. These events should be investigated M e r to better define how and when they impact pathogen passage. C. parvzirn removals by filtration were rnoderately (-0.5-1 log) lower during ripening than during stable operation, these less substantial dserences occurred over a relatively short duration of the npening period. During most of these operating conditions, oocystsized polystyrene microspheres appeared to be reasonable surrogates for C.parvum removal by filtration; however, they should continue to be evaluated relative to oocysts to better define the lirnits of îheir applicabiIity as surrogates. The pilot-scale investigations resulted in several operational and design implications and strategies for maximizing C.parvum removal by granular media filtration, the most notable of which were the importance of optïmized chernical pretreatrnent (coagulation pnor to filtration) and the potential for increased pathogen passage during end-of-run operation. n i e fina1 cornpletion of this thesis would have been impossible without the support of several individuals and institutions. First, 1 would like to acknowledge the support of my supervisor, Dr. Peter M. Huck. 1 greatly appreciate the opportunities and support that you have given me durùig the Iast five years. 1 would also like to thank the other members of my advisory committee for their continued support and interest in this work. My sincere thanks go to Dr. Robin Slawson for providing microbiological advice. 1 greatly appreciate Dr, Bill Lennox's perspective on multiple issues and his willingness and abiiity to discuss statistics on short notice. There are no words for the mentorship and renewed joy in learning that working with Dr. Park Reilly has brought me over the last few years - 1 hope that 1 can be a fraction of the teacher that he is. 1would also like to thank my extemal examiner Dr. John Tobiason whose comments and suggestions were truly appreciated. The £inancial support of the partners in the NSERC Chair in Water Treatrnent at the University of Waterloo, the Amencan Water Works Association Research Foundation, and the University of Waterioo made this research possible. This work would have also been impossible without the support of Ian Douglas and John Van Den Oever fiom the City of Ottawa and Saad Jasim fkom the Windsor Utilities Commission. I am indebted to you for your cornmitment to this work; particularly John, who was willing to stay up with me into the wee hours of the rnorning at the pilot plant, just so we could catch the onset of breakthrough. Your insights and perspectives were instrumental to the fiil1 completion of our research objectives. 1would also like to express my appreciation to al1 of the members of the NSERC Chair in Water Treatment. Whenever I had a technical or administrative question, Bill Anderson was always willing to help out. Dr. Sigrid Peldszus' assistance with issues ranging fiom chemistry to pottery is much appreciated. Janis Zimmer has an extraordinary work ethic and was a godsend - this work would not be done were it not for your assistance in the micro lab. 1 would also like to acknowledge other graduate students fiom the NSERC Chair, particularly Chdy Lang, Dr. Graham Gagnon, Dr. Daniel Urfer, Dr. Xibo Liu, and Katarina Pintar. 1 truly enjoyed o u . numerous discussions regarding our respective research and fiiture endeavors. 1 would also like to acknowledge the support of several individuals from the Department of Civil Engineering. Dr. Neil R. Thomson is a valued mentor who is generous in his willingness to provide assistance. 1 sincerely appreciate your perspective and friendship. 1am also indebted to Terry Ridgway for his hours of dedication to helping build the pilot plant and get it running and to Mark Sobon and Bruce Stickney who are always willing to help out. The completion of a Ph.D. is a long and occasionally Fustrating and isolating journey. Good fiends have tremendously brightened that journey. 1 sincerely appreciate the perspective and fnendship of Franklyn Smith of the Region of Waterloo, fiom whom 1 have learned a great deal about water treatment operations and management. 1 am grateful to Stefano Nomani for his contïnued willingpess to provide advice on a variety of issues ranghg fiom random number generators to home construction. It's also nice to have someone to waIk out with when everyone else has akeady gone home. Hopefdly one of these days it won't be us, eh? Nurnerous late night discussions with Shayne ("Giovanni-Sthpy") Giles and Sharla Howard have brought balance to this whole process. Those evenings always came at the right t h e (as did hallway wanderhgs and drop-ins) and are much appreciated. 1 would also like to extend my gratitude to Leah MacKinnon for her insighmil perspectives on graduate school, academia, and the consulting world; al1 fiom across the potters wheel. 1 am particularly thankfil for Sarah Anderson and Jenna Hemebry who are hue fiiends. You've been here through it al1 - without your support this thesis would not have been completed. 1 am tnily indebted to Greg Da Re for his love and encouragement. Your drive and pursuit of excellence continue to inspire me. 1 would like to express my sincere appreciation and thanWulness to Eric Hood for his love, kindness, and understanding. Without your support and patience during seemingly endless trips to perform experiments, conferences, and late night thesis writing sessions, 1 would not have made it to this p o i ~ t .Thank you for believing in me. Finally, my sincere gratitude is extended to my parents Anna and Michael Emefko for their unfailing support and encouragement. vii CHAPTER 3 MATERIALS AND METHODS ........................................................77 ....................................................................................... 7 7 3.1 EX~ERIMENTAL DESIGN 3.2 EXPER~MENTAL SET-UP ...................................................................................... 8 3 3.2.1 Bench-Scale Filtration Apparatus........................................................... 85 3 - 2 2 Ottawa and Windsor Pilot Plants ..................... . . . .................................. 86 .............................................................. 3.2.3 Ukiversity of WaterlooPilot Plant. 8 8 3.3 SEEDINGAND SAMPLING ........................................................................................ 88 3.3.1 Pilot Plant Coagdation and Jar Coagulation Protocol ............................... 89 3.3.2 Calculation of Microorganisrn and Microsphere Concenhatbn and Removal ....................................... .9 1 3.3.3 Bench- and Pilot-Scale Seeding Protocol ..... ..............................................9 2 3.3.4 Bench- and Pilot-Scale Sampling ProtocoZ. ................................................. 92 3.3.5 Micruorganism Losses to Seeding Apparatus ............................................... 96 3 -4 MICROBIOLOGICAL PAEUMETERS ..........................................................................97 .................... 97 3.4.1 C.pawum ............................................................... , . 3.4.2 B- swbtilis ................. ..................................................................................... 102 3.5 MCROSPHERES .................................................................................................... 102 PARA~IETERS .......................................................... 103 3.6 PHYsIcAL AND CHEMICAL 3.6.1 Headloss ...................................................................................................... 106 3.6.2 ParticleCotcnts........................................................................................... 106 3.6.3 Tzrrbidr'ty...................................................................................................... 107 107 3.6.4 pH ............................................................................................................... . . . . CHAPTER 4 QUANTIFYING THE RELXABILITY OF PATHOGEN DATA.................................................................................................. 108 .....................................108 4-1 CRI~PTOSPORIDTUM DISTRIBUTION IN NATUMLWATERS ................................................ 4.2 NECESSITYOF RELIABLE STATISTICAL METHODS 109 . . , ............................................................................... 110 4.3 EXAMPLE DATA ................ OF POISSON ESTIMATION OF CONFIDENCE INTERVALS ..................... 113 4.4 ADEQUACY 4.4.1 Calculating Poisson Confidence Intervals .................................................. 112 ...........-. 113 4 - 4 2 Limitations of Poisson Confidence Intervals ..................... . . . 4.5 SOURCESOF ERRORDURINGCRYPTOSPORTD~M CONCENTRATION AND ENLTMERATION ................... ,. ........,,,................................................................... 120 4.6 APPROXIMATION OF BETAPAUMETERS............................................................. 120 4.7 CALCULATING CONFIDENCE INTERVALS ............ . ................................................ 125 CHAPTER 6 PAR VU2MAND POTENTLAC SURROGATE RXblGVAL BY FILTRATION ............................................................................. 162 6.1 LNTRODUCTION ..................................................................................................... 162 6.2 OTTAWAPLOT PLANTINVESTIGATIONS .............................................................. 165 6.2.1 Stable (Optimized) Operation .............................. .......................................165 6 - 2 2 Ripening ......................................................................................................177 6 - 2 3 Breakrhrough ...............................................................................................18.5 6.2.4 Coagulant Eflects ....................................................................................... 204 6.2.5 Hydraulic Step ............................................................................................ -219 6.3 UNIVERSITY OF WATERLOO 0 PILOT PLANTINVESTIGATIONS ......................228 6.3.1 Stable (Optimized) Operation .................................................................... 2 2 8 6.3.2 WdrauZic Step ............................................................................................ 7 3 3 6.4 WINDSOR PILOTPLANTINVESTIGATIONS ............................ ..,..........................239 6.4.1 Stable (Optimized) Operation ..................................................................... 239 6 - 4 2 Discussion .................................. . . . ........................................... 2 4 3 6.4.3 Rate Effects ............. ...... ...................................................................... 7 4 4 CHAPTER 7 INTEGRATED C PAR VU32 AND POTENTLAL SURROGAm DATA ........................................................................ 248 7.1 OVERALL c.PARWMJXEMOVAL ...................................................................... 248 7.2 OVEML ASSESSMENT FOR POTENTIAL SURROGATES FOR C. PAR W M ............... 253 CHAPTER 8 CONCLUSIONS. IMPLICATIONS. AND RECOMMENDATIONS ..................................................................261 ..................................................................................................... 2 6 1 8.1 CONCLUSIONS 8.2 LMPLICATIONS ................................................................................................ 2 6 3 ......................................................................................... 8.3 RECOMMENDATIONS 264 8.3.1 Water Treamtent Plant Operations and Management .............................. 264 8.3.2 Water Treatment Research .............. .......................................................... 2 6 6 APPENDIX A .....268 ANALYTICAL METHOD DEVELBPMENT: C PAR W M OBJECTIVE ........................................................................................................... 268 ...............2 6 9 A.2 AVAILABLE METHODS: ADVANTAGES AND LIMITATIONS . ........ .... A.3 ORIGINAL M E ~ OOF D YATESE T L .(1997) ........................................................ 274 A .1 ................................................... 275 A.4 C. PAR WM INACTIVATION AND PRESERVATION .............................................. 275 A S ENUMERATION OF C. PARWM STOCK SUSPENSION ....................................................... 2 7 6 ................ A.6 METHOD OPTIMTZATION ..,.. A.6.1 Membrane Type (Polycarbonate vs. Cellulose Acetate) ............................. 276 ..... .......... 277 A .6.2 Direct Vacuum Filfration (vs. Svn'nge Filtration) .................... .................................................... 282 A.7 OPTIM~ZED C. PAR WMMËTHODPROTOCOL .............................................. 285 A.8 METHoD RECOVERY IN DIFFERENT WATERTYPES ......................... 389 A.9 EFFECT OF COAGULANT ON RECOVERY............................ . . APPENDIX B APPENDIX C APPENDIX D REFERENCES PAR VUlM QUALITY ASSURANCE AND QUALITY CONTROL DATA ........................................................................... 292 ANALYTICAL RECOVERY OF PAR VUBf AND MICROSPHERES ...........................................................................297 DETAILED P M CrU2M7B.SUBTEIS7 AND MICROSPHERE DATL4.................................................................311 .............................................................................................................. 342 LIST OF TABLES Waterborne Outbreaks of Cryptosporidiosis..................................................... 15 Operational Factors and Related Particle Removal and Passage Mechanisms .40 Surrogate Parameters for Removal of Cryptosporidium...............................48 Removal of CryptosponnBiurnDuring Stable Operation ...................................54 Methodological Factors and Relevant Questions to Consider When Evaluating Pathogen Removal Data ................................................................59 Table 2.6 Summary of Critical Information fkom Studies Evaluating Process Removals of C.p a m m ................................................................................... 60 Table 3.1 Summary of bench- and pilot-scale experiments .......................................... 78 Table 3.2 Operating conditions examined during pilot-scale experiments at Ottawa ...... 81 Table 3.3 Process configurations at the various research platfoms .................................83 Table 3.4 Nominal raw water quality at the various research platforms........................ 84 Table 3.5 Jar coagulation protocol .................................................................................... 90 Table 3.6 Pilot-scale seeding and sampting specifics .......................................................95 Table 3.7 C.p a m m QNQC data comparing slides read at the University of Waterloo and CH Diagnostic and Consulting Services Inc. (Loveland, CO)......................................................................................... 101 Table 3-8 Hemocytometer Enurneration of Microspheres .............................................. 104 Table 4.1 Example C.p a m m Recovery Data .................. ....... .............................. 1 1 1 Table 4.2 Example C.parvum Experirnental Data.......................................................... 112 Table 4.3 Simulated Data Using Nahrstedt and Gimbel (1996) Mode1......... ................. 115 Table 4.4 Calculated Data Used for Describing the Overall Recovery Profile . . . .................... 123 (Beta pdf) for Recovery Data in Table 4.1 ................... . Table 5.1 Dual- and Tri-Media Filter Innuent and Effluent C.pawum Concentrations 142 and Effluent Turbidity D u h g Stable Operation..................................... Table 5.2 Dual- and Tri-Media Filter M u e n t and Effluent C.panmm Concentrations and Effluent Turbidity During Ripening....................................................... 143 Table 5.3 Dual- and Tri-Media Filter Influent and Effluent C.pawurn Concentrations and Effluent Turbidity During Coagulation Failure...................................... 144 Table 5.4 Theoretical and Measured Filter Influent C.p a m m Concentration Data...... 153 Table 6.1 Surnrnary of Stable Operation Experiments at the Ottawa Pilot Plant ............ 167 168 Table 6.2 Filter Performance DuMg Stable Operation at Ottawa ............................... Table 6.3 95% Confidence Intervals and C.pawurn Removal Ranges During Stable Operation at Ottawa .............................. . . . . ................................................ 176 Table 6.4 Summary of Ripening Experiments at the Ottawa Pilot Plant ........................ 178 Table 6.5 Filter Performance During Ripening at Ottawa .............................................. 179 Table 6.6 95% Confidence Intervals and C-palvurn Removal Ranges During Ripening at Ottawa ........................................................................................ 183 Table 6.7 Summary of Breakthrough Experiments at the Ottawa Pilot Plant.................187 Table 6.8 Filter Performance During End-of-Run and Breakthrough at Ottawa ............ 188 Table 6.9 95% Confidence Intervals and C.pawum Removal Ranges During End-ofRun, Early Breakthrough, and Late Breakthrough at Ottawa ....................... 201 Table 6.10 Sumrnary of Coagulant Eflects Experiments at the Ottawa Pilot Plant ........ 208 Table 6.1 1 Filter Performance During No Coagulant and Sub-Optimal Coagulation Expiments at Ottawa .................................................................................. 209 Table 2.1 Table 2.2 Table 2.3 Table 2.4 Table 2.5 xii Table 6.12 95% Confidence Intervals and C.p a m Removal Ranges During the No Coaadant and Sub-Optimal Coagulation Experiments at Ottawa..........217 Table 6.13 Summary of Hydraulic Step Experîments at the Ottawa Pilot Plant ............ 22 1 Table 6.14 Filter Performance During Hydraulic Steps at Ottawa ......................... ........ 221 Table 6.15 Summary of Stable Operation Experiments at the UW Pilot Plant ..............229 Table 6.16 Filter Performance During Stable Operation at UW ................. ........... 230 Table 6.17 95% Confidence Intervals and C.p a m m Removal Ranges D u ~ Stable g Operation at UW .......................................................................................... 233 Table 6.18 Silmmary of Hydraulic Step Experiments at the UW Pilot Plant ................. 235 Table 6.19 Filter Performance D u ~ Hydraulic g Steps at UW ......................................236 Table 6.20 Surnrnary of Stable Operation Experiments at the Windsor Pilot Plant .......240 Table 6.21 Filter Performance During Stable Operation at Windsor ..............................240 Table 6.22 95% Confidence Intervals and C.parvurn Rernoval Ranges During Stable Operation at Windsor ...................................................................................2 4 4 Table 6.23 Sumrnary of Rate Effects Experiments at the Windsor Pilot Plant ............... 245 Table 6.24 Filter Performance During Rate Effects Experiments at Windsor................346 Table A.1 Summary of Common Methods Used in Detecting Cwptospondium ........... 271 Table A 2 C.parvurn Recovery fiom Cellulose Acetate Membranes ........... ........... 278 Table A.3 C.parvum Recovery from Polycarbonate Membranes ..................................279 Table A.4 Statistical Analysis of Cellulose Acetate and Polycarbonate Membranes ..... 280 Table A S Statistical Analysis of Direct Vacuum (Manifold) and Syringe Filtration.....281 Table A.6 C.p a m m Concentration and Enurneration Protocol ................................ ....283 Table A.7 C.parvum Recovery from Ottawa Water.................................................... 286 Table A.8 C.parvtmz Recovery fiom UW Water (1-5 NTU) .......................................... 287 Table A.9 C.parvum Recovery from UW Water (3.5 NTU) .................................... 2 8 8 Table A.10 ANOVA Analysis of C.pawum Recovery fiom Various Water Matrices..289 Table A .11 C.parvum Recovery fiom Ottawa Water with 30 mgL AIum .................... 290 (A12(S04)3'1 8HzO) ...................................... . . , Table A .12 ANOVA Analysis of C.parvurn Recovery fkom Various Water Matrices Including Ottawa Water with a Hi& Coagulant Dose ................. . .............291 Table B.I Methodological Positive and Negative Control and Filter Influent and Effluent Control Results for Pilot-Scale Experiments at Ottawa ..................295 Table B.2 Methodological Positive and Negative Control and Filter Influent and Effluent Control Results for Pilot-Scale Experiments at Windsor................ 296 Table B.3 Methodological Positive and Negative Control and Filter Influent and Effluent Control Results for Pilot-Scale Experiments at UW ....................... 296 Table C.1 Calculation of Beta Parameters a and b for C.parvum Recovery fiom Ottawa Filter Influent ................................................................................ 3 9 8 Table C.2 Calculation of Beta Pararrieters a and b for C.parvum Recovery fiom Ottawa Filter Effluent....-...... ..... .............................-....-...--..-......-...--.*. 2 9 9 Table C.3 Calculation of Beta Parameters a and b for C.pawum Recovery fiom Ottawa Water (Influent and Effluent) ......................... . . ............................. 300 Table C.4 Calculation of Beta Parameters a and b for C.pamrm Recovery f?om Ottawa Filter Influent with 40 mg/L Alum ..................................... ........3 0 1 Table C S Calculation of Beta Parameters a and b for C.p a w m Recovery from Ottawa Filter Effluent with 40 mg& Alum ................... . , ......,.....-.............302 - ... Xlll Table C.6 Calculation of Beta Parameters a and b for C.parvurn Recovery fiom Ottawa Water (Influent and Effluent) with 40 mg/L Alum ............ ..........*.303 Table C.7 Calculation of OveralI Beta Parameters a and b for C.pawurn Recovery . .................304 from Ottawa Water (AU Influent and Effluents) .................... Table C.8 Calculation of Beta Pararneters a and b for C.pamrrn Recovery from UW (1.5 NTU) Filter M u e n t ...............................................................................305 Table C.9 Calculation of Beta Pararneters a and b for C.parvtrm Recovery fiom UW (1.5 NTU) Filter Effluent ...................................... . .306 Table C .I O Calculation of Beta Parameters a and b for C.parvurn Recovery from UW (1.5 NTU) Water (Idluent and Effluent) ............... ...................... 3 0 7 Table C.11 CalcuIation of Beta Parameters a and b for C.pawum Recovery fiom U W (3.5 NTW) Filter M u e n t ............... ...... .................................................... 308 Table C.12 Calculation of Beta Parameters a and b for pawum Recovery from UW (3.5 NTO) Filter Effluent .................... . . . .................................................309 Table C.13 Calculation of Beta Parameters a and b for C.parvztrn Recovery fiom UW (3 -5 NTU) Water (Influent and Effluent) ...................................................... 310 Table D.1 C.parvum Removal Data fiom Bench-Scale Experknents............................ 312 Table D.2 Calculation of C&oretical for Bench-Scale Experiments ..................................314 17 Table D.3 Experimental Schedule.............................................................................. 3 Table D.4 Filter Performance at Ottawa ........................................................................ 320 Table D.5 Turbidity and Total Particle Data fiom Ottawa ...................... . . .................. 322 Table D.6 Microorganism Data fiom Ottawa ................................................................ 327 Table D.7 Microsphere Data fiom Ottawa ....................i ................................................. 331 Table D.8 Microorganism and Microsphere Removal (Filtration) and Total Particle Reduction (Plant) .Summary fiom Ottawa ..................... ...... .........3 3 3 Table D.9 Filter Performance at Windsor ....................................................................... 336 . ............... -337 Table D.10 Total Particle Reduction (Plant) Data fiom Windsor ................ Table D.1 1 Microorganism Data fkom Windsor ...................................... ...............3 3 8 Table D.12 Filter Performance at UW ...................................................................... 3 39 Table D.13 Turbidity and Total Particle Data fiom W ...........................................3 4 0 Table D.14 Microorganism and Microsphere Data fiom UW ........................................ 341 . . xiv LIST OF FIGURES Figure 1.1 Conceptual relationship between pathogen passage and traditional operational parameters such as turbidity and partide counts .......................... 2 Figure 1.2 Research approach.......................................................................................... 6 Figure 2.1 Life cycle of Cryptospondium (after Fayer and Ungar, 1986)...................... 12 77 , Figure 2.3 Sampling fkom a water body (afrer Nahrstedt and Gimbel, 1996)..... ...........Figure 2.3 Effect of sample preparation on number of observed oocysts (after Nahrstedt and Girnbel, 1996)......................................................................... 33 Figure 2.4 Effect of Beta parameters a and b on recovery probability density function 26 Figure 2.5 Filtration mechanisms (after Amirtharajah, 1 988) ....... ............ ... .............. 30 Figure 2.6 Particle transport mecfianisms (after Ives, 1982; Amirtharajah, 1988).........30 Figure 2.7 Mechanisms affecting attachent during filtration .......................................33 Figure 2.8 Modes of detachment during filtration .................... . ................................. 35 Figure 2.9 Conceptual mode1 of attachent and detachment during filtration............... 38 Figure 3.1 Bench-scale filtration apparatus.... ............ . . ..........................O................... 85 Figure 3.2 Pretreatrnent at the Ottawa Pilot Plant .................................................... 87 Figure 3.3 Filter columns at the Ottawa Pilot Plant.......................... . . . . ......... 87 Figure 3 -4 Filter influent sampling location at Ottawa................................................94 Figure 3.5 C.p a m m Iosses to seeding apparatus and equipment during no coagulant and no media control experiments-............................................................ 97 Figure 3.6 Direct vacuum filtration apparatus for processing C.parvtrrn ....................... 99 Figure 3.7 C.parvum oocysts and YG polystyrene microsphere ( 4 0 0 ~magnification, Nikon Labophot 2A, Nikon Canada Inc., Toronto, ON)........... ,.................105 Figure 3.8 BB polystyrene microspheres ( 1 0 0 ~magnification, Nikon Labophot 2A, Nikon Canada Inc., Toronto, ON).............................................................. 106 Figure 4.1 Overall Recovery Profile (Beta pdf) for Recovery Data in Table 4.1. ...,.... 124 Figure 4.2 Overall Recovery Profile (Beta cdf) for Recovery Data in Table 4.1. ........ 121 Figure 4.3 Probability defisity functions (pdfs) for C.parvum Removal Example Data in Table 4.2 (Stable Operation at Ottawa)......................................... 130 Figure 4.4 Effect of observations (counts) and replicate samples on confidence interval range . ............................................. . . ..... . 131 Figure 5.1 Impact of coagulant on colloidal zeta potential (Arnirtharajah, 1988)........135 Figure 5.2 Bench-scale experimental configuration .....................................................138 Figure 5.3 Summary of sampling tirnes and coagulation conditions during benchscale experiments........................................................................................ 139 Figure 5.4 Filter eflluent turbidity, seeding period, and sampling times during dualmedia filter experïments with formalin-inactivated oocysts.......... .............. 141 Figure 5.5 Filter effluent turbidity, seeding period, and sampling tirnes during dual........----.. media filter experiments with viable oocysts..*............. ........ 145 Figure 5.6 Filter effluent turbidity, seeding penod, and sampling times during trimedia filter experiments with formalin-inactivated oocysts........................ 146 Figure 5.7 Filter effluent turbidity, seeding period, and sampling times during tri-media filter experiments with viable oocysts ............................. ........................146 Figure 5.8 Dual-media filter removals of viable and inactivated Cyptosporidiurn during stable operation, ripening, and coagulation failure....................... ...148 Figure 5.9 Tri-media filter removals of viable and inactivated Cqptospon'dium during stable operation, ripening, and coagulation failure. ......................... 149 Figure 5.10 Pooled dual- and tri-media filter removals of Cryptosporidium during stable operation, ripening, and coagulation faiture. .....................................150 Figure 5.11 Dual-media filter enluent oocyst concentrations and oocyst log removals as a fùnction of filter effluent turbidity during stable operation and coagulation failure. ...................................................................................... 158 Figure 5.12 Dual-media filter effluent oocyst concentrations and oocyst log removals as a function of filter effluent turbidity during stable operation and 158 ripening. ....................................................................................................... Figure 5.13 Tri-media filter emuent oocyst concentrations and oocyst log removals as a function of filter effluent turbidity during stable operation and coagulation failure. ...................................................................................... 159 Figure 5.1 4 Tri-media filter effluent oocyst concentrations and oocyst log removals as a function of filter efnuent turbidity during stable operation and ripening -159 Figure 6.1 Filter effluent turbidity and particle concentration during May 3 1, 1999 stable filter operation experiment at Ottawa.,..,. ................ ................ 169 Fi,oure 6-2 Filter effluent particle, C.pamrn, and B. subtiZis concentrations during May 3 1, 1999 stable filter operation experiment at Ottawa. ....................... 169 Figure 6.3 Filter effluent turbidity and particle concentration during January 19, 2000 stable filter operation experiment at Ottawa....................... . . .... 171 Figure 6.4. Filter effluent particle, C.parvum, and microsphere concentrations during January 19,2000 stable filter operation experiment at Ottawa. ...... .171 Figure 6.5 Relationship between C.pamirn, B. subtilis, and microsphere removals by filtration during stable operation at Ottawa. ............... .. .............................. 173 Figure 6.6 Relationship between C. pamrn removal by filtration and total particle (22 pm) reductions through the plant during stable operation at Ottawa. ... 173 Figure 6.7 Filter effluent turbidity and particle concentration during November 10, 1998 ripening experirnent at Ottawa. ........................................................... 180 Figure 6.8 Filter effluent particle and C.pczrvum concentrations during November 10, 1998 ripening experiment at Ottawa............... .................. ........ 180 Figure 6.9 Filter effluent particte and B. subtilis concentrations during November 10, 1998 ripening experiment at Ottawa...................................................... 181 Figure 6-10 Relationship between C.parvum and B. szibtilis removals by filtration and total particle (22 pm) reductions through the plant during ripening at Ottawa. ........................................................................................................ 182 Figure 6.11.Filter effluent turbidity during December 22, 1999 late breakthrough ...................................... 189 experiment at Ottawa. ................................. Figure 6.12.Filter effluent turbidity, C.panum, and microsphere concentrations during December 22, 1999 late breakthrough experirnent at Ottawa. .........190 Figure 6.13.Filter efnuent particle counts, C.parvum, and B. subtilis concentrations during December 22, 1999 late breakthrough experiment at Ottawa. ......... 191 Figure 6.14.Filter effluent turbidity and particle concentration during March 3,2000 early breakthrough experirnent at Ottawa.................................................... 192 Fi,we 6.1S.Filter particle, C.parvurn, and microsphere concentrations during March 3, 2000 early breakthrough experhnent at Ottawa................................... 193 . . . . xvi Figure 6.16.Fiiter effluent turbidity and partîcle concentration duruig March 9,2000 . , ...................................... 195 end-of-- experiment at Ottawa. ......................, Figure 6.17.Filter efnuent particle, C. p a m , and microsphere concentrations during March 9, 2000 end-of-run experirnent at Ottawa. ............................195 Figure 6.18 Filter effluent particle, C. parvum, and B. subtilis concentrations during January 2 1, 1999 end-of-nin experirnent at Ottawa. .....,.... .........................196 Figure 6.19.Box-and-whisker plot of C.pawum, B. subtiZis, microsphere removals by filtration and total particle (2 2pm) reductions by the plant during stable operation, end-of-run, and breakthrough at Ottawa................................... 197 Figure 6.20.Relationship between C.pawum and microsphere removals by filtration during the end-Of-mn, ear1y breakthrough, and late breakthrough at . . ................................................................ 198 Ottawa. .................................... Fieve 6.2 1.Relationship between C. pawum and B. szrbtiZis removals by filtration and total particle (22pm) reduction through the plant during end-of-m. early breakthrough, and late breakthrough at Ottawa. .............. 198 Figure 6.22 Box-and-whisker plot of C.p a m rernovals by filtration during the no coagulant and sub-optimal coagulation experiments at Ottawa. ................. 2 10 Figure 6.23 Box-and-whisker plot of B. subrilis removals by filtration during the no coagulant and sub-optimal coagulation experiments at Ottawa. ................. 310 -. Figure 6.24 Box-and-whisker plot of total particle (22pm) reductions through the plant during the no coagulant and sub-optimal coagulation experirnents at Ottawa. ............................................................................ ................. 2 11 Figure 6.25 Filter effluent particle, C. parvrcm, and microsphere concentrations durhg March 10,2000 sub-optimal coagulation experiment at Ottawa. ....2 13 Figure 6.26.Relationship between C.parvzirn and microsphere removals by filtraiton during the no coagulant-extended duration, no silicate, and sub-optimal .............................2 14 coagulation experiments at Ottawa. .................... ..... Figure 6.27.Relationship between C.parvum and B. subtilis removals by filtration and total particle (22pm) reductions through the plant during al1 no coagulant and sub-optimal coagulation experiments at Ottawa. ................. 2 15 Figure 6.28 Turbidity and particle response of filter during hydraulic step experiment on June 7, 1999 at Ottawa pilot plant. ......................................................... 3 3 3 Figure 6.29 Particle and microorganism response of filter during hydraulic step experiment on June 7, 1999 at Ottawa pilot plant. ......................................223 Figure 6.30 Turbidity and particle response of filter during hydraulic step experïment ........................... 223 on June 15, 1999 at Ottawa pilot plant. .................... . . Figure 6.3 1 Particle and rnicroorganism response of filter during hydraulic step experiment on June 15, 1999 at Ottawa pilot plant. ................................... .224 Figure 6.32 Relationship between C.parvum, B. subtilis, and microsphere removals by the pilot-scale dual- and tri-media filters during stable operation at UW. ..332 Figure 6.33 Relationship between C. p a m m and microsphere removals by the piiotscale dual- and tri-media filters during hydraulic steps at UW. .................. 237 Figure 6.31 Relationship between C. parvum and B. subtilis removals by filtration and total particle (Xpm) reductions through the plant at Windsor. ........... 242 Figure 7.1 Box-and-whisker plot of C.pamïm removals by filtration during al1 operating periods (except hydraulic steps) investigated at Ottawa.............. 248 --- xvii Figure 7.2 95% Confidence intervals and adjusted ranges of pawum removals by filtration d u k g al1 operating periods (except hydraulic steps) investigated at Ottawa, ................... .-.......................................................................... 2 5 1 Figure 7.3 Box-and-whisker plot of C. p a m m removals by dual-media filters during stable operation at Ottawa, Windsor, and UW research platforms. -252 Figure 7.4 95% codïdence intervals for C. parvum removals by dual-media filters during stable operation (pooled data) at Ottawa, Windsor, and UW ...........253 Fi,oure 7-5 Box-and-whisker plot of B. subtilis removals by filtration during a11 operating penods (except hydraufic steps) investigated at Ottawa..............254 Figure 7.6 Box-and-whisker plot of polystyrene microsphere removals by filtration during al1 operating periods (except hydraulic steps) investigated at . Ottawa....................................... . . . .2 5 4 Figure 7.7 Box-and-whisker plot of total particle (22 pm) reductions through the plant during alI operating periods (except hydraulic steps) investigated at Ottawa. ..................................................................................................... 255 Figure 7.8 Relationship between C.parvum removals by filtration and total particle (22 pm) reductions through the plant during al1 operational periods . investigated at Ottawa. ................................................................ 256 Figure 7.9 Relationship between C.parvum and B. subtilis removals by filtration during al1 operational penods investigated at Ottawa. ................................257 Figure 7.10 Relationship between C.panmm and polystyrene microsphere removals by filtration during al1 operational penods investigated at Ottawa. ............ 257 Figure 7.1 1 Relationship between C.parvum and polystyrene rnicrosphere removals by filtration dunng al1 operational periods investigated at W. ...................... 258 Figure 7.1 2 Relationship between C.parvum, B. subtilis, and pol ystyrene microsphere removals by filtration during al1 operational periods investigated at Ottawa, Windsor, and tTW......................................... ........ .................... 258 . .............. 275 Figure A. 1 C.parvum analytical method of Yates et al. (1997) ............... . xviii The pnmary goal of drinking water supply systems is to protect public health by supplying water with acceptably low concentrations of microbial and chemical contarninants. Increasingly stringent regulations for drinking water quality necessitate a multi-barrier approach for providing protection fkom waterborne pathogens. Widely found in surface waters, the pathogen C~yptospon'diump a m m is particularly resistant to chemical disinfectants cornmonly used in dnnking water treatrnent. It has been suggested that C. pntum may require up to ten times the ozone dose required for effective inactivation of Giardia lamblia cysts (Owens et al., 1994), a level which cm push the limits of economic feasibility and disinfection by-product cornpliance for many utilities. The considerable costs and practical limitations of adequate inactivation of C. p a m m by traditional disinfection processes have underscored the importance of muItiple treatment strategies for inactivation or removal of C. parvum from druiking water. Although alternative treatment processes such as pressure-driven membranes offer excellent removal capabilities of parasitic pathogens, they are not currently economical for treatment of flows greater than I to 5 MGD (Wiesner e t al., 1994). A common component of conventional water treatment operations, filtration has demonstrated efficacy as a Sarrier against C.parvum. Full-scale C. p a m m removals fiom 2 to >4 log have been reported in the literature (Baudin and Laîné, 1998; Nieminski and Ongerth, 1995). This range of reported removals may be in part explained by differences in operating conditions which can greatly affect oocyst removal by filters (Patania et al., 1995). To date, most of the reported investigations of C.p a m m removal by filtration have focused on optimal operating conditions. Pathogen passage during vulnerable penods in the filter cycle when particle removal processes are challenged (e.g., ripening, breakthrough, etc.) has been less thoroughly investigated. Originally described by Huck et al. (200 l), a conceptual representation of the relationship between pathogen passage and traditional filter performance parameters such as turbidity and particle counts is presented in Figure 1.1. The left vertical axis shows turbidity or particle counts while the nght vertical axis shows pathogen concentration in filter effluents. The hoL=izoneal axis represents the time of approximately cne filter cycle. Hydraulic Surge -Turbidity Pathogens Coagulation Upset Figure 1.1 Conceptual relationship between pathogen passage and traditional operational parameters such as turbidity and particle counts (modified fiom Huck et al., 200 1). As shown in Fiapre 1.1, regulatory -pidelines and treatment objectives ofien necessitate that filter emuent turbidity must always remain below a specified goal or standard. Afier backwash, both turbidity and particle counts increase during fdter ripening. Pathogen passage through the fllters may increase during this time for the same reasons. Events such as coagulation upsets result in sub-optimal pretreatment and may consequently cause filter effluent concentrations of turbidity, particles, and pathogens to increase as a result of non-attachent. Similarly, events such as hydraulic surges rnay result in the release or detachment of particles including pathûgens. As the filter becomes loaded, particles and perhaps pathogens may also break through the Nter, resulting in the initiation of a backwash cycle. The primary focus of this thesis research is to address the passage of C. parvum through filters relative to memements of turbidity, particle counts, and potential surrogaîes (polystyrene microspheres and B. subtilis spores) during various phases and events occurrhg throughout typicai filter cycles. A fbll understanding of the ability of filters to remove C. parvurn (or any other pathogen) can only be attained when the reliability of experïmëntally obtained data is understood and quantified. Coliecting C. pawum data and expressing the reliability of those data has been one of the greatest challenges associated with studying and optimizing water treatment processes for the rernoval of this pathogen. Reliable C. parvum removal data are necessary so that they c m be related to on-Iuie performance parameters that c m be measured, such as turbidity and particle counts, that can be responded to in real time. To that end, this thesis research also endeavored to provide a reiiable analytical method for evaluating C.p a r w m concentration and removal during treatment process challenge studies. The reliability of these data was demonstrated with a quantitative tool developed for assessing the reliability of C. parvum data. The relationships between on-line performance parameters and oocyst removal were evaluated and contributed to the development of practical treatment strategies for maximizing C. p a m m removal by filters. Since surrogate parameters could also be very useful in developing such operational strategies, potystyrene microspheres and B. subtilis spores were also assessed as potential surrogates for C. pawum removal by filtration. The overall goal of this research was to evaluate the impact of filter design and operational parameters on pathogen removal during drinking water treatment so that conceptual mechanistic models for maxuniPng pathogen removai by filtration could be developed and incorporated into practical treatment strategies. Specific objectives in pursuit of this goal were: 1. To implement a reliable, state-of-the-art method for concentration and enurneration of C. p a m m oocysts fiom water during treahnent process challenge studies d u ~ g wliich oocysts are spiked into the treatment process. 2. To deveiop a quantitative statistical tool that describes the reliability of, or uncertainty associated with, C.parvurn (and other discrete particle) concentration and rernovai data. 3. To determine if the removal of formalin-inactivated C. pamrm oocysts by filtration is a reliable surrogate for the removal of viable oocysts. 4. To evaluate the impact of several design and operational factors on C. parvurn oocyst removd by granular media filters. 5. To evaluate the removal of polystyrene microspheres as surrogates for the removai of C.pawum by ganular media fdtration. 6 . To pmvide practical design and operational strategies for maximizing the removal of C.p a m m by granular media filters. Physico-chemical water treatment processes for the removal of parasitic pathogens such as C-parvtcm have been receiving increased attention because of the diffculty of chemically inactivating such microorganisms. Although numerous investigations have examined the removal of these and other pathogens through fdtration processes, those studies have generally been conducted under optimized operating conditions. Several studies of pawum and surrogate removal by filtration have included some examination of vulnerable operating penods; however, they have been limited in scope and have not focused on thoroughly assessing vulnerable periods of operation and providing design and operational strategies for maximizing C.p a m m removal by filtration. The approach used for the completion of this thesis research was based on defining C.p a m m and potential surrogate removds during vuherable periods of filter operation, relating them quantitatively to removals during stable operation, and i d e n t i w g general strategies for maximizing C.pawum removal by filtration. This research approach, as it addresses the experimental objectives outlined in Section 1.2, is shown in Fiame 1.2 and siimmarized below. After a thorough review- of the relevant filtration and particle removal literature, several key studies were identified which focused on analytical methods for concentration and enurneration of C.p a m m , approaches for expressing C. p a m m concentration data, and removals of C. parvtlm and potential surrogates by granu1a.r media filtration. A wide range of methods and recoveries were identified, demonstrating the need for a statistical k e w o r k for incorporating analytical uncertainty into assessrnents of C. p a m m removal by water treatment processes. Limited information regarding C. pawum removal at various points in the filter cycle or during sub-optimal operating conditions was available, emphasizing the need for investigation of design and operational effects on C.p a m m rernoval by filtration. \ Development of Design and Operational Strategies for Maximizing C. parvum Removal by Filtration (OBJECTIVE #6) Figure 1.2 Research approach. Prior to actuaily investigating C. pawum removals by fdters, it was necessary to develop anci implement a reliable oocyst concentration and enurneration method. Given the ultimate goal of assessing oocyst removal effxciencies of filters at several operating conditions, it was considered critical that oocysts be present in both the filter influent and effluent in reliably countable concentrations. Kigh influent oocyst concentrations were particularly critical for evaluating the oocyst removal capacity of fdters under optimal operating conditions when, as indicated by the literature, filter effluent oocyst concentrations were expected to be low. Therefore, it was essential to have high concentrations of oocysts in the filter influent during the thesis investigations. Analyticai methods for the concentration and identification of C.parvurn are generally laborious and fraught with uncertainty. In the present research, a method reported in the literature was optimized so that it could be used in conjunction with a seeding protocol that provided high enough filter influent oocyst concentrations so as to ensure reliably counîable numbers of oocysts in the filter effluent. The burden of implementing a reiiable analytical method for enumerating C. parvzrrn during such studies was considerably Iess than it would have been for evaiuation of indigenous concentrations. Seeding C. parvurn oocysts into the filter at a hi& concentration (several orders of magnitude higher than indigenous concentrations) allowed for the collection and processing of small water sample volumes. Although this approach necessitated the assumption that filters would remove comparable levels of oocysts regardiess o f influent concentration, it was the on1y approach feasible given the available state-Of-the-art analytical techniques for C, p a m m enurneration. Large-scale filtration investigations with viable C.pawum were impractical because of the heaith risks associated with the handling and discharge of such a pathogen. It has typicdy been assumed that inactivated oocysts are suitable surrogates for viable oocysts during Ntration investigations, however, it has also been demonstrated that chemical inactivation changes the surface charge properties of oocysts, thereby possibly affecting removal by filtration. The fïrst experimental phase of this thesis research was designed to demonstrate whether or not comparable removals of viable and formalin-inactivated oocysts could be expected by filtration at various operating conditions. The second experimental phase of this research represented a large majority of the experimental work and was designed to investigate design and operational strategies for maximizing C.panrum removal by fdtration. This research component was crucial to establishing the extent to which elevared turbidity or particle counts in filter effluents might be accompanied by increased C. parvum passage through filters. Experiments were conducted at the Ottawa, Windsor, and University of Waterloo pilot plants, The multiple research platfoms permitted investigation of different types of raw waters, water temperatures, coagulation regimes, and filter designs. Formalin-inactivated C-parvurn oocysts were seeded at ail of the experimental locations. Polystyrene microspheres were evaluated as potential surrogates for C.parvum because they were similar to oocysts in size and easy to ident* and enurnerate. Microspheres were seeded at Ottawa and the University of Waterloo. Microsphere rernovals by filtration were cornpared to B. subtilis spore removals by filtration and total particle reductions through the treatment process, both of which are indicators of treatment performance, but not qualitative surrogates for C. panum removal by filtration. The experimental data were evaluated using a statistical approach that was developed during the course of this thesis research. The statistical approach incorporated analytical recovery and uncertainty of recovery and yielded confidence intervals on both measured C. parvum concentrations and rernovals by filtration. These data were then evaluated relative to traditional performance data (e-g.turbidity and particle counts) to provide clear design and operational strategies for maximiziog C. p a m m removal by filtration. The overall research approach employed in this thesis was designed to provide clear outcomes of practicai value to the water industry while elucidating some of the fundamental mechanistic processes g o v e r h g the removal of rnicrobiological particles such as C. parvum by filtration processes. A more detailed discussion of the experimental design and statistical fkarnework can be found in Chapters 3 and 4, respectively. Chapter 8 provides a summary of the practical outcomes and recommendations for the water treatment uidustry. Chapter 2 presents background information regardhg the significance of the proposed research and the theoretical concepts employed therein. An extensive review of previous research regarding C. pamrm occurrence, detection, and removal during dnnking water treatment has been included in that chapter. Chapter 2 concludes with a surnmary of the limitations of current knowtedge regarding the removal of C. parwrrn by filtration. Chapter 3 outlines the experimental design, research platforms, and analytical methods employed dirnng the course of this research. A statistical h e w o r k for describing the reliability of C. parvurn concentration and rsmovai data is developed in Chapter 4. Bench-scale data focused on establishing that formalin-inactivated oocysts are adequate surrogates for viable oocysts are discussed in Chapter 5; a preliminary evaluation of design strategies for maximizing C. p a m m removal by filtration (e-g.,by comparing trimedia to dual-media fdtration) is also incfuded in that chapter. The pilot-scale experiments address the main objective of this research by investigating C.p a m m and potential surrogate (B. subtilis and polystyrene microsphere) removals by various filter designs during several operating conditions; these experiments are discussed in Chapter 6. In Chapter 7, the bench- and pilot-scale C. p a m m and microsphere data are discussed relative to traditional performance measures (turbidity and particle counts) and surrogates in the broader context of water treatment optimization and practice. h Chapter 8, the experimentd data are then integrated to provide design and operational strategies for maximizing C.pavvtrrn removal b y filtration. This chapter also summarizes the conclusions, implications, and contributions of this thesis research to the water industry. 2.1.1 Importance and Epidemiology Pathogenic microorganisms interact with hosts to cause disease. Arnong the most cornmon pathogen-associated illnesses are acute gastrointestinal infections that are typically manifested as diarrhea (Ryan et al., 1994). Only exceeded in fiequency by respiratory tract infections such as the cornmon cold, most gastrointestinal infections do not receive medical attention because they are usuaily self-limiting, often within hours or days (Ryan et al., 1994). However, diarrheal illness is of profound worldwide significance as it is a major cause of morbidity and mortality among infants and children. As many as five billion episodes of diarrhea and five to ten million diarrhea-associated deaths have been reported annually in Asia, Afnca, and Latin Arnenca (Current, 1988). Depending on socioeconomic and nutritional factors, diarrhea-associated death rates of children under the age of 7 can be as high as 50% in these areas (Current, 1988). Mortality associated with diarrhea in developed nations is considerably lower, but still significant (Ryan et al., 1994). Recognized as a major worldwide heakh problem in 1976, Cryptosporïdium infections are the third or fourth most cornrnon cause of diarrhea worldwide (ASM News, 1996). Diarrhea from Cryptosporidium infections can 1 s t fkom three to twelve days in most well-nourished persons (Current, 19S8), however in some cases symptoms can last for more than one year afier the acute phase (ASM News, 1996). Persistent diarrhea can lead to poor absorption of foods and severe malnutrition in children. It has been estirnated that in developing nations such as Brazil, C~ptosponndium infection rates in children may exceed 90% (ASM News, 1996). Cryptosporiàium oocysts have been detected in 14% of gastroenteritis cases in young children in western countries whereas these rates have ranged fkom 4- 11% in developing nations; adults suffer fkom gastroententis at rates approximately one third of those reported in children (Ryan et al,, 1994). Infection rates increase in immunocompromised populations. Cwptospondium has been identified in 15% of American A I D S patients suffering fkom diarrhea and as many as 50% of such individuais in Haiti and Africa (Ryan et al., 1994)- Other enteric pathogens such as Giardia lamblia have been recovered only fiom a minority of these patients (Ryan et al., 1994). Cryptosporidiosis is an important and dangerous disease for both immunocompromised and immunocompetent individuals since it has no effective treatment (ASM News, 1996). 2.1.2 Life Cycle Crptosporidium is an obligate parasite that replicates in the intestinal tract of a wide range of host mammals including humans. Like other sporozoans, these 3-6 p m (in diameter) sphencal parasites exhibit alternating cycles of sexud and asexual reproduction that are completed within the gastrointestinal tract of a single host (Ryan et al., 1994). Parasitized anïmals excrete Ctyptospondium oocysts, which are envkonmentally resilient. Infection is acquired by ingesting oocysts which can release four infective sporozoites that attach to epithelial cells of the gastrointestinal tract and transfonn into trophozoites that subsequently go through merogony (asexual reproduction). The reproductive process involves multiple fission (schizogony) of the sporozoites to form schizonts containing eight daughter cells known as Type 1 merozoites. A second generation of Type 1 merozoites is produced when the daughter cells are released fiom the schizont, attach to fbrther epithelial cells, and repeat schizogony. Figure 2.1 depicts the life cycle of Cyptosporidium. ~porazolte Trophomite Q4 -.--.* ... ---.C Excystatbn A . - Type l meront ---' exit from Figure 2.1 Life cycle of Cryptosporidium (aBer Fayer and Ungar, 1986) Oocysts typically undergo one cycle of asexual reproduction and produce a self-limited diarrhea (Schaechter et al.. 1993). After asexual reproduction, schizonts containing four Type 2 merozoites incapable of asexual reproduction are formed. During gametogony, the Type 2 merozoites undergo sexual reproduction. Following fertilization, the resulting zygotes develop into oocysts that are subsequently excreted (Ryan et al., 1994). Approximately 80% of the oocysts formed from zygotes are of the environmentally resilient nature that, upon finding a new host, undergo the same life cycle described above. The remaining 20% of oocysts form only a thin wall and initiate an autoinfective cycle within the original host. The immunity of imunocompetent hosts dampens the production of merozoites and thin-walled oocysts, stopping autoinfection and terminating the acute infection (Ryan et al., 1994). h the irnmunocompromised, oocysts undergo numerous cycles of sexual and asexual reproduction that may result in severe infections that last indefinitely (Schaechter et al., 1993). 2.1.3 Sources and Occurrence of parvrrm in Water C~ptosporidiumhas woridwide distribution and occurs in several host species including mammals, birds, and fish. Cwptospovidizim p a m m is associated with most human infections and is also common in livestock (Rose, 1988). C.pawtrrn oocysts have been detected in surface waters in concentrations as hi& as 104 /100 L and as Iow as 0.Y 100 L (LeChevallier and Norton, 1995; Lisle and Rose, 1995; Smith et al,, 1991), regardless of whether the waters are pnstine or impacted by h u m a . and animal activity. Several studies have indicated that watershed character and protection influence parasite contamination (Ong et al., 1996; Hansen and Ongerth, 1991). Human cryptosporidiosis is believed to involve both anthroponotic and zoonotic cycles of transmission (Casemore et al., 1997). Transmission in humans is either dircct through the fecal-oral route or indirect through water contamination. runoff are often siWcant Treated wastewater effluents and agricultural sources of oocysts in surface waters (Rose, 1988). At least two distinct genotypes of C. p a m m are currently linked to human cryptosporidiosis (Peng et al., 1997; Patel et al., 1998; Widmer et al., 1998). Peng et al. (1997) demonstrated that genotypes 1 and 2 were infective to humans, but only genotype 2 isolates were infective to mice and cattle under routine laboratory conditions; this demonstrated the occurrence of two distinct C. pamirn transmission cycles in humans. Waterborne outbreaks of cryptosporidiosis have been historically attributed in significant part to agricultural runoff and were considered to be associated with C. p a m m of genotype 2. Several recent investigations have challenged this assumption by demonstrating that a majority of human infections of cryptosporidiosis from waterborne outbreaks were linked to genotype 1 (Patel et al., 1998; Sulaiman et al., 1998; Widmer et al., 1998). In one study from the United Kingdom, however, similar fiequencies of genotype I and 2 isolates from sporadic human infections were reported (McLauchlin et al., 1999). The compelling evidence suggesting that C. pamrm oocysts of genotype 1 are the predominant cause of waterbome outbreaks of h a n cryptosporidiosis in North Amenca may result in increased focus on domestic wastewaters and may lead to some reevaluation of watershed protection strategies aimed at minimizing oocyst contamination of drinIcing water supplies. Methodologies used to evaiuate treatment effkacy for removing or inactivating oocysts may also have to be re-examined. Despite the apparent epidemiological significance of genotype 1, (the predominant cause of human cryptosporidiosis), type 1 isolates are not typically used in removai, viability, or infectivity investigations because of the inability of maintaining them in the laboratory. Although the genotype 1 oocysts may be the predominant cause of waterborne outbreaks of cryptospondiosis, the results of investigations utilizing genotype 2 are not necessarily invalidated. Since isolates of the same genotype can vary significantly in their infectivity in cultured cells and in animal models, it is possible that results obtained with genotype 2 may be similar to those that might be obtained with genotype 1 (Chappe11 et al., 1999). In the future, investigations with genotype 1 may be possible. Recent research has demonstrated the e s t successful serial propagation of genotype 1 C - p a m r n ; type 1 isolates w&e successfülly adapted to propagate in gnotobiotic pigIets (Wïdmer et al., 2000). 2-1.4 Waterborne Outbreaks of Cryptosporidiosis Outbreaks of waterbome cryptosporidiosis have been weU documented in North America and abroad (Fox and Lytle, 1996; Roefer et al., 1996; Welker et al., 1994); drinking water has been implicated in several worldwide outbreaks. The most famous outbreak was the 1993 Milwaukee outbreak, during which more than 400,000 people were affecte& making it the largest outbreak of waterbome disease ever recorded in the United States (Solo-Gabnele and Neumeister, 1996). The Las Vegas outbreak was also significant since 20 imrnunocompromised individuals died as a result of infection (Roefer et al., 1995; 1996). Outbreaks of waterborne cryptosporidiosis are of interest to the water treatment industry because of their regulatory and engineering implications as well as their public heaith consequences. During the documented outbreaks in the United States, each of the implicated water treatment plants was in compiiance with federal and local regulations (Solo-Gabriele and Neumeister, 1996). Table 2.1 summarizes some of the documented waterborne outbreaks of cryptosporidiosis. Table 2-1 Waterborne Outbreaks of Cryptospondiosis Location Braun Station, TX Est, # of People Affected (Confirmed) 5900 (2006) Source Water Water Treatment Strategy ground chlorination Cobham, Surrey, UK 1983 1985 spring chlorination, softening Sheffield, UK 1986 surface conventional chlorination, filtration minirnaHy NA Sucface Loch Lornmond, UK 1990 surface Isle of ?hanet, UK 1990 surface D'Antonio et al. Barer and Wright ( 1990) cattle feces in runoff ( 1995) Lisle and Rose raw senage. mttle runo% ( 1989) Hayes et al. pipe cross connection, cattle feces Smith et al. ( 1989) fiIter BW recycle. cattle feces Richardson et al. (1991) NA Barer and Wright ( 1990) Lisle and Rose ( 1 995) ground Jackson County, OR !992 sewage contaminated well Reference Lisle and Rose ( 1995) river South London, UK 1991 Suspected Source of Oocysts chlorination treatment deficiencies Joseph er al. (1991) septic tadi effluent. creek Moore et al. ( 1994) - Magoire et al. ( 1995) river, runoff, wastewater vrin%' river lake conventional Kitchener-Waterloo, ON 1993 river, pond conventional + biofiltration Yakima County, WA 1993 well - Leland er al( 1993) cattle /slaughter waste, sewage runoff - elk, cattle, sheep Weiker er al. ( 1994) Solo-Gabriele and Neurneister (1996) Table 2.1 Waterbome Outbreaks of Cryptosporidiosis (Continued) Est, # of People Location Source Water Cook County, MN 1993 NA (108) lake Las Ve,p, NV NA ( 13W) Iake 1994 Waila Walla Conty, WA 1994 86 well (15) South and West Devon, UK (575) 1995 Water Treatment conventionai, pre- and postchlorination - river and prave1 wells Suspected Source of sewage or septic tank effluent treated wastewater. boat sewage Reference Solo-Gabnele and Neumeister (1996) Roefer et al. ( 1995. 1996) treated wastewater Solo-Gabriele and Neumeister ( 1996) treated wastewater C~ptosporidium Capsule ( 199Sa) Patel et al. ( 1998) - C~ptosporidium Capsule ( 1997a) Pozio et al. ( 1997) C~ptosporidiitm Capsule (1996a,b) Ogose, Japan 1996 C'ptosporiditrrn Capsule ( 19 9 6 ~ ) Cryptosporidium Capsule (1996e) London and Hertfordshire, (345) UK treated wastewater Patel et al. ( 1998) 1997 North Thames, England 1997 244 ground filtration and ozonation unknown C~prospovidium Capsule ( 19966 1997b) Shoal Lake, ON 1997 1O0 (Il) lake chlorination &own Cryprosporidiurn Capsule ( 1997~) Bmshy Creek, TX 1998 1300 (32) ground Sewage contamination of creeWweIls Cntprosporidium Capsule ( l998b) NW England UK 1999 360 &ce unfiltered - C~ptosporidium Capsule ( 1999a.b) 2.2 RECOVERYAND DETECTION OF CRWTOSPOR~>IUM FROM WATER Analytical methods for quantification of C.parmm oocysts fiom water were originaliy developed fiom those for Giardia. Most of these methods require three steps: concentration, purification, and enurneration. Oocyst concentrations are typicaliy low in natunl waters, necessitahg the concentration of oocysts fiom large volumes of water to achieve consistently countable numbers. Typicaliy at least 100 L of source waters and >1000 L of finished waters are recommended for C. p a m m 1996). anaiyses (Jakubowski et al., Oocyst concentration is typicdy achieved via some form of cartridge or membrane filtration under mild vacuum pressure (ASTM 1993; USEPA, 1999a). Tt can be difficult to distinguish between oocysts and other particies. Since oocysts represent a srnall fkaction of the particles and microorganisms found in natural waters, purification steps are often incorporated to separate oocysts fiom the remaining debris in water concentrates. One of the challenges associated with detection of C.parvrrrn has been poor and highiy variable oocyst recovery (CIancy et al-, 1994). Oocyst purification becomes additionally important because the analytical recovery of seved common detection methods is greatly impacted by the amount and type of debris present (LeChevallier et al., t 990; 1991a,b,c). Flotation by density gradients (Hansen and Ongerth, 1991; LeChevallier et al., 1995) and flocculation (Vesey et al., 1993a) have been used for oocyst separation from concentrated water samples. More recently, immunornagnetic separation has been dernonstrated as a more reliable method of oocyst purification (Rossomondo et al., 1994; Bukhari et al.,1998). Once oocysts are concentrated Immunofiuorescence assays (IFAs) and purified, commody they used must to be detect enumerated, and quant@ Cwtosponndiumoocysts allow good discenunent between oocysts and other particles. Both monoclonal and polyclonal antibodies (LeChevallier et al., 1995; Ongerth and Stibbs, 1987) have been used successfùlly in direct and indirect IFAs for C. pavvurn (Musial et al., 1987; ASTM, 1993). In these IFAs, the FITC (fluorescein isothiocyanate) label appears bright apple green during epifluorescence microscopy (Nieminski et al., 1995). Some of the difficulties associated witb IFAs can include background fluorescence and false positives (Rodgers et al., 1995). FITC-IFAs only identm oocysts; they do not indicate whether or not the oocysts are viable and infectious. Vital dye (Roberston et al,, 1992) and in vitro excystation (Robertson et al., 1993) assays have been developed to indicate oocyst viability, but they do not measure infectivity. Finch et al- (1993) demonstrated that the analyticai methods with which oocyst inactivation and infectivity are detennined could potentially underestimate assessments of inactivation. Their work contributed to the general acceptance of in vivo b a l infectivity assays as the standard method for assessing C.pawum inactivation. Several cornparisons of in vitru methods such as excystation and DAPUPI (4',6-diamidino-2-phenylindole/propidiumiodide) to in vivo animal ùifectivity have revealed that in vitro assays often over-estimate oocyst viability compared to in vivo infectivity data (Bukharï et al., 1999). One of the first C.pawtrrn concentration and identification rnethods was proposed by the Amencan Society for Testing and Materials (ASTM) and involved the passage of large volumes of water through a polypropylene yam-wound carûïdge filter, notation on a Percoll-sucrose gradient, and IFA enumeration (ASTM, 1993). The literature strongly and frequently indicates that the ASTM rnethod is inadequate because it typicaily results in h i m y variable and Iow oocyst recoveries (Hargy et al., 1996). LeChevallier et al. (1995) examined the ASTM method and revealed that the centrihgation and clarification steps could each result in losses as high as 30%. Fuaher difficulties with the ASTM rnethod can result fiom the presence of algae because numerous species can fluoresce and result in false positive counts (Rodgers et al., 1995). Despite the lack of a more accurate and consistent method, the ASTM method was incorporated into the United States Environmental Protection Agency's (USEPA's) Information Collection Rule (ICR) as the standard protocol for C. p a m m detection and enumeration from water (USEPA, 1996). Several alternatives to the ASTM protocol have been exarnined (Nieminski et al. 1995; Whitmore and Carrington, 1993; Vesey et al., 1993a). Most notably, the USEPA introduced Method 1622 which requires filtration, immunomagnetic separation of oocysts, and an IFA for detennination of oocyst concentrations, with c o ~ a t i o n through DAPI stainllig and differential interference contrast (DIC) microscopy (USEPA, 1999a). Almost identical to Method 1622, Method 1623 was introduced for simultaneous detection of Cryptosporidiurn and Giardia (LISEPA, 1999b). Method 1622 is a significant improvement over the K R method and demonstrated mean oocyst recovenes of >70% in early trials (Clancy et al., 1997; Bukhari et al., 1999). Other studies of Method 1622, such as the USEPA validation experiments, have yielded recoveries of approxhately 35% with 13% relative standard deviation (Clancy et al., 1999). More detailed discussions of analytical methods for the detection and enumeration of C. parvum are available in the Ziterature (Jakubowski et al., 1996; Vesey et al., 199 1). Due to continued Iow and non-constant analyticd recovenes, evaluations of C.parvurn removal by water treatment processes usually involve seeding the water to be treated with oocyst concentrations higher than those typicdy present in raw water. This approach is analyticalIy advantageous because it allows for s m d e r sample voIumes that do not necessarily require purification. Seeding hi@ concentrations of oocysts can also be critical to accurately assessing the oocyst removal capacity of treatment processes, which requires oocysts present in reliably countabIe concentrations in the effluent. Log removal estimates may be Iimited by influent concentration when effluent concentrations are low or based on non-detects because treatment processes cannot remove more oocysts than those with which they have been spiked- Cornparisons between studies should be evaluated with caution because oocyst recovery is sensitive to seeded concentration (Straub et al., 1996; Musial et al., 1987), oocyst quality (Dawson et al., 1993), sample volume processed (Nieminski et al., 1995), flow rate (Musial et al., 1987), and water q d t y (Shepherd and Wyn-Jones, 1996). Ensuring that spiked oocyst concentrations are accurately measured contributes to the consistency of analytical recoveries (Straub et al., 1996). Several methods for deteminkg spike doses of Cryptusporidiurn and Giardia have been evaluated. Arnong the methods evaluated by Straub et al. (1996) were well-slide IFA and hemocytorneter counts. Theoretically, hemocytometer counts should provide the most accurate and consistent results because sample processing is not necessary and detection does not rely on staiaing. Despite initial parity between the two methods, the authors concluded that well-slide IFA provided the most accurate results; however, they speculated that the unexpected relative performance of the hemocytometer was likely due to an additional sample transfer and holding step (Straub et al-, 1996). A second study with consistent sample handling indicated that hemocytometer counts generally yielded higher recoveries and better precision than well-slide IFA counts; for any gîven trial, no significant differences among analysts were observed, regardless of the counting method (Jackson et al., 1997). A drinkùig water limit of approximately one Cgptosporidium oocyst per 34,000 L (one per 34 m3) has been suggested (Lisle and Rose, 1995). Such a low limit, or even one several orders of magnitude higher, would be diffîcult to implement because curent methods for measuring Cqptosporidium concentrations are unreliable, Iabonous, and expensive. Methodological difficulties are likely more pronounced when attempting t O quanti@ indigenous concentrations, since high concentrations of oocysts (typically 103- 106oocysts/L) are commonly used in experimenta.evaluations of treatment processes. To date, few studies of Cqptosporidiurn concentration and removal efficiency have addressed the issue of data reliability beyond stating average analyticai recovery. The lack-of information regarding the reliability of Ctptospondium data is likely associated with the expense and difficulty of processing samples. It is also commonly understood that means and standard deviations calculated fiom replicate samples do not account for al1 of the uncertainty in the data. For example, they do not account for variability resulting fiom sarnpling strategy or analytical methodoiogy. The complexity of incorporating these additional sources of variability into a statement of data reliability such as a confidence interval is increased by the fact that the normal distribution is often inappropriate for describing distributions of microorganisms such as Cwptosporidium in water samples. Haas and Rose (1996) showed that naturally occurring oocyst densities could be described adequately by the Poisson distribution, as would be expected for a sample ftom a uniform suspension of such microorganisms in water that was enumerated by an ideal method, Atherholt and Korn (1999) presented Poisson methods for sample counts in the context of the ICR protocol and suggested that a more complex distribution than the Poisson distribution may be requùed to account for the various errors in the analytical process. Nahrstedt and Gimbel (1996) developed such a statistical fhmework for caiculating confidence intervals by assuming a Poisson distribution for the true sample counts, a binomial distribution for modeling the recovered fiaction of oocysts, and a Beta distribution for describing the uncertainty of recovery. They also noted that the time and location of sampling influence error, but suggested that this contribution cannot be statistically deterrnined. As defmed primarily by Nahrstedt and Gimbel (1996), the statisticalli describable errors that influence the concentration and enurneration of Cryptosparidium (and other discrete particles) are summarized below. 2.3.1 Representative Sampling The nurnber of oocysts in a sample fiom a unifonn suspension of oocysts in water should have a Poisson distribution because Poisson events are those occ&g in time or space where the probability of success (or occurrence) is independent and constant in each unit of t h e or space (BIom, 1989). According to the Poisson model, !te probability of N oocysts occuning in a sample is, where A is the expected value or mean of the distribution. nie parameter ;irepresents the tnie mean count for the population and can be thought of as the product of the true concentratiorr of oocysts (c) in the water body and the volume of water sample evaluated (VI given by, The Poisson distribution (Equation 2.1) describes random sampling error resdting fkom the fact that the entire water body is not evaluated. In other words, A oocyst count that wodd be expected if an i d h i t e number of. -dom represents the samples were collected. This was descnbed by Nahrstedt and Gimbel (1996) and is illustrated in Figure 2.2, where a true concentration of 1 oocyst/500 L of water is assumed. If three random 500 L samples were collected, the actual number of oocysts (N) in each sample might be different; in this case, NI = 1 oocyst, N2 = 2 oocysts, and %1 =O oocysts. As the number of samples approaches infinity, the expected value (calculated by Equation 2.2) approaches 1. The Poisson approach may be Limited in applicability because the tnte oocyst concentration (c) is not known and rnethodological losses cannot be accurately incorporated because loss is sample specific. Deviations fkom the Poisson distribution cm occur due to factors such as non-representative sampling or imperfect analytical methods. A relevant example of deviation from a Poisson distribution is non- independence of observations due to oocyst aggregation. - c = 1 oocyst150o L Cmwaw*mmmmm.mmmmm.mmmmmmmmmmmmmmmmmmmmmmmmmmm' Figure 2.2 Sarnphg from a water body (ajier Nahrstedt and Gimbel, 1996) Observed Crptosporidium counts have been adequately represented by a Poisson distribution in at least two cases (Haas and Rose, 1996; Parkhurst and Stem, 1998). However, it has also been suggested that the Poisson mode1 may not be directly applicable for real data sets where methodologica! error may play a substantiai role (Nahrstedt and Gimbel, 1996). Since analyticd methods for the detection of Cryptosporidium are ofien highly uncertain over a range of recoveries, several researchers have suggested that a distribution more complex than a Poisson distribution rnay be necessary to account for methodologicd error (Nahrstedt and Gimbel, 1996; Atherholt and Kom, 1999). 23.2 Random Analytical Error Given an imperfect analytical method, the oocyst counts observed afier Cryptospodium processing represent a portion of a sample's true counts. Only the recovered h c t i o n of oocysts @) is observed under the microscope (Figure 2.3). The number of observed oocysts is a sub-sample from the original Poisson distribution of oocysts in the water sarnple. The probability of observing oocysts in the sub-sample follows a binomial distribution with probabilityp that each of the ori,@nal oocysts wili be observed (Fisher and van Belle, 1993; N h s t e d t and Gimbel, 1996). concentration abelling Figure 2.3 Effect of sample preparation on nurnber of obsenred oocysts (after Nahrstedt and Gimbel, 1996) The probability of detecting X oocysts from a water sample containing N oocysts by using an ana.iytica.1method with recoveryp is described by the binomial distribution Parkhurst and Stem (1998) demonstrated that a binomially distnbuted sub-sample drawn fiom a Poisson-distributed sample also has a Poisson distribution. This observed Poisson distribution has a mean (&,) described by, which is similar to Equation 2.2. Based on this analysis, the mean count of observed oocysts can be scaled to account for a constant recovery ofp. Numerous investigators have reiterated the fmding that oocyst recovery can be afTected by numerous factors such as water quality for several +xi.lytical methods (Shepherd and Wyn-Jones, 1996). Therefore, recovery studies to determine p should always be performed under conditions comparable to experimental conditions. accounting for ma& This includes effects (e-g., turbidity, presence of coagulant, etc.), using the same sample volumes, and q u a n m g the recovery of anticipated counts. When performing recovery studies, the initial mode of oocyst quantification must also be considered reliable and reproducible. Precautions should be taken tu minimize oocyst losses during dilution steps pnor to the processing of the samples used in the recovery study. The most common methods for quantifying oocysts (when they are present in high concentrations) are hemocytometer counts and weU-slide counts employing an IFA (Straub et al., 1996; Jackson et al., 1997). Due to the numerous and complex individual steps associated with oocyst concentration and enurneration, it is often difficult to maintain highly reproducible oocyst recovery. The non-constant analpical recovery of oocysts may therefore require description beyond that of a single value. 23.3 Non-Constant Analytical Recovery Nahrstedt and Ghnbel (1996) addressed the issue of non-constant analytical recovery by using a Beta distribution for the uncertainty (or scattering) of recovery. They introduced two additional parameters ( v and w) to account for the scattering of recovery. Atherholt and Kom (1999) pointed out that the experimental and mathematicai determination of these parameters was not developed in Nahrstedt and Gimbel's (1996) research. In the present discussion, the Beta distribution parameters are referred to as a and 6. These parameters are constants that affect the spread and height of the Beta distribution. The probability of recovery @) depends on these parameters and is described by, where l? represents the gamma fiinction. The Beta distribution is useful for describing oocyst recovery because the recovery is bound between O and 1 (i-e., OS@) and the distribution is very flexible, allowing for the description of a variety of recovery profdes. Figure 2.4 describes the effects of two combinations of a and b on the recovery probability density function; the Beta parameters are those that Nahrstedt and Gimbel (1996'1 used to describe the recovery profiles of LeChevailier et al. (1991~)and Vesey et al. (1993a). As can be seen in Figure 2.4, larger values of a and b result in more narrow distributions. When a>b, the distribution is skewed to the right; when a<b, the distribution is skewed to the lefi. Therefore, analyticai methods that consistently demonstrate high recoveries will be described by Beta distributions with large values of a and b where a%. O 10 20 30 40 90 60 70 80 90 1O0 Recovety (96) Figure 2.4 2.4 Effect of Beta parameters a and b on recovery probability density h c t i o n TREATMENT OPTIONSFOR CRYPTOSPOMDZUM Studies of C. parvum inactivation have indicated that, compared to Giardia cysts exposed to the same disinfection conditions, oocysts were 30 times more resistant to ozone and fourteen times more resistant to chlorine dioxide disinfection, while chlorine and monochloramine were found to be considerably less effective (Korkh et al. 1990). Other studies have concluded that C ~ t o s p o r i d i u moocysts were approxirnately ten times more resistant to ozone inactivation than were Giardia cysts (Owens et al. 1994). Regardless of the magnitude of the increased level of disidection necessary for oocyst inactivation relative to cyst inactivation, it is generally accepted that ozone is the most effective oxidant used against C. p a m m in water treatment disinfection. As a sole barrier, however, ozone does not provide adequate inactivation of C.p a m m oocysts at practical operathg conditions. Finch et al. (1994) and Gyürék et al. (1996) indicated disinfectant syner-q when the sequential application of ozone and chlorine species was employed. These studies concluded that previous research based on the disinfection capabilities of chemicals used singly grossly underestimated the disuifection efficacy of chemicals used sequentially, which is often done in water treatment practice. The data indicated th& the current practice of post-ozonation chloramination rnight provide a partial disinfection barrïer against C. pawurn. UV disinfection technologies have demonstrated over 3-log (99.9%) inactivation of Cryptusporidium (Campbell et al., 1995; CIancy et al, 1996; Clancy et al., 2000), but their operation is most effective in very clean waters such as filter effluents (Bukhari et al., 1999). The use of UV for disinfection of unfiltered waters, however, may not be consistently effective (Bukhari et al., 1999). Despite progress in the development of dissection technologies that can achieve reasonable C. parvum inactivation, such as UV, the traditional physico-chernical barriers used in drinking water treatment rernain critical to achieving desirable Ievek of Cryptosporidiurn removal. Raw water storage reservoirs are often the first physico-chernical barrier against Cïyptosponditrm and other microorganisms during drinking water treatment. Bertolucci et al. (2998) demonstrated that approximately 0.7-log removal of oocysts could be achieved in a reservoir with a theoretical detention t h e of 18 days. Studies of reservoirstored river water indicated 1.3- and 1.7-log removal of oocysts stored for 10 and 24 weeks respectively (van Breemen et al-,1998). Such studies indicate that reservoirs cm act as a barrier against Crptosporidium passage into water treatment plants. Coagulation and clarification processes provide another physico-chemical barrier against Cryptosporidium. Several bench-, pilot- and full-scale studies have demonstrated that conventional coagulation and sedimentation processes can achieve approximateIy 0.5- to 2-log removal of oocysts, with average removals in the 0.5- to 1-log range (KeUey et al., 1995; Baudin and Laîné, 1998; Fox et al., 1998; Dugan et al., 1999). Coagulation coupled with dissolved air flotation clarification has demonstrated 0.6- to >3.1-log removal of oocysts (Edzwald et al., 1996; Edzwald and Kelley, 1998; EdzwaId et al., 1999). Coagulation combined with other clarification systems such as solids blanket and bdlasted clarifiers has respectively yielded 1.8- to 3.7- and 2.5- to 4.2-Iog removal of oocysts (Alvarez et al., 1999). Whde optimized coagulation and clarification contribute a substantial barrier agaulst Clyptospondiurn, fdtration is one of the most critical and successfül physico-chernical barriers against Cwptosponndiurnpassage through water treatment. Although pressuredriven membranes offer excellent removal capabilities of parasitic pathogens, they are not necessariiy economical for treatment of more than 1 to 5 MGD (e-g.,Wiesner et al., 1994). A component of conventional water treatment operations, rapid granular media fdtration has demonstrated eEicacy as a barrier against Cptosporidium. Full-scale Crptosporidiurn removals from 2- to >4-log have been reported in the literature (Baudin and Laîné, -1998; Nieminski and Ongerth, L995). Consideration of the particle removal mechanisrns of fdtration is a reasonable starting point for evaluating pathogen rernoval by filtration because the particles found in naturd water supplies occur in a wide variety of shapes and sizes; they typically include microorganisms, algae, and clay (Montgomery, 1985). Particle removal during filtration occurs when particles deviate fkom the fluid streamlines due to gravitational forces, difision gradients, and inertial effects of momentum (OfMeliaand Stumm, 1967). The relative impact of these effects depends on water quality and physicd charactenstics of the particlles and filter media. Particle removal can occur either by strainuig or attachrnent (Montgomery, 1985). Particles larger than the pore space of media can be removed by straining. ParticIes smaller than the typical pore distances between media grains can also be removed by straining as particles collect and close the pores. Upon consideration of the relative dimensions of representative particles, collectors, pore sizes, and particle-collecter separation distances, Amirtharajah (1988) concluded that particle removal during filtration is predominately a function of non-straining, rather than straining, mechanisms. The non-straining mechanisms of w u l a r media filtration involve three distinct steps: transport, attachment, and detachment. These mechanisms highiight the physico- chemical nature of the filtration process. The physical nature of filtration .requires the transport of suspended particles to the vicinity of the filter grain (collector). The chemical nature of filtration is evidenced in the ability of particles to attach to the surface of collectors. Particles detach when adhesive forces are exceeded by shear forces (that increase as a filter clogs). Attachent of a given particle can occur more than once, however, and particles may reattach at M e r depths within the filter. Particle capture during granular media fdtration can be estimated fkom knowledge of particle mechanics under the influence of hydrodynamic and physicochemical forces in porous media. The mechanistic basis for removal by particle deposition requires particle transport to distances close to collectors (i-e.,media grains, with or without previously deposited particles) where attachment can occur (O'Melia and Stumm, i 967; Elimelech, 1991). Particles deposited in filters c m also detach and be transported to the bullc fiuid. Detached particles rnay re-attach at M e r depths or pass through the filter and appear in the filter effluent (Amirtharajah, 1988). Figure 2.5 schematically describes the mechanisms of filtration. 2.5.1 Transport O'MeLia (1985) and Amktharajah (1988) provided thorough summaries of fltration theory. The mechanistic basis for particle removal fiom suspensions requires that particles transported put collectors must deviate fiom the fluid s~eRmlinesin the b d k of the fluid to distances close to the collector surfaces where attachment is possible. Described schematically in Figure 2.6, particle transport mechanisms include hydrodynamic action, difision, sedimentation, and inertia. Though sometimes listed as a transport mechanism and included in this figure, interception is more appropnately considered a boundary condition that results in deposition. Iuid Streaniline Fiawe 2.5 Filtration mechanisms (a@ Hydrodynamic Diffiision Amirtharajah, 1988) Sedimentation Inertia Interception D f i s i o n and sedimentation are the dominant transport mechanisms during water filtration (Ives, 1982). Diffusion resu1ting fiom Brownian motion is relevant to particles less than 1 pm in size, wnereas sedimentation fiom gravity and the associated particle settling velocity is more significant for particles greater than 1 pn in size (Amirtharajah, 1988)- The combination of these hvo transport mechanisms yields a minimum net transport efficiency for particles that are approximately 1 CM in size (Amirtharajah, 1988). Given that Cwptosponditrm oocysts are typically 3-6 pm in size, they- are near the minimum net transport effrciency. Amiaharajah (1988) incorporated interception as a boundary condition for attachent resulting fiom diffusion and sedimentation. Interception occurs when a particle dong a streamline is close enough to a collector for attachment to occur. Although hydrodynamic action has not yet been quantitatively described, it involves particle movement across streamhes and is a function of particle shape and interaction in the fluid field. Inertial effects are negligible during water fdtration (Ives, 1982). O'Melia and Stumm (1967) f ~ spresented t the concept of trajectory analysis for water filtration. The fundamental principle of trajectory analysis views granula beds as an assembly of collectors ont0 which deposition occurs as a suspension flows past or through them. Tien and Payatakes (1979) specified that the analysis requires knowledge of the geometry and size of the collectors, the flow field around and thruugh the collectors, the nature and magnitude of the forces acting on the particles in the suspension, and the criteria for particle attachment. The concept of paaicie collection efficiency (q) (for a single collector) is defined as the total rate of particle contact with a single collector divided by the rate of particle flow toward the projected area of the collector. n i e transport mechanisms of diffusion, sedimentation, and interception can be described in terms of their single-collecter removd efficiency that is the surn of the individual collector efficiencies. Physical parameters impacting particle transport (e.g., particle size and density, media size, fluid temperature, and filtration rate) were incorporated into several theoretical trajectory theory models (Levich, 1962; Yao, 1968; Yao et al., 1971; Rajagopolan and Tien, 1976; etc.) which have been supported with experimental data (Rajagopolan and Tien, 1976; Tien and Payatakes, 1979). These transport models (and others) are not alone able to precisely predict iilter performance, horvever, because they assume particle destabilization such that there is no repulsive potential between the particles and collectoa. Vaidyanathan and Tien (1988) indicated that collection efficiency under a repulsive double layer potential shows a gradua1 decline rather than the sudden decline predicted by trajectory theory (Payatakes et al., 1974), emphasiang that trajectory analysis is Limited in its ability to predict the results of filtration with unfavorable surface interactions. 2.5.2 Attachment Particle attachent to either a media grain or previously retained particies depends on the surface properties of these materials. O'Melia and S t u i n m (1967), Tobiason and O'Melia (1988), and Raveendran and Amirtharajah (1995) specified seveml mechanisms that affect attachent during filtration. Summarized in Figure 2.7, they include: London-van der Waals (LVDW) forces, electrical double-layer (EDL) forces, hydrodynamic forces, stenc forces, Born forces, structural forces, and chernical or brid-&g forces. The LVDW and EDL forces are considered long-range forces (Raveendran and Amirtharajah, 1995). Due to interactions between electronic dipoles of the surfaces and the solution, the LVDW force is typically attractive in aqueous systems (Tobiason and O'Melia, 1988). Spielman (1978) showed the increased importance of London-van der Waals forces at srnail separation distances between particles and collectors. As particles approach collectors, hydrodynamic retardation slows them down and rvould preclude attachent without these attractive forces. Van der Waals O / media +Ac Elestrical -7- Double-Layer + elecirun douds approach and overlap resufting in repuision result of interactions m e e n electronicdipoles of surface and soluüon moleailes ,E{e.g. solvation or 9 , hydratiw forces - Sîrucbral cause cepulsion H\ CI \ atbactionlrepulsion due to diffuse ion atmospheres hydrautic shearing forces can result in detachment Hydrodynamic esuk of chernical interactiw ion exchange. hydrogen bonding, and formation of coordinative bonds and Iinkages. Bridging ... media / Steric media Figure 2.7 Mechanisms affect ing attachent during filtration interactions beîween The EDL force resuits fiom the two difise ion atmospheres surrounding suspended particles and collectors; these atmospheres interact as particles approach collectors. This interaction is attractive when the double layers are of opposite charge, otherwise the interaction is repulsive (O'Melia and Stumm, 1967). The magnitude of the EDL force depends on the separation distance between the paaicle and coliector, the ionic stren,&, and the potential or charge at each surface (Tobiason and O'Melia, 1988). The steric interactions of adsorbed macromolecules can also result in similar repulsive forces (Tobiason and O'Melia, 1988). Water passing through filter media exerts hydrodynamic forces, which c m result in the detachment of particles. Hydrodynamic forces consist of lift and drag forces (Raveendran and Amirtharajah, 1995). Detachment occurs when these forces exceed adhesive forces. Specific modes of detachment are discussed below (Section 2 - 5 3 ) . Born and structural forces are considered short-range forces that are necessary to explain the initial events of detachment (Raveendran and Amirtharajah, 1995). Born forces are repulsive and result f?om the overlap of electron clouds; they determine how close two atoms or molecules can ultimately approach one another. Between particles, this repulsive interaction c m be determined by summing the individual interactions between molecules or atoms (Raveendran and Amirtharajah, 1995). At short ranges, structural forces are often much stronger than LVDW and EDL forces; they arise fiom disruption of the ordering of Iiquid molecules during the approach to a second surface (Raveendran and Amirtharajah, 1995). When water is the medium, they are termed hydration forces and are repulsive whenever water molecules strongly bind to surfaces containing hydrophilic groups (Raveendran and Amirtharajah, 1995). Raveendran and Amirtharajah (1995) demonstrated that the inclusion of such short-range forces in theoretical force calculations helped to qualitatively explain experimental results describing attachment and detachment behavior during filter backwashing. O'Melia and Snimm (1967) pointed out that chemical bridging forces often outweigh electrostatic forces. The forces might include interactions such as ion exchange, hydrogen bondhg, and the formation of coordinative bonds and linkages (O'Meiia and Stumm, 1967). An example of chemical bridging in water treatment is the adsorption of anionic polymers on negatively charged surfaces (BIack et al.,1965). The- detachment process is analogous to attachment in that it can be considered as a detachment step followed by a transport step into the flowing liquid. As particles are removed fiom the bulk flow and attach to collectors, they accumulate and can act as additional collectors (O'Melia and Stumm, 1967). When the magnitude of the hydrodynamic forces (consisting of Lift and drag) on particles exceeds that of adhesive forces, the particles are detached and reach lower depths withh the filter; an avalanche effect of arriving particles may be an alternative mechanism of detachment (Figure 2.8). Attachent and detachment occur simultaneously in filter layers that have reached a saturated but metastable configuration of deposited particles (Arnirtbarajah, 1988). Many studies have demonstrated that optimized particle destabilization minimizes 'premature" detachment (Roebeck et al.,1964; O'Melia, 1985)- > adhesive - f~rces avalanche of amving particles Figure 2.8 collecter (media) Modes of detachment during filtration Ionic strength and solution pH have been identified as significant factors affecting particie detachment when a constant hydrodynamic forcc is applied for detachment (McDowell-Boyer, 1982; Ryan and Gschwend, 1994). By varying the hydrodynamic force, Hubbe (1984) and Sharma e t al.(1992) theoretically and experimentally addressed physical aspects of particle detachment. These investigations demonstrated that particle detachment increased with flow rate and particle size. Interstitial velocities increase as deposits build up in filters, resulting in less effective removal. As deposits build up, detachment ùicreases. Though particles may re-attach at lower depths, ultimately, the filter bed depth is inadequate for providing desirable filter effluent quality and particle breakthrough occurs. During breakthrough there are several possible sources of particles in filter effluents. Particles may pass through the fdter directly from the influent, attach and detach fkom the fdter, or enter the filter at a smaller size, form flocs on the surface of the media, and then detach (Lawler et al., 1995). Graham (1988) defined pore flocculation as inter-particle aggregation an attempted to quanti@ it with a cornputer model, and so concluded that fdter pore flocculation, while appreciable, was less important than particle-grain attachment and particle-particle attachent by previously retained particles. This investigation only addressed pore fiocculation resulting primarily from polymer addition, however. Ginn et al. (1992) speculated that filter pore flocculation can Iikely be ignored in systems exclu.sively using an inorganic coagulant such as alum because particle-grain and particle-particle bonds are weaker In such systems. The particle size distribution data of Ginn et al. (1992) dernonstrated a decrease in removal efficiency of larger particles as particle deposits in the bed increased however, this trend \vas not evidenced in ~ b i d i t measurements. y This fmding was consistent with that of Moran et al., (1993a) who concluded that ripening and breakthrough were strongly dependent on particle size. That investigation demonstrated that while rernoval of smaller particles increased (or npened) for the Iongest duration, removal of intermediate sizes (such as those of Cqptosponndizm)ripened early but decreased substantially as deposits in the bed increased (Moran et al., 1993a). Moran et al., (1993b) also concluded that particle detachment was predominant in intermediate and large particle size ranges. Two mechanisms could have possibly caused the observed particle size increases in the filter effluents: the detachment of deposits from collectors andlor pore flocculation of suspended particles ( G i . et al., 1992; Moran et al., 1993a,b). Based on a substantial investigation of particle size distribution data, Ginn et al. (1992) concluded that detachment was the major source of particles in the studied filter effluents and that the size distribution shifted to larger particle sizes. This conclusion was generally consistent with those of M o r a et al., (1993b), however, the later investigations suggested some detachment of smaller particles as parts of larger flocs. The mechanistic studies conducted by Gllin et al. (1992) and Moran et al. (1 993b) concluded that detachment of previously retained particles or flocs contributed substantially to breakthrough. Moran et al. (1993b) additionally speculated that the input of new particles may have been necessary for breakthrough to occur (Le., avalanche effect described in Figure 2.8). Ginn et al. (1992) proposed a conceptual model of filtration consisting of four phases (Figure 2.9). From a *mechanisticstandpoint, this model was generally consistent with the findings of M o r a et al. (1993a) that indicated that npening and breakthrough were not distinct stages of £ilter operation, but occurred simultaneously for different sized particles. The fmt phase of the G h et al. (1992) model is fdter ripening during which attachment increases as particles are deposited and subsequently act as collectors. Assuming a clean filter with no deposits, there is essentially no detachmznt at the start of a filter cycle; detachment slowly increases as deposits begin to accumulate in the filter. Effective filtration is the second phase of the rnodel during which increasing interstitial flow velocities result in increased detachment and decreased attachent; however, attachment is sufficient to remove many of the influent particles. Effluent turbidity begins to breakthrough during the third phase of the filtration model, as a result of increasing interstitial flow velocities caused by clogging of pores. Attachent continues to decrease while detachment increases. Preferential tubular passages (wormholes) of flow begin to form; because of their hi& pore velocities, these passages result in the significant decrease of particle attachment. Detachment peaks and begins to decrease because of the continued concentration of flow in the wormhole passages in which little new deposition occurs. The fourth phase of the model occurs afler breakthrough, when womhole flow begins to dominate and the filtering capacity of the bed is almost exhausted. At this tirne, attachment and detachment do not occw and fdter effluent turbidity and particles rapidly approach their influent values. Several important ramifications for water treatment may be drawn fiom the reported experimental and theoretical work regarding particle detachment during filtration. The experimental fmduigs of Ginn et al. (1992) and Moran et al., (1993a,b) underscored the inadequacy of particle count data as a sole tool for assessing filter removals of pathogens such as Cvptosporidiurn. The demonstrated deterioration in removals (or early breakthrough) of oocyst-sized particles in the later portions of filter cycles suggested that end-of-nui operating conditions rnay be particularly vulnerrtble in terms of Cryptosporidiurn removal, or more specificaliy, C>?.ptosporidilm detachment (Ginn et al., 1992; Moran et al., 1993a,b). 1 1 1 Effective Filtration Ripening I 1 I i i l Figure 2.9 Breakthrough W o m hole Flow Filter Cycle Conceptual mode1 of attachment and detachment during filtration (modr$edfiom Ginn et al., 1992) Lùnited experimental data suggested the potential for pore flocculation of smaller particles to form intermediate sized particles (Moran et al., 1993a,b), suggesting that Cryptosporiditrm oocysts in filter effluents would not necessarily be found exclusively in the oocyst-sized range of particles (i.e., they could dso flocculate to form larger particles). As pointed out by Ginn et al. (1992), these data also indicated the possibility of passage of destabilized particdate matter through well-operated deep bed filters. Microorganisms bound in such aggregates could be more difficult to inactivate in subsequent disinfection processes that relied on direct contact with the individual microorganisrns. It is very difficdt to determine the source and history o f particles exithg filters. Regardless of the degree to which pore flocculation may occur for certain types of particles, several studies have demonstrated that particle detachment occurs contuiuously and influences particle removal during filtration (Ginn et al., 1992; Moran er al., 1993a,b). Since detachment and attachment are critical to understanding particle and microorganism removal, their respective roles during various operating conditions must be considered when optimizing removal of specific particles such as Cptospondiurn oocysts. The filtration mode1 proposed by Ginn et al. (1992) provides a conceptual fkunework for attachent and detachment mechanisms during filtration that can be generally applied to understanding and predicting operational effects on particle, and more specifically, Cryptosponditrm removal by filters. The operating mode and conditions of water treatment processes can affect particle removd. Both pretreatment conditions (e-g.,coagulation) and fdtration (e.,s, filter aid, hydraulic changes, backwash strategy, media type, ripening filtration, breakthrough filtration) conditions directly impact the particle removal efficiency of filters. Table 2.2 summarizes operational conditions and their impacts on particle removal and passage mechanisms. Table 2.2 Operational Factors and Related Particle Removal and Passage Mechanisms Operational Factors Coagulation Particle RemovaVPassage Mechanisrn Transport d Attachent d Detachent d d Filter Aid Key References Roebeck et al,, 1964 OtMelia and Sturnrn, 1967 O'Melia 1985 Black et al., 1965 Zhu et al., 1996 Hydraulic Changes d Cleasby et al., 1963 Tuepker and Bauescher, 1968 Logsdon et al., 198 1a Sharrna et al., 1993 Backwash Stratew d Amirtharajah, 1988 Raveendran and Amirtharajah, 1995 Coiton et al., 1996 Temperature d Ives and Sholji, 1965 Ives, 1982 O'Melia and Ni, 1978 Amirtharajah, 1985 Moran et al., 1993a Ripening d Media d Ginn et al., 1992 Moran et al.. 1993a,b LawIer et al., 1995 IvesandSholji,1965 TrusseIl et al., 1980 2.6.1 Coagulation Although considerable reductions in both particles and turbidity can occur during filtration without chemical preeeatment, it has been repeatedly demonstrated that proper coagulation is critical in maintaining good particle removal during filtration (O'Melia and Crapps, 1964; Roebeck et al., 1964; Trussell et al., 1980). As discussed previously, several particle attachment mechanisms occur during filtration (O'MeLia and Stumm, 1967; O'Melia, l985), including particle destabilization forces, which are directly affected by coagulation. The relative strength of attachment forces affects the degree of non-attachent that occurs in filters; however, the relative ratio of attachment to detachment (hydrodynamic shear) forces dictates detachment. Chernical parameters such as pH and ionic strength, which are impacted by chemical pretreatment, have been experimentdy identified as significant factors affecting particle detachment in packed bed filters when a constant hydrodynamic force is applied (McDowell-Boyer, 1992; Ryan and Gschwend, 1994; Nocito-Gobe1 and Tobiason, 1995). The impact of coagdation on filtration is considerable because coagulation affects the balance between fundamental aitachment and detachment mechanisms that dictate whether or not particles wiiI be retained in fïiters (attachent), retained and subsequendy released fiom filters (detachment), or pass through fdters (non-attachment). The difficulty in understanding and optimizing coagulant effects during filtration is compiicated by the different roIes of attachment and detachment mechanisms occurring during different periods of the filter cycle such as ripening and breakthrough, which occur simultaneously but at diflerent rates for different sized particles (Moran et al., 1993a,b). In s m e cases, chernical pretreatrnent can affect competing objectives such as extent of ripening period, extent of backwash period, etc. It stands to reason that coagulation would have a similar impact on the removai of pathogenic colloidal particles such as Cryptosporidiurn oocysts. Tne importance of coagulation processes for improving filter removal efficiencies of Giardia cysts has been well documented. Several investigations have demonstrated little ( c l -log) to no removal of Giardia cysts by GAC filters (Patania et al., 1995), sand and dual-media filters (M- Ani et al., 2986), and tri-media faters (Hom et al., 1988) during no coagulation conditions. Even sub-optimal coagulation conditions have been shown to afTect the pathogen removal efficiency of filters. In a pilot-scale direct fdtration plant, Logsdon et al. (198 la) fomd that mean Giardia mun's removals decreased by approximately 1- to 2.5-log durkg periods of sub-optimal and minimal coagulation compared to optimal operating conditions. Similar decreases in cyst removal as a result of sub-optimal coagulation conditions have been demonstrated at other direct (Ongerth and Pecoraro, 1995) and conventiond pilot plants (Patania et al., 1995). 2-6-2 Filter Aid Filter aids can considerably improve the qualitty of filter effluents. When a polymer is used as a filter aid, both transport and attachment mechanisrns c m be facilitated by the formation of polymer-particle flocs which are generally larger and stronger than those achieved with conventional chemicai pretreatment alone (Zhu et al., 1996). Studies have demonstrated that polymer addition during filtration is similar to coagulation in that particle destabilization must be optimized (Zhu et al., 1996). The use of fiter aids has provided improvements in filtrate quality (Conley and Hsiung, 1969), resistance to eariy breakthrough (Conley and Pitman, 1960), and reduction in the magnitude and duration of ripening (Tuepker and Bauescher, 1968), however, it can also result in significant increases in head loss. Appropriate addition of filter aids should result in higher attachment forces and therefore should be concurrent with optimized backwashing (Amirthâratjah, 1988). Patania et al. (1995) investigated of the impact of filter aid addition on Giardia removal by GAC/sand filters and concluded that the use of filter aid did not improve cyst rernoval under the conditions studied; this conclusion was consistent with the fmdings of Ongerth et al. (1989). This apparent discrepancy between improved fdtrate turbidity but unchanged cyst/oocyst removal may suggest that filter aid improved filtrate quality by enhancing the removal of non-oocystkyst sized particles. It is possible that filter aid induced particle bridging considerably increases the size of particles originally in the 1 pm range, which are known to have a minimum transport efficiency (Yao et al., 1971), thereby increasing their transport and removal efficiencies wMe the removaIs of larger particles remained largely unafTected, 2.6.3 Hydraulic Changes Severai studies have concluded that different filtration rates do not necessarily result in different protozoan removals by fdters. Al-Ani et ai. (1986) perfomed pilot-scale sand and dual-media filtration experiments of G.lamblia removal at rates ranghg fiom 2 to 8 gpm/fi' (5 to 19 d h ) and concluded that the range of filtration rates had Little impact on cyst removal by the filters. Similar h d i n g s were reported by Hom et al. (1988) who investigated G. lamblia rernovals by tri-media füters operated at rates fiom 5 to 10 gpm/fiZ (12 to 24 mh); removals ranging fiom 0.7 to 3-logs were reported for two different water types. Hydraulic conditions can significantly impact the quaiity of fïiter effluents. It is generally recognized that filter performance is adversely affected by non-steady flow (Trussell et al., 1980). Hydraulic changes can occur suddenly or gradually and at the extreme they encompass sudden starts and stops in operation. In general, hydraulic changes disrupt the equilibrium between particle attachent forces and hydraulic shear (detachment) forces (Logsdon et al., 1999). Tfierefore, it is reasonable to sunnise that the relative impacts of hydraulic changes are inextricably Iinked to other operational factors such as coagulation conditions; however, some general conclusions regarding the impact of h y h d i c changes on filtrate quality c m be drawn. Cleasby et al. (1963) and Tuepker and Bauescher (1968) showed that large flow rate changes cause deterioration of filtered water quality by the detachment of previously retained particles. The degree of deterioration was related to the magnitude and rapidity of the rate change and independent of the duration of the disturbance. These relationships support observations that declining rate filters may provide better performance than constant rate filters (Hudson, 1959; DiBemard0 and Cleasby, 1980). Subsequent experiments by Hilmoe and Cleasby (1986) found no significant differences between declining rate and constant rate filters, however. The authors speculated that the previously reported poorer effluent quality achieved by constant rate filtration rnight have been caused by the constant rate control system used by DiBernard0 and Cleasby (1980), which might have inadvertently resulted in continuous flow rate fluctuations or surges. Cleasby et al. (1963) also revealed additional complexity in speculating on the risk of pathogen passage through jïlters when flow rate changes or hydraulic steps are applied because they demonstrated that particle passage through filters following a disturbance was dependent on the composition of the filter influent. This result underscores that the balance between attachment and detachment forces is affected by multiple factors such as coagulation, the ongoing ripening and break-uirough of different sized particles, floc strength, and hydraulics. Fitzpatrick et al. (1999) demonstrated that large and sudden changes in flow dramatically deteriorated particle removal by filters while smaller changes that were implemented gradually did not necessanly increase particle counts. Filtration rate increases of 50-150%, when flocs were not strengthened with filter aid, resulted in increased Giardia passage that was considerably higher than the increases in turbidity (Logsdon et al., 1981a). These results suggested that increases in Cqptosponndizmpassage îhrough filters could likely be expected under similar operating conditions. 2-6.4 Backwashing Strategy During backwashing, hydrodynamic shear is primarily responsible for detachment of particles. Raveendran and Amirtharajah (1995) empirically and theoretically described the interactions between suspended particles and media grains coated with previously deposited particles under fluidization conditions. A calculation of the interaction forces that control particle detachment indicated that solution chemistry affects particle detachment during backwashing; however, optimization of operating conditions for stronger attachment during normal filter operation may possibly make it more difficdt to remove particles during subsequent filter backwashing. If media cleaning is ineffective it will lead to poorer quality effluent during the initial stages of filtration, mudball formation, and other long-term problems in the filter bed (Amirtharajah and Wetstein, 1980; Amirtharajah, 1993). Ensuring thorough removal of pathogenic particles during backwashing may also prevent them ftom release during subsequent ripening or normal mter operation. The simultaneous application of air scour with subfluidization water wash (collapse pulsing) has been identified as an optimal backwashing strategy (Amirtharajah and Wetstein, 1980; Amiaharajah er al., 1991; Amirtharajah, 1993). The use of collapse-pulsing backwash strategies has been shown to reduce the number of oocyst-sized particles in filter effluents during ripening (Colton et al., 1996). 2.6.5 Temperature Since particles move relative to water to reach a grain surface, attach, and be removed, they expenence a viscous drag proportional to water viscosity (Ives and Sholji, 1965). The more viscous the water (e.g., colder temperatures), the more slowly particles move relative to the water to reach the grain surface, reducing the probabifity of removal (Ives and Sholji, 1965). Experiments conducted by Ives and Sholji (1965) demonstrated that colder water temperatures resulted in considerable decreases in particle removal when ail other factors (eg., raw water quality, chernical pretreatment, etc.) remained constant. In n o m d water treatment practice the effects of water temperature on particle transport and subsequent removal by filters are not easy to delineate because other operations such as clarification are also affected by temperature. Seasonal changes also often result in considerably different raw water qualities making it almost impossible to identie certain effects as exclusively temperature-related. The importance of evaluating the role of temperature as it affects pathogen passage through fdters is twofold. First, for given water treatment plants it is important to deterrnine stable operation removals across the range of seasonal conditions experienced. Second, it is important to determine if other operational conditions or events such as coagulation upsets, ripening, and breakthrough have simiiar implications for pathogen passage during warrn and cold water conditions. 2.6.6 Ripening It has been suggested that 90% of the particles that pass through a well-operated filter do so during ripening (O'Melia and Ali, 1978; Amirtharajah, 1985). The two peaks of particle passage that occur during npening result fiom backwash water remuants (Amirtharajah, 1988) and the penod that particles are being retained and subsequently act as coiiectors (OTMeliaand Ali, 1978); during this second peak particle passage is primarily due to non-attachent. It is likely that these mechanisms also affect pathogen passage through filters during this period. A variety of findings regarding protozoan passage through granular media filters duing ripening have been reported. Logsdon et al. (198 la) reported that Giardia cyst passage higher during ripenuig than during stable operation, even through filters was ~i~nificantly at low effluent turbidities. Similar fmdings were obtained at two pilot plants studied by Patania et a[. (1995); the differences between cyst removais during stable filter operation and ripening were less dramatic, however. At a third pilot plant studied by Patania et al. (1995), Giardia rernovals during ripening were comparable to those achieved d h g stable filter operation. These data suggest that multiple factors may affect pathogen removal during ripening. 2.6.7 Breakthrough Possible sources of breakthrough during filtration include particles that pass through directly fi-om the influent (non-attachment) or particles that become detached (Lawler et al., 1995). According to the mode1 presented in Figure 2.9 and other studies (Moran et al., 1993b), non-attachment and detachment occur d u ~ breakthrough g conditions. As particle detachment and non-attachent increase, increased pathogen passage through filters would also be expected. Logsdon et al. (198 la) demonstrated that turbidity breakthrough at the end of a filter cycle (when fdter effluent turbidity was above 0.4 NTU) could be accompanied by a tremendous passage of Giardia cysts, even if they were not present in the filter influent. A considerable increase in cyst passage was also observed during early breakthrough conditions when filter effluent turbidity was just above 0.2 NTU. Patania et al. (1995) investigated Giardia passage through filters during breakthrough when effluent hirbidities increased fiom 0.1 NTU to 0.2 M'U or higher. They found that Giardia removal was approximately 0.5-log lower during breakthrough than during stable operation. These data suggested that increased Cryptosporidiurn passage could also be expected during breakthrough, especially at frlter effluent turbidities above 0.2 NTU. 2.6.8 Media Imporhnt aspects of media specifications are size, shape, depth, unifonnity, and the choice of single-, duai-, or multi-media (Trussell et al., 1980). Media size affects the length of time to turbidity breakthrough and limiting headloss (Tmssell et al., 1980). Media shape affects head loss (rate of head luss build up) and is associated with the ability of filters to rernove particles (Tmssell et al-, 1980). Angdar media have demonstrated better turbidity and particle removal compared to smoother media (TrusseIl et al., 1980). Lower uniformity coefficients of anthracite media have dso demonstrated ïmproved rernoval of oocyst-sized particles (Yohe et al-, 1999). Filtration theory and practice has indicated that reverse-graded muiti-media typicdly provide better particulate (Rimer, 1968; Conley, 1972) and protozovz removal (Logsdon et al., 1385) at the cost of faster headloss buildup. Other studies, however, have indicated comparable particulate removal by dual- and tri-media filters (Tate and Tmssell, 1978, Biilica et al-, 1999), suggesting that tri-media may not necessarily offer a C. p a m removal advantage over dual-media. It is M e r possible that media selection rnay offer C. p a m m removal advantages only under specific operating conditions. 2.7 SURROGATES FOR CRWTOSPORTDILLM REMOVALDURING WATERTREATMENT Several different types of surrogates for viable C.pawum oocysts have been evaluated and include surrogates for occurrence (indigenous concentrations), disinfection, and removal. An ideal surrogate for cyst or oocyst removal should provide at Ieast a semiquantitative indication of cyst/oocyst removal by the process during a range of operating conditions that can be encountered during typical operation- Potentiai smogates for cyst and oocyst removal by drinking water treatment processes that have reported in the literature include: turbidity, particle counts, heterotrophic bacterial counts (HPC), aerobic spores (typically BaciZZzis szrbtilis), UV,,, dissolved organic carbon (DOC), polystyrene microspheres, and chemically inactivated C. p a m m oocysts. Table 2.3 lists and summarizes the main obsentations £kom several studies that evaluated potential surrogates for the removal of C. parvum oocysts during water treatment. Table 2.3 ~urrog'ateParameters for Removnl of ûyptospur-ma,, Study Surrogate Parameters LeChevallier et al., 1991c d d LeChevallier and Norton, 1992 d d dl" Major Observations Turbiditylparticles not indicative of oocyst removal. Giardia and Cryptosporidium removed similarly. d di,n~ Turbiditylparticles not Indicative of oocyst removal, Turbiditylparticles indicative of cystloocyst removal. Cryptospon'dium more difficult to remove than Giardia. HPCs not good surrogate for cysUoocyst removal. Nieminski, 1994 P 00 West et al., 1994 J Charles et al., 1995 J Kelley et al., 1995 d J Nieminski and Ongerth, 1995 J J Ongerth and Pecoraro, 1995 J d Patania et al., 1995 J J d Particle removal underestimated oocyst removal. dC Cryptosporidium removed at similar, but lower levels than Giardia, Turblditylparticles not indicative of oocyst removal, dd d J Turbiditylparticles not lndicative of oocyst removal. Particles more lndicative of oocyst removal than turbidity, HPCs not reliable surrogates for oocyst removal. Cryptospoddium removed at similar, but lower levels than Glardia. Similar Clyptosporidlum and Giardia removals. J~,M J Turbiditylparticles not indicative of oocyst removal. Similar Cryptosporidium and Giardia removals, not directly related, Table 2.3 Surroogate Parameters for Removal of Cryptosporidium (Continued) Study Plummer et al., 1995 Surrogate Parameters d Coallier et al., 1996 Lytle et al., 1996 d d d d Spore removals more sensitive than turbidity and similar to particles. d d4' d d Spore removals closely parallel particle removals, Spores rernovals decllne more rapidly than partlcles and turbldlty. Oocyst-sized particles easiest to remove, not good surrogate. d dC II d Spore removals useful for determining treatment efficiency, Spore removal closely and conservatively match particle removal. I/ dCth P Swaim et al., 1996 d Turbidity, UV264, and DOC not indicative of oocyst removal. d m Rice et al., 1996 Major Observations Scott et al., 1997 db Yates et al., 1997 V d Baudin and Lûiiié, 1998 d d Edzwald and Kelley, 1998 d d dk i/ Tu~biditylparîiclesnot indicative of oocyst fernoval, Spores conservative indicator of oocyst removal. g~~ d Spore removal not directly related to paiticle and turbidity removal. IJ dl1 Cyst and oocyst removals not well correlated with particle removals. Removal of cyst sized particles lower than cyst removal. Comparable Cryptosporidium and Giardia removals. d d Turbiditylparticles not indicative of oocyst removal. Turblditylparticles not indicative of oocyst removal. Table 2.3 S~irrogateParanieters for Reinoval of Csyptosporidiuin (Contiaued) Fox et al., 1998 d g Dugan et al., 1999 d d Edzwald et al., 1999 d Nieminski and Bellamy, 1998 I/ J d k I / J i Swertfeger et al., 1999 d d ,/" df dJ d Trends in turbidity and particle removal through clarifier generally indicative of oocyst removal, dj d Positive correlation between oocyst removals and particle and turbidity reductions through coagulation and settling. Turbidity removal most consenrative lndicator of oocyst removal through filtration. Vi O 1-2 vln slze range. 2-5 pm size range. 2-6 pm slze range. 2-15 pm size range, 3-5 pm size range. 3-6 pm size range. 4-7 l m slze range. h 5-15 Fm size range, a ' Particle and cyst removal trends generally consistent across clarification and filtration. Particle rernoval provides a conservative estimate of oocyst removal, I/ Turbidity, particles and spores indicative of water quality, not oocyst removal. Mostly non-detects of cystsloocysts in finished water, Microsphere and spore removals were more similar to cyst and oocyst rernovals than were turbidity and particle removals, Turbiditv and article removals lower than cvst~oocvstremovals, ' Other particle slze measurements were also made. Cumulative particles 2 1 pm. Cumulative particles 2 2 Fm. Cumulative particles 2 3 pm. Cumulative particles 2 5 Fm, NS Nol stated, ' Particle counting aIiows for red-tirne monitoring and is more sensitive than turbidity for treatment optimization (Hargesheimer er al., 1992; Arora et al., 1992). However, particle counting lacks the ability to discem between particles in the same size range and to detect very small changes in concentration that wodd likely be associated with Cryptospon'a'iurn concentrations (Hargesheimer et al., 1992). Several studies have indicated that while turbidity and particle removds are good indicators of general water quality, they are not quantitatively indicative of Crvptosponditirn removal by treatment processes (LeChevallier et al., 1991c; Nieminski and Ongerth, 1995; Swaim et al., 1996). Severd pilot- and MI-scale studies have demonstrated that orgaaism-sized particles and turbidity are approximate indicators of pathogen removal by dnnking water treatment processes, but not reliable surrogates (LeChevallier and Norton, 1992; Nieminski and Ongerth, 1995). Plummer et al. (1995) reached similar conclusions about turbidity, as well as UV,, and DOC. Patania et al. (1995)indicated îhat achieving a goal of 0.1 NTU was Indicative of effective cyst/oocyst removal. Although the risk of Cwp~ospon'dizm passage appeared to increase with increasing filtrate turbidity in several studies (Hall et al,, 1995; Nieminski and Ongerth, 1999, Fuller et al. (1 995) did not observe sipificant oocyst passage during the fnst hour of operation after backwash when filter effluent turbidity was high (filter npening); the sarne result was reported at one of three pilot plant plants studied by Patania et al- (1995). S M a r discrepancies between increases in turbidity durhg breakthrough and increases in Giardia cyst passage at some plants (Logsdon et al., 198la) and not others (Patania et al,, 1995) have also been reported. These data suggest that turbidity is an inadequate surrogate for predicting cyst and oocyst passage through filters. HPCs yield an estimate of the total number of viable bacteria that can be successfblly grown on plates and have been used to determine treatment efficiencies and distribution system integrity in several studies (Ferguson et al., 1990). Evaluating treatment processes for pathogen removal based on HPCs may be difficult because of the varying composition of bacterial populations that constitute HPCs. Several pilot- and full-scale studies have demonstrated that removal of heterotrophic bacteria were not an effective surrogate for predicting cystloocyst removal (Nieminski; 1994; Nieminski and Ongerth, 1995; Nieminski and Bellamy, 1998). Aerobic spores, prixriariIy Bacillus, have been used for evaluating treatment efficiency (Coallier et al., 1996; Rice et al., 1996). They are present in most surface waters, pose no public health threats, and are not indicative of fecai contamination (Jakubowski et al., j m in size, close to the size range of C~ytospon'dium oocysts (Rice et al., 1996). They are also füghiy resistant to 1996). The spores are approximately 1 disuifection and their removals closely parallel particle removal (Coallier et al., 1996; Rice et al,, 1996). Studies performed by Scott et al. (1997) evaluated BaciZZus spores, turbidity, and particle c ~ u n t sas suzrogates for pathogen removal. Turbidity and particle counts yielded conclusions similar to those discussed above (Nieminski and Ongerth, 1995; Patania et al., 1995) whereas spores demonstrated a significant correlation with CryptosporzMdium removal at both pilot- and fiill-scale. Lytle et al. (1WQ), Nieminski and BeiIamy (1W8), and Swertfeger et al. (1999) also concluded that aerobic spores were indicative of treatment eEciency, however, they did not conclude that the spores were adequate surrogates for oocyst removd. These studies suggested that although spores could be considered a more conservative indicator of treatment efficiency than particles or turbidity, they were not necessarily indicative of C~ptosporidizmremoval during water treatment. Fluorzscent polystyrene microspheres in the size range of Cryptosporidium oocysts were tested by Li et al. (1997) in field scde bag filtration systems. A nearly linear correlation between log removals of microspheres and oocysts was established. During these studies the relationships between Crvptosporidium removal and other parameters such as particle counts and turbidity were also very linear. Fox et al. (1996) included poiystyrene microspheres in evaluations of Cgptosporidium removals by granular media filtration processes; however, neither microspheres nor oocysts were found in the filter effluents. Sweafeger et al. (1999) had more success with microspheres in one experiment that indicated polystyrene microspheres might show promise as a çurrogate for Ctypfosporiditrrn removd. Due to the lack of adequate surrogates for the removal of viable C. pawrrnz oocysts, chemically inactivated oocysts have been comrnonly used for treatment evaluations. The use of chemically inactivated oocysts is preferable to that of viable oocysts because of the potential health risks associated with the use and release of viable C. panrum. It has been suggested that chemicaiiy inactivated oocysts may not be ideal surrogates for viable oocysts due to differences in surface charge, described by zeta potential (Lytle and Fox, 1994). Change in colloidal zeta potential can affect removal during granular media filtration because zeta potential is indicative of particle destabilization (Amirtharajah and MUS, 1982). Particles are most readily removed when zeta potential is near zero, corresponding to the particles isoeleciric point (Amirtharajah, 1988). Chemical inactivation can change oocyst zeta potential in a rnanner that might affect coagulation (Lytle and Fox, 1994; Ongerth and Pecoraro, 1996) and therefore filtration. D d g water treatment, however, oocyst zeta potential is af5ected by multiple factors such as water quality, coagulant type and dosage, and pH in addition to chemicai inactivation prior to treatment . Filtration is one of the most critical and successful physico-chernical barriers against C-ptosporidiurn passage through water treatment. Several studies have assessed C. parvum oocyst removai by granular media filters operated at or near to optimized stable operating conditions. Summarized in Table 2.4, these studies include investigations of both C. parvum removal by granular media filtration alone (with removals based on filter influent and emuent oocyst concentrations) and as part of the water treatment process (with removals based on upstream, but not filter influent, and filter effluent concentrations). Unless specified in the table footnotes, the C. parvum removals in Table 2.4 are removals by granula media filtration alone. Table 2.4 Removiil of Ciyptosporidiirrir During Stable Operation Study Type of Treatment - Filter Type Filter Loading (gpmlft2) C,parvum variad ~2.4' p5.3 Rapid sand and GAC efluents had hlgher oocyst concentrations than dual- and tri-media eMuents posslbly related to different raw water quality, Rapld sand more effective than GAC for oocyst removal, Many non-detects in filter effluent samples, Oocystlcyst occurrence related to raw water levels. Many non-detects In filter effluent samples. Log Removal Seeded Conc. (#IL)' Observations of Relevance to Presenl Study LeChevallier et al,, 1991c Varied 66 full scale plants 2 pilot-scale conventional plants varied GACIsand LeChevallier & Norton, 1992 3 full scale, conventional plants anth.Isand GAClsand anth.lsand >2.4 >2,5 >2.3 0.5-gpm conventional pilot-scale 05-gpni direct filtration pilot-scale 600-gpm conventional full-scale 600-gpm direct filtration fult-scale anth.lsand anth.lsand anth,/sand anthhand avg. 2.8 avg, 2.9 avg, 2.1 avg, 2,7 No difference between conventional and direct filtratlori for cysl and oocyst removal, Dlrect filtratlon yielded better cyst and oocyst removal, but influent quality was very different. - West et al., 1994 Pilot-scale dlrect filtralion anth, anth. 2,1 - 3.3 1.9 - 3.2 Many nondetects In filter effluent samples, Charles et al., 1995 Bench-scale dlrect filtration 2-gpm conventional pilot plant mixed mixed 2,3 - 3.3' 2,2 4.5' Lack of coagulant decreased removal by >2 -legs, Rapld mlx conditions impacted removal by +log, Stableloptlrnal removals not specified. Hall et al., 1995 sand Pilot-scale dissolved air flotatlon (DAF) followed by rapid gravlty anthhanci filtration (RGF) GAC sand anth,lsand GAC sand sand sand Pilot floc blanket clarification and RGF - 2,9 - 3.4' 3.9' 4.4' >3.6' 3 , l - >4,4' 3,2- 4,l' >3.4' 3.5" 3.9' No differences behveen single, dual, and GAC filters, Lower removal during rlpenlng (-0.4-log) - Haif the effluent samples wete non-detects. Table 2,4 Removnl of Ciyptosporidirrin During Stable Operation (Continued) Study Type of Treatment KeIiey el al., 1995 1 MG0 conventional plant 3 MGD conventional plant Niernlnskl & Ongerîh, 1995 0.5-gpm conventional pllot-scale 0,5-gprn direct filtration pilot-scale 600-gpm conventional full-scaie 600-gpm direct filtrationfull-scale Ongerth & Pecoraro, 1995 1-gpm direct filtration pilot plant Patanla et al., 1995 Conventional pilot-plant Filter Type Filter Loading (gpmlft2) C. pawum Log Rernoval saiid sand NS >1.O anth.lsand anth,lsand anth./sand anth.lsand 1.9 - 4.0 1.3 - 4.0 1,9 - 2.8 2,6 - 2.9 tri-rnedia 2.5 - 3.2 Seeded Conc. (#IL)' - -lo4" -1 04" -lo7" -10'~' Obsewations of Relevance to Present Study Many non-detects in filter effluent sarnples, Fluctuatlng turbldity resulted in varlable cyst and oocyst concentrations, Direct filtration yielded better cyst and oocyst removal, but Influent quality was very dlfferent. -400-3000 hadequate coagulation decreases rernoval to -1.5-log Cheniical pretreatrnentcritlcal for oocyst removal. Media design, filter ald, filtratlon rate: iess important. During ripening, cyst loocyst removals -0,5- to 1-log lower than during stable operation. Rlpenlng effect observed at 2 pilot plants. At a lhird pilot pilot plant, no difference behveen ripeninglstable. Stable and breakthrough reniovals comparable. Many lnfluent and effluent non-detects, Large range of fiiter lnfluent concentrations. GAClsand anth.lsand Pilot-scale in-iine filtration anth./sand sand Conv. pilot wl low-rate surface filtratlon - slow sand Timms et al,, 1995 Pilot-scale slow sand filtration Edzwald et al., 1996 Pilot-scale contact filtration anthhand Pilot-scale dissolved air fiotation (DAF) anth.lsand anth.lsand followed by RGF Swaim et al., 1996 Pilot-scate direct filtration Scott et al,, 1997 6-gpm conventional pilot plant -. deep anth. deep anth. deep dual deep dual dualltri >4.5 avg. 4.7 -2 -4.1 -2 - 2.2 No oocysts found in filter effluent. Filter lnfluent concentrationschanged due to chemical pretreatment conditions and affected log removals. 3.7 - >4.3' 3.6 - >4,3' 3.5 - 4.1' 3.6 - >4,3' During ripening, cyst loocyst removats -0.5- to 1-log lower than during stable operation, Cyst and oocyst removals ~ 3 - i a gduring ripening, - 4.3 Did not dlstlngulsh between dual- and tri-media removals. -1.7 Table 2.4 Removnl of C~~ptospoi*idiiii~ During Stable Operation (Coiitinued) Study Swertfeger et al., 1999 Type of Treatment Conventional pilot plant (summer) Conventional pilot plant (winter) - 4 Cn Filter Type sand anthhand deep dual sand anth./saiid deep dual Filter Loading (gpmlf12) 2.5 5,O 50 25 5,O 5,O C. parvum Log Removal 1.8 - 3.3 1.6 - 3.4 3.4 - 4.2 2.5 - 3.0 - 3,l 3.2 2,9 - 4.0 Seeded Obsen/ations of Relevance Conc. to Present Study (#IL)' -10~-10~"Al1 media configurations statlstlcally comparable, - 1 0 ~ - 1 0 Removals ~~ adjusted (empty column removals subtracted), - I O ~ - I O ~ ~ Almost entlre effluent processed. - 1 0 ~ - 1 0 ' ~Polystyrene bead rernoval somewhat consistent with -IO~-IO~" oocyst removal ln one experlment. -10~-10~"Oocyst were not coagulated, only pretreated with a small amount (0.1-05 mglL) of ferric sulphate as filter aid. ot auulicable. "Nol stated. Measured concentration ln filter influent unless otherwise stated. Seeded concentration in raw water. 9~nclearas to whether measured or calculated filter influent concentrations were used. Theoreticai influent concentration based on hernocytometer counts and dilution calculatlons. Filter influent concentration was not stated. I t ~ o t anurnber l of cystsloocysts seeded at filter Influent. * ~ o t a lnumber of cystsloocysts seeded at clarifier influent. Removals based on raw water concentration (not fiiter influent concentration), #* Removals based on influent clarifier concentration (not filter influent concentration). ' Full-scale C.p a m m removals have been reported anywhere fiom 2- to 3-log (e.g.,Kelley et al., 1995; Nieminski and Ongerth, 1995) to >4-log (e.g-, Baudin and Laîné, 1998). Pilotscale C ~ t o s p o n d i u mremoval data have suggested that filters can achieve anywhere fiom 2- to 3-log (e.g., West et al., 1994; Kelley et al., 1995; Fox et al., 1998), 3- to 4-log (Yates et al., 1997), 4- to 5-log (e-g..Patania et al., 1995), and >5-log (e-g., LeChevallier et al., l99lc) removal of oocysts. Several studies that have evaluated C.parvurn removai by filtration are surnmarized in Table 2.4. Aithough this table includes a wide range of fütration scenarios (eg., different raw waters, different modes of filtration, different scales, etc.), the listed C. parvum removds predominantly correspond to optimized filtration; results from non-optimal operating conditions are noted and sllrnmarized when applicable. When considering the removal of C. p a m m (as in Table 2.4) or any other pathogen by filtration, experimental conditions and methods must be considered. Differences in analytical reiiability, processed sample volume, method detection limits, and influent microorganism concentrations c m al1 contribute to inter-study differences in C. p a m m removal by filters. The range of oocyst removals described in Table 2.4 underscores the need for accurate and thorough description of experimental conditions and methodologies so that C. parvurn removals by filters and other processes can be thoroughly assessed. Several of the methodological factors that can cntically affect the interpretation of C. parvum (and other microorganism) removal data are tisted in Table 2.5 dong with relevant questions that should be considered when evaluating pathogen removal data. The C.p a m m removal studies included in Table 2.4 were evaluated in the context of the methodological factors m d questions listed in Table 2.5; several of the methodological factors relevant to those studies are listed in Table 2.6. Many of the points h i w g h t e d in Table 2.6 are referred to in the critical discussion of cianiently available C. pawurn removal data below. Although some of the points seem obvious or redundant, they underscore the need for clear interpretation and understanding of limitations of the currently available C. parvum removal data in the literature. Table 2.5 Methodologid Factors and Relevant Questions to Consider When Evaluating Pathogen Removal Data Methodo10,qicd Factor Method Recovery Relevant Questions What was the analyticd method recovery? How was recovery described? How was recovery determined? Was recovery the sarne for al1 of the waters studied? Data Handling Were the data adjusted to account for recovery? if yes, how? How were nondetects handled? How did non-detects affect maximum possible removals? How was pathogen removal calculated? What are the implications of reporting geometric vs. arithmetic means? Operating Conditions Were several pretreatment conditions evaluated? Did pretreatrnent affect influent concentrations of pathogens? Did preûeatment change in response to varying raw water conditions? Were operating conditions optimal? Were various process confiptions/types cornpared? Were al1 relevant operating conditions specified? Pathogen Concentrations Were many non-detects found? Did processed sample volume affect the results? Were influent and effluent concentrations high enough to evaluate maximum removal (i.e., no non-detects)? Were pathogens seeded into the treatment process? How were seeded concentrations determined? Were filter influent and effluent concentrations determined by the same analytical method? Were results obtained fiom experiments with atypically hi& seeded concentrations or concentrations comparabIe to indigenous levels? Table 2.6 Suiiiiiiary of Critical Iiifoniiatioii from Studies Evaluating Process Removals of C, y a n w i Study FE Sample Volume Removal Adjusted for Recovery 13-59% no no LeChevallier & Norton. 1992 RP no many Nieminski, 1994 81% no NS NS West et al., 1994 1 O-40% not clear many LeChevallier et al., 1991c Charles e l al., 1995 NS NS - - Seeded Conc, (#IL)' FI Sarnple Volume Recovery 3785 L FE Non-Detects Observations of Relevance In Filter to Present Study Effluent rnany Geometric means reported for oocyst denslties. Stableloptimal reniovals not specified, NS no no m O Hall et al., 1995 IO-30% no no some Geometric means reported for oocyst densltles. Stableloptimal removals not specified, Only used data for samples in whlch oocysts were found. Different influent quality between experiments. Stableloptlrnal rernovals not specified. Unclear whether nor not recovery corresponds to speclfic waters studied. SOM0 no Many non-detects In filter effluent samples. 6-14% Yes many avg, 9% raw avg. 12% FE 535% FE no NS no NS Different influent quality between experiments. Only used data for saniples in whlch oocysts were found, Lower recovery durlng higher turbidlty conditions, Ongerih and Pecoraro, 1995 avg. 17% FI avg. 29% FE Yes NS Variable filter Influent and effluent recovery profiles. Patanla et al., 1995 avg, 15-35% no many Timrns et al,, 1995 NS - all Kelley et al., 1995 Nieminski & Ongerth, 1995 Recoveiy ln al1 waters assumed equivalent Recovery only measured from raw waters, Many influent and effluent non-detects, large influent range. No oocysts were found ln the filter effluents. Table 2.6 Suiiiinary of Critical Iiifosmntion froiii Studies Evaluaiing Psocess Reinovals of C, porwria (Continued) Study O! CL Recovery Removal Adjusted for Recovery 7210 2 - 631 FI Sample Volumo NS NS FE Sample Volume NS NS Seeded Conc. (#a)' Non-Detects Observations of Relevance in Filter to Present Study Effluent NS Filter Influent concentrations changed due to chemlcal NS pretreatmenl and affected log removals, Edzwald et al., 1996 NS - Lyile et al,, 1996 NS - 6900' 1L 1L al1 No calculatlons of removal were made due to nondetects, Swaim et al., 1996 NS no -lo3' NS NS some Recovery ln raw and filtered waters assumed equivalent. Scott el al,, 1997 NS - -10'" Yates et al., 1997 NS - Baudin and Laine, 1998 NS Edzwald and Kelley, 1998 _ 0.1-10 mL 5 - 500 mL NS 100" 1,101nL 300mL NS - - NS NS many NS - 2510' NS NS NS Pretreatment affect filter influent concentrations. Fox et al., 1998 NS - -lo5 250 mL 5L no Focused on clarifier rernovals, filtration generally mentioned. Stableloptimal removals not specifled, Dugan et al,, 1999 NS - -84480 250 mL 5L no Stableloptimal removals not specified. Edzwald et al,, 1999 NS YeS NS NS al1 Pretreatment affect filter influent concontratlons. Swerîfeger et al., 1999 NM - 53-106L 24 L no Reniovals adjusted (empty column removals subtracted), -1 07§5 -IO~-IO~" -Nol applicable. N S ~ staled. ol '"eference provided. N M ~ omeasured, t 'concentration unlts unless olherwise slaled. l ~ e e d e concentralion d In raw waler, %nclear as Io whellier measured or calculated filler influenl concenlrations were used. "~heoreticalinfluenl concentration based on hernocytometer counts and dllulion calculatlons. Fllter Influent concenlration was no1 staled. "Total number of cyslsloocysts seeded al filter Influent. % ~ o t anumber l of cystsloocysls seeded al clarifier influent. Rernovals based on raw waler concentration (1101filler lnfluenl concenlrallon). Rernovals based on clarifier Influent concentration (not filter influent concentrallon). " Ripening and breakthrough turbidities not specified, 2.8.1 Effect of Analyticd Recovery on Interpretation of C p a m m Removal Data Perhaps the most commonly cited source of uncertahty associated with C. p a m m removal data is the lack of consistency in analytical recovery of oocysts fiom water. Althou& standard deviations and coefficients of variation can b e used to describe the uncertainty associated with oocyst recovery, tools that incorporate this information into C. p a m m removd data are current1y lacking. Therefore, most reported oocyst removal data are analyzed with little if any consideration of andytical recovery and the resulting conchsions are essentially based on the assumption of consistently hi& recovery (ie., not much uncertainty associated with the data), hproved detection rnethods for pawum will undoubtedly increase the reliability of oocyst removal data collected during future investigations. Many of the currently available methods are adequate for drawing general conclusions about process removals of oocysts, however. Various analytical methods for C. pawum exist; the less consistent the analytical recovery, the more difficult it is to attribute differences in removal data to differences in treatment processes because it is unknown whether differences in treatment or uncertainty associated with recovery cause the observed differences in removal. Presentation of al1 of the recovery data (not just means and standard deviations) relevant to a particular study is therefore the optimal form of recovery description because it will allow for subsequent reassessment of data reliability once tools are in place to integrate that information with measwed removal data. It is worth noting that a lack of recovery information does not completely invalidate findings fkom challenge studies. For exampie, when investigations comparing two filtration scenarios treating the same water are conducted, it is reasonable to assume that the filter influent waters O : each Nter have similar oocyst recovery profiles (since the filters treat exactly the same water). If the filter effluents produce waters of similar quality (e-g., turbidity and particle counts), it is also iikely that the filter effluents will have similar oocyst recovery profiles, although these profdes will not necessarily be the same as those obtained Çom filter influents. If statistical differences are not obsenred when recovery is ignored (i.e, when recovery is assurned ideal), the incorporation of increased uncertainty of recovery will likely rnake it more dificult to demonstrate statistical differences. Therefore, inferences drawn fiom such data are essentially the same; regardless of recovery which only affects the magnitude of the reported removais. When investigated water qualities are similar, it is o d y when diReremes in removal are observed that the investigators must question whether or not they orïginate fiom inherent differences between the oocyst removal capacities of the different processes or fiom the uncertainty and inconsistencies associated with the analytical methodologies. Regardless of whether or not merences between removals are observed, real differences in oocyst removals between cornpared processes may exist. In such cases, the onIy way that the dserences cm be determùied is by repeating the experiments and processing samples with more accurate analytical methods. If oocyst recoveries Vary between influents andor effluents fkom treatment processes (perhaps due to the presence of algae, turbidity, etc), recovery can affect both the magnitude of reported oocyst removals and the conclusions of cornparisons drawn between different processes or operating conditions. The reporting of al1 relevant recovery data under these conditions c m be critical to identifying the limitations of inferences that can be drawn from the removal investigations. Furthemore, the recovery data can be subsequently integrated with the removal data when tools are developed to incorporate methodological uncertainty into removai data. 2.8.2 Mode of Filter Operation Pilot- and fûll-scale studies performed by Nieminski (1994) and Nieminski and Ongerth (1995) indicated comparable pilot-scale removals of C-ptosporidium by conventional and direct filtration processes. DBerences were observed at full-scale where direct filtration removals surpassed those of conventional filtration. The authors speculated that the differences between the full-scale oocyst removals were attributable to differences in filter irfluent quality and the associated difference in recovery fiorn higher turbidity waters (such as those encountered during conventional treatment experiments). The authors also clearly stated that non-detects were not included in the calculation of log removals; however, it is possible that if larger samples had been processed that oocysts would have been found eventually, thereby increasing mean removals. The inclusion of recovery and sample handling information was critical to appropriately evaluating the C-pawum removal capacity of the two modes of filtration. C. parvurn removal data by various modes of nitration were resrted by pitania et al. (1995). The authors studied two conventional pilot plants, including a pilot plant with inline filtration and a conventional pilot plant with low rate surface filtration. Although very hi& (4-and often >5-log) removals of oocysts were observed at both pilot plants, the filters alone achieved anywhere fiom essentially no removal to 4.7-log removal of oocysts (Table 2.4). Considered without methodological information, these data rnight have suggested that the conventional pilot plant with low rate surface filtration and the pilot plant with in-line filtration offered superior oocyst removd when compared to the conventional pilot plants (Table 2.4). The authors provided critical methodological data that precluded the conclusion that one form of filtration was superior over another, however. As is pointed out in Table 2.6, the C.parvum recoveries in ail of the waters were assumed equivalent, which was not necessady accurate, as demonstrated by Nieminski and Ongerth (1995). In addition, the experiments described by Patania et al. (1995) were conducted over a wide range of influent C. parvum concentrations associated with seeding oocysts into the raw water, pnor to pretreatment. While the filter influent concentrations were as high as 10' oocysts/L during the in-line and low rate surface filtration experiments, they were as low as c0.22 oocystsL during the conventional filtration experiments, with several influent and effluent non-detects. Although the inappropnateness of concluding that in-he filtration was superior to conventional fdtration for oocyst removal is perhaps obvious, the methodological information clearly provided justification for the Limitations of the conclusions that could be drawn fkom the C.panwm removd data. 2.8.3 Media Type and Design Investigations of media type and design have demonstrated that these parameters have Little impact on oocyst removals by filters. Hall er al. (1995) did not find performance differences between sand and dual media filters when the filters had similar filtrate quality (measured by turbidity). This fmding was not incon~overtibledue to the presence of non-detects in some of the filter effluents, however; in addition, no analytical recovery information was specified by the authors, allowing for the possibility of different recoveries fkom different waters. Pilot-scale studies conducted by Patania et al. (1995) indicated that GAC, GAC/sand, and anthracite/sand (with both 1.0 mm and 1.5 mm effective size of anthracite) filters also achieved essentially comparable oocyst removals by the various types of filters. The authors emphasized that the filter influent oocyst concentrations were below detection limitts in one of the media cornparisons (GAC/sand versus anthracite sand), thereby precluding a true cornparison of oocyst removals by the two filters. h addition, Patania et al. (1995) also assumed that oocyst recoveries from the difEerent waters were equivdent- Such a difference could have possibly, though not necessarily, impacted the interpretation of their oocyst removal data. Cornparisons of C. p a m removals by sand and dual-media frlters by Dugm et al. (1999) also resulted in similar oocyst removals; the presence of oocysts in al1 of the filter ef'fluents M e r substantiated the conclusion that media type did not have a large effect on oocyst removaI by filtration. Swertfeger et al. (1999) drew the same conclusion by demonstrating that oocyst removals by fiters of sand and dual-media could not be statistically disthguished fiom one another. This conclusion was also based on experiments that were conducted to yield oocysts in ail of the filter effluents. Although the Dugan et al. (1999) and Swertfeger et al. (1999) investigations produced reliably countable numbers of oocysts in both the fdter influents and effluents, the lack of malytical recovery information allowed for some speculation regarding the reliability of the data. However, the comparative conclusions (between media types) that could be drawn fiom these investigations were essentially unaffected by the recovery specifics because, as was discussed in Section 2.8.1, the m e r s in each study treated the same influent and produced effluents of similar water quality (based on turbidity and particle counts). Dugan et al. (1999) and Sweafeger et al. (1999) also investigated the effects of relative media depth on C.parvum removal by filters. Both studies investigated dual-media filters and concluded that media depth did not substantially impact oocyst removal. Dugan et al. (1999) specified that their data failed to consistently demonstrate increased oocyst removal with Uicreased media depth, but noted that their filter cycles were stopped at 36 hours. The authors speculated that differences in oocyst removal between filters with different media depths might have been observed more consistently if the filters had been operated beyond 36 hours. 2.8.4 - Filtration Rate Pilot-scale investigations of sand filtration conducted by Hall et al. (1995) did not demonstrate any difterences in oocyst removal when the filter was operated at 2 and 4.1 gpmlft' (5 and 10 mh). The authors noted, however, that there were several non-detects at the iower rate and no non-detects at the higher rate, suggesting that differences mi& have been observed between the two rates if larger samples had been processed. Patania et al. (1995) also investigated the effects of filtration rate on oocyst removal in several types of filters. The authors studied GAC/sand filters operated at rates of 3 and 6 gpm/ft2(7 and 15 m/h) and anthracite/sand filters at 5 and 8 gpm/ft' (12 and 20 mm). They concluded that C.parvum removals were not greatly afXected by the range of fdtration rates studied; these data were strengthened by the presence of oocysts in alrnost al1 of the filter effluent samples. The authors speculated that as long as pretreatment conditions were optimized, the different filtration rates would have a minimal impact on oocyst removal by the filters. hcreases in filtration rate dso showed no adverse effects on oocyst removals by slow sand filters (Timms et al., 1995); however, these data were inconclusive because no oocysts were found in any of the filter effluent samples. Dugûn et al. (1999) investigated the effects of filtration rate over time in dual-media filters. The authors demonstrated that oocyst removals by filters operated at 2 and 4.1 gpm/ft' (5 and 10 mh) were comparable for the fxst 30 to 40 hours of operation. Afterward, oocyst removal decreased by >l-log during filter operation at the higher rate. Although recovery information was not specified, oocysts were found in ail of the filter eflluent samples. These fbdhgs did not contradict those of Hall et al. (1995) and Patania et al. (1995), but rather suggested that additionai operational factors must be considered when evaluating the impact of filtration rates on C. parvum removals by fdtration. They also might have considerable implications for filter operation practices, particularly if the analytical rnethod employed by Dugan et al. (1999) was considered relatively reiïable; however, speculation on this issue was not possible because no recovery information was provided. Minimally, these data suggested that the effects of different filtration rates on oocyst removal over t h e should be M e r investigated. 2.8-5 Filter Aid Patania et al. (1995) investigated of the impact of filter aid addition on Cryptosporiditrrn and Giardicr removal by GAC/sand filters and concluded that the use of anionic polymer filter aid did not improve the removal of either microorganism under the conditions studied. The authors noted that the fiIter infiuent Crptospondiurn concentrations were inconclusive since they were below their rnethod detection limit. Swertfeger et al. (1999) also discussed the impacts of filter aid on oocyst removal. During these investigations, pilot-scale filters received settled water fiom a full-scale treatment plant. Non-pretreated (non-coagulated) oocysts were seeded into the filter influent and femc sdphate was added as a filter aid at doses of 0.5 and 1.0 mgK. Oocyst removals of > 2.5-log were achieved by the pilot-scale filters. Although Sweafeger et al. (1999) did not investigate oocyst removals by filtraion without filter aid, they speculated that filtration with filter aid might be capable of substantial pathogen removal in the event of a limited breakdown in the coagulation process or if cysts/oocysts were introduced into the treatment process after coagulant addition. The latter speculation was supported by the authors' data, although the impact of filter aid on the achieved oocyst removals was unclear. The combined impacts of sub-optimal coa,dation and the use of filter aid on oocyst removal by filtration were not discussed during this study; therefore, the speculation on mitigating oocyst passage through filters with the use of filter aid during periods of sub-optimal coagulation represents a potential area of m e r research that was not substantiated by the data presented by the authors. 2.8.6 Pretreatment A variety of pretreatment options pnor to filtration are possible. Hall et al. (1995) conducted pilot-scale investigations that demonstrated that good C. parvum removals could be achieved by rapid gravity filtration following either dissolved air flotation or floc blanket clarification. Recovery information was not cleariy specified and non-detects occurred in some of the mter effluents, however, thereby allowing for some speculation regarding the amount of unceaainty associated with these data. Patania et ai. (1995) similady demonstrated that various pretreatment stmtegies in the f o m of different coagulant combinations (femc chloride or alum, with and with cationic polymer) applied to and optimized for the same water could produce equally effective removal of oocysts by GAC/sand and anthracite/sand fdtration. It should be noted, however, that the resulting oocyst concentrations in the filter effluents included several nondetects. The authors indicated that their conclusion was supported by other investigations that demonstrated that different coagulation schemes applied to the same water could produce equally effective removal of turbidity and cysts (Logsdon et al., 1985; Al-Ani et al., 1986; LeChevallier et al., 199lc). The authors also speculated that different coagulation conditions did not greatly impact oocyst removal through the entire treatment process because each coagulation condition had been optimized with jar tests prier to application. These findings and speculations were reasonable given that effective filtration is inextncably iinked to effective pretreatment duing typical water treatment conditions. Several studies have indicated the importance of coagulation processes for improving filter removal efficiencies of C. parvum. The difficulty in understanding and optimizing coagulant effects during filtration is complicated by the different roles of attachment and detachment mechanisms o c c w i g dwing different periods of the filter cycle such as ripening and breakthrough. It has been demonstrated that a complete lack of chemical pretreatment can result in very poor to no Cwptospondiurn removal by conventional fdters (Charles et aL, 1995; Patania et al., 1995); this result underscored the link between effective fdtration and pretreatment during water treatment. Although non-ideal analytical rnethods can make it difficult to measure low concentrations of oocysts, Patania et al. (1995) found relatively high oocyst concentrations in filter effluents during Cqptosporidium seeding studies conducted when no chemical pretreatment preceded filtration. This result was demonstrated regardless of fdter pre-conditioning with effectively coagulated water prior to the no coagulation conditions. Although a l l of the recovery S o m a t i o n was not explicitly provided, the oocyst removal data were convincing for even a moderately consistent analyticai method because the filier effluent concentrations were so high relative to those obtained during typical stable operation (it difficult to imagine somewhat consistent recovery differences of several log). The cntical importance of adequate coagulation for effective filtration of oocysts was M e r supported by the relativeiy consistent fmdings in replicate experiments conducted by the authors. Charles et al. (1995) concluded that rapid mix conditions were less critical than coagulation and impacted oocyst removals by -4-log. The authors also dernonstrated considerable deterioration in C. pawum removals when no coagulant was used prior to filtration; however, they did not s p e c e if the fsllters had been pre-conditioned with effectively coagulated water prior to oocyst seeding during the no coagulation investigations. Although Patania et al. (1995) did not observe any differences between oocyst removals obtained during no coagulation filtration by pre-conditioned and non-pre-conditioned füters, it is likely that pre-conditioning effects can be somewhat site-specific, much like coagulation conditions. Sub-optimal coagulation (non-optimal chemical pretreatment) occurs due t O factors such as mis-dosing or sudden water quality changes. The fiequency of sub-optimal coagulation events varies fiom plant to plant. Although not as severe as complete coagulation failure, sub-optimal coagulation conditions have consistently demonstrated >l-log decreases in C. pamrrn removal by fütration. Ongerth and Pecoraro (1995) conducted pilot-scale direct filtration investigations during optimized and sub-optimal coagulation conditions. Oocyst removals averaged only 1.5-log during the sub-optimal coagulation experiment, a BI-log decrease relative to stable (or optimized) operating conditions. This experiment was only performed once; given that sornewhat different mean oocyst recoveries were obtained fiom the filter influent and effluent waters, replication would strengthen this result. The data of Ongerth and Pecoraro (1995) strongly suggested that oocyst removals detenorated during sub-optimal coagulation since the three oocyst removals measured during the sub-optimal coagulation experiment were the lowest of a i l eleven measured removals. Dugan et al. (1999) reported similar findings when they Ulvestigated pilot-scale C. punurn removal during penods of sub-optimal coagulation. The authors demonstrated that undercoagulation of moderately and hlghly turbid waters and moderate coagulation of low turbidity waters resulted in at least 1-log deterioration in oocyst removals relative to those obtained d u k g optimized operating conditions. Mthough no recovery information was provided, oocysts were detected in all of the fïiter effluents; oocyst removals by filtration were also consistently lower during non-optimal coagulation experiments than during experiments performed at optimized operating conditions. The various coagulation scenarios that these authors examined underscored that the general term "sub-optimal coagulation" can be interpreted to mean a variety of operating conditions that result in a large range of filter infiuent and effluent water qualities. The effects of sub-optimal coagulation may also be site specific and result in a range of impacts on C.parvtirn removal by filtration. In general, aU of the reported investigations of C.panmm removal by filtration during sub-optimal coagulation and no coagulant conditions have demonstrated that optimal chernical pretreatment is cntical to rnaximizing oocyst removal during filtration. 2.8.7 Ripening Several pilot-scale studies have demonstrated oocyst removais approximately 0.5- to 1-log lower during frlter ripening than during stable fdter operation. Lower removals during filter ripening or maturation could be expected because ripening is a penod during which the initial improvement of filtered water quality occurs in typical filter cycles. H& et al. (1995) demonstrated slightly lower oocyst removal during Nter ripening (-0.4-log lower) than during stable operation by sand filtration preceded by dissolved air flotation. The authors only presented data for one experiment, however. The generalized recovery data and a lack of replicate data made it difficult to speculate whether or not this small difference between oocyst removals during npening and stable operation was due to methodological variability. Patania et al. (1995) also uivestigated C. parvurn removals by filtration at three different pilot plants. Two of the pilot plants yielded moderately lower (-0.4- to 0.9-log lower) oocyst removals during ripening than during stable filter operation, a result similar to that reported by Hall et al. (1995). No discemable difference between oocyst removals dluing ripening and stable filter operation was observed at the third pilot plant. The authors speculated that the deep filter media at the third pilot plant resulted in virtually no maturation with respect to microorganisms. It was also possible that this result was due to methodological variability9 rather than media effects, however. Swaim et al. (1996) also reported slightly lower mean oocyst removals d-g ripening compared to stable flter operation; their data were additionally complicated by several non-detects in the filter effluents during both ripening and stable operation experiments. Data such as those of Hall et al. (1995), Patania et al. (1995), and Swaim er (2.1.(1996) emphasize the need for a tool to quanti@ the uncertainty associated with C.parvum analytical methods. These fmdings also emphasize that multiple operational and design factors must always be considered when evaluating C. p a m m removal data fiom filtration processes. Findings similar to the pilot-scale results discussed above were presented by Baudin and Lainé (1998) who demonstrated oocyst removals -1-log lower during filter ripening than during stable operation at two fidl-scale plants. Only one stable operation and one r i p e h g experiment were conducted at each plant; no methodological recovery data were presented. The authors specified that oocyst removals during filter ripening depended on raw water concentrations at one of the plants. Fluctuating filter influent oocyst concentrations at one plant and lacking recovery information made it difficult to speculate about the cause of the observed -1-log difference in oocyst removais during npening and stable operation. LeChevallier et al. (1991~)also reported slightly lower oocyst removals during fdter ripening than during stable operation at two full-scale plants. Numerous non-detects in the filter effluents and somewhat variable analytical recoveries also made it difficult to conclude whether or not the observed differences were due to experïmental and methodological vâriability. 2.8.8 Breakthrough Pilot-scale investigations of C.parvzlm removal by filtration during turbidity breakthrough were performed by Patania et al. (1995). The authors studied turbidity breakthrough when filter effluent turbidities increased fiom 0.1 NTU to 0.2 NTU or higher. Substantial differences between oocyst removals during stable operation and breakthrough were not obsenred, although the authors speculated that oocyst removal might have detenorated if sampling had continued beyond the point at which fdter effluent turbidities were -0.2 NTU. Oocysts were detected in almost alI of the samples collected during the stable operation and breakthrough experiments at the particular pilot plant where ~ b i d i t ybreakthrough was investigated; replicate experiments strengthened the findings. Sinular results were obtained during two full-scale investigations reported by Baudin and Laîné (1998). Little deterioration in oocyst removals was obsemed during filter breakthrough at either plant. The authors indicated that oocyst removal during breakthrough at both plants depended on the filtration rate. Fluctuating filter influent oocyst concentrations at one plant during the stable operation experiment, unspecified filter effluent turbidities during the breaktbrough experiments, and a general lack of recovery idonnation, made it difficdt to draw inferences fkom the oocyst removal data collected during the breakthrough and stable operation experiments. Despite progress in the development of disinfection technologies that c m achieve reasonable C.pamîm inactivation, the traditional physico-chemical processes used in drinkkg water treatment remain critical to achieving desirable Ievels of CrptosponXum removal. One of these important barriers is filtration. Currently, the amount of oocyst-removal protection provided by filtration processes is not M y known (Solo-Gabriele and Neumeister, 1996). Full- and pilot-scale C. pawum removals for a variety of filtration regimes have been reported anywhere in the range fiom 2- to 3-log (Nieminski and Ongerth, 1995; Fox et al., 1998) to X-log (Baudin and Laîné, 1998; LeChevallier et al., 1991~). Although several challenge studies evaiuating the ability of specific filtration regimes to remove C.pamrrn during optimized treatrnent conditions exist in the literature, bmited data are available regarding the C. p a m m removal capability of granular media filters duMg vulnerable periods of operation. Moreover, a strategy or framework for assessing and describing the overall ability of granular media fïlters to remove C.p a m m (and other pathogens) during periods of both optimized operation and process upset is lacking. Such a fiamework would be instrumental in providing guidance to water treatment professionals and regdators. Many of the limitations of available C- parvum removd data are due to the limitations of analytical methods for the concentration and identification of oocysts fkom naturd waters. Differences in analytical reliability, processed sample volume, rnethod detection limits, and influent microorganism concentrations can d l contribute to inter-study Merences in C.parvum removd by filters. Many of the C. pantum removal studies sunmarized in Table 2.4 employed analytical methods that were similar to the proposed ASTM (1993) method. The limitations of this method were underscored by a blind survey of the accwacy and reproducibility of the method which revealed that 6 of 12 laboratories failed to detect any oocysts in spiked sarnples and that the recovery among laboratones ranged fiom only 1.3 to 5.5% (Clancy et al., 1994). Several alternatives to the ASTM protocol have been examined, the most notable of which is USEPA Method 1622. Method 1622 offered significant improvements over the proposed ASTM and ICR methods; it demonstrated mean oocyst recoveries of >70% in early trials (Clancy et al., 1997; Bukhari et al., 1999). Subsequent studies of Method 1622, such as the USEPA validation experiments, yielded recoveries of approximately 35% with 13% relative standard deviation (Clancy et al., 1999)- Aithough this method is a clear improvement over previously used rnethods, these data indicate that considerable analytical limitations continue to challenge the evaluation and interpretation of C. pawwm concentration and removal data. Untii a reliable method for C.p a m m analysis is available, the challenges associated with the concentration and enurneration of oocysts fkom natural waters will undoubtedly result in the continued application of numerous methods for C.p a m analysis. The development of new methods is supported within the framework of Method 1622. While method development is critical, the use of various methods with variable and low to moderate recoveries makes it difficult to interpret and synthesize the results fiom multiple C. parvzrm removal studies. The lack of reliable analytical methods particularly necessitates the need for accurate and thorough description of experimental conditions and methodologies so that C.parvurn concentrations and removals by filters and other processes can be thoroughly assessed. Describing the methodologicai limitations of C- parvum removal data is critical to both interpreting the data and to quantitatively describing the uncertainty associated with the data, Although it is commonly accepted that mean concentrations or rernovals used to calculate cadidence intervals do not account for ail of the uncertainty in the C. p a m m data (Le. they do not account for sarnpluig strategy, method recovery, variabiiity associated with the anaiytical method), few studies of C r p r o s p ~ ~ d i uconcentration m and removal efficiency have addressed the issue of data reiiability beyond stating average analytical recovery. Incorporation of this uncertainty into a statement of data reliability like a confidence interval can be quite complicated because the normal distribution is often inappropriate for describing distributions of microorganisms such as C~ptosporidizrmin water samples. Haas and Rose (1996) showed that naturally occurring oocyst densities could be described adequately by the Poisson distribution. ParMiurst and Stem (1998) also suggested Poisson confidence intervals for describing oocyst data. Atherholt and Korn (1999) presented Poisson methods for sample counts in the context of the K R protocol and suggested that a distribution more complex than the Poisson distribution may be required to account for the various errors in the analytical process; Nahrstedt and Gimbel (1996) developed such a statistical framework. The disadvantage of the Nahrstedt and Gimbel (1996) model is its increased level of complexity, which resulted in a computer program rather than a simple statistical table; the scope of this approach was also Iunited to calculating confidence intervals on Cqptospondiurn concentrations, not removals. Developments such as the Nahrstedt and Gimbel (1996) model represent tremendous progress toward handling C . p a m m concentration data in a statistically rigorous mamer. Accessibility to and application of such tools by water treatment professionals is still lacking, however, perhaps due to a need for tools to evaluate oocyst removals by treatment processes, rather than concentrations. The C. p a m m removal efficiencies of treatment processes can be evaluated by monitoring the removal of indigenous oocyst concentrations through the treatment processes or by seeding the processes with C. parvurn oocysts or surrogates. Mthough most representative of real operating conditions, monitoring the removal of indigenous oocysts is not cwrently feasible due to methodological limitations that preclude the processing of the very large water volumes (often »IO00 L) necessary to achieve reliably countable concentrations of oocysts in waters such as filter effluents. The other alternative for assessing treatment process removals of parvum is to seed influent waters with concentrations of oocysts hi& enough to yield reliably countable numbers of oocysts in the treated waters; such evaluations can also be performed with surrogates for oocyst removal. Several different types of surrogates for the removal of viable C. parvzrrn oocysts by drinking water treatment processes have been evaluated- These include türbidity, particle counts, HPCs, aerobic spores (typically Bacillus subtilis), UV,,, DOC, polystyrene - DOC) and microspheres. Of these potential surrogates, some are inadequate (HPCs, W,,, others are indicative of treatment efficiency but not oocyst removal (turbidity, particle counts, aerobic spores), the remainder need to be M e r evaluated (polysiyrene microspheres). Due to the lack of adequate surrogates for the removal of viable C. panlzrm oocysts, chemically inactivated oocysts have been commonly used for treatment evaluations. Differences in surface charge, described by zeta potentiai, exist between chemicdly inactivated and viable oocysts of C. p a m m (Lytle and Fox, 1994); it is not presently known whether or not these differences are substantial enough to result in significantly different removals by grandar media filtration processes. Information related to the relationship between viable and inactivated oocyst removals during filtration is necessary for proper interpretation of research evaluating process removals of C.pawurn. Unfortunately it is usually impossible to evaluate the C. parvrrm removal efficiency of granular media filtration processes with indigenous oocyst concentrations because existing analytical methods lack the ability to process the volumes of water typically necessary to achieve reiiably countable nurnbers of oocysts in filter effluents. Process evaluations that involve seeding (or spiking) atypically high concentrations of oocysts must therefore assume that the removals achieved with high Ilifluent concentrations are the sarne as those that would be achieved with indigenous influent concentrations. The conclusions that c m be drawn fiom C. parvum removal studies are therefore limited in that they are not necessarily representative of those that would be obtained under similar operathg conditions with indigenous oocyst concentrations. The work described in this thesis endeavored to address some of the research needs discussed above. The primary focus of this thesis research was to assess the removal of C.parvum oocysts by granular media fi1tration processes relative to rneasurements of turbidity, particle counts, and pot ential surrogates (polystyrene microspheres) during various phases and events occurring throughout typical filter cycles. A full understanding of the ability of filters to remove C.pawum (or any other pathogen) could only be attained when the reliability of experimentally obtained data was understood and quantified. To that end, a statistical tool was developed to describe the reliability of C. pamrrn concentration and removal data collected during treatment process challenge studies by integrating methodologjical uncertainty with oocyst removd data. The reliability of the data collected during this thesis research was demonstrated with that quantitative tool. The relationships between on-Line performance parameters and oocyst removal were evaluated and applied to the development of practical treatment strategies for maximizing C. parvurn removal by granular media filtration. Polystyrene microspheres were also assessed as potentiai surrogates for M e r studies of C.pavvum removal by filtration. The experimental approach employed d u ~ this g research inchded defining the removals of pathogens and surrogates during vuluerable periods of filter operation, relating them quantitatively to removals during stable operation, and investigating design and operational strategies for maintainhg removals as hi& as possible during potable water production. The research approach inciuded three experirnental tasks designed to provide concrete outcomes of prôctical value to the water industry while elucidating some of the fundamental mechanistic processes governing the removal of rnicrobiological particles such as C.parvurn by filtration processes. These tasks included: bench-scale experiments investigating the applicability of formalin-inactivated C.p a m m as sunogates for viable C-parvum during studies of oocyst removal by filtration, pilot-scale experiments investigating the influence of design and operational effects on C.parvum and potential surrogate (polystyrene microsphere) passage in benchmark systems, and pilot-scale experiments investigating mechanistic issues associated with C. parvurn removal by filtration. experimental components of this research are surnmarized in Table 3.1. This table is organized according to the research platform at which the experiments were perforrned. Table 3.1 Surnmary of bench- and pilot-scale experirnents Research platform bench-scale - . - Experhentai objectives compare removd of viable and iuactivated oocysts by fiItratïon Operational conditions stable operation ripening compare dual- and tri-media coaggation failure Ottawa pilot-scale BASE CASE compare several operating conditions to stable fiiter operation stable operation (cold and t conditions) m water stable operation d h g spring runoff filter ripening evduate microsphere removal as surrogate for Cr~ptosporidiumremoval by filters sub-optimal coagulation coagulation failure (no alum/silicate) after a period of stable operation coagulation failure (no aldsilicate) since backwzh partial coagulation failure (no silicate) hydraulic step end-of-nin (frlter effluent turbidity c0.1 NTU) early breakthrough (filter effluent turbidity 0.1-0.3 NïU) late break-through (filter effluent turbidity XI-3 NTU) Windsor pilot-scale compare effects of different raw water qdty stable operation compare constant and declining rate filtration U W pilot-scale compare dual- and tri-media stable operation evaluate microsphere removal as surrogate for C.parvurn removal by filters hydraulic step The fÏrst experimentai phase of this thesis research was designed to demonstrate whether or not comparable removals of viable and formalin-inactivated oocysts could be expected by filtration at various operating conditions encountered during typical water treatment. Given the requirements associated with the handling and disposai of viable oocysts, these experiments were planned for bench-scale at the University of Waterloo. Specifically, the experirnents were designed to examine the relative removals of oocysts by dual- and tri-media filters during stable operation, ripening, and coagulation failure. Both dual- (anthracitekmd) and tri-media (anthacite/sand/garnet) filters were studied to examine media effects on removals of viable and inactivated C. p a m m oocysts. Stable operation was evaluated first because it represented baseline removals at optimal operating conditions (e.g., optimized chernical pretreatment, particle removal, etc.), to which other operating conditions with possibly compromised C.p a m m removal would be compared. Coagulation failure was selected as one such compromised operating period It was considered a critical vulnerable period for study because oocyst surface charge is least affected by the presence of coagulant during this period. Any differences between filter removals of viable and inactivated oocysts would most likely be demonstrated during this period. Ripening was evaluated because it typically represents a brief penod of hi& turbidity and particle passage through most filters and it occurs during most filter cycles. It is also a penod during which differences in the attachment ability of oocysts may affect their passage through filters; however, this effect was considered less likely to result in a difference between removals of viable and inactivated oocysts since coagulant is typically present during ripening. Particle and turbidity breakthrough also would have been interesting to study; like rïpening, this penod is associated with a relatively rapid detenoration in filter effluent quality. Unfortunately, availabie head limitations precluded the study of this operating condition during this phase of the experiments. The second experimental task evaluated the removals of C. pawurn, potential surrogates (polystyrene microspheres), and performance indicators (turbidity, particle counts, and B. szrbtilis spores) in benchmark systerns ( i e . , typical design and operating conditions). As is outlined in Table 3.2, they were conducted at the Ottawa, Windsor, and University of Waterloo (UW) pilot plants in Canada. Inactivated Cryptosporidizrm parvum oocysts and pure-cultured Bacillus subtiïis spores were seeded at these locations; in addition, fluorescent polystyrene microspheres were seeded during some of the experiments at Ottawa and UW. The experirnents were designed to assess the C. parvum removal that could be reasonably expected fiom typical rapid filtration systems and to evaluate the sensitivity of C.p a m m removal during typical filter cycle events and conditions of process stress. Experiments were developed to capture most of the common (and some less common), dynarnic events that occur during typical water treatment operation. Seven basic conditions were investigated duruig these experiments. They were: 1. stable filter operation, 2. no coagulation, 3. sub-optimal coagulation, 4- ripening, 5. hydraulic step, 6. endofrun, and 7. breakthrough. Each condition was studied at least in triplicate; several sub-categories of the seven general conditions were also examined. Most of the experiments were conducted at Ottawa, allowing for the reiative cornparison of operating conditions; they are summarized in Table 3.2 along with their rationale and specific experimental conditions. As was noted in Table 3.1, additional experiments were performed at UW and Windsor that permitted the evaluation of different filtration regïmes (constant vs. declining rate at Windsor) and different media 0 and raw water (Windsor) types. Table 3 2 Operating conditions examined during pilot-scale experiments at Ottawa Operatinp Condition Stable operation Sub-optimal coagulation Specific ~xperiments* warm water Evperimental Rationaie defines highest attainable removal and acts as baseline to which other operating conditions are compared cold water (T < 3°C) defines highest attainable removal under cold water conditions spring nmoff gauges effect of sudden water quality change seediag at rapid mir establishes if pre-coagulation in a jar followed by subsequent seeding to filter influent is representative of pilot-plant coagulalion and subsequent filtration no co@ants baclcxash deterinines extent of coagulant impact on filtration (Le,. worst case scenario) since temporary coagulation failure addresses effect of temponry coagulation failure (e-g., pump failure) no pre-coagulation of pathogen seed suspension determines if coagulant in seed suspension (fed into filter influent) impacts pathogen removal pre-cos-dation of seed suspension only (no plant coagulation) determines if coagulant in seed suspension impacts pathogen removal without plant-level coagulation coagulant underfeed demonstrates the role of coagulant loss of coagulant aid addresses the relative importance of coa-dant aid in pathogen removal (Le., settling aid) pezk filter effluent turbidity determines pathogen passage associated with the operational period that potentially represents one during ripening of the most si-pificant types of deterioration in filtrate quality Hydraulic step sudden increase in flow determines any deterioration in pathogen removd associated with rapid changes in filtration rate (e.g., filter out of service) that are h o w n to cause deterioration of fiïtrate qwlity End of nin first sign of change in filter effluent turbidity (c0-1 NTU) determines when pathogen passage increases relative to b a s e h e at the end of a filter cycle and relative to indicators like turbidity Breakhrough early break-through (0.1 - 0.3 NTU) late breakthrough detennines degree of pathogen passage at the end of a filter cycle relative to small changes in indicators like turbidity The filters at the Ottawa and Windsor pilot plants contained media depths and sizes typical of the utilities' full-scale plants (and typical of many existing treatment plants). The operational mode chosen was essentially conventional treatment with inclined plate sedimentation and dual-media filtration. The filter media specifics at the UW pilot plant were also typical of many existing treatment plants. These filters were operated in an inline floccuiation (contact filtration) mode and included both dual- and tri-media filtration. Further operating details regarding the research platforms are specified in Section 3.2. It shodd be noted that this thesis research was funded in part by the American Water Works Association Research Foundation (AWWARF) - without this funding, such an extensive investigation as described in this thesis would not be possible. A subs-tantial portion of the C.parvum and B. subtilis experiments at Ottawa concurrently contnbuted to the AWWARF report "Filter Operation Effects on Pathogen Passage," by Huck et al. (2001). The collaborative effort associated with that project contributed directly to the experimental design at Ottawa, as well as the seeding and sampling protocols descnbed in Section 3.3. Collaboration with the Metropolitan Water District of Southern California provided a starting point for the C. parvurn analytical method refinement descnbed in Appendix A (Yates, 1997; Yates et al., 1997; Huck et al., 2001). The majority of the pilot-scale experiments were perEormed at Ottawa where the raw water required a relatively hi& coagulant dose designed for combined TOC and particle removal. Virtually identical to the pilot plant in Ottawa in design and construction, the Windsor pilot plant also employed a relatively high coagulant dose; however, the raw water quality was substantially different. In comparison to Ottawa, the raw water at Windsor had a considerably lower TOC and higher alkalinity and turbidity (which could spike as high as 350 NTU during the spring). The raw water treated at UW was synthetic, comprised of dechlorinated tap water with kaolinite-induced turbidity. The coagulation regime at UW required a low coagulant dose designed for particle removal. n i e pilot-plant process configurations and nominal raw water qualities are listed in Table 3.3 and Table 3.4 respectively. Table 3.3 Process configurations at the various research platforms Location Experimentai Scale Ottawa Windsor pilot pilot pilot 16 (8/train) 0.5 Design capacity (gpm) Preoxidation Chlorine Waterloo da Chemicals AIum Sodium Silicate Perchol LT-24 yes no yes 5 mg1-L variable in-line da nia Rapid Mis G.sec-' Hydrauiic detention time (min) in-line 0.02 1.8 FIocculation G for stages 1.2. and 3 (sec-') Hydraulic detention time (min) nia da Sedimentation Hydraulic detention time (min) Filtration Loading (-1 (@y*) SurIàce area (ft-) Filter 1 operathg mode media (mm) ES (mm) Constant rate anthracitetsand 3501275 anthracite 1.07 sand 0.5 15 anthracite 135 sand 1-32 constant rate anthracitefsand 430/300 declining rate anthracite/çand 450/300 anthracite sand anthracite sand declining rateœ an?hracite/sand anthracite 0.98 sand 0.5 anthracite 1.5 sand 1.5 deciining rate* anthracite/sand 650/350 anthracite 0.98 sand 0.5 anthracite 1.5 sand 1.5 d e c h h g rate* anthracitelsandlgamet 3701i6W6 anthracite 098 sand 0.5 gamet 0.32-038 anthracite 1.5 sand 1.5 garnet 1.4 declining rate. anthracite/sand/garnet 600/300/100 anthracite 0.98 sand 0.5 gamet 0.32-03 8 anthracite 1.5 sand 1.5 gamet 1.4 508/203 Fiiter operation was essentiaily constant rate durulg the relatively brkfeqerimental periods. Table 3.4 Nominal raw water quality at the various research platEorms Location Experhental Scale Source water Qualitative description of source water Ottawa Windsor pilot pilot pilot bench Ottawa River Detroit River tap water tap water Few upstrearn inputs, some Iogging in past Waterloo Great Lakes Dechlorinated D e c h l o ~ a t e d tap water SUPP~Y tap water with with kaolinite kaolinite induced induced turbidity turbidity Temperame (OC) nominal value range TOC/DOC (mg/L) nominal value range Turbidity (NTU) nominal value range 14 c 1-40 spikes to 350 PH nominal value range Alkalinity (mg& as CaC03) nominal value range -37 15-40 . 90 85-120 - - 300-330 300-330 3.2.1 Bench-Scale Filtration Apparatus A schematic of the bench-scale filtration apparatus is presented in Figure 3.1. It included a g l a s filter column (50 mm in diameter) containhg one meter of media. The filter was operated in a constant head, declinutg rate mode during bench-scale experiments. The dual-media filter consisted of 700 mm of anthracite over 300 mm of sand- The tri-media filter consisted of 650 mm of anthracite over 250 mm of sand over 100 mm of gamet. Three layers of grave1 supported the media; each layer was 5 cm thick. Media specifics such as effective size (ES) and uniforrnity coefficient WC) are available in Table 3.3. AI1 of the media were al1 riffled to ensure uniformity between e x p e b e n t s . raw water seed suspension h re-circulating filter influent fdter effluent Figure 3.1 d dual-media or Bench-scale filtration apparatus. tri-media Penstaltic purnps (Cole Parnier, distributed by Labcor Inc., Ste. Anjou, Québec) were used to introduce the coagulated raw water and microorganism seed suspension to the filter. To ensure reasonable rnixing of the seed suspension with the influent water prior to filter intluent sarnpling, the seed suspension introduction point was situated approximately 75 cm (2.5 ft) above the filter media. A peristaltic pump was also used to re-circulate water fiorn approximately 5 cm (2 in) above the media; this acted as the filter influent sampling location. The fiiter effluent sampling location was located at the outlet of the column, direct1y after the support gravel. The process confi,wation and raw water quality details were iisted in Table 3.3 and Table 3.4 respectively. 3.2.2 Ottawa and 'Windsor Pilot Plants The Ottawa and Windsor pilot plants are vimially identical. The pilot plants are constructed only of stainless steel, g l a s and inert fluorocarbons. Raw water was typically pumped at 50-60 L/rnin to a constant heed tank. From the constant head tank, the flow was split between twc identical process trains. During this research, Side 1 was operated with pre-chlorination. Metering pumps injected treatment chernicals into the feed line; the chernicals were mixed in-line. The water then entered a 3-ce11 underjover flocculation tank. Flocculated water passed into an inclined plate sedimentation tank. The pretreatment units are shown in Figure 3.2. Settled water was collected in a settled water storage tank, before being sampled and fed to a dual-media tilter (anthracite/sand). The filter could operate in either the constant rate or declining rate mode. The filter columns are shown in Figure 3.3. Filtrate was collected in a backwash water storage tank with dedicated storage cells. Water for backwashing was then pumped fiom the dedicated ce11 back to the filter on the side of the plant fiom which it was collected. An air compressor and injection port allowed for air-scour during backwashing. On-line data (turbidity, particle counts, flow rates, and headloss) were recorded at I-minute (Ottawa) or 10-minute (Windsor) intervals by a SCADA data-logging program. The process configuration and raw water quality details were listed in Table 3.3 and Table 3.4 respectively. Figure 3.2 Pretreatment at the Ottawa Pilot PIant. Figure 3.3 Filter columns at the Ottawa Pilot Plant. 3.23 University of Waterloo Pi10t Plant The University of Waterloo pilot plant is constructed o d y of stainless steel, lucite, glass, and inert fluorocarbons. During this research, the University of Waterloo pilot plant was operated in direct filtration mode with in-Iine flocculatioa (contact filtration). It treated synthetic raw water comprised of dechlorinated tap water with kaolinite-induced turbidity. The raw water was coagulated in-Iine with aIum and then filtered by both dualand tri-media filters. The fibers were backwashed by simuItaneous air scour with subfluidization water wash (collapse pulsing) with dechlorinated tap water. The pilot plant is equipped with continuous on-line turbidity and particle count measurements that are recorded at 1-minute intervals. The process configuration and raw water quality details were listed in Table 3-3 and Table 3.4 respectively. Al1 of the bench- and pilot-scale seeding experiments employed continuous seeding of microorganisrns during specific points of the filter cycle. Almost al1 of the seeding experùnents conducted during this study involved microorganism seeding at the filter iduent; a very limited number of experiments exarnined microorganisrn seeding at the rapid mix. Prior to seeding, C. pawurn and B. subtilis concentrations in the stock suspensions were determined by triplicate counts on a hemocytometer (Petroff-Hausser Bacterial Counting Chamber, Hausser Scientific Corporation, Horsham, PA); this rnethod is descnbed in Sections 3-4.1 and 3A.2. Polystyrene microspheres were determined by the method provided by the manufacturer (Polysciences Inc., Warrington, PA); this method is discussed in Section 3.5. At Ottawa, the feedstock microorganism/microsphere suspension was prepared by adding the microorganism stock suspensions (typically -8 mL volume) and microsphere aliquots (typically -150 pL volume) to a 1.5 L seed suspension of chlorinated and quenched raw water. The suspension was added to raw water at Windsor and UW. Each via1 contaùiing the concentrated microorganism stock was rinsed into the feedstock ten times with chlorhatedlquenched raw water (Ottawa) or raw water (Windsor and UW). Additional samples were collected fiom the feedstock to confïrm microorganism concentration. At Ottawa and Windsor, the targeted filter influent concentrations of the seeded microorganisms were typically l 0 ~ - 1 0oocysts ~ per liter (C. panum) and 10"-1 o6 C N s per liter (B. subtilis). At UW, the targeted filter influent concentrations of the seeded microorganisms were typically 10' oocysts per liter (C.panrum) and 10'-106 CFUS per liter (B. subtihk). When microspheres were seeded at Ottawa and UW, they were seeded at concentrations comparable to those of the C.p a m oocysts. Microorganisrn pre-treatment was achieved via a jar-coagulation method described below (Section 3.3.1). In most cases, the jar coagulation conditions were identical to those in the pilot plant during the specific experirnents. Some experiments were perforrned to specifically examine the eEects of the coagulant added to the seed suspension (e-g., potential filter aid effects). During these experiments, the jar coagulation conditions were different than those in the pilot plant, The feedstock rnicroorganism suspensions were added at the filter influent (except in a few experiments where they were added at the rapid mix) immediately afier jar coagulation. 3.3.1 Pilot Plant Coagulation and Jar Coagulation Protocol During pilot-scale testing of stable operation, liquid aluminurn sulfate (alum, A12(S04)3-18H20)was used either alone or in combination with cationic polyrner (Windsor) or activated silica (Ottawa) to achieve the desired conditions for turbidity and particle removal. n i e cationic polymer or activated silica was dosed during rapid mix simultaneously with alum to improve charge neutralization. Chlorine was added at rapid mix at a dosage of -2 mg/L (ûtîawa). The chlorination dosage was to achieve the benefits of pre-oxidation, but was not necessarïly sufficient to rneet disinfection requirements such as those required by the Surface Water Treatment Rule (USEPA, 1989). The microorganism jar coagulation protocol mimicked the pilot-scale coagulation conditions; it is provided in Table 3.5. It included the use of sodium thiosulphate (Na2S2Q3)at a final 2:l molar ratio of NazSzQ3:C1to quench the chlorine residual in the microorganism/microspherestock suspension (to prevent any potential disinfection). Table 3.5 Jar coagulation protocol 1. While stirring pilot plant raw water ( 4 . 0 L) on a stir plate, add NaOCl to a final chlorine concentration of 2 mg/L (Ottawa). 2. Stir chlorinated sample on stir plate for 15 minutes (Ottawa). 3. Quench chlorine by adding sodium thiosulphate (Na2S203)to a final 2: 1 molar ratio of Na2S203:Cl,adding 2x more Na2S203as a safety factor (Ottawa). 4. Stir chlorinated/thiosulphate-quenchedsample on stir plate for 5 minutes (Ottawa). 5. Pour 1.0 L of chlorinated/quenched raw water (Ottawa) or raw water into a 2-L jar. 6. Vigorously shake and add the microorganism stock (-8 mL) to the 1.0 L of chlorinatedquenched raw water (Ottawa) or raw water. 7. Using an additional 0.5 L of chlorinatedlquenched raw water (Ottawa) or raw water, rime the microorganism spike container 10 times, adding a11 of the rime water to the 2-L jar (for a final volume of -1 -5 L). 8. Jar-coagulate the samples with the jar test apparatus. This step is site specific. Ottawa: - add alum - 5 sec @ 100 rpm - add silicate - 15 min @ 30 rpm - 15 min @ 15 rpm - 60 min settling Windsor: - add d u m 2 min @ 100rpm - add polymer - 10 sec @ 100 rpm 10min@20rpm - 10 min settling - W: - add alum - 60 sec @ 100 cpm - 15 min @ 50 rpm - 15 min @ 15 rpm 60 min settling - 9. M e r settling, gently mix jar coagulated suspension to break up any flocs. 90 33.2 Calculation of Microorganism and Microsphere Concentration and Removal Filter influent and effluent microorganism concentrations were determined by the same p a m m , if a reasonably analytical method for a given rnicroorganism. In the case of countable nurnber of oocysts was present (< -3000 per slide), the C. pawrcm concentration was simply calculated as the number counted per volume processed. During several of the initial experiments at Ottawa, filter influent C- panmm concentrations were estimated based on the examination of fifty random fields of view because the number of oocysts on the slides was typically hi& (-10' to 10' oocysts/slide). In these cases, the C. pawum concentration was calculated by the following general equation: microorganisms - microorganisms fields of view X volume processed field of view filter membrane X filter membrane (3.1) volume processed in which the n e b e r microorganisms per field of view is the average of fi@ fields of view. random The nurnber of fields of view per membrane is a function of the microscope, optics, and mapification utilized during the specific experiment. The last component of this equation is the sample volume passed through the filter membrane. This method of enurnerathg filter influent C.parvum concentrations was only used for a few experiments because it only provided estirnates of filter f l u e n t oocyst concentrations and required the assumption that the oocysts were unifomily distributed on the filter membrane. For most of the experiments, srnaller sample volumes of filter idluent were processed, allowing for enumeration of oocysts on the entire filter membrane. The approaches used to enurnerate the microspheres were the same as those used for C.pawum. Microorganism and microsphere removals (logio)were calculated by subtracting the log of the filter effluent concentration fiom the log of the innuent concentration. In the subsequent discussion, microorganisms are used as an example. When no rnicroorganisms were detected, the concentration was reported as 0; however, removal was calculated by using a concentration of I microorganism per sample volume processed. For example, a value o f 1 oocyst per L would be used in the Jog removal calculation if no oocysts were found in a 1-L sample. A value of 2 oocysts per L would be used in the calculation if no oocysts were found in a 500-mLsample (1 oocyst per 0.5 L equals 2 oocysts per L). 3.3.3 Bench- and Pilot-Scale Seeding ProtocoI A peristaltic pump was used to add the feedstock to the pilot plant filter influent water (and in some cases the raw water). The seed suspensions were introduced into the filter influent water approximately 2.5 feet above the filter media so that some mixing with the filter influent water would occur prior to filtration. Microorganisms were typically seeded into the filter i d u e n t for one hour at a rate of 25 ml/rnin. A limited number of experiments required microorganism seeding for longer periods of t h e (5 hours). This was achieved by reducing the seed flow to 5 mL/rnin. The microorganism seed suspension was continuously fed to the bench- or pilot-scale plant filter influent until the last set of filter influent and effluent samples was collected. In some experiments (e-g.. those exarnining detachment-related phenornena), additional samples were collected after microorganism seeding had ceased. A second peristaltic pump was continuously recirculated filter influent water fiom a few inches above the surface of the filter media. To prevent canyover of microorganisms between experiments, both this line and the microorganism feed line were flushed for at least ten minutes afier the completion of an experiment. These lines were also flushed for several minutes while the flow rates were confmed prior to the start of each experirnent. 3.3.4 Bench- and Pilot-Scale Sampling Protscol Microorganism samples were collected fiom the filter influent and effluent. The filter influent location was approximately 5 cm (2 in) above the surface of the filter media (Figure 3.4); the effluent was collected at the column exit immediately after passage through the support grave1 (upstream of the turbidimeter and particle counter). Prior to the seeding experiments, 1-L negative controls were collected fiom the filter influent and effluent. Additional QA/QC samples were collected from the microorganism feedstock suspension, The samples were collected in 250-rnL and 1-L glass Wheaton bottles respectively. AIiquots fiom the microorganism feedstock were collected in 5-mL glass chromatography vials. Al1 sampiing containers were washed, autoclaved, and rinsed with a few milliliters of a buffered detergent solution ( l x phosphate buffered saline PBS] with final concentrations of: 0.1% sodium dodecyl sulphate, 0.1% Tween 80, and 0.01% Sigma Antifoam A and final pH of 7.4) prier to use. The excess surfactant solution was discarded and sodium thiosulphate (Na2S203)was added to each sample bottle for a final concentration of 0.0 1%. The bench-scale experiments were performed during the fïrst four hours of filter operation. Either viabie or formaiin-inactivated C.parvum oocysts were seeded for one h o u , with samples collected 15, 30, 45, and 55 minutes after the start of seeding. Additional seeding and sampling details for the bench-scale experiments are provided in Chapter 5. Most of the pilot-scale experiments were performed during the early to mid portion of the filter cycle after at least four hours of filter operation; ripening and breakthrough conditions were an exception to this criterion. With the exception of ripening, al1 experiments were conducted after a period of stable operation during which filter effluent turbidities were continuously below 0.1 NTU. Microorganisms (and sometimes rnicrospheres) were typicaily seeded for one hour, with samples collected 15, 30,45, and 55 minutes after the start of seeding. Seeding and sampling information for the pilotscale experirnents at Ottawa, Windsor, and UW is summarized in Table 3.6. Additional seeding and sampling specifics for the pilot-scale experïments are presented in Chapter 6. Figure 3.4 Filter influent sampling location at Ottawa. Table 3.6 Pilot-scale seeding and sampling specifics Condition Stable operation 1-hour seeding period (0-60 minutes) seeding during first haIf of filter cycle after at least 4 hours of operation filter effluent turbidiw c 0.1 NTU Sub-optimal coagulation 1-hour seeding period (0-60 minutes) samples collected at i5,30.45, and 55 minutes (Ottawa and Windsor) sarnples collected at 20.40. and 55 minutes (UW) samples coIlected at 15,30.45, and 55 minutes (Ottawa) seeding during first haIf of filter cycle after at least 4 hours of operation 1-hour seeding period (0-60 minutes) samples collected at 15,30,45. and 55 minutes seeding during first halfof filter cycle after at Ieast 4 hours of operation Ripening 30-minute seeding period capturing both samples collected at 5, IO, 15.20, before and afier peak filter effluent and 25 minutes turbidity during ripening Hydraulic step seeding during fmt balf of filter cycle after at least 4 hours of operation Ottawa rnv hydraulic step initiated at 300 minutes samples collected at 280,295. 300,305,3 10.320, and 360 minutes 1-hour (0-60 minutes) C.parvrlrn seeding C.parvum collected at 15. 60,70, and 80 minutes 1-hour (60- 120 minutes) B. subtilis seeding B. subrilis collected at 60,70, and 80 minutes 5-hour (300 minute) seeding period hydraulic step initiated at 65 minutes End of nui 1-hour seeding period (0-60 minutes) samples collected at 15,30,45, and 55 minutes BreActhrough 1-hour seeding period (0-60 minutes) samples collected at 15,30,45, and 55 minutes 1-hour (-60-0 minutes) B. subtilis seeding samples collected at 15,30,45, and 55 minutes December 1999 experiments 1-hour (0-60 minutes) C.parvum seeding 33.5 Microorganism Losses to Seeding Apparatus Control experiments were conducted to determine microorganism losses to the seeding apparatus and pilot plant equipment. Performed at Ottawa and UW, these experiments consisted of removing the media and seeding the columns with microorganisms (C.p a m m and B. subtilis) to determine system losses. Al1 of the media were removed at UW; at Ottawa 2 to 3 cm of grave1 could not be removed. No coagulants were used in the pilot plants or the jar coagulation step during these experiments. Given that the Windsor and Ottawa pilot plants are identical in construction, it was assurned that the Ottawa control results were applicable to Windsor. Similarly, the UW pilot- and benchscale columns were aIso almost identical and the pilot-scale control data were considered applicable to the bench-scale system. The control experiments at UW demonstrated 0.03 * standard deviation) of * 0.06 and 0.04 * 0.05 log loss (mean C. panrurn and B. subtifis respectively. Despite the few remauiing centimeters of grave1 at Ottawa, the C. p a m m and B. szrbrilis losses were only 0.09 *0.12 and 0.15 k 0.06 log respectively. Overall, these results suggested that the oocyst and spore removals observed during the subsequent pilot scale experiments were due to filtration, with only minimal system losses. Combined with the analytical method recovery data discussed below, these data demonstrated considerable reliability of the microorganism removal data presented in this thesis. The probability density functions for C.p a m m losses at Ottawa are presented in Figure 3.5. The loglooocyst loss was based on four replicate samples; however, since one of the samples resulted in negative removals, it was treated as O-log removal. confidence intervals (with highest postenor density The 95% - HPD) for each of the data points depicted in Figure 3.5 were calculated. The range between the lowest and highest of these values represents the range of removals, or in this case losses, that could be expected during this operating period given the uncertainty resulting fiom the anaiytical - method. Based on this analysis, it can be concluded that system losses were minimal, approximately O- to 0.25-log. I I-t l = 15 minutes / ---__-t = 45 minutes : !- -t = 55 minutes: Range Behrveen Highest and Loweçt HPD Regions: O- to 025-109. 30 C.parvum Removal (loglo) Figure 3.5 3.4.1 C. p a m m losses to seeding apparatus and equiprnent during no coagulant and no media control experiments. C. parvum C parvuni inactivation and preservation Samples of C. parvum were preserved in a penicillin/streptomycin solution because recent research has indicated that oocysts stored in this preservative may more closely represent oocyst behavior in the natural environment (Li et al. 1997). Obtained from a commercial laboratory (Waterbome, Inc., New Orleans, LA. or University of Arizona, Department of Veterinary Science, Tucson, AZ.), the oocysts were bovine in ongin and were provided in a clean, purified form. For each experiment, -10' (Ottawa and oocysts were obtained; they were inactivated with 5% formalin Windsor) or 107 0 (final concentration) for C.p a m m in approximately 8 mL of 1x PBS with 0.01% Tween 20 to prevent oocyst clurnping. Al1 microorganism stocks were refiigerated at 4OC in the dark until use. Enurneration of spike suspension Prior to C.p a m m seeding, the stock suspension was bnefly vortexed and a small portion of the suspension (< 100 pL in total) was removed to enurnerate the oocyst concentration. The stock concentration was determined by averaging triplicate counts with a hemocytometer (Petroff-Hausser Bacterial Counting Chamber, Hausser Scientific Corporation, Horsham, PA). The entire grid (1 mm2) was used for oocyst enumeration at 4 0 0 ~magnification (Nikon Labophot 2A, Nikon Canada Inc., Toronto, ON or Zeiss Axioskop 2, Empix Imaging, Mississauga, ON). The oocyst concentration in the spike suspensions was detennined by the equation below. number of oocysts/L = 1 1mm3 nurnber oocysts counted x x1 mm' grid SOmmdepth ImL (3 -3) Filter influent and effluent concentrations of C. parwm were rneasured. Filter influents were analyzed in 100, 10, 5, and 2.5 mL volumes. Filter effluents were analyzed in volumes ranging fiom 5 mL to 1 L, depending on the operating condition studied. Sample volumes were chosen to yield between 10 and 2000 oocysts per membrane. Al1 pipettes and glassware were pre-rinsed with the buffered detergent solution (described in Section 3.3.4) to prevent oocyst losses. Samples were filtered through 25 mm, 0.40 pm polycarbonate membranes (Coming, Acton, MA). The filter membranes were placed on top of 25 mm diameter, 8.0prn nitrocellulose support membranes (Millipore Canada Ltd., Nepean, ON) placed on a manifold (Hoefer Scientific, San Francisco, CA) and maintained at a vacuum of 5 in Hg. Weights held the membranes in place. Two milliliters of 1% bovine serum albumen @SA) were passed through the filter membranes, then the samples were filtered. The glassware that had contained the sarnples was then rinsed with the buEered detergent solution. This was followed by filtration of 2 mL of BSA and then the immunofluorescence assay ( F A ) described below. The membranes were kept wet with l x PBS and covered until mounting on slides. The concentration and enumeration protocol is presented in Appendix A. A schematic of the direct vacuum filtration apparatus is provided in Figure 3.6. Figure 3.6 Direct vacuum fdtration apparatus for processing C. parvum. Al1 C. parvum identification was performed using the Hy&ofluorm Combo Cryptosporidium and Giardia Kit (Strategic Diagnostics, Newark, DE). Presumptive magnification at microscopic analysis for C.pawurn enurneration was perforrned at 4 0 0 ~ the University of Waterloo (Nikon Labophot 2A, Nikon Canada Inc., Toronto, ON or Zeiss Axioskop 2, Empix Imaging, Mississauga, ON); the fluorescein isothiocyanate @TTC)-stained oocysts did not fluoresce with sufficient intensity for enmeration at lOOx magnification (Nikon Labophot ZA, Nikon Canada inc., Toronto, ON or Zeiss Axioskop 2, Empix Imaging, Mississauga, ON). A limited number of slides were shipped to a commercial laboratory (CH Diagnostic and Consulting Services, Inc., Lovehnd, CO), for presumptive rnicroscopic analysis as a component of the quality assurance and quality control (QA/QC) program. Recovery The C. parvum data reported throughout this thesis are presented in both raw and adjusted (for method recovery) foms. The statistical method for adjusting concentration and Iog removal data for analytical recovery and calculating the endpoints of the associated confidence intervals is developed in Chapter 4. Recovery experiments were performed to determine the parameters ( a and b from the Beta distribution) necessary for descnbing the methodologicai recovery profile and integrating this information into a confidence interval that describes the reliability of the C. p a m m removal and concentration data. C.pawum recovery experiments were performed on water fiom the Ottawa and UW pilot plants. Recoveries were based on at least five replicate samples of oocysts spiked into filter Muent and effluent fiom each pilot plant; spike concentrations were deternined with a hemocytometer (as descnbed above). The processed sample volumes were those typically used during the experiments (1-10 mL of innuent and 0.5-1 L of effluent). C-parvum recoveries for Ottawa and UW waters respectively ranged fiom 56 to 86%, and 51 to 93% with overall mean recoveries of 74 and 73%. The detailed C.parvurn recovery data are presented and discussed in Appendix A. The hi& oocyst recoveries were likely attributable to the direct filtration rnethod that did not require elution steps. The use of this method was possibIe due to the high seeded oocyst concentrations and associated small sample volumes. There were no substantial deviations in recovery between filter influent and effluent samples, so the resulting log removals were essentially unaffected by methodological recovery. Sample handling, identification, preservation, transportation and storage were completed according to estabkhed procedues as described in Section 3.4.1 and following the ICR Methods for Protozoa Analysis (USEPA, 1996). Samples were transported on ice and shipped via overnight courier (if necessary) for processing. The samples were processed immediately or refkigerated at 4OC as specified in the ICR methodology. The quality assurance/quality control ( Q N Q C ) program ensured that accurate protozoan data were produced. The program iacluded recovery studies and the analysis of negative controls fiom the waters studied, method bianks, and positive controls. Analyses of method blanks and positive controls were performed each tirne samples were processed for C.parvum. Filter influent and effluent negative controls were collected during each experiment. No C. p a m m oocysts were found in any of the method blanks or filter influent and effluent negative controls. The control data suggested appropriate sample handling during the C. p a m m analyses. The lack of oocysts in the filter influent and effluent negative control samples suggested no substantial outside (non-seeded) sources of oocysts. Detailed C.pawurn QA/QC data are available in Appendix B. To ensure comparable slide readhg of processed C. panwm samples, several samples were read at both the University of Waterloo and a commercial laboratory (CH Diagnostic and Consulting Services hc., Loveland, CO). Lirnited cornparisons included sending slides twice to each laboratory. Table 3.7 sumrnarizes the results of the blind inter-Iaboratory slide reading cornparison; these data suggest general agreement between both laboratones over the entire range of counts observed during this study. Although some differences were observed between samples counted at the two laboratories, the replicated counts at a given Iaboratory were generally intemally consistent. As is demonstrated in Table 3.7, the relative impact of the inter-laboratory differences in counts typically had a negligible impact on logto counts which were used to calculate the loglo removals that fonned the basis of the conclusions in this thesis. Table 3.7 C.parvum QA/QC data comparing slides read at the University of Waterloo and CH Diagnostic and Consulting Services Inc. (Loveland, CO). Sample Filter Filter effluent UW Count 1" count zndc o u t (#/slide) (#/siide) 6542 -- 76 1 14 24 20 19 43 O O 23 19 45 - CH Diagnostics Count l* count zndcount (#/slide) (#/dide) 6780 - 23 23 - 3 53 O 45 O O -- 790 9 14 DBerence to UW 5~veragelog difference is calculated for samples read mice by each laboratory. ( *%) 4 Relative Log DEerence (log UW Iog CH). -0.02 - -4 36 21 -15 92 -11 O O O B. szrbtilis HundZing Samples of B. subrilis spores were stored in superclean (Milli Q TM) water. The spores were derived fiom vegetative ceEs of B. subtilis grown in a nutrient broth consisting of 25% soil extract (77 g afican violet soil and 0.2 g NarC03 to200 mL water). The B. szrbtilis stocks were refkigerated at 4OC in the dark until use. Enunteration of Spike Suspension Prior to pilot-scale B. mbrilis seeding, the stock suspension was briefly vortexed and a small portion of the suspension (< 100 pL in total) was removed to enumerate the oocyst concentration. The stock concentration was determined by averaging triplicate counts with a hemocytometer (Petroff-Hausser Bacterial Counting Chamber, Hausser Scientific Corporation, Horsham, PA). The entire grid was (1 mm2) used for spore enurneration at 4 0 0 ~magnification (Nikon Labophot 2A, Nikon Canada Inc., Toronto, ON or Zeiss Axioskop 2, Empix Ima-@ng, Mississauga, ON). The hemocytometer formula used to calculate the C. parvum concentrations (Equation 3.1) was also used to calculate the B. subtiZis concentrations. The analysis for B. subtilis (ATCC 605 1) was performed according to a method described by Rice et al. (1996). This method consists of filtration of samples ont0 47 mm diameter, 0.45 p m gridded cellulose acetate membranes (Pal1 Gelman Corporation #66278, Ann Arbor, Mi) and growth at 37OC for 24 hours on plates of nutrient agar with trypan blue (0.OljgL). Spores were identified by their blue color. Typically, duplicate sample volumes of 90 mL and 1.0 L were used to enumerate filter influent and effluents respectively. Serial dilutions were performed as necessary. FluoresbriteTM carboxylated YG fluorescent-dyed polystyrene microspheres (Polysciences Inc., Warrington, PA.) were used as non-biological surrogate indicators for C. pawum removal. In a limited number of experirnents evaluating detachment durhg breakthrough, Fluoresbrite- caboxylated BB fluorescent-dyed polystyrene microspheres (Polysciences Inc., Warrington, PA.) were also used. The microspheres had an average diameter of 4.675 I 0.208 p m and a density of 1.045 p/mL. The YG dye is a proprietary chernical that is hydrophobic (to prevent dye leaching fiom the particles into the aqueous phase) and matches the fluorescence filter settings of fluorescein (i-e., maximum excitation at 458 nm and maximum emission at 540 nrn; same as FITC for C. p a m m ) . The BB dye is also a proprietary chernical that is hydrophobic and matches the fluorescence filter settings of coumarin (Le.. maximum excitation at 365 nm and maximum emission at 468 nm). The Material Safety Data Sheets (MSDS) provided by the manufacturer indicated that neither of the microsphere products contained any hazardous components. The manufacturer provided the polystyrene microspheres in suspensions of 2.5% aqueous solids in de-ionized water; neither biocides nor stabilizers were added to the suspensions. The microspheres were stored at 4°C until their use. The wei&t to volume packaging allowed for the calculation of the particle concentration per milliliter by the followfng equation ~olysciencesInc., 1995): Number of particles per mL = 6Wx10E pxxx03 W = grams of polyrner per rnL in latex (0.025 g / mL for 2.5% solids) 0 = diameter in microns of latex particles p = density of polymer in grams per mL (1.045 g / mL for polystyrene). (3-3) According to Equation 3.3, the concentration of a stock suspension of 4.675 Pm rnicrospheres was 4.45 x 108 spheres/mL. This concentration was also confirmed with hemocytometer enurneration (Petroff-Hausser Bacterial Counting Chamber, Hausser Scientific Corporation, Horsham, PA) and Equation 3.2. Due to the hi& microsphere concentration, 1/10 of the grid (0.1 mm2) was used in the enurneration process. The average microsphere concentration was 4 . 5 5 ~10' spheredml afier fifteen replicate analyses with the hernocytometer; these data are presented in Table 3.8. A two-tailed ttest (a = 0.05) failed to demonstrate a statistical difference between hemocytometereniunerated and weight to volume calculated microsphere concentrations; therefore, the manufacturer's weight to volume method was used to determine the volume of stock microsphere suspensions that was added to the feedstock suspension during seeding experiments. Table 3.8 Hemocytometer Enurneration of Microspheres Trial 1 Number of Microspheres in 1/10 ofgrid (0.1 mm') 936 Average Concentration Standard Deviation Mcrosphere Concentration (spheres/mL) 4.68 x 10' 4.55 8.88 x x 108 10' Microspheres were concentrated and enumerated by the same method used for C. parvum (described in Section 3.4.1), with the exception that antibody staining was not n e c e s s q for samples that did not contain oocysts. FITC-stained C. p u m m oocysts and YG microspheres shown at 400x magnif~cationin Figure 3.7, were clearly distinguishable. Although approximately the same size and shape, the microsphere appears larger than the oocyst in Figure 3.7, as a result of the halo-effect associated with the strong intensity of the YG dye. This intensity pemitted microsphere enumeration at Z OOx magnification, as demonstrated in Figure 3.8, which depicts the BB microspheres. Microspheres were enumerated at lOOx magnification in samples that were collected prior to C. p a m m seeding, othenvise they were enumerated concurrently with C. p u m m oocysts at 400x magnification (FITC-stained oocysts did not fluoresce with enough intensity for enumeration at 1OOx magnification). C.parvum Figure 3.7 C. purvum oocysts and Y G polystyrene microsphere (400~ magnification, Nikon Labophot 2A, Nikon Canada Inc., Toronto, ON). polysryrene microspheres Figure 3.8 BB polyswene microspheres ( 1 0 0 ~magnification, Nikon Labophot SA, Nikon Canada Inc., Toronto, ON). 3.6 PHYSICAL AND CHE~~ICAL PARAMETERS 3.6.1 Eeadloss Differential pressure transducers continuously measured headloss. 3.6.2 Particle Counts A standard protocol was used at al1 sites to ver@ the calibration of the particle counters using commercially available, calibrated, mono-disperse polyrner microspheres (Duke Scientific Corp.; Palo Alto, CA.). Each particle counting instrument was calibrated (by the manufacturer) according to ASTM F 658-87 and met the resolution requirements of USP 788. The particle counters measured total particles from 2-150 Pm, with the data reported as total particles 22 jm- B R particle counters (BR, Grass Lake ML) were used at Ottawa and UW; Windsor used a Chemtrac mode1 PC2400D particle counter (distributed by SUMMA Engineering, Mississauga, ON). At pilot-scale, turbidity was monitored using on-lhe turbidimeters that were calibrated using dilute formarin solutions as specified by the manufacturer. Caiibration was routinely checked by cornparison with a bench-top turbidimeter (Ottawa, Windsor, and UW) with an accuracy of +2%, using standards of 0.80 and 6.6 W . Hach model 172OC turbidimeters (Hach Co., Loveland, CO.) were used at plant influent, filter influent, and filter effluent locations at Ottawa and Windsor; they were also used at the fi!ter effluent at UW. An additional turbidity meter was used at the filter effluent sampling location in Ottawa (ABB Model 7997/201, Calgary, AB). Filter influent turbidity was measured by g a b samples analyzed with a Hach Model 2100P hand held turbidimeter at UW (Hach Co., Loveland, CO.). This turbidimeter was routinely caiibrated with formazin standards of O S , 1, and 5 NTU. Sample pH was measured with grab sarnples analyzed by a pH meter that was calibrated daily, using pH 7.00 and 9.18 buffer solutions. In recent years, the protozoan pathogen Cryptosporidium parvum has been the causative agent of several outbreaks of waterborne illness, the largest of which resuIted in over 400,000 cases of cryptosporidiosis in Milwaukee in 1993 (MacKenzie e t al., 1994). Currently there is no effective treatment for crypto sporidiosis (ASM News, 1996). Consequently, these organisms are of significant concem to public health authorities and the drinking water industry. The environmental resiiience and ubiquitous nature of Cryptosporidizm oocysts have resulted in intensive international research on its occurrence in different water sources and its removal andlor inactivation by various water treatment processes. In the United States, the information collection rule (KR)required large utilities to monitor their surface water for pathogens such as Cryptosporidizmz and Giardia, as weii as viruses (USEPA, 1996). The i.iformation collected contributes to setthg potential dnnking water standards for these microorganisms. The surface water treatrnent rule (SWTR) addresses GiQrdia and viruses (USEPA, 1989). The interim enhanced surface water treatment rule (IESWTR) is a first step toward addressing Cryptosporidium (USEPA, 1998). Methods for quantifjhg the reliability of pathogen data are critical to the long-term success of such regulatory prograrns. This chapter proposes guidelines for reporting pathogen, and in particuIar Cryptosporidium, data. Some researchers have suggested strategies for reporting Cryptosporidium concentration data; however, these techniques are not directly applicable to the reporting of oocyst removals by treatment processes (as will Likely be useful for determining regdatory requirements). Statistical approaches for the calculation of confidence intervals, more appropriately called probability intervals, for both Cryptosporidizrrn concentrations and removals by treatment processes are developed below. For the purposes of discussion, the term "confidence intervaIsmwill be used. The strategies described are relevant for any microorganisms or discrete particles; however, they are discussed in the context of Cryprosporidi~mzbecause the diff?culty in sarnpling and measurement of this pathogen necessitate a rigorous approach to quantifying the reliability of such data. Cryptosporidium oocysts can be fkequently isolated f?om surface waters as well as occasionally from finished waters. It is difficult to precisely measure the occurrence and removal of oocysts fkom natural waters because they are present in varied and often low concentrations. Oocysts have been found in surface waters in concentrations as high as 10'' 1100 L and as low as 0.31100 L (LeChevallier and Norton, L995; Lisle and Rose, 1995; Smith et al,, 19911, when detected at dl. Traditional water treatment processes can decrease oocyst concentrations by several orders of magnitude, necessitating relatively large sample volumes of treated water so that statistically meaningful data can be O btained. A drinking water iimit of approximately one Cqptosporidi~rrnoocyst per 34,000 L has been suggested (Lisle and Rose, 1995). Such a low M t , or even one several orders of magnitude higher, would be difficult to implement because current methods for measming Cryptosporidiz~mconcentrations are unreliable, laborious, and expensive. Recovery efficiencies for these analytical rnethods are often low and highly variable (Nieminski et al., 1995). Increasing the volume of processed water to obtain higher counts of oocysts is often Wtcult, particularly in raw waters, due to the presence of other particles. The presence of other rnicroorganisms such as algae can further hterfere with oocyst identification (Rodgers et al., 1995). AIthough these methodolo~caldifficulties are relevant to sampling indigenous and seeded populations of oocysts, they are iikeIy more pronounced when quantifying indigenous concentrations, since high concentrations of oocysts (typicalIy 103-1o6 oocystsL) are commonly used in experirnental evaluations of treatment processes (Patania et al., 1995; Nieminski and Ongerth, 1995). Although it is also commonly accepted that mean concentrations or rernovds used to calculate confidence intervals do not account for all of the uncertainty in the C. parvzrm data (e.g., they do not account for sarnpiing strategy, method recovery, variability associated with the analytical method, etc.), few studies of Cryprospondizrm concentration and r e m v a l eficiency have addressed the issue of data reliability beyond stating average analytical recovery. The lack of information regardhg the reliability of C. parvurn data is likely associated with the expense and difficulty of processing samples. The complexity of incorporating these additional sources of variability into a statement of data reliability such as a confidence interval is increased by the fact that the n o d distribution is often inappropriate for describing distributions of rnicroorganisrns such as Cryptosporidi~imin water samples. 4-3 EXAMPLE DATA To illustrate the statistical methods developed here, several of the recovery data described in Chapter 3 and part of a data set collected during the course of experimental evaluations of Cryptospondirtm passage through drinking water filters were used. Table 4.1 includes three sets of recovery data originating from two different raw water sources. During the recovery studies, a hemocytometer was used to enurnerate stock oocyst concentrations so that they could be diluted to concentrations similar to those that would be expected during subsequent investigations (Le., evaluations of oocyst rernovai by filters). The recovery study sarnples were processed in the same volumes as were processed during the challenge studies described in Chapters 5 and 6. The samples were processed according to the C. parvum protocol described in Chapter 3. Table 4.1 Exampie C pawurn Recovery Data Water Seeded Sample Volume Seeded Type Concentration Volume Processed Number (oocysts/L) (L) (%) (oocysts) 1.OE+06 O, 1 0.5 500 Ottawa Filter Influent Observed Number (oocysts) 35 1 Recovery 70 Ottawa Filter Effluent Ottawa Filter Effluent Illustrative experimental data are summarized in Table 4.3. These data are part of a data set collected after pre-coagulated, formalin-inactivated Cryptosporidi~tnzoocysts were seeded into an anthracite/sand filter. In this example, stable filter was evaluated aî four time intervals. During each time interval, a filter influent and efftuent sample were collected. The influent sample volume (K) was 2.5 mL and the effluent sarnple volume (V,) was 1000 or 700 rnL. The oocysts were concentrated and enumerated by the smne direct vacuum fiiation/immunofluorescence assay described previously. Table 4 2 Example C.parvrlm Experimental Data Water Sample Type Volume Ottawa Filter Influent Ottawa Filter Effluent 4.4.1 3.5 Sample Tirne (minutes) 15 30 45 55 1000 1000 100 700 15 30 45 55 2.5 2.5 2.5 Observed Observed Number in S a q l e Volume Concentration (oocystsL) (oocystsj 1062 4.2E+5 1104 4.4E-t-5 1249 5,OE+5 1215 4.9E+S Operating Condition Stable Stable Calculating Poisson Confidence Intervals Haas and Rose (1996) showed that the number of Cyptosporidirrrn oocysts in a sample could be described adequately by the Poisson distribution (Equation 3.1), as would be expected for a sample fiom a uniforrn suspension of such microorganisms in water that was enumerated by an ideal method (Blom, 1989). Parkhurst and Stem (1998) suggested Poisson confidence intervals for describing oocyst data. Atherholt and Kom (1999) presented Poisson methods for sarnple counts in the context of the ICR protocol. This approach consists of deterrnining the mean observed oocyst counts for a given sample and then sirnply determining the endpoints of the confidence interval fiom the following equations, P ~ ( X5 CI,)= %= - P ~ ( X= O ) +P ~ ( X= I ) + . * - + P ~ ( x= CI,) (4-1) where CIu is the upper endpoint of the confidence intervai, is the lower endpoint of the confidence interval, and a is the desired significance level. The endpoints represent values of h that just produce significance at the cr/7% level for the nuU hypotheses &CIU and h = testing = against the respective alternative hypotheses ~ C I and L k C I U ; these values c m also be readily looked up in statistical tables. In repeated sampling, (100-a)% of the intervals calculated in this way include the true value of h. Therefore, at the 5% significance level, 95% of the intervals calculated with Equations 4. L and 4.2 include the true value of h. Some of the advantages of calculating foisson confidence intervals on Cryptosporidizrrn concentrations are that the necessary statistical tables are r e a a y available and that observations of zero microorganisms (non-detects) c m be handled. Naturally, non- detects are not desirable because they lead to large confidence intervals. Caution should be taken when analyzing data that includes non-detects because they are specific to the sample volume examined. A better approach may be to take advantage of the additive property of the Poisson distribution by combining all appropriate replicates and nomaiking to the total sample volume examined. This can be done because when individual sarnples fkom a Poisson distribution are cornbined, the resulting distribution is also Poisson with a mean equal to the sum of the means of individual sarnples (E3ox et al., 1978). This is essentially what Parkhurst and Stern (1998) recomrnended when they proposed surnrning oocyst counts for several sarnples and dividing this sum by the sum of the effective volumes (the definition of which is developed in their paper). 4.4.2 Limitations of Poisson Confidence Intervals The limitation of approaches such as that of Parkhurst and Stem (1998) is that they do not incorporate uncertain analytical recovery into the confidence interval calculation. Nahrstedt and Gimbel (1996) used several reported analytical recovery data and demonstrated that a distribution of recoveries rnay be considerably more appropriate than a single estimate. The disadvantage of the Nahrstedt and Gimbel (1996) mode1 is its increased level of complexity that resulted in a computer program rather than a simple statistical table. During this study, Monte Carlo simulations were performed with the Nahrstedt and Gimbel (1996) model to generate data sets and examine them for agreement with the Poisson distribution. This was done to determine whether or not such additional complexity was necessary. The Monte Carlo simulations generated a set of sirnulated counts of oocysts that one would expect to count on a microscope slide given method recovery information described by a and 6, the Beta distribution parameters that describe the uncertainty associated with recovery. To complete these simulations, a cornmon algorithm for random number generation was used. The algorithm is based on three congruential generators and is f d l y descnbed in Press et al. (1989). Three levels of expected values or true oocyst counts ( k 1 0 0 0 , 100, and 10) were simulated. These true counts correspond to the nurnber of oocyst one would expect to count on a microscope slide given an ideal method that always achieves 100% recovery. Only a &action of these true counts would be enumerated when using an imperfect method, the profile of which is described by the Beta distribution. The first pair of Beta parameters (a = 6.739, b = 19.9) were those calculated for the LeChevallier et al. (1991~)data by Nahrstedt and Gimbel (1996). The second set of Beta parameters (a = 92.43, b = 33.79) corresponded to a recovery profile simila to that descnbed for Vesey et al. (1993a) by Nahrstedt and Gimbel (1996). Triplicate data sets of 100 sainples (n) were simulated for each set of parameters (A, a, and b). Fisher's chi-squared (X') test (Fisher, 1938) was used to assess whether or not the simulated data sets (9 were consistent with the Poisson distribution. This test was applied by calculating the index of dispersion described by, where x represented the observed counts and F represented the mean observed counts. The simulated data and calculated values of X; are presented in Table 4.3. Table 4.3 Sirnulated Data Using Nzhrstedt and GimbeI(1996) Mode1 Beta Parameters a = 6.739 b = 19.9 h 1000 261 1000 308 1000 268 100 19 100 30 100 34 10 10 O 10 2 Table 4-3 Sirnulated Data Using Nahrstedt and Gimbel(1996) Mode1 (Continued) Beta Parameters ?L Io00 1000 Io00 278 236 320 a = 6.739 b = 19.9 100 100 100 10 - 10 10 3 100.000 100.000 100.000 44.1 23 99.587 95.461 89.61 2 33.586 99.965 p r k 2<= S) X Reject Poisson at a = 0.05 X X X X 100.000 100.000 99.954 40.944 88.963 13.581 65.012 87.617 55.219 p r k 2<= S) Reject Poisson at a = 0.05 X X X Simulated Data Usino Nahrstedt and Gimbel(1996) Mode1 (Continued) Beta Parameters A, Io00 702 Io00 1000 711 790 a = 92.43 b = 33.79 100 100 100 74 73 49 10 9 10 10 9 Table 4.3 Simulated Data Using Nahrstedt and Gimbel( 1996) Mode1 (Continued) Beta Parameters a = 92.43 b = 33.79 p r k 2<= S} 93.220 94.570 99.060 18.058 79.940 86.805 19.039 86.758 35.028 X Reject Poisson at a = 0.05 p t k 2<= S} 92.859 98.541 99.956 78.303 88.656 98.139 97.415 '89.662 Reject Poisson at a = 0.05 X X X 2.1 89 X Using the values of x:, it can be determined whether or not the expected Poisson distribution has k e n obtained. For a data set containhg a uumber (n) of samples, the probability that 'X for a @en number of degrees of fieedom (11-1) is less than or equal to X; can be calculated by the integral (Blom, 1989). in which r is the gamma function. The calculation of this function is readily available in most statistical software packages and spreadsheets. Table 4.3 includes probability calculations (using Equation 4.4) for each of the sirnulated sample sets. For a selected significance level (a), the nuU hypothesis that a disuibution is consistent with the Poisson distribution cannot be rejected when, fa< P&' s x:) < (1 -+a) (4-5) If a distribution is Poisson at a significance level of 5% (a = O.OS), the probability calculated using-Equation 4.4 should be between 2.5% and 97.5% based on Equation 4.5. This general approach has k e n previously described (Haas and Rose, 1996; Parkhurst and Stem, 1998). Using this method, one would expect that approximately 1 in 20 distributions (or sets of oocyst counts) would have a probability (caiculated with Equation 4.4) outside of the 2.5 - 97.5% range. A 5% significance level was used to determine whether the simulated data sets were consistent with the Poisson distribution. Data sets that were not consistent with the Poisson distribution were marked with an "X" in Table 4.3; several of these scenarios were inconsistent with the Poisson distribution. If the number of simulated samples were substantially increased to 1000 or 10000, then likely even more of the resulting distributions would be inconsistent with the Poisson distribution. When enumerating C. p a m m , it is uncornmon to collect 100 or more samples with the same expected value and recovery profile. Therefore, a second analysis was perfonned ushg only the first thirty data points fiom each sample set- The results of the Fisher's chi-squared test indicated that several of the 30-sarnple data sets were also inconsistent with the Poisson distribution at the 5% si30nificance level (Table 4.3). Overall, the data suggest that Poisson assumption might be more appropnate for some recovery profiles than others; however, in the general case, it does not adequately describe data obtained ushg an analytical method with non-constant recovery. Therefore, the above exercise suggests that a distribution more cornplex than the Poisson distribution rnay be required to account for the various analytical errors associated with parvzmz enurneration. Although Atherholt and Korn (1999) presented Poisson methods for describing sample counts, they also suggested that a distribution more complex than the Poisson distribution may be required to account for the various anaiytical errors. Nahrstedt and Gimbel (1996) developed such a statistical mode1 by assuming a Poisson distribution for the m e sample counts, a binomial distribution for modeling the recovered fiaction of oocysts, and a Beta distribution for descnbing the uncertainty of recovery. They also noted that the tirne and location of sampling influence error, but suggested that this contribution cannot be statistically determined. These sources of error were described in Chapter 3. The Beta distribution is useful for describing oocyst recovery because the recovery is bound between O and I (Le., OIpIl) and the distribution is very flexible, allowing for the description of a variety of recovery profdes. The Beta probability density function (pdf) was presented in Equation 2.5. With more detail, the Beta pdf is described by a and b, the parameters of the Beta distribution, such that: where, mean or expected value of p variance of p = Var(p) = = E(p)= a ab (a+ b)'(a + b + 1) Recovery studies must be performed to determine the Beta parameters a and b that describe the overall recovery profile. Each replicate sample represents a trial Tom which each counted oocyst represents a success. Button and Reilly (2000) presented a method for determining o v e r d Beta distributions for multiple trials. The authors explained that o v e r d Beta pdfs were necessary because each set of ni& provides a dfierent estimate of the Beta distribution for the variabIe of interest (in this case p, the analytical recovery). The authors emphasized that simply summing all of the data (in the case of oocysts, using the total number of oocysts counted kom all of the slides and the total volume of water processed to make all of the slides) does not reflect the uncertainty between data sources and tirnes perïods. For example, different analysts increase uncert ainty, potenhally resulting in different recovery profiles (Beta pdfs). Button and Reilly (2000) presented an overall Beta distribution that had the same mean and variance as the mixture of Beta distributions korn the different sets of trials;it was described by where P Pi k xi Yi pdf@i, xi, yi) - overall probability of success probability of success on each trial (9 number of trials number of successes in trial i number of failures in trial i Beta pdf for the ith trial weighting factor for the ith Beta distribution, where the value of each Ai is proportional to its variance-'. k i= 1 Ai = 1 and This method weighted the Beta distribution fiom each trial in proportion to its precision. The mean and variance of the overall Beta pdf were calculated using Equations 4.9-4.12 as follows: where 6; '= xiYi + y,)2 (xi+ yi + l) Once the mean and variance of the overall Beta pdf were calculated, the values of the parameters a and b for the overall Beta distribution were calculated fiom the standard expressions for the rnean and variance of the Beta distribution; these expressions were provided in Equations 4.6 and 4.7 respectively. The o v e r d Beta distribution was fit to the exarnple recovery data f?om Table 4.1. The overall mean and variance were calculated using Equations 4.9-4.12; some of the key data are surnrnarized in Table 4.4. Using the overall mean and variance, Equations 4.6 and 4.7 were used to calculate the Beta pararneters a and b, 28.12 and 8.43 respectively. The corresponding pdf and cumulative density function (cdf) for the recovery data fiom Table 4.1 are presented in Figure 4.1 and Figure 4.2. There are fifteen mals in this example (k = 15). This approach was used to calculate the Beta pararneters used in this thesis (Appendix C). Table 4.4 Calculated Data Used for Describing the Overall Recovery Pronle (Beta pdf) for Recovery Data in Table 4.1 Seeded Number of Oocysts 500 Observed Number of Oocysts 35 1 overall mean = 0.7693 overall variance = 0.0047 A;- .Yi E(P) 35 1 149 0.70 V ~ P ) 4.176 E-4 Fi,pre 4. i Overall Recovery Profile (Beta pdf) for Recovery Data in Table 4.1. Fi,gure 4.2 Overall Recovery Profile (Beta cdf) for Recovery Data in Table 4.1. 4.7 CALCULATING CONFIDENCE INTERVALS The approach described in this chapter is similar to that previously described since a Poisson mode1 is used to descnbe the original dismbution of microorganisrns in the water body. The issue of analytical recovery is also addressed by incorporating the binomial and Beta distributions. This work is unique in that strategies are presented for q u a n t m g the reliabiIity of both concentrations and removals of microorganisrns. In most cases where oocysts (or any other discrete particles) are king enumerated, a number of oocysts (X) is obsemed from a processed volume of water (V). The value of the parameter N is an estimate of the m e number of oocysts in the water (A). If the distribution of tnie counts of oocysts in the water can be described by a Poisson distribution and if a constant fraction @) of the oocysts are recovered from the water, then the probability of observing X oocysts is described by, In other words, Equation 4.13 is the product of the Poisson distribution (Equation 2.1) and the binomial distribution (Equation 2.3) and describes the probability of observing X oocysts under given experimental conditions described by A, N, and p. N is an unknown parameter representing the me count of oocysts in the sample drawn from the water body. Information regardhg p helps to estirnate N. This information is necessary to detemillie the hue count of oocysts in the water (A)and the parameter of interest. Recovery studies can be performed to describe the uncertainty of p by distributions such as the Beta (Equation 2.5), however, they must somehow be incorporated when drawing inferences about the tnie count of oocysts (A) based on the observed count (X). Bayesian inference aIlows for explicit incorporation of assumptions describing the complexity of recovery @) that is fiee fkom impediment from purely technical limitation (Box and Tiao, 1973). While relatively simple solutions can be obtained from other inferential theories when assumptions such as nonnality and independence of errors are possible, the solutions are often intractable (Box and Tiao, 1973) in cases such .as describing Cryptospondiurn concentrations and removals, where such assumptions are not necessanly valid. Bayes' theorem can be applied to make inferences about c (the true concentration of oocysts in the water body) and therefore  (the me count of oocysts in the sample), given information about V;, pi (in the forrn of a and b fÏom the Beta distribution described by Equation 2.5) and Xi- According to Bayes' theorem, where i = 1,2,---,n According to Equation 4.14, in order to describe the density function (Df)of the m e oocyst concentration c given the observed oocyst count X, prior informatio* regardkg the density function of c. N, and p is required ( ~f(c, N i ,pi) ) for aU values of i. If n replicate samples of volume V are take fiom a water body of oocyst concentration c, substituting Equation 2.2 into Equation 4.13 yields, where is known. Equation 4.15 does not require that all of the water samples have the same volume (Le., VI does not necessarily equal Vz,etc). It should be noted that Equation 2.4 could be used which would eliminate N (Lobs = A?), but subsequent sampling would be awkward. The likelihood function describes information about c, N, and p that has been derived from the experimental data and is proportional to Equation 4.13. The prior density function describes what is known about c, N, and p prior to looking at the data. It c m be described by, According to Equation 4.16, no information regarding c is known in advance. Since cLO (negative oocyst counts and concentrations are not possible) a Jeffkeys prior (lfc)is used to indicate that all outcomes are equaily likely (Jeffreys, 1961). .A uniform prior of 1 is used for N, indicating no prior information about N. Finally, the Beta distribution from Equation 2.5 is used as an informative prior that describes the recovery profile of pi. Using Bayes' theorem from Equations 4.14, Equations 4.15 and 4.16 are combined to yield the posterior density function, which describes c, N,and p @en knowledge of the data X where 00, K X , OSpSl, and i = l,2,---,n. Monte Car10 methods often estirnate features of the posterior or predictive distributions such as the one presented in Equation 4.17 by u s k g samples drawn fiorn that distribution. As described by Smith and Roberts (1993)- however, ''generating sarnples from an arbitrary, often highly dimensionaljoint distribution is not often possible, thus seemingly making sample-based approaches of Limited use." The Gibbs sarnpler is based on Markov Chahs and represents an indirect approach to the required sampling that can overcome that problem (Smith and Roberts, 1993). Introduced by Geman and Geman (1984), the Gibbs sarnpler, as described by Caseila and George (1992) is "a technique for generating random variables from a rnar30inal distribution indirectly, without having to calculate the density." The authors explain that supposing a given a joint density f(x, yl, ..., y,,) and an interest in obtaining the characteristics of the marginal density [ f ( x ) = j- - -Jf (x, y,, -,y,) dy, -- dy, ] such as the mean or variance, the typical approach would be to calculate f(x) and use it to obtain the desired characteristic. In the present context, Equation 4.17 describes the joint density for the true concentration of oocysts in the water body (c), the oocyst number in the sarnple (N, an estimate of the me number of oocysts in the water body, A), and the probability of recovery (p) given a count on the slide (X). The integrations in f(x) are extrernely difficult to perform either analytically or numencally in many cases, such as Equation 4.17; in such cases, the Gibbs Sampler provides and alternative method for obtaining f(x) (Caseh and George, 1992). The Gibbs sampler does not require the direct cornputation or approximation of f(x), rather, it effectively allows for the generation of a sample Xi, ...,X, -f(x) without requiring f(x), the marginal density (Casella and George, 1992). This is achieved by generating a sampIe by r o m f(x) by sampling fkom the conditional distributions; in the case of a pair of random variables (X, Y), the conditional distributions are f (XI y ) and f (y(x) (CaseLia and George, 1992). The mean, variance or any other characteristic of f(x) can be calculated to the desired degree of accuracy by simulating a large enough sarnple (Casella, and George, 1992). Several detailed discussions of the Gibbs sampler are available in the lrterature (Casella, and George, 1992; Smith and Roberts, 1993). The Gibbs sarnpler can be used to estimate values of c, N, and p given knowledge of the data This is accomplished by sampling fiom the distributions described by the conditional probability density functions (WC) of each of the parameters. The conditional probability density functions of c, Ni, and pi are respectively described by, Although each of the conditional probability density functions of c, N, and p c m be described, it is the oocyst concentration (c) that is of interest. Confidence intervals on oocyst concentration can be calculated by binning the generated data to yield the pdf. The two unique points with equal height (density) that are a distance (P) apart on the pdf are the endpoints of the ( IOOP)% confidence interval for oocyst concentration. This confidence interval represents the region of highest posterior density in which no point outside of the intervd has a higher posterior density than the points inside the interval. Confidence intemals for log removals can be calculated using the same method used for oocyst concentrations. This is achieved by concurrently simulating the conditional density functions on influent (ci) and effluent (cc)concentrations. The log10 of cJci is calculated and stored for each iteration, resulting in the density function of log removal through the treatment process k i n g evaluated. The 95% confidence intervals for the C. pawum removal data listed in Table 4.2 were cdculated. These intervals were calculated for each of the individual data points as well as the overall (combined or pooled) data set. The intervals for the t = 15, t = 30, t = 45, and t= 55 samples and the overall data set ranged fkom 4.79- to 6.12-log, 4.8 1- to 6.13- log, 4.97- to 7.03-log, 4.39- to 5.17-log, and 4.89- to 5.42-log respectively. Of the four samples, the lowest of the highest posterior density @PD) region endpoint was 4.39-log, while the highest endpoint was 7.03-log. This range of removals is practically relevant because it represents the range of removals that could be reasonably expected during the conditions investigated, based on the measured oocyst counts and the uncertainty associated with the analyticd method. This range of removals is also indicated in Figure 4.3, which depicts the pdfs for the evaluated data points and combined data set. The pdfs in Figure 4.3 demonstrate that the combination of data points to yield an overall pdf resuits in a less diffuse result with Iess uncertainty. The pooled data resulted in a 95% confidence (probability) interval with a range of 4.89- to 5.42-log removal of C. pawzrm, a considerably smaller range than the overall range of 4.39- to 7.03-log described above. The pooling of replicate data results in a smaller confidence interval because of the increased number of observations (oocyst counts) associated with the replicate analyses. Since Poisson distributions are additive it is the number of actual observations (counts) nther than the number of replicate experiments that decreases the uncertainty associated with the enumeration of discrete particles such as oocysts. 3 35 4 15 - low - 439 high - 5.32 6 low 439 C.purvum Removai (loglo) Fi,we4.3 4 65 hi& 75 8 - 7.03 Probabilitydensityfunctions(pdfs)forC.parvumRemovalExampteData in Table 4 2 (Stable Operation at Ottawa). The relative effect of observations (counts) and replicate samples is demonstrated in Figure 4.4. Two types of data are presented in this figure- The data represented by the black bars correspond to three fictional pairs of filter influent and effluent counts that all resulted in an unadjusted (for recovery) rernoval of 4-log. The three pairs of data Vary in that they each correspond to different counts of oocysts in the filter influent and effluent. The effluent counts indicated in Fi,we 4.4 correspond to counts of 1. 10, and 100 oocystsL. The respective influent concentrations were IO', los, and 106 oocystsL to yield 4-log removal fkom each influent-effluent pair. From Figure 4.4, it is clear that the range of the confidence interval is much smaller in the sarnple with higher counts (100 oocystsL in the filter effluent). While counting 1 oocystL in the filter effluent resulted in a confidence interval that spanned approximately 2.1-log, counts of 10 and 100 oocysts/L resulted in confidence intervals that spanned approximately 0.6- and 0.2-log respectively. This result clearly demonstrates that a considerable amount of uncertainty can be elirninated by enumerating >10 oocysts in filter effluent samples. Fùter Effluent Counts in 1 SampIe 1 0 Number of Repliate Samples of 1 oocysdL in Fdrer Effluent L 0 - 10 1 100 Oocysts/Lor Number of Replicate Samples Figure 4.4 Effect of observations (counts) and replicate sarnpks on confidence interval range. The data represented by the white bars in Figure 4.4 correspond to replicate sampies containing 1 oocyst/l in the filter effluent and 10"oocystsL in the filter influent, again to yield an unadjusted (for recovery) removal of 4-log. The three bars correspond to pooled sets of influent-effluent pairs. The first bar £rom the left (1 replicate samplej represents the confidence interval calculated for one sa.rnple containing 1 oocyst/l. in the filter effluent and 10' oocystsL. The second bar corresponds (10 replicate samples) corresponds to ten replicate pairs of data in which the influent and effluent C. parvlrrn concentrations are each 10" oocystsll and 1 oocyst/L respectively. The third bar similarly represents the confidence interval for one hundred replicate pairs. A cornparison of the black and white bars in bars in Figure 4.4 clearly demonstrates that the number of observations (counts) affects the uncertainty. Simply stated, counting at least ten oocysts in filter effluent samples (assuming a higher concentration in the filter influent) will substantially decrease the uncertainty associated with calculations of log removals, regardless of whether the ten oocysts are fkom one sample containing ten oocysts or from ten samples each containing one oocyst. This result is a practical demonstration of the additivity of Poisson data. Given the tirne and cost associated with processing water samples containing C. p a w r n , the desirability of analytical methods that have the ability to process large enough volumes to yield at least ten oocysts is obvious f?om Fi,we 4.4. Another important question also aises fi-orn the discussion of pooled data. Although statistically valid, the pooling of data averages out differences in C. parvum removal that occur during a given samphg period, with samples with higher counts skewing the pdf to warci lower removals. So when is it appropnate to pool data or consider data points as replicates? In this work, the removals calculated fkom influent-effluent data pairs were considered replicates when the independent variables (settled water turbidity, filter loading rate, pretreatment conditions, etc.) were constant during the sarnpling period. In these cases, the removal data were pooled. When the independent variables were not constant (e.g., sub-optimal coagulation conditions when settled water turbidity was changing, end-of-run operation when the balance between anachment a n d detachment of particles within the filter was changing and obviously affecting water quality, etc.), the removal data were not pooled. In these cases, the range of removals was based on the lowest and highest HPD endpoints calcuhted tkom each of the individual pdfs. This approach for defining replicates and pooling data will be used in subsequent chapters for evaluating subsequent C. pawum and polystyrene rnicrosphere removal data. Traditional filtration performance parameters such as turbidity (USEPA, 1989) and particle counts are comrnonly considered adequate surrogate measures for evaluating the removal of microbial pathogens such as Giardia by filtration (USEPA, 1991). Although turbidity and particle counts are reliable indicators of treatrnent efficiency, they are not reliable quantitative surrogates for Cryptosporidiwn removal (Nieminski and Ongerth, 1995; Patania et al., 1995; Fox et al., 1998). While removals of other microbial parameters, such as aerobic spores of BaciIZus subtilis, have been correlated with the filter removal efficiency of C.p a m m in some studies, no adequate surrogate for removal of the pathogen has yet been identified. Accordingly, water treatment process evaluations of C. parvurn removal are often perfonned using inactivated (non-viable) C. paniurn oocysts. High concentrations (Nieminski and Ongerth, 1995; Fox et al., 1998) of chernically inactivated oocysts are typically used during these studies because of the potential health risks associated with the use and release of viable oocysts. Several researchers have noted that chemically inactivated C. parwtrn oocysts may not necessarily be ideal surrogates for viable oocysts due to differences in surface charge (Lytle and Fox, 1994). A change in the original surface charge (described by the zeta potential) of colloida1 particles Iike C. parvurn might affect their rernoval during granular media filtration because zeta potential is indicative of the degree of particle destabilization (Amirtharajah and Mills, 1982). Particles are most readily removed when their zeta potential is near zero, corresponding to their isoelectric point (biaharajah, 1988). It has been demonstrated that chemical inactivation can change oocyst zeta potential (Ongerth and Pecoraro, 1996). As a result, some researchers have speculated on how this change rnight affect coagulation and subsequent filtration (Lytle and Fox, 1994). During water treatment, the zeta potential of C. parvrrrn oocysts is affected by multiple factors such as water quality, coagulant type and dosage, and pH, in addition to chemicai inactivation prior to treatment. Amirtharajah (1988) described charge neutralization by demonstrating that colloidal zeta potential is different in coagulated and non-coagulated waters (Figure 5.1). Following chemical pretreatrnent (coagulation), the surface charge of oocysts and other colloids becomes a function of the pH, coagulant type, and coaguiant concentration rather than the specific colloidal zeta potential pnor to coagulation, as the positively charged hydrolysis species attach to the particles and neutralize or reduce the net particle surface charge. Similarly, during hi& coagulant dosing and sweep floc coagulation, oocysts become enmeshed in the precipitating hydroxide solids and the surface charge relevant for oocyst removal becomes a function of the entire floc rather than a single oocyst Differences in oocyst zeta potential pnor to coagulation do not likely impact overall coagulation chemistry because oocyst surface area is essentially insignificant relative to that of other particles present in water. In many cases, this is m e for the artificially high oocyst concentrations used in treatment optimization investigations (including the present study). Uncoated colloid Figure 5.1 PH Impact of coagulant on colloidal zeta potential (Amirtharajah, 1988). As filtration is a dynamic process, it is critical to identie key operational and design strategies for rnaximizing C.parvunt removal during filtration, especially during periods when the process is challenged. As described in Chapter 2, numerous investigations have been conducted to evaluate the removal of C. parvum through filtration under normal operating conditions where chemical pretreatment and filtration processes are performing at optimal or near-optimal conditions. Full-scale C.ptosporiditcrn removals fiom 2 to >4log have been reported in the literature (e-g.. Kelley et al.., 1995; Nieminski and Ongerth, 1995; Baudin and Laîné, 1998). Pilot-scale oocyst rernoval data have sugpested that filters can achieve anywhere fi-om 2 to >5-log removal of oocysts (e.g., Patania et al., 1995; Fox et ai., 1998). Pathogen passage through filters during vulnerable points in the filter cycle (e.g., ripening, breakthrough, etc.) or when particle removal processes are challenged (eg.. hydraulic surges, coagulation upsets, etc.), however, has been less thoroughly investigated. It has been suggested that maintainhg optimal chemical pretreatment is critical to maximiPng C. parvum removals during filtration. Patania et al. (1995) demonstrated that filtration was ineffective for oocyst removal without chemical pretreatment. Other pilot-scale studies also indicated that sub-optimal coagulation decreased oocyst removal by filters by at least 1-log (Charles et al., 1995; Ongerth and Pecoraro, 1995; Dugan et al., 1999). These findings concur with theoretical arguments that suggest proper coagulation conditions are necessary to achieve adequate particle destabilization for attachment by charge neutralization (Figure 5.1) or enmeshment in precipitates that can be subsequently removed by sedimentation or attachent during filtration. When coagulation is not optimized, filter performance deteriorates resulting in higher filter effluent turbidities and particle concentrations compared to stable or optimized operation. Differences between removaIs of viable and inactivated oocysts by filtration would be rnost Iikely to occur under operating conditions such as coagulation failure, when the differences in oocyst zeta potential are least influenced by factors such as coagulant type and concentration. Filter ripening is another potentially vulnerable penod of filter operation because Nter effluent turbidities and particle concentrations are also hi& during ripening compared to stable operation. Amirtharajah (1998) suggested that >90% of the particles that pass through a well-operated filter do so during ripening. SeveraI pilot-scale studies have demonstrated that although filter effluent turbidities can be quite high during ripening, log removals of C. parvzm o d y deteriorated by -0.5 to 1-log compared to those obtained d u h g stable filter operation (Patania et al., 1995; Charles et al., 1995). These findings were confinned at fuli-scale by Baudin and Laîné (1998) who demonstrated an -1-log deterioration in oocyst removals durùlg filter ripening. Despite considerable filter effluent turbidities and particle concentrations during this period, these reported fmdhgs suggested that oocyst removals are not particularly vulnerable during filter ripening. Non-attachment of particles partially contributes to the higher filter effluent turbidities and particle concentrations that occur during ripening (Amirtharajah, 1988). Filter ripening represents a relativeIy brief operating period during which an increase in aonattachent of particles is expected relative to stable operation. If they exist, subtle differences between attachment efficiencies of viable and inactivated oocysts could potentially be revealed during this penod. Although end-of-run breakthrough represents another period of increased non-attachment relative to stable operation, this operating condition couid not be achieved in the bench-scale filter due to available head limitations. Investigations of media type and design have demonstrated that these parameters have little impact on oocyst removals by filters. Hall er al. (1995) did not fmd performance differences between sand and dual-media filters when the Elters had s i r d a r filtrate quality (measured by turbidity). Other pilot studies have also failed to demonstrate statistically significant differences between C. pamim removals by sand anthracite/sand, and GAC/sand filters @ugan et al., 1999; Swertfeger et al., 1999). In this study, evaluations of various media types were performed at bench-scale durkg operating periods when an increase in non-attachment of particles was expected relative to stable operation. These experiments were performed to elucidate any differences between attachment efficiencies of viabie and inactivated oocysts that resulted fiom the media specifically, rather than fiorn the operating conditions. The &st objective of the bench-scale experiments described in this chapter was to determine if formalin-inactivated C.parwrn oocysts were adequate surrogates for viable oocysts during both optimal and vulnerable penods of filter operation. Coagulation failure was selected as a critical vulnerable p e h d for study because oocyst zeta potential was least likely affected by the presence of coagulant during this period; any differences between filter removals of viable and inactivated oocysts would most likely be demonstrated during this period. Despite Limited data suggesting only marginal deterioration of oocyst removal by filtration during ripening, this penod was studied because hi& filter effluent turbidities, particle passage, and non-attachrnent of particles relative to stable operation were expected. Subtle differences between attachment efficiencies of viable and inactivated oocysts could potentially be revealed during this period. The second objective of this shidy was to identify operational and design strategies for minimizing oocyst passage through filters. In addition to the operational impacts of coagulant and operation during ripening, dual- and tri-media filter oocyst removals were compared to elucidate any potential media type (design) advantages during either stable or vuherable operating conditions. Bench-scale experiments investigated dual- and tri-media filter rernovals of viable and formalin-inactivated CryptosponXurn parvum oocysts during stable operation, ripening, and coagulation failure. Specific information regarding the seeding conditions and filtration apparatus is provided in Chapter 3 (Tables 3.3, Section 3.3, and Figure 3.5). The filters were operated at a loading rate of -7.5 m/h (3.1 US gpm/ft') in a constant head, declining rate mode during these evaluations. To ensure that no oocysts were carried over between experiments, the media were replaced between al1 experimentsFigure 5.2 generally describes the experimental configuration and conditions. 7.5 m/h raw water -3.5 NTU 1 - I O 7 Cryptospondium (coagulateci in raw water) SAMPLlNG turbidimeter I duai-media Figure 5.2 of ?+media Bench-scale experimental configuration Only one operating condition (stable operation, ripening, or coagulation faihre) was evaluated per experiment. During the ripening experiments, the filters were seeded with oocysts during the f ~ s30 t minutes of operation; ripening was defmed as the period fiom the start of filter operation to the point when filter effluent turbidity decreased to -0.2 NTU afier peaking. The seeded oocysts were jar-coagulated at the same conditions as the raw water (5 mg/L alum (A12(SO&-18H20) at pH 6.9). The coagulation conditions were detennined using the lowest coagulant dose that achieved a filtered water quality of <0,2 N'TU (-0 -05NTU). Raw water coagulated at the same conditions as described for the ripening experiments pre-conditioned the fiIters prior to oocyst dosing during the stable operation and coagulation failure experiments. The filter pre-conditioning period was defmed as the 2hour penod after filter effluent turbidity peaked during ripening. The seeded oocysts were jar-coagulated at the same conditions as the raw water during the stable operation experiments. During the coagulation failure expenments, raw water coagulation ceased afier two hours of filter pre-conditionhg and non-coagulated oocysts were seeded into the non-coagulated raw water. The filters were dosed with the oocyst suspension for one hour, during the third hour of operation. Filter performance was evaluated by monitoring filter influent and enluent turbidity every three minutes during ripening and every five minutes subsequent. Coagulation conditions and sampling periods during the benchscale experiments are summarized in Fi-pre 5.3. Stable Operation l Raw water Oocyçt seed 60 Filter Ripening ++ Sampling Raw water Oocyçt seed Coagulation FaiIure O 60 120 180 240 Time Since Filter Effluent Turbidity Peaked During Ripening (minutes) Figure 5.3 Summary of sampling times and coagulation conditions during benchscale experiments. The filters were seeded with a total of -10' viable or formalin-inactivated oocysts. The oocyst concentration of the original spike suspension was determined directly with a hemocytometer and phase contrast Mcroscopy as described in Chapter 3. Oocysts in samples of the seed suspension, filter influent, and filter effluent were concentrated by direct vacuum filtration and enumerated on the membranes using the immunofluorescence assay descnbed in Chapter 3. During the stable operation and coagulation failure experiments, filter influent and effiuent samples were collected for C.parvum analysis during three consecutive 15minute periods beginning at 10, 25, and 40 minutes afier the start of the 1-hour seeding period. During ripening, filter influent and effluent samples were collected during four 5minute periods beguuiing at 5, 10, 15, and 20 minutes after the start of the 30-minute seeding period. Al1 of the sampling times are indicated on Figure 5.3. 250-mL samples of filter influent wcre collected during each sampling period; of these samples, 10 mL were analyzed for C.parvum. Almost al1 of the filter effluent during the oocyst dosing period was coilected (-3 x x 3.75 L during stable operation and coagulation failure and -4 1.25 L during ripening) and analyzed for C. panalm. Turbidity data were collected to confirrn comparable filter performance between experiments. Turbidity data fiom the dual-media experiments using formalin-inactivated oocysts are presented in Figure 5.4. These data indicated that several filter runs (each containing new media) produced sllnilar turbidity removals during the fust three hours of operation when seeding occurred. The filter effluent turbidity during the dual-media filter experiments conducted during stable filter operation ranged from 0.03-0.07 NTU, with a mean of 0.05 NTU. Filter effluent turbidity trends during the ripening experiments were relatively reproducible and ranged fiom 0.04-0.87 NTU. The turbidity trends were also generally consistent between replicate coagulation failure experiments; filter effluent turbidities ranged fiom 0.03-0.69 NTU during these experiments. - StableOperation Seedmg Duration (Stable Operation and Coaguhtion Failure) -- Ripeniog - Coa-dation Faiiure -20 O 20 JO 60 80 100 120 140 160 180 200 Tïme (minutes) Figure 5.4 Filter efnuent turbidity, seeding period, and sampling times during dualmedia filter experiments with formalin-inactivated oocysts. The filter effluent turbidity data in Figure 5.4 indicated that turbidity removal was relatively consistent during replicate filter runs of the three studied conditions, suggesting that media replacement between nuis did not result in substantial differences in particle removal trends. Dual-media £ilter effluent turbidity data collected during the stable operation, ripening, and coagulation failure experiments are summarïzed in Table 5.1 to Table 5.3 respectively. Similar filter performance and reproducibility of turbidity removal was observed during the dual-media experiments using viable oocysts (Figure 5.5). During the experiments conducted with viable oocysts, the dual-media filter turbidity during stable filter operation ranged fiom 0.04-0.06 NTU, with a mean of 0.05 NTU. Filter effluent turbidity trends during the ripening experiments ranged fiom 0.05-0.89 NTU. The turbidity trends were also generally consistent berneen replicate coagulation failure experiments; filter enluent turbidities ranged fiom 0.04-0.60 NTU during these experiments. Table 5.1 Dual- and Tri-Media Filter Influent and Enluent C. parvum Concentrations and Effluent Turbidity During Stable Operation. Experiment and oocyst Sample Time ( h l Dual-Media Cyptosponiiiurn Turbidity (oocysts/L) @TU) Tri-Media Cryptospon'dium (0ocystdL) Turbidity WU) Viabiliîy FI FE FE Inactivated 1 FI FE FE 5.9 x los 5 0.03 3.3 x 10' 4 0.06 Inactivated 2 Inactivated 3 Viable 1 Viable 2 Viable 3 40 "Filter influent. "Filter effluent. 4.1 x lo5 3 0.04 Table 5.3 Dual- and Tri-Media Filter Muent and Effluent C.pamrm Concentrations and Effluent Turbidity During Coagulation Failure. Experhnent and Oocyst Viability Sample Time (min-) hactivated 1 10 DuabMedia Cryptospo~dium Turbidity (oocysts/L) m FI FE 5.1 x 10' 4-8 x 10' FE Tri-Media Ctyptosporidium (oocysts/L) FI FE 4.0 x IO* 8.8 x lo4 Inactivated 2 hactivated 3 Viable 1 Viable 2 Viable 3 "Filter innuent. Filter effluent. FE - -- - Turbidity m FE -- Seeding Duration (Stable Operation and Coagulation Faiiure) - := SamplingTimes Seediug Duration -!$ -20 Figure 5.5 O 20 40 60 80 100 Time (minutes) 120 140 =& 160 3 180 200 Filter effluent turbidity, seeding period, and sampling times during dualmedia filter experiments with viable oocysts. Despite considerably less consistent turbidity profiles during the k s t 60 minutes of operation (ripening), the tri-media stable operation and coagulation failure experiments achieved filter effluent turbidities comparable to those fiom the dual-media experiments (Figure 5.6 and Figure 5.7). During the experiments conducted with inactivated oocysts, the tri-media filter effluent turbidities ranged fiom 0.03-0.08 NTU, with a mean of 0.05 NTU during stable filter operation. The htrbidity profdes during ripening were not h i a y reproducible in the hi-media filter. DUnng ripenùiç, the filter effluent turbidities ranged fiom 0.08-0.79 NTU. Filter effluent turbidities were also less reproducible during coagulation failure in the tri-media filter, ranging fiom 0.04-0.52 NTU during those experiments. Although the filter effluent turbidity data were less reproducible between tnplicate experiments in the tri-media filter than in the dual-media filter, the filter effluent turbidity trends were generally consistent between replicates and media types during the oocyst seeding periods. The tri-media filter effluent turbidity data for stable operation, ripening, and coagulation failure experhnents are summarized in Table 5.1 to Table 5.3 respective1y. -- Stable Operation - Seedmg Duration (Stable Openuon and Coagulation Failure) -- Ripenuig Sampling Times jeedmg Dmtion (Ripenmd Filter effluent turbidity, seeding period, and sampling times during trimedia filter experiments with formalin-inactivated oocysts. Figure 5.6 Seeding Duration (Stable Opention and Coagulation Failure) - Coagulation Failure 1 Sampling T k /r >r t .. .: ? . -20 Figure 5.7 O Coaguiation Faiiure 20 40 60 80 LOO Time (minutes) 120 140 160 180 fOO Filter effluent turbidity, seeding period, and sampling times during trimedia filter experirnents with viable oocysts. The filter effluent turbidity data collected durhg the tri-media filter experiments conducted with viable oocysts are presented in Figure 5.7. During these experiments the filter effluent turbidities during stable operation ranged from 0.03-0.12 NTU, with a mean of 0.07 NTU. As was experienced during the experiments with inactivated oocysts, the ripening periods were difficult to reproduce in the tri-media filter. During the one npening experiment performed with viable oocysts, the filter effluent turbidities ranged fiom 0.17-0.89 NTU. Filter effluent turbidities were also less reproducible during the trimedia coagulation failure experiments with viable oocysts, ranging fiom 0.04-0.37 NTU. The tri-media fiiter effluent turbidity data for stable operation, ripening, and coagulation failure experiments are surnmarized in Table 5.1 to Table 5.3 respectively. Each of the stable operation and coagulation failure experiments was performed in triplicate and consisted of three pairs of filter influent and effluent data per experiment The ripening experiments were also performed in triplicate (with the exception of the trimedia ripening experiment with viable oocysts which was perfomed only once) and consisted of four pain of filter influent and effluent data per experiment. Measured filter s u e n t oocyst concentrations were consistentiy maintained at -105 oocysts/L, regardless of experirnental conditions and media type. Regardless of the media type, filter effluent C. parvurn concentrations were typically -1 - 10 oocystsll during stable filter operation. Marginally worse oocyst rernovals were observed during ripening in both dual- and trimedia filters, when filter effluent oocyst concentrations were -1-50 oocystsL Filter effluent oocyst concentrations of -1 04-1o5 oocystsL indicated substantial deterioration of C.parvurn removal by both filter types during coagulation failure. Filter influent and effluent C. parvum concentrations, measured during stable operation, ripening, and coagulation failure, are presented in Table 5.1 to Table 5.3 respectively. Loglo removais of C. parvum were calculated fkom the individual pairs of measured filter influent and effluent oocyst concentrations that were presented in Table 5.1 to Table 5.3. Calculated log removals for the experiments discussed in this chapter are available in Appendix D (Table D.l). The dual-media filter removals of viable and inactivated oocysts were surnrnarized in a box-and-whisker plot (Figure 5.8). In these plots, the square in the center represents the median removal(50" percentile). The lower and upper portions of the box respectively indicate the 25h and 75" percentile removals. The lower and upper portions of the line (whisker) respectively indicate the minimum and maximum rernovals observed. Coagulated, formalin-inactivated oocyst removal by the dual-media filter ranged from 4.7 to 5.7-log during stable operation, with a mediaa removal of 5.3-log; coagulated, viable oocyst removal also ranged fiom 4.7 to 5.7-log with a median removal of 5.1-log. Dualmedia removals of coagulated, formalin-inactivated oocysts decreased somewhat during ripening wheo removals ranged from 4.0 to 5.1-log, with median removal of 4.6-log; viable oocyst removals similarly ranged fiom 4.0 to 5.4-log with a median removal of 4.8-log. nie most subçtantial decrease in oocyst removal by the dual-media filter occurred during coagulation failure when removal of non-coagulated formalin-inactivated oocysts ranged from 0.6 to 1.3-log, with a median rernoval of 0.8-log. Dual-media removals of non-coagulated viable oocysts ranged fiom 0.4 to 1-2-log during coagulation failure. with a median removal of 0.6-log. l 1 LA- Stable operation inactivated Stable operation viable Ripening inactivated Rfpening viable Coaylation mure Coagdauon f;lilure inactivated viable Operatmg Condmon and C. parvum Viabiliw Fiagure 5.8 Duat-media filter removals of viable and inactivated C~yptospo*dium during stable operation, ripening, and coagulation failure. Tri-media filter removals of viable and inactivated oocysts were also summarized in a box-and-whisker plot (Figure 5.9). CoaguIated, formalin-inactivated oocyst removal by the tri-media filter ranged ffom 4.9 to 5.8-Iog during stable operation, with a median removal of 5.4-Iog; coagulated, viable oocyst removal ranged fiom 4.6 to 5.8-log with a rnedian removal of 5.3-log. Similar to dual-media filtration, the removal of coagulated, formalin-inactivated oocysts by tri-media filtration was lower during ripening when removals ranged fiom 4.2 to 5.3-log, with a rnedian removal of 5.0-log; viable oocyst rernovals sirnilarly ranged fkom 4.4 to 5.7-103 with a median removal of 5.1-log. During coagulation failure, 0.6 to 2.2-log removal of non-coagulated formalin-inactivated oocysts was achieved by the tri-media filter, with a median removal of 1-1-log; removals of non-coagulated viable oocysts ranged fiorn 0.5 to 2.5-log in the tri-media füter, with a median removal of 0.8-log. Of the operating conditions studied, coagulation failure resulted in the greatest detenoration of Cwptosporidium removal by tri-media filters, a result that was consistent with the fïndings of the dual-media experiments. Stable operation inactiviifed Stable operation viable Ripening inactivated Ripe* viable Coagulation Mure Coagulation tàilure inactivated viable Operaring Condition and C.patvum Viability Figure 5.9 Tri-media filter removals of viable and inactivated Cwptosporidiurn during stable operation, ripening, and coagulation failure. Since both the dual- and tri-media filters did not demonstrate substantial differences between removals of viable and formalin-inactivated oocysts, Cryptosporidium removals during dual- and tri-media stable operation, ripening, and coagulation were also compared by pooling the viable and inactivated oocyst removal data. The pooled dualand tri-media data in the box-and-whisker plot in Figure 5.10 ilIustrated that tri-media filter removal of oocysts was only marginally better than dual-media removal of oocysts during stable fùter operation. The same result was found during filter ripening. During coagulation failure, tri-media removal of oocysts was considerably more variable than in the dual-media fiIter. The pooled data did not demonstrate substantial differences between the dual- and tri-media removals of C. parvzrm during coagulation failure (~0.05);however, the tri-media removals were sIightly higher. Stable operation duai-media Stable operation ai-media Rqenhg dual-media Ripening m-media Coagulation fàiiure Coaylation failure dual-media ai-media Operatïng Condition and Media Type Figure 5.1 0 Pooled dual- and tri-media filter rernovals of Cryptosp~~dium during stable operation, ripening, and coagulation failure. The theoretical filter influent O O C ~ Sconcentration ~ (Ctheorericaf) was calculated for each experiment by the equation in which Cwikeis the original oocyst concentration (detennined with a hemocytometer) present in a volume of Kpike- The oocyst spike was added to a water volume of Keed;this constituted the seed suspension that was added to the filter f l u e n t at a volumetric flow rate of Specific seeding conditions and detailed information necessary for the calculation of Crheo~en'CCIl are available in Appendix D,Table D.2. Equation 5.1 describes the expected filter influent oocyst concentration by accounting for dilution of the seed suspension as it is introduced into the filter influent flow Stream. Measured filter influent oocyst concentrations (CF[)during the experiments were consistently -75% of the theoretical filter influent oocyst concentration (CrheoreriCa~)These concentrations ranged fiom 50% to 91% of the theoretical filter influent oocyst concentration. Consistent with the analytical recovery studies described in Appendix A, an assurnption of 75% analytical recovery for the direct vacuum filtration/IFA method used to enurnerate the measured filter influent concentrations (CF/) and inspection of the Crheorericai/CFI ratio suggests essentially no loss of oocysts to the seeding apparatus. From these data it can be concluded that the observed C. p a m m removals were due to filtration, rather than systern losses. These data also demonstrated that the seeding and coefficients of practices consistently resulted in comparable ratios of CF~Crheoren'Caf variation, suggesting that oocyst seeding was comparable in the dual- and tri-media filters during al1 of the operational conditions investigated. The theoretical and measured filter influent data for al1 of the experiments discussed in this chapter are reported in Table 5.4. Table 5.4 TheoreticaI and Measured Filter Muent C.p a m Concentration Data Stable Inactivated 1 Stable Inactivated 2 StabIe Inactivated 3 StabIe Viable 1 Stable Viable 2 Stable Viable 3 Ripening Inactivated 1 Ripening Inactivated 2 Ripening Inactivated 3 Table 5.4 Theoretical and Measured Filter Influent C. pannun Concentration Data (Continued) Ripening Viable 1 Ripening Viable 2 Ripening Viable 3 Coagulation Failure Inactivated 1 Coagulation Failure Inactivated 2 Coagulation Failure Inachvated 3 Coagulation Failure Viable 1 Coagulation Failure Viable 2 Coagulation Failure Viable 3 AVERAGE COEFF. VARIATION The >5-log pilot-scale removals achieved during the stable filter operation experiments discussed in this chapter are on the higher end of pawurn removals that have been reported in the literature. Full-scale oocyst removals have typically been reported in the range of 2 to 3-log (e-g, KelIey et al., 1995; Nieminski and Ongerth, 1995). Pilot-scale p a m m removal data have suggested that filters can achieve a~ywherefiom 2-3 log (e-g., Ongerth and Pecoraro, 1995; Kelley et al., 1995; Patania et al., 1995), to 3-4 log (Yates et al., 1997), or >5-log (e-g.. Patania et al., 1995; LeChevallier et al., 1991~) removal of oocysts. Several expIanations are possible for this range of differences in C.pantum removal by filtration. When d i s c k i n g oocyst removal data, it is important to note that log rernoval estimates are often limited by influent concentration. Assuming no background levels of oocysts, a filter cannot remove more oocysts than are naturally present in the raw water or have been spiked into the water. Furthemore, the removal efficiency of a filter can only be fiilly evatuated when oocysts are present in the effluent in reliably countable concentrations (a count of at least 10 oocysts per sample was suggested in Chapter 4). When no oocysts are found in filter effluent samples, detection limits are commonly used to calcdate log removals; several different practices for reporting detection limits have been used in the literature. Reported log removals may be highly inaccurate in such cases and larger sample volumes or higher idluent concentrations may have been necessary for achieving adequate nurnbers of oocysts in filter effluent samples. Ensuring high seeded oocyst concentrations can be particularly critical when evaluating the oocyst removal capacity of filters under optimal operating conditions where filter effluent oocyst concentrations are often 1ow (e-g.,Table 5.1 to Table 5.3). Analytical differences may also contribute to reported variability in the C. parvuin removal capacities of filters between studies. Several analytical methods with varied reliabilities have been reported in the literature (Vesey et al., 1993a; Nieminski et al., 1995; Clancy et al., 1999). When comparing C. parvum data fiom different studies, it is cntical to consider such factors as analytical recovery (and whether or not it is incorporated into the Enal assessrnent of oocyst removal) and uncertainty associated with recovery. To date, very few studies have incorporated analytical recovery into calculations of oocyst removal. Although a few authors have. suggested methods for incorporating analytical recovery and uncertainty associated with recovery (Nahrstedt and GimbeI, 1996; Parkburst and Stem, 1998), the suggested approaches have been rarely, if ever, applied in the iiterature. The considerable differences in the reported range of parvclrn rernovals in the literature are not surprising given the lack of reliable analytical methods and statistical tools for describing data uncertainty associated with analytical inconsistencies. In addition to analytical issues, C. parvum removal data could also be substantially affected by the operating conditions during which removal assessments were performed; the Limited reporting of such operational data has in part prompted the research descnbed in this thesis. The present stiidy concurs with others (Patania et al., 1995; Charles et ai., 1995) that suggested that some periods during the filter cycle (e.g., sub-optimal coagulation) have a higher likelihood of pathogen passage than others (e-g.. optimized operation). Deviations fiom optimal operating conditions might have contributed to the range of Cryptosporidium removals reported in the filtration literature. Filter media depth and size distribution rnight also play a role. The differences in oocyst removals during the operating conditions discussed in the present study underscored the need for accurate and thorough description of experimental operational and design conditions and M e r investigation of C. pawum removals by filters; this type of research was performed and is discussed in Chapter 6. The present bench-scale study indicated that formalin-inactivated oocysts were reliable surrogates for viable oocysts during removal studies. The stable operation and ripening experiments did not demonstrate substantial differences between removals of jarcoagulated, viable and formalin-inactivated C. parvum when the filters treated optimally coagulated water. Without coagulation of either the raw water or the oocysts, inactivated oocysts were rernoved somewhat more readily than viable oocysts, possibly due to differences in surface charge; however, a larger range of removals was observed for the viable oocysts. This result was particularly important because differences in oocyst zeta potential were likely most pronounced d u ~ coagulation g failure. The relatively smaI1 difference between zeta potentials of viabIe and inactivated oocysts that has been reported in the literature (Ongerth and Pecoraro, 1996) may contribute to differences in oocyst removal by filtration with increased replication. Although such a fhding rnight be usehl in understanding laboratory results obtained from filters challenged with noncoagulated oocysts, it is udikely to be of practical sipificance at full-scale where oocyst surface charge is impacted by multiple factors including coagulation. It is important to note that during the present investigation the filters were conditioned with opthalIy coagulated water for two hours prior to the stable operation and coagulation failure experiments. In accordance with filtration theory, it was expected that the conditioning penod prior to oocyst seeding resulted in the capture of particles that subsequently acted as collectors within the filter. The captured particles were expected to contribute to continued particle (and oocyst) removal during seeding. Thus, the stable operation experiments realistically represented the optimal filtration conditions. During the coagulation failure experirnents, the two-hour conditioning period likely favored more oocyst removal than would have occurred if the filter media had not been exposed to coagulated water pnor to oocyst seeding. The relative impact of the conditioning period on oocyst removal in a specific water type and coagulation regime not currently understood, however. The relative importance of filter conditioning at stable operating conditions prior to coagulation upsets is discussed in greater detail in Chapter 6. If the filtered water had not been coagulated at all, differences in oocyst zeta potential rnight have resulted in diEerent removals of viable and inactivated oocysts by filtration. Although coagulant-fkee conditions do not occur during normal water treatment, investigations of coagulant-fiee conditions may be warranted. The coagulation failure experiments conducted during the present research were representative of a coagulant feed failure. These experiments successfully demonstrated a worst-case assessrnent of the adequacy of using formalin-inactivated oocysts as surrogates for viable during challenge studies. Coagulation failure represented the period of greatest oocyst passage of the conditions investigated, regardless of media type (Figure 5.10). C. p a m m removal by filtration decreased by >3-log as a result of coagulation failure; oocyst removals during this period were dramatically lower than during either ripening or stable operation. The tremendous deterioration in oocyst removal during coagulation failure concurred with other studies that also demonstrated that optimized coagulation was critical for achieving good C.pamrm removal by filters (Patania et al., 1995; Charles et al., 1995; Ongerth and Pecoraro, 1995; Dugan et al., 1999). Perhaps rnost importantly, the coagulation failure data presented in this chapter revealed that sutistantial C.pawum passage through filters codd occur at turbidities below 0.3 NTU (Fiayre 5.1 1 to Figure 5-14), which is the current filter effluent turbidity standard mandated by the U S . EPA's lnterirn Enhanced Surface Water Treatment Rule (USEPA, 1998). These data emphasized that prompt response to coagulation upsets can be an important strategy for rninimizing C p t o s p o n X u m and likeIy other pathogen passage through filtration. In addition, these data also underscored the generally accepted idea that while traditional performance measures such as turbidity are indicative of treatrnent performance, they are not necessarily quantitative indicators of C.parvum removal by filtration. This point is emphasized when coagulation failure is contrasted with rïpening in dual-media filten (Figure 5.1 1 to Figure 5.14); indicating that the same filter effluent turbidities could correspond to very different oocyst removals by filtration. It should be noted that the data in Figure 5.11 to Figure 5.14 were collected when filter effluent turbidities were rapidly changing. Although the oocyst concentrations and log removals originated fkom samples collected over 15 minutes during coagulation failure, the filter effluent turbidity values were those measured at the start of the 15-minute sampling period. This is less of an issue when examinhg the ripening data because C p a m m samples were collected every £ive minutes while turbidity was measured every three minutes. A more detailed assessment of the relationships between traditional performance measures (turbidity and particle counts) and oocyst removal by fiiters is presented in Chapters 6 and 7. Figure 5.1 1 Dual-media filter effluent oocyst concentrations and oocyst 103 removals as a fùnction of filter effluent turbidity during stable operation and coagulation failure. Figure 5.12 Dual-media filter effluent oocyst concentrations and oocyst log removals as a function of filter effluent turbidity during stable operation and ripening. Stable opeiation O Figure 5.13 Tri-media filter effluent oocyst concentrations and oocyst log removals as a function of filter effluent turbidity during stable operation and coaguf ation failwe. Log Removal O OocyswL f Stable operation l i Filter Effluent Turbidity (NTU) Figure 5.14 Tri-media filter effluent oocyst concentrations and oocyst log removals as a fiction of filter effluent turbidity during stable operation and npening. Srimmarized in Table 5.4, measured filter influent oocyst concentrations (CF[)d u ~ g each of the replicate experiments were consistently -75% of the theoreticai filter influent oocyst concentration (Cthmreticd). The recovery data presented in Chapter 3 indicated that the analytical method for C.parvum consistently achieved approxïmately 75% recovery of oocysts fiom both filter influent and effluent samples fiom the water matrix studied during the investigations described in this chapter. Assurning that an analytical recovery of 75% contributed to the difference between the theoretical (Crheore~cal) and measured (CFI) filter influent C. parvzrrn concentrations presented in Table 5.4, it could be concluded that almost a11 of the seeded oocysts reached the filter (Le,, negligible loss of oocysts to the seeding equipment). This was considered critical for achieving consistent results and reliably countable oocyst concentrations in the filter emuent, especially during stable operation. Unaffected by losses to the seeding apparatus, log removal calculations were based on measured influent oocyst concentration rather than theoretical secded concentration. Loglo removals of C. parvum were calculated h m the individual pairs of rneasured filter influent and effluent oocyst concentrations. As discussed in Chapter 3, the analytical method for C.pawum yielded comparable recovery profiles fiom the filter influent and effluent water matrices studied during the present investigation. Therefore, the recovery profiles did not greatly impact confidence intervals for oocyst removal because the inherent methodological uncertainty of recovery was consistent. This investigation has yielded several important insights into C-ptosporidizrm p a m m removal by drinking water filters- The following conclusions can be drawn fiom the research presented in this chapter: Fonnalin-inactivated C-pawtrm oocysts were found to be reliable surrogates for viable oocysts during filtration studies conducted during stable operation, ripening, and coagulation failure in both dual- and tri-media filters, suggesting that formalininactivated oocysts are good surrogates for viable oocysts. This is a very important k d i n g because virtually al1 filtration studies investigating C. p a w m removal are conducted with inactivated oocysts for practical reasons. C. panurn removals by filtration were moderately lower (-0.5 to 1-103) during ripening than during stable operation. During the coagulation failure conditions investigated, the C. parvurn removal capacity of both dual- and tri-media filters was severely and sipificantly compromised relative to both filter ripening and stable (optirnized) operation. Coagulation failure decreased C. parvzrm removal by >3-log relative to stable operation. C.parvum removals were not substantially different in dual- and tri-media filters during stable filter operation, ripening, and coagulation failure. Increased replication rnay reveal any significance of the marginally higher median removals achieved by tri-media filtration. Traditional filtration performance parameters such as turbidity (USEPA, 1989) and particle counts are good indicators of filtered water quality, however, they are not reliable surrogates for C. parvum removal during drinking water treatrnent (Nieminski and Ongerth, 1995; Patania et al., 1995; Fox et al., 1998). While removals of other microbial parameters such as aerobic spores of Bacillus subtilis have correlated somewhat with the removal of C. pamrm by physico-chernical processes such as granular media filtration, no adequate quantitative surropate for removal of the pathogen has yet been identified. The research described in this chapter investigates design and operational strategies for maxirnizing C.p a m m removal by filtration. A secondary objective of this work was to evaluate the adequacy of oocyst-sized polystyrene microspheres as surrogates for the removal of C.pamrn by filtration and to demonstrate how their removaI compares to that of B. subtilis spores. The experimental approach employed dunng this research consisted of: 1. identifying benchmark removals of pathogens and surrogates during stable (optimized) operation; 2. investigating pathogen removals during vulnerabk penods of filter operation and reIating them quantitatively to removals during stable operation; and, 3. investigating design and operational strategies for maximizing pathogen removals during potable water production. D e t a s regarding the experimental approach, experimental facilities, raw water qualities, and analfical methods are provided in Chapter 3. Key points are summarïzed below. Pilot-scale experùnents were conducted at three research platforms: the Ottawa pilot plant, the Windsor pilot plant, and the University of Watdoo (UW) pilot plant. The multiple research platfonns allowed for the investigation of different raw water types, water temperatures, coagulation regimes, and filter designs. Formalin-inactivated C. panrum oocysts were seeded to a final filter influent concentration of -105 oocysts& at each experimental location. Microspheres w-ere evaluated as potential nirrogates for C. p a m m because they were oocyst-sized, easy to identiQ, and relatively easy to enurnerate; they were seeded at Ottawa and the UW. Spores of B. subtilis were also examined as potential surrogates for C p a m removal by filtration because they had previously demonstrated some success as surrogates for C. pamtm removal (Scott et al., 1997) and could act as benchmark surrogates to which the microsphere removals could be compared. Turbidity, particle counts, flow rates, and head loss were monitored during each experiment. The optimized treatment conditions were selected to rneet the 0.1 NTU turbidity goal of the Partnership for Safe Water, a voluntary treatment optimization program sponsored by the U.S. Environmental Protection Agency and the American Water Works Association. Except for one experiment specified later, microorganisms seedùig occurred at the filter influent. In many experiments this transpired over a one-hour period, although in a few experiments the seeding period was extended. During seeding, several (typically three or four) pairs of filter influent and effluent samples were taken for the enumeration of the seeded microorganisms and microspheres; removals were calculated fiom these pairs. Each experimental condition was replicated several times. The non-optirnal operating conditions investigated were sub-optimal coagulation and z sudden increase in flow to the a t e r (hydradic step). Both of these operationa1 conditions may occur to varying degrees in real treatment plants. The ripening and breakthrough portions of the fifter cycle were also investigated because of the potential for decreased rnicroorganisrn removals during these periods. Ripening is an essentially unavoidable part of every filter cycle; in some plants the water produced during this period is mtered to waste rather than sent to the distribution system. In principle, potable water production during end-of-- turbidity or particle breakthrough can be avoided by temiinating filter operation sufficiently earfy. Of course when this is done, shorter filter cycles result and increase costs due to an ùicreased number of backwashes. The breakthrough experirnents focused on identifying how long fdters could be lefi in senice without compromising C.pawurn removal while filter effluent qualiîy, though deteriorating, might still be considered acceptable by traditional turbidity or particle criteriaAs discussed in Chapter 3, samples (three, four, or five) were collected at the filter influent and filter effluent locations during each of the experiments. Sampling typicalfy occurred during the period when the seed suspension was added at the filter influent. The statistical approach described in Chapter 4 was used to calculate the 95% confidence intervals and removal ranges for each experiment and operating condition. Statistical cornparisons were made by calculating the upper and Iower limits of the 95% confidence interval for each sample; ciifferences between data sets were statistically different at the 5% significance level when the confidence intervals did not overlap. Overall removal ranges (described in Chapter 4), rather than overall confidence intervals were calculated for the experïxnents for which data could not be pooled due to dynamic operating conditions. For both the confidence interval and removal range calcuIations, the beta distribution parameters used to describe the C.parvzrm recovery profile were the overall Ottawa (a= 23 -91, b = 8.26) and overall UW- 1-5 NTU (a= 11.OS, b = 4.42) parameters presented in Appendix C . The Ottawa recovery parameters were also used in assessing the Windsor data. Al1 of the figures and tables containing data that incorporate the uncertainty associated wiîh analytical recovery are clearly identified. Although mode and mean removals are presented in the tables, rnoderemovals are used and presented in the figures because they represent the most likely removals (i.e.,the highest probability of occurrence). The detailed experimental schedules, raw water quality data, performance data, microorganism and microsphere data, 95% confidence interval endpoints, and adjusted ranges (for non-pooled data) are provided in Appendix D. The majority of the pilot-scale C. p a m m , B. subtiiis, and polystyrene microsphere removal investigations were conducted at the Britannia Pilot Plant in Ottawa, Ontario, Canada. Described in Chapter 3, ail of the seven general operating conditions investigated during this thesis research (stable filter operation, no coagulation, suboptimal coagulation, ripening, hydraulic step, end-of-m, and breakthrough) were investigated at Ottawa, allowing for the relative cornparison of C.parvum removal during the various operating conditions. Each condition was studied at least in tripkate; several sub-categories of the seven general conditions were also examined. The rationale and specific seeding conditions for each of the experiments were surnmarized in Chapter 3 and are elaborated upon below. 6.2.1 Stable (Optimized) Operation Pilot-scale C.pamrn removal data reported in the literature have suggested that fdters can achieve anywhere fiom 2 to >5-log removal of oocysts (Patania et al., 1995; Fox et al., 1998). To detemine the C.p a m m removal capacity of the pilot filter in Ottawa, eight stable operation experiments were performed throughout the approximate two-year period of C.parvurn removal investigations at the pilot plant. experiments represented optimized pretreatment The stable operation (coagulation, flocculation, and sedimentation) and filtration conditions (Le., no perturbations or upsets). 6.2.1.1 Experimental Design The purpose of these experiments was to document the highest C.pawurn (and B. subtilis and polystyrene microsphere) removals that could be obtained by the pilot-scale filter during optimal operating conditions. As mentioned previously, seeding and sampling were conducted after at least four hours of filter operation, during the early to middle portion of the filter cycle. These experiments were repeated severd tuiles because they provided a baseline against which the other operating conditions were compared. Eight stzble operation experiments were conducted at Ottawa. Jar-coagulated C.parvum oocysts were seeded h o the filter influent for one hour during these experiments. B. strbtilis spores and yellow polystyrene microspheres were also seeded dunrig some of these experiments. Samples were coUected at 15, 30,45, and 55 minutes after the start of seeding. Stable operation at Ottawa was considered as the period during which frlter effluent turbidities and total particle concentrations 22p.m were consistently c0.1 NTU and <5 particleslml respectively during the entire filter cycle (after ripening). One additional stable operation investigation was conducted in which seeding occurred at the rapid mix location. The purpose of this experiment was to document filter performance when the seeded rnicroorganisms had been exposed to the sequence of coagulation, floccuIation and sedimentation in the pilot plant. As discussed in Chapter 3, this approach could not be used on a regular basis because it would iikeIy result in fdter effluent C.pawurn concentrations below the detection limit of the analytical method used during this research. A puise-input of 10' C.p a m m oocysts and 10' B. srrbtiZis spores was used during this experiment. Though no rnicroorganisms were expected in the fdter effluent, this experiment was conducted to detemiine if there were gross ciifferences in C.pamirn removal by the pilot-filter as a result of jar-coagulation and direct seeding into the filter influent. Two additional stable filter operation experiments were conducted during spring runoff conditions when changes in raw water quality impacted the chernical pretreatment process prior to filtration. Since the filters were able to maintain excellent filtered water quality (consistently <0.1 NTU filter effluent turbidity) during spring runoff, it was speculated that the stable operation microorganisrn removals during runoff would be similar to those achieved during non-runoff periods. The runoff experiments were conducted in a manner identical to the other stable operation experiments, with the exception that B. subtiZis spores were not seeded. Al1 of the stable operation experiments and seeding conditions are sumrnarized in Table 6.1. Table 6.1 Summary of Stable Operation Experirnents at the Ottawa Pilot Plant Experiment Date C.p a m m B. szrbtilis Blue Spheres Stable Operation 08/06/98 J Stable Operation During Runoff 04/05/99 04/ 12/99 J Stable Operation 07/20/99 Seeded at Rapid Mix J Yellow Spheres J 4 Filter effluent turbidity and particle counts were consistently low (-0.04 NTU and <5 particles/ml respective1y) duMg the stable operation experiments. Filter effluent turbidity and particle counts, C. parvum, B. sztbtilis, and polystyrene rnicrosphere removals, and total particle reductions through the treatment process during stable operation at Ottawa are summarized in Table 6.2. The general filter performance data and detailed instantaneous turbidity and particle data are available in Appendix D (Tables D.4 and D.5 respectiveiy). The detailed instantaneous C.parvum and B. strbtilis data are available in Table D.6 while detailed microsphere data are available in Table D.7. Table 6.2 Filter Performance During Stable Operation at Ottawa Date 08/06/98 0veraII- LogloRemovaI Loglo~ e d . ' Effluent Concentration (mean 4 standard deviation) C.p a m m B. srrbtilis Blue Yellow Particles Particles Turbidity Spheres Spheres 22pm (#/rnL) 0 4.9 0.2 1 - 5.5 2 037 3.9 5 1.13 -- - -- 3.2 + 0.29 4.9 10.22 3.6 _+ 0.63 3.7 + 3.9 0.02 2 0.00 3.1 _+ 3.5 0-035 0.01 * L O reduction ~ of partides œ* through treatment process (plant infhent to filter effluent). non-dete& - this is indicated by &%" symboi here. Ail experiments except for 04105'99 (nino@, 04/12/99 (runoff), and 07/20/99 (seeded at mpid mix). AU of filter effluent samples w& *O* Traditional performance measures (turbidity and total particle counts) and filter effluent microorganism and microsphere concentrations were relatively constant d d g the stable operation investigations and consistent between replicate experirnents. Filter effluent turbidity and total particle counts rarely exceeded 0.06 NTU and 5 partic!es/ml during stable operation at Ottawa. The turbidity and particle responses of the Ottawa filter durhg the May 3 1, 1999 stable operation experiment are shown in 6.1, which demonstrates the excellent and consistent filter effluent turbidity and particle concentrations achieved by the pilot filter duMg stable operation. This figure also indicates the seeding period and sampling times. The filter effluent turbidity and particle counts were generally consistent between the eight replicate stable operation experiments at Ottawa. The uistantaneous turbidity and particle data and microorganism data are available in Appendix D (Tables D.5 and D.6, respectively). S e e d i n ~Period Filter effluent turbidity and particle concentration during May 3 1, 1999 stable filter operation experiment at Ottawa. 10 - j 100000 Seeding Period r C. p m u m toocystslL) 15 Figure 6.2 -- Filter effluent particle, C. p a m m , and B. szlbtilis concentrations during May 3 1, 1999 stable fdter operation experiment at Ottawa. C.p a m passage tbrough the Ottawa filter during stable operation was consistently low, with fiiter effluent concentrations often near the method detection lirnit. Filter effluent particle, C.parvum, and B. subtilis concentrations fiom the Ottawa Nter during the May 31, 1999 stable experiment are shown in Fi,gre 6.2- While fiiter influent C.p a m m concentrations were typically 10' to 106 oocysts/L,, the filter effluent concentrations were consistently less than 10 oocysts/l and ofien <Z oocyst/L. Though they were more variable between experiments, the filter influent B. szibtiZis concentrations were roughly comparable to those of C. parvurn, particularly in the later experiments foliowing the implementation of improvements to the B. szrbtilis analytical method, B. subtilis spores were invariably detected in the filter effluent samples at higher concentrations than the C. parvurn oocysts, however, as shown in Figure 6.2, B. szrbtilis spore passage through the Ottawa filter was generally consistent with the C.p a m m data. Similar C. parvurn oocyst and polystyrene microsphere removals were observed during the January 19, 2000 stable operation experiment at Ottawa. The filter effluent turbidity and total particle concentrations during the stable operation experirnent investigating oocyst and microsphere removals are presented in Figure 6.3. The filter influent oocyst and microsphere concentrations during the stable operation experiment were similar (means of -4.6 x 10' oocystsL and -6.5 x 10' microspheresL respectively). Both C.p a m m and microspheres were found in all of the effluent samples collected during this experiment (Figure 6.4). The filter effluent C. pawurn concentrations during the Jmuary 19, 2000 experiment were less than 10 oocystsil, which was consistent with the other stable operation experiments conducted at Ottawa. Though the filter effluent microsphere concentrations were more variable than those of C. parvunz, they were generally of the same order of magnitude, wiîh effluent concentrations of less than 13 microspheresL These data indicated that polystyrene microsphere passage through the Ottawa filter was also genedly consistent with the C. parvum data. However, since only one experiment examined microsphere removals during stable operation, m e r investigations would be necessary to address the relationship between C. pawurn and microsphere removals during this operating condition. The instantaneous C. parvum, B. subrilis, and microsphere data are available in Appendix D (Tables D.6 and D.7). Seeding Period Samphg Points O O O Tie Figure 6.3 1; Figure 6.4. Filter effluent turbidity and particle concentration during January 19,2000 stable filter operation expe&ent at Oîîawa. s Microspheres (spheres/L)I Filter effluent particle, C. parvum, and microsphere concentrations during January 19,2000 stable filter operation experiment at Ottawa. parvzrm removds during stable (optimized) operation ranged f?om 4.7 to 5.8-log, with a mean oocyst removai of 5.5-log (32 pairs in total). The experiments were conducted at several raw water temperatures (ranging f?om I to 24°C; Table C.4); no deterioration in C.pamrrn removals was observed at temperatures as low as 1 O C . B.subtiiis removals ranged fiom 2.0 to 5.1-log, with a mean rernoval of 3.9-log (20 pairs in total); microsphere removals ranged £rom 4.7 to 5.1-log, with a mean removal of 4.9-log (4 pairs in total). The removal data are siImmarized in Table 6.2. Although there was some variation in removals caiculated on the basis of individual influent-effluent sample pairs of microorganisms or spheres, the removals during a given experiment were fairly reproducible (Table 6.2). The single exception to the consistency of microorganism removds was an experiment (9/22/98) in which B. sz~btilisremovals demonstrated a particularly high standard deviation; however, tbis variability was attnbuted to analytical difficulties. The subsequent standard deviationç of stable operation experiments involving B. subtiiis spores were considerably tower (Table 6.2). B. szrbtilis removals by filtration were generally lower than C.pamrrn removals. This result is not surprishg a v e n that the B. subtilis spores are -1 p m in size which is at the minimum transport efficiency for particle removal by filtration. regression and comparing the Perforrning linear pawum and B. subtilis removds during stable operation demonstrated that they were only weakly correlated, with a coefficient of determination (R2) of only 0.16 or 0.19, dependhg on whether or not the 9/22/98 data set was inciuded. (Figure 6.5). While spore removais were generally lower than oocysts removals, the lack of a relationship between oocyst and spore removals suggested that B. subtilis removal data were not quantitatively indicative of the filter's ability to remove C.parvzcm. Polystyrene microsphere removals were generally lower than C.pamrrn removals during the stable operation experiment perforrned with microspheres. Though the negative slope of the best fit hne achieved by linear regression suggests that microsphere removals were not a good surrogate for C. parvum removal by filtration, however, only one such experiment was perforrned during stable operation. More data would be necessary to better elucidate the relationship between C.pawlrm and microsphere removals during this operating period. 6.0 y = -0.48~+ 7-43 5.0 --- 'R = 0.63 -- & - - - --- 4.0 -0 ---a- * A -AF & A f - O =O C a y = 1.51~-435 0 C 2 =O y = 1.45~-4-10 R2= 0.19 3.0 -- I VI A -s 2.0 - & A & A 3 B. subtilis (without 9/?7/98 data) Microsphere -Lin- (B. subtilis (\vithout 9 P 2 9 8 data)) 0.0 4.0 Figure 6.5 3.5 5.0 5J 6.0 Relationship berneen C. p a m m , B. subtilis, and microsphere removals by fdtration during stable operation at Ottawa. C. parvum Removal (loglo) Figure 6.6 Relationship between C.parvum removal by filtration and total particle (22 pm) reductions through the plant during stable operation at Ottawa. Although total particle reductions (2 2,m) through the treatment process were calcdated (Table 6.2), it should be recded that these removds were based on raw water rather than fdter influent values. Filter influent particle concentrations could not be measured due to floc breakage and accumulation on and in the influent particle counter sensor chamber. As with B. subtilis, examination of the data in Table 6.2, might suggest that particle reduction through the treatment process was a conservative indicator of a filter's ability to remove C. p a m m during stable operating conditions. As would be expected, a cornparison of C. parvum removds and particle reductions during stable operation indicated that these parameters were not correlated (Figure 6.6). This result is not surprishg given that particles can change in size in, be removed by, and attach and detach fiom media in the various treatment processes between the raw water and filter effluent sampling locations. The two stable operation experiments conducted C W g spring runoff conditions were critical fiom an operational perspective because they were conducted during a period when chernical pretreatment wôs challenged due to a si@~cant variation in the raw water quality. This ultimately resulted in considerably higher settled water turbidities (34 NTU during stable operation during spring -off as compared to 0.5-2.5 NTU during non-runoff periods). Despite elevated raw and settled water turbidities (Table C.3), filter effluent turbidities and particle concentrations remained low during runoff operations and were comparable to other periods of stable filter operation (Table 6.2). C. p a m m removals by filtration during spring nuioff also appeared comparable to those achieved during the other stable operation periods (Table 6.2). C.parvtrrn and B. subrilis removal data for the experiment durùig which seeding occmed at the rapid mix are also presented in Table 6.2. The C. parvum and B. subtilis removals during this experiment were clearly lower than during the other stable operation experiments that involved microorganisrn seeding into the fdter influent. The lower microorganism removals were associated with the considerably lower filter influent concentrations during this experiment (Table C.6) that resulted fkom microorganism removal during coagulation, flocculation, and sedimentation. In the case of C.parvum, no oocysts were detected in the filter effluent samples during this experiment. The handling of non-detects is descnbed in Chapter 3. 6.2.1.3 Statistical Analysis Confidence intervals for the individual C. p a m m log removals during stable operation were calculated using the method described in Chapter 4, and incorporated the uncertainty of analytical recovery. The endpoints of the 95% confidence intervals for oocyst removal are summarized in Table 6.3. Almost ali of these codïdence intervals overlap, thereby failhg to demonstrate statisticaliy significant differences between most of the data collected d w stable operation (Table 6.3, cc=0.05). Similady the stable operation during spring runoff experiments also failed to demonstrate significant ciifferences. in C. p a m n z removals between the replicate experiments (Table 6.3, 0 ~ 4 . 0 5 ) .The paired removal data for each of these conditions were pooled because the data from each were expected to be replicates (since operational conditions and water quality were essentially consistent throughout the experiments) and failed to demonstrated significant differences in microorganism/microsphereremovals (when the individual confidence intervds were compared). The overall95% confidence interval for C. pczmrm removal d-g stable operation was 5.4- to 5.6-log (32 influent-effluent pairs of data), during spring runoff it was 5.1- to 5.4-log. The endpoints of the range indicate the lowest and highest endpoints of the individual 95% confidence intervals for the C.parvum removds observed during the operating period. The overall 95% confidence interval for the pooled microsphere removals was 4.6- to 4.9-log. The C.p a m m removal range was 4.0- to 7.2-log during stable operation and 4.4- to 6.6log during stable operation durkg runofK The range for the individual experiments is summarized in Table 6.3. Cornparison of the removal ranges and the overall confidence intervals (calculated with pooled data) demonstrates how increased counts (achieved by either larger processed volumes of water or replicate samples) substantially decrease uncertainty. Although pooling was possible during stable operation, it would not be possible during more dynamic operating conditions. A comparison of the confidence intervals for the stable operation and stable operation during nuioff data also f d e d to demonstrate statisticaliy sigdkmt differences between C. parvurn removals during these penods (Table 6.3, 01=0.05), suggesting that the filter could compensate for the chailenged pretreatment conditions during this period. A comparison of the overaiI confidence intervaIs for C.pawurn and microsphere remvals did demonstrate significant differences at the 5% significance level, suggesting that the microspheres were removed to a lesser degree than oocysts. Table 6.3 95% Confidence Intervals and p u m m Rernoval Ranges During Stable Operation at Ottawa Table 6.3 95% Confidence Intervals and C. parvum Removal Ranges During Stable Operation at Ottawa (Continued) Time 01/19/00 Stable Operation D u ~ Runoff g 6.2.1.4 03/05/99 (log,o) (logio) 4.79 6 -12 439 15 4.42 4.82 4.83 5.0 1 5-70 4.42 t=30 t=45 t=55 6-16 5.01 6.33 t= 15 t= (lo~ro) 7.03 633 Discussion The stable filter operation experiments demonstrated that C. parvurn removals of >5-log could be consistently achieved by the Ottawa fdter at water temperames ranging fiom 1 to 24°C. The fdter was also able to compensate for higher seîtled water turbidities that occurred during stable operation during spring runoff conditions. These data provided the baseline removals to which other operating conditions at Ottawa were compared. Although the rnicrosphere removals during stable operation at Ottawa were significantly lower than the oocyst removals, this result does not necessarily suggest that microspheres are poor surrogates for C.parvurn removal by fdtration. It should be recalled that only one stable operation experiment was performed with microspheres and that a 1:l relationship is not necessarily required for a good surrogate. n i e issue of microspheres as overall surrogates for C. parvum removal by filtration is discussed in Chapter 7. 6.2.2 Ripening It has been suggested that 90% of the particles that pass through a well-operated filter do so during ripening (07MeLia and Ali, 1978; Amirtharajah, 1985; Cranston and Amirrharajah, 1987). To assess the passage of C.p a m and B. subtilis through filters during ripening, several experiments were conducted during filter ripening at the Ottawa pilot plant after periods of stable operation. Relative to stable operation, lower microorganism removais were expected during the period of particle passage that is typically associated with non-attachment during ripening (O'Melia and Ali, 1978). Experimental Design 6.2.2.1 During ripening at Ottawa, the fdter effluent turbidity and particle concentration typicdy increased during the first 10 minutes of the experirnent and then decreased toward the end of the seedùig penod. The seeding period during the rïpening experiments was only 30 minutes because füter water quality typically improved to filter effluent turbidities < 0.1 NTU md particle concentrations < 5-10 particies!mL (such as those observed during stable operation) during the first thuty minutes of operation after backwashing. Samples were collected at 5, 10, 15, 20, and 25 minutes afîer seeding cornmenced. Sample collection was designed so that it was essentially continuous, resulting in composite samples representing five-minute intervals. The ripening experiments and seeding conditions are summarized in Table 6.4. Table 6.4 Surnmary of Ripening Experiments at the Ottawa Pilot Plant Experiment Ripening 6.2.2.2 Date 10/27/98 11/03/98 1 1/10/98 C. p a m m B. szrbtilis J J J J Blue Spheres Yellow Spheres 4 J Results Filter effluent turbidity and particle counts, C. p a m m and B. subtilis removals by filtration, and total particle reductions during npening at Ottawa are summarized in Table 6.5. The general filter performance data and detailed instantaneous turbidity and particle data are available in Appendix D (Tables D.4 and D.5 respectively). The detaiied instantaneous C. parvurn and B. subtilis data are also in Appendix D (Table D.6). Filter effluent turbidity, total particle counts, and microorganism concentrations varied throughout the ripening period. Peak filter effluent turbidity and particle counts during ripening at Ottawa ranged fiom 0.41 to 0.69 NTU and 91 to 840 particles/mL respectively. The turbidity and particle response on November 10, 1998 is shown in Figure 6.7. This figure depicts a sharp fdter effluent turbidity and particle peak approximately 5 to 10 minutes after the filter returned to service following backwashingThe seeding period and sampling times are noted in Figure 6.7 and indicate that microorganism seeding occurred during the turbidity and particle spikes associated with ripening while sampling captured most of the spikes, which lasted less than 30 minutes. The filter effluent turbidity and particle counts peaked at approximately 0.67 NTU and 840 particles/ml during the November 10, 1998 experiment (Figure 6.7). On October 27, 1998, the filter effluent turbidity and particle counts peaked at 0.69 NTU and 174 particles/mL respectively; they peaked at 0.41 NTU and 91 particIes/mL during the November 3, 1998 experiment. Although the magnitude of spikes varied somewhat between experiments, the duration of the ripening period was generally similar between the three experiments. The instantaneous turbidity and particle data (which did not necessarily indicate the peak turbidity and particle concentrations that occurred during ripening) are available in Appendk D (Table D.5 j. Table 6.5 Filter Performance During Ripening at Ottawa Date LogloRemoval Log ~ e d . ' Effluent Concentration (mean fstandard deviation) C.pamrm B. subtik Blue YelIow Particles Particles Turbidity - - + 5.1 10.66 2.1 2 1.36 2.9 0.55 22 2 37 * L O ~reduction of partides through treatment process (plant influent to filter effluent). Overali 0.13 _+ 0.07 1 .O 0.5 1 -5 Fiiter Run Time (houn) Figure 6.7 Filter effluent hirbidity and particle concentration during November 10, 1998 ripening experiment at Ottawa. Maximum Effluent Oocyst Concentration Observed (n = 32) During Stable (Optimized) Opention -: LOO. E h \ 3= Y 1 0.0 0-5 I 1 .O I 1 -5 I 2-0 Filter Run Time (hours) Figure 6.8 Filter effluent particle and C.parvum concentrations during November 10, 1998 ripening experirnent at Ottawa. Fûcer Run Time fiours) Figure 6.9 Filter effluent particle and B. subtilis concentrations d W g November 10, 1998 npening experiment at Ottawa. C. p a m m passage through the Ottawa filter during ripening appeared to have a dynamic pattern similar to that of turbidity and particles. Consistent with the moderately elevated turbidity and particle counts, more oocysts were detected in the fdter effluent during ripening relative to stable operation. Filter effluent C.parvzrm and parîicle concentrations during the November 10, 1998 experirnent are shown in Figure 6.8. This figure indicates that the moderate spike in filter effluent oocyst concentration that occurred at 5 minutes was concurrent with the spikes in turbidity and particles (Figure 6.7). The 110 oocysfi spike during ripening on November 10, 1998 accompanied the highest instantaneous particle concentration during microorganism sampling (145 particles/mL) and one of the highest turbiciity values (0.19 NTU) experienced during the ripening experiments. During the October 27, 1998 experirnent the highest fdter effluent total particle counts during sampling were considerably lower (3 1 particles/mL) than on November 10, with a generally comparable turbidity (0.09 NTU). Accordingly, the peak fiter effluent C . p a m m concentration was lower (82 oocystsL). The lowest peak filter effluent oocyst concentration during ripening at Ottawa (62 oocystsll) occurkd during the November 3, 1998 experiment when the highest filter effluent particle counts and turbidity during sampling were 58 partic1edmL and 0.32 NTU respectively. B. subtilis passage through the fdter was somewhat consistent with the C.p a m m data, however, trends in filter effluent spore concentration spikes were not necessarily cornmensurate with the C. pamrm, particle, and turbidity spikes during the ripening experiments (Figure 6.9). The detailed p a m m and B. subtilis data are in Appendix D (Table D.6). As during stable operation, B. subtilis removals by filtration were lower than C-pawurn removds (Table 6.5) and were not indicative of the fdter's ability to remove C.pawum (Figure 6.10). In this case, the slope of the best fit line by Iinear regression describing the relationship between oocyst and spore removals was negative, suggesting higher oocyst removals corresponding to lower spore removals - an unlikely scenario in reality. Of course, reductions in total particle counts (2 2 p ) were also lower than C.pamrrn removals during ripening (Table 6.5). They did, however, appear to be somewhat simiiar ' of 0.76 (Figure 6.10). to oocyst removals during these experiments, as indicated by the R 45 Figure 6.10 5 .O C.p a m m Removal (log,,) 5.5 Relationship between C. parvum and B. subtilis removals by filtration and total particle (22 pm) reductions through the plant during ripenùig at Ottawa. 6.2.2.3 Statistical Analysis Confidence intemals for C.pawum removals during filter ripening were calculated using the method described in Chapter 4. The endpoints of the 95% confidence intervals for oocyst removal are srrmmarized in Table 6.6. During each experiment, the C. parvum removal at the beginning of the ripening period, corresponding to the highest level of filter effluent turbidity and particles, was significantly iower (Table 6.6, a=0.05)than the removals observed during the subsequent samples when flltered water quality was improving lie., turriidity was decreasing). M e r ripening, the water quality and C.parvum removals improved and were consistent with those observed during the stable operation experiments (Table 6.3). Overall confidence intervals for C.parvzim removal could not be calculated due to the dynamic changes in attachment efficiency within the filter during this period. The endpoints of the range of parvum removals observed during ripenuig and adjusted to incorporate analytical uncertainty are 3.7- and 7.2-log, the higher endpoint being affected by the length of the sampling period (a longer samphg approaches or extends into stable operating conditions). The removal ranges determined for the individual experiments are summarized in Table 6.6. Table 6.6 95% Confidence Intervals and C.p a m m Removal Ranges Durhg Ripening at Ottawa Experiment Ripening Date 10/17/98 Sarnple Time t=5 CI I O W ~ CI u (Ioglo) 3.93 (~o~Io) 3.2 1 p ~ ~ R ,01 R uppcr (lo~io) 3.93 m%o) 7.17 6.2.2.4 Discussion Relative to stable operation, moderately higher filter efÏluent C.parvum and B. subti2i.s concentrations were observed during ripening (Table C.6). These data suggested lesser or weaker and more transient attachent of these microorganisms during ripening than during stable fdter operation. The filter microorganisrn concentration trends were generally consistent with spikes in filter effluent turbidity and particle counts. Although the instantaneous turbidity and particle count data (Table CS)were generally indicative of filter performance, the actual values were not quantitative indicators of C.p a m m passage through the flters. While raw water total particle counts and filter influent oocyst concentrations were relatively constant during the experimental penod, Nter effluent total particle counts or turbidity values were not necessarily correlated with specific filter effluent C.parvurn concentrations (Figure 6.10). There appeared to be some relationship between the reduction of total particles through the treatment process and parvurn removals b y filtration during filter ripening (Figure 6.10). It should be recalled, however, that the particle reductions are instantaneous concentrations that continuously changed during the C.parvum sampling period as a result of dynamic raw and filter effluent water quality at the time of sampling. Since the detention time between the r a w water and filter influent locations was approximately three hours, the particle reductions were not necessarily indicative of those that occmed during sampling. The raw water data were not adjusted to account for detention time since substantial diEerences in raw water quality did not occur during these, or any of the other experiments. Potential surrogates for C.pannrrn removal during filtration are M e r discussed in Chapter 7. Dunng the first few minutes of ripening, the removals of C.p a m m were lower than during stable fdter operation. The range of C. parvum removals during r i p e h g is Large because, in addition to incorporating anaIytical uncertainty associated with oocyst recovery, the statistical comparisons and overd C. parvzim removal ranges included al1 of the removals obsemed during the sampiing periods. The twenty-five minute sampling period (thirty-minute seeding period) at Ottawa included the very transient spike in particle counts and turbidity as well as periods of improved filter effluent turbidity and particle concentrations. If the ripening period were defined differently (e-g.,as the period when filter effluent turbidities were 0.2 NTU or higher) and did not include the last few samples collected when water quality had substantially improved, the overall C. pawum removal ranges during rïpening would have been smailer (Table 6.6). In general, these experiments suggested a brief and minimal-to-moderate increase in C.pawurn passage tbrough filters concurrent with spikes in filter efluent turbidity and particle counts during ripening. Overall, the C.panntrn removals during rîpening were approximately 0.5-log lower than during stable operation, consistent with Patania et al. (1995), Hall et al. (1995), and Swaim et al. (1996) who demonstrated oocyst removals during npening that were 0.5 to 1.O-log lower than during stable operation. These results were generally consistent with the findings of LeChevallier et al. (199 1), which failed to demonstrate statistically different oocyst removais between ripening and stable filter operation. 6.2.3 Breakthrough Both non-attachment and detachment occur during breakthrough conditions (Ginn et al., 1992; Moran et al., 1993b). As particle detachment and non-attachment increase, h c r e a s i n g pathogen passage through filters would also be expected. To assess the effect of end-of-run and breakthrough conditions on the removal of C.pawurn, B. subtilis, and polystyrene microspheres by filters, several experiments were conducted at Ottawa. Endof-run, early breakthrough, and late breakthrough filtration conditions were investigated. 6.2-3.1 Experimental Design End-of-nui operation was defined as the period during which subtle changes in filter effluent turbidity and particle counts were noticed and increased to approximately 0.1 NTU of fdter effluent turbidity. Early breakthrough filtration was defined as the period when filter effluent turbidities were between approximately O. 1 and 0.3 NTU. Though non-optimal due to changing filter effluent turbidity and particle counts, both of these operating conditions were at the upper Mt of compliance with the IESWTR fdter effluent turbidity requirement of 0.3 NTU or less in greater than 95% of measured samples. The late breakthrough experiments investigated C-p a m m and surrogate removal when filters were operated for a period shortly d e r reaching effluent turbidities of approximately 0.3 NTU oust out of cornpliance with the IESWTR). Jar-coagulated C.parvurn were seeded into the filter influent for one hour during these experïments; samples were coiiected at 15, 30, 45, and 55 minutes d e r the start of seeding. B. subtilis spores and polystyrene rnicrospheres were also seeded during some of these experiments. When microsphere seedùzg was included, yellow and blue polystyrene microspheres were sequentially seeded into the filter influent. The seeding of blue rnicrospheres was planned for a one-hour period of stable operation approximately five h o m pnor to the ônset of breakthrough (when filter effluent turbidities started noticeably changing). However, it was very difficuIt to predict when turbidity breakthrough would occur at the end of a filter cycle, so blue microsphere seeding occurred anywhere fiom one to five hours before the onset of turbidity breakthrough. This was done to elucidate mechaaistic behavior duing early breakthrough (detachment versus non-attachent) by accumulating the spheres on the filter just before the filter effluent quality started substantially deteriorating. Yeilow microspheres, when included, were always seeded concurrently with rhe C-parvum oocysts during the one-hour seeding period when filter effluent turbidities were deteriorating. The C.pawum and yellow microsphere concentrations in the fdter effluents were considered indicative of either non-attachment or very poor, transient attachment and subsequent detachment. It should be noted that these experiments were not designed to conclusively detemilne whether non-capture or release of previously deposited oocysts was the dominant mechanism of oocyst passage hto filter effluents during breakthrough. Rather, they were designed to determine the effects of filter operation on pathogen passage through filters while providing some insight into the mechanistic behavior of pathogen passage through filters during breakthrough. B. subtilis spores were typically seeded concurrzntly with the C.p a m m oocysts; the only exceptions to this were two early breakthrough experiments (Decernber 20 and 22, 1999). During these experiments, jar-coagulated B. subtilis spores were seeded for one hour and then jar-coagulated C. parvum oocysts were seeded for one hour as breakthrough cornmenced. Samples were only collected at 15, 30, 45, and 55 minutes after the start of oocyst seeding. The staggered seeding was planned to help elucidate mechanistic behavior during eady breakthrough by accumulating B. subtilis spores on the filter, just before the filter effluent quality started substantially deteriorating. As with the blue rnicrospheres, if spores were subsequently detected in the filter effluent when influent concentrations were low (ideally near O c m ) , they would be indicative of detachment. The breakthrough experiments and seeding conditions are summarized in Table 6.7. Table 6.7 Summary of Breakthrough Experiments at the Ottawa Pilot Plant Date C.p a m m B. subtilis End-of-Run O I/2 1/99 J J Early Breakthrough 03/0 1/00 03/03/00 03/OU00 J J J Late Brcakthrough 11/25/98 t 2/09/98 01/13/99 12/20/99 12/22/99 J J J J J Experirnent - %seededbefore p a m to elucidate mechanistic behavior. Blue Spheres Yellow Spheres J J J J* J* 4 J Results 6.23.2 Filter effluent turbidity and particle counts, C. pawurn, B. szrbtilis, and polystyrene microsphere removals, and total particle reductions during end-of-- early breakthrough, and late breakrthrough at Ottawa are siimmarized in Table 6.8. The general fdter performance data and detailed instantaneuus turbidity and particle data are available in Appendix D (Tables D.4 and D.5 respectively). The detailed instantaneous C. parvurn, B. subtiiis, and microsphere data are also available in Appendix D (Table D.6 and D.7 respectively). Table 6.8 Filter Performance Durhg End-of-Run and Breakthrough at Ottawa Date Log10 Removal Loglo~ e d . ' Effluent Concentration (mean 4 standard deviation) C. p a m m B. subtilis Blue Yellow Particles Particles Turbidity Spheres Spheres 22 p (+/rnL) End-of-Run 01/21/99 3.0 10.53 03/08/00 2.3 2 0.53 03/09/00 2.7 0.44 Overall 2.7 & 0.54 + Early Breakthrough 0310 1/00 2.2 + 0.50 03/03/00 2.3 + 0.34 03/04/00 2.1 2 0.33 Overail 2.1 _+ 0 3 6 Late Breakthrough 1 1/25/98 1.8 0.03 12/09/98 1.7 0.06 01/13/99 1.5 2 0.68 12/20/99 1.5+0.20 12/22/99 1-4 0.05 Overall 1.6 0.32 1.0 _+ 0.34 1.6 0.25 2.2 5 0.67 106 f 107 0.52 _+ 0.26 .LO~ reduction of particles through treatment process (plant influent to filter effluent). + + + + **&ta not available. + Of the end-of-nui and breakthrough investigations, the largest deterioration in oocyst, spore, and microsphere removals was expected during late breakthrough filtration when fdter effluent turbidities were hi& (-0.3 NTLT). Two experiments investigating C.parvum and microsphere removal during late breakthrough were performed. Three other late breakthrough experïments investigated C.parvum and B. subtilis removal by filtration. The filter effluent turbidity and total particle counts were typically -0.25- 0.3 NTU at the start of these experiments. The fdter effluent turbidity, seeding period, and sampling points during one of the late breakthrough experiments investigating C.pamim and polystyrene microsphere removal by filtration are presented in Figure 6.1 1;unfominately, total particle counts were not available during this experiment. 0.7 -Seeding Penod Seeding Penod 1200 AM 11:OO PM Time Fiawe 6.1 1. Filter effluent turbidity during December 22, 1999 late breakthrough experiment at Ottawa. The elevated filter effluent turbidities during late breakthrough at Ottawa were accornpanied by high filter effluent C.p a m m and microsphere concentrations relative to those obtained during the stable operation experiments. C. p a m m d d g the late breakthrough experiments ranged fkom 1.3 to 1-8-Iog, with a mean oocyst removal of 1.4log. Microsphere removals during late breakthrough ranged fkom 1.3 to 2.0-log, with a mean microsphere removal of 1-8-log. Relative to stable operation, these experiments clearly demonstrated a >3-log reduchon in C. parvtrrn and microsphere removals during Iate breakthrough. The frlter influent oocyst and microsphere concentrations during these experiments were similar, -6.9 x oocystsL and -6.8 10' 10' microspheresll on x average. Both C. p a m m and microspheres were found in dl of the filter effluent samples durùig the Iate breakthrough experiments (Figure 6.12). DetaiIed microorganism and microsphere concentration and removal data are provided in Appendix D (Table D.6 and Table D.7, respectively). 0.8 100000 a - 0: -- IO000 0.6 - 0 5 -- 8 2- :-IO00 2 C- 5 2? 0.4 .3 -- 2- -- Maximum EfliuentOocyst Concentration Observed (n = 32) n I During Stable (Optimized) Operation ?fi* ...............................--....--..-......... ! * II..------ *..*CCCCC--. 1- 10 0.1 -1 I 1 :O0P M 1200 A M C -- 62 III. . 03 -- 1:ûû AM Tiie Figure 6.12. 3m Cs X f gu 5 g 2- 0.3 -- 0.0 7 10:OO PM q-3 C> i-i Filter effluent turbidity, C. parvurn, and microsphere concentrations during Decernber 22, 1999 late breakthrough experiment at Ottawa. The elevated filter effluent particle counts and turbidities during late breakthrough at Ottawa were accompanied by hi& flter efnuent aerobic spore concentrations relative to those obtained during the stable operation experiments. B. subtilis removals by the pilot filter ranged fiom 0.1 to 1.5-log during the late breakthrough experiments, with a mean spore removal of 1.O-Iog. Relative to stable operation, these experiments clearly demonstrated an -3-log reduction in B. subtilis removals during late breakthrough. This result was similar to the decrease in C.p a m m removals during this period relative to stable operation. Filter effluent B. sztbtilis spore concentrations were similar to the C. p n m m concentrations; similarly, spores were detected in al1 of the filter effluent samples during the Iate breakthrough experiments (Figure 6.1 3). Detailed microorganisrn concentration and removal data are provided in Appeadix D (Table D.6). 8 ! Maximum Effluent Oocyst Concentration Observed in = 32) Durhg Stable (Optunized) Opention :' . .&&& *-. - -? &d&-= ..................................... j. Figure 6.13. Filter effluent particle counts, C. p a m m , and B. strbtilis concentrations durkg December 22, 1999 late breakthrough experiment at Ottawa- The early breakthrough filtration performance data demonstrated a very dynamic penod during which filter effluent turbidity and total particle counts changed considerably at the Ottawa pilot plant. Three experiments investigating C.parvum and microsphere removal during early breakthrough were performed. The filter effluent turbidity was low (-0.04 - 0.08 NIU) at the start of these experiments and increased to approximately 0.2 NTU by the end of the experiments, a level stiil in cornpliance with the IESWTR requirements for filter effluent turbidity. The filter effluent turbidity and particle concentrations during one of the early breakrthrough experiments (March 3,2000) are presented in Figure 6.14. j Figure 6.14. - Particles >= 2 p Seeding Filter effluent turbidity and particle concentration during March 3,2000 early breakthrough experiment at Ottawa. The increased filter effluent turbidities and particle concentrations during early breakthrough were also accompanied by increased filter effluent C-parvum and microsphere concentrations relative to those obtained during stable operation. C.parvum removals by the pilot Hter during early breakthrough ranged from 1.7 to 2.8-log, with a mean oocyst removal of 2.1-log. Microsphere removals during early breakthrough also ranged nom 1.7 to 2.8-log, with a mean microsphere removal of 2.1 -log. The observed ranges of C.parvum and microsphere removais were 22-log lower than those observed during stable opeïation. The fdter influent oocyst and microsphere concentrations during the early breakthrough experiments were similar and averaged -6.6 -5.7 x 10' microspheresL respectively. x 10' oocystsL and C. pamcm oocysts and microspheres were found in all of the fdter effluent samples during these experiments (Figure 6.15). -0 .C 1 Z y IO -- Maximum Emuent Oocyst Concentration Observed (n = 32) During Stable (Optimized) Opention O E C 100 5 -- l 10 ii C Tiie Figure 6.15. Filter particle, C.parvum, and microsphere concentrations during March 3,2000 early breakthrough experiment at Ottawa. Considerable detenoration in C. parvum and polystyrene microsphere removal d-g end-of-nin filtration when filter effluent turbidities were increasing but still below 0.1 NTU was not expected given other data in the literature (Patania et al.,1995; Baudin and Lahé, 1998). Two experiments investigating C. parvum and microsphere rernovai during end-of-nui filtration were perfonned. The filter effluent turbidity was low (-0.04 NTU) at the start of these experiments and increased to approxirnately 0.13 NTU by the end of the seeding period in the first experiment and 0.06 NTU at the end of the seeding period in the second experiment. The filter effluent turtbidity and particle concentration data f?om the second (March 9, 2000) of three end-of-ri experlments are presented in Fi-oure 6-16. Although the filter effluent turbidities and particle concentrations increased only slightly during the end-of-run experiments at Ottawa, they were accompanied by considerably elevated filter effluent C. p a m and rnicrosphere concentrations relative to those obtained during stable operation- C.p a m m removals by the pilot filter during end-ofnin filtration ranged from 1.8 to 3.3-10% with a mean oocyst removal of 2.5-log. Microsphere removals during early breakthrough ranged from 1.8 to 3.1 -log, with a mean microsphere removal of 2.4-log. The filter influent oocyst and microsphere concentrations during the end-of-nui experiments were sirnilar, -6.8 -5.6 x x oocystslL and 10' rnicrospheres5 on average. Both C. parvzrrn oocysts and microspheres were found in al1 of the filter effluent samples duruig the end-of-nui experiments (Figure 6.17). -- An increase in filter effluent B. subtilis was also observed during the end-of-run filtration experiments. Filter effluent particle counts and C. parvzrrn and B. szibtilis removai data for the January 21, 1999 experhent are presented in Figure 6.18. These data demonstrated that the increase in filter effluent spores during the experimental period was generally consistent with the increase in füter effluent oocyst concentrations. The M a c h 8 and 9,2000 data were generally consistent with the January 2 1, 1999 data and indicated considerable passage of oocysts through the filter during end-of-= conditions. Figure 6.16. Filter effluent turbidity and particle concentration during March 9,2000 end-of-run experiment at Ottawa. C Maximum Eftluent Oocyst Concenmtion Observed (n= 32) During SnbIe (Opamired) Opmuon g IO 3 9 0 PM $00 PM 5:00 PM 600 PM 7:00P M 8:Oû PM a - -- 9:00 PM 10:OO PM Time Figure 6.17. Filter effluent particle, C.parvum, and microsphere concentrations during March 9,2000 end-of-= experiment at Ottawa. - 1: ~anicies>= 2gm / Seeding Penod I ! 1 1 1 Maximum Efiluent O o q a Concentration Obsewed (n = 32) During Stable (Optimized) Operation l r i a h P 6 -- ----- 1200 PM 1 1:O0AM Time Figure 6.18 Filter effluent particle, C.p a m m , and B. subtilis concentrations during January 2 1, 1999 end-of-experirnent at Ottawa. - The C. parvum; B. subtilis, and polystyrene microsphere removals and total particle reductions through the treatment process are sumxnarized in a box-and-whisker plot (Figure 6.19). This plot includes the entire range of removals observed during the end- of-run and breakthrough experimental periods (defined in Section 6.2.3 :1). Removais during stable operation are also included for cornpanson. These data clearly indicated severe deterioration in oocyst, spore, and microsphere removals during end-of-run (< 0.1 NTU filter effluent turbidity), early breakthrough (0.1 - 0.3 NTU filter effluent turbidity), and late breakthrough (> 0.3 NTU füter effluent turbidity) fùtration. The deterioration in total particle reduction through the treatment process did not appear as severe as the deterioration in oocyst, spore, and microsphere removals. Not surpnsingly, microorganism and microsphere removals generally continued to decrease as filter effluent turbidities and total particle concentrations increased duriag these successive operating periods. Stable Operation Figure 6.19. End-of-Run Eariy Breakthrough Late Breakzhrouph Box-and-whisker plot of C.parvum, B. stibtilis, microsphere removals b y filtration and total particle (2 2pm) reductions by the plant durhg stable operation, end-of-nin, and breakthrough at Ottawa. The box-and-whisker plot indicates considerable similarity between C.parvum oocyst and polystyrene microsphere removals d u ~ the g range of end-of-run and breakthrough operational conditions investigated (Figure 6.19). As indicated in Figure 6.20, the relationship between oocyst and microsphere removais by the pilot frlter was linear with a coefficient of determinahon (R~) of 0.92. Nthough oocyst and microsphere removals during stable operation (Figure 6.5) were not as similar as those obtained during the endof-run and breakthrough conditions, the data in Figure 6.19 and Figure 6.20 suggest that polystyrene microsphere removals were good, and under some circumstances (stable operation), conservative indicators of C. parvtirn removals b y filtration. Further investigations are necessary to determine if this relationship is consistent during stable operation and other non-optimal operating periods. Potential sumogates for C. p a m m removal during filtration will be discussed in M e r detail in Chapter 7. 1 .O O .O 2.0 3.0 1.0 C.parvum Removal (logio) 5 .O 6.0 Relationship between C.parvtrm and microsphere removais by filtration during end-of-run, early breakthrough, and late breakthrough at Ottawa. Figure 6.20. 6.0 a - B.subtilis Pilr?.ïcles(>='plTl) Lineu (B. subtiiis ) Linear (Particles ( X Z p n ) ) _ --, 5-O-- , 0 0.0 1.O 2.0 I 1 , 3.0 4.0 5.0 6.0 parvum Rmovai (logio) Figure 6.2 1. Relationship between C.p a m m and B. subtilis removals b y filtration and total particle (22 p ) reduction through the plant during end-of-run, early breakthrough, and late breakthrough at Ottawa. B. subtilis spore removals and total particle reductions during end-of-ru, early breakthrough, and late breakthrough were compared to C. p a ~ r mremovals in Figure 6.2 1. These data indicated some correlation between C.parvzrm and B. strbtilis removals. R') of 0.62. The Least-squares linear regression indicated a coefficient of determination ( relationship between C.parvum removal by filtration and total particle reductions through the treatment process was considerably weaker, with a coefficient of determination (R') of only 0.22. As mentioned above, a more detailed discussion of these relationships is provided in Chapter 7. The end-of-= and breakthrough experiments also included the seeding of blue microspheres to elucidate the mechanisms associated with C.p a m and polystyrene microsphere passage through the filter during this operational period. The blue microspheres were seeded several hours ( w i c w -3 to 5 hours) before the end-of-nui and breakùirough conditions were anticipated and before the seeding of oocysts and yellow microspheres. Ideaily, filter influent concentrations of blue microspheres near O spheresa were targeted during the end-of-run and breakthrough conditions so that fdter effluent concentrations could be attributed to detachment. Although the filter influent concentrations of blue microspheres steadily decreased after seeding, the desired low concentrations of blue microspheres were rarely achieved during these experiments (Table C.7), due to non-plug-flow conditions in the standing water above the filter media. Some examples of possible detachment occurred during the March 1 and March 4, 2000 early breakthrough experiments when filter effluent sphere concentrations exceeded filter influent concentrations. However, regardless of influent concentration, the filter effluent concentrations of yellow microspheres were higher than the filter effluent concentrations of blue microspheres, typically by one order of magnitude or more (Table C.7). This result suggested that the passage of oocysts through the filter d u ~ end-of-run g and breakthrough filtration was largely a furiction of non-attachment or weak, transient attachent and subsequent detachment. 6.2.33 Statistical Analysis Confidence intervals for C. p a m log removals for the individual samples during endof-run, early breakthrough, and late breakthrough were calculated using the method described in Chapter 3. The endpoints of the 95% confidence intervals for oocyst removal are summarized in Table 6.9. A cornparison of the individual confidence intervals for C.pawum removal during end-of-run, early breakthrough, or late breakthrough at Ottawa, relative to those obtained during stable operation, demonstrated that the Iower removals during these conditions were significantly different fiom those obtained during stable operation (Table 6.3 and Table 6.9, a=0.05). As during other dynamic operating periods, the overaii confidence intervals for end-of- run and break-through operating conditions could not be calculated. During these conditions, the attachment efficiency of the filter was likely decreasing (or detachment increasing), resulting in the observed deterioration in treated water (Nter effluent) quality. Therefore, the individual data could not be pooled, as they were not true replicates, precluding the calculation of the overall confidence intervals. The endpoints of the range are the lowest and highest endpoints of the individual 95% confidence intervals for C.p a m m removal observed during the given operating period. The C.pawum removal range was 2.0- to 3.6-109 during end-of-run operation, 1.7- to 2.9-log during early-breakthrough operation, and 0.00 1- to 2.0-log during Iate breakthrough operation. The range for the individual experiments is summarized in Table 6.9. In general, these data dernonstrated a clear deterioration in C.p a m m removal as water quality deteriorated (Le., fiter effluent turbidity was increasing) as the filter cycle progressed. Table 6.9 95% Confidence Intervals and C.parvzlm Removal Ranges During End-of-Run,Early Breakthrough, and Late Breakthrough at Ottawa End-of-Run Early Breakthrough Late Breakthrough Table 6.9 95% Confidence Intervals and C.parvum Removal Ranges During End-of-Ru, Early Breakthrough, and Late Breakthrough at Ottawa (Continued) Experiment 6.2.3.4 Date Sample CI lowc a UPW -,01 RUPP~ Discussion The data collected during this research indicated that end-of-run, early breakthrough, and late breakthrough filtration ail represent operating penods where C. p a m m removals can be severely compromised relative to those obtained durhg optimized filtration (Figure 6.19). During stable filtration, the pilot-scale dual-media filter approximately achieved 25-log removal of both oocysts and microspheres. During end-of-run operation when filter effluent turbidities demonstrated the first signs of increasing (but were still below 0.1 NTU), C. parvurn removal decreased to approximately 2-3-log; oocyst and microsphere removals decreased even more during early and late breakthrough. The early and late breakthrough findings were in general agreement with earlier fmdings that demonstrated that turbidity breakthrough could be accompanied by considerable passage of Giardia cysts (Logsdon et al., 198 la). The early breakthrough results were also different fiom those obtained during other investigations of Giardia and Ctypt~sponXurnpassage through fdters during breakthrough when effluent turbidities increased fiom 0.1 NTU to 0.2 NTU or higher. While Giardia removal was approximately 0.5-log lower during breakthrough, no difference between C. p a m removals during stable operation and breakthrough was observed (Patania et al., 1995). Nonetheless, C. parvzim oocysts were clearly present in the filter effluents during these experiments and at high concentrations. It is possible that other factors such as chernical pretreatment, which is critical for optimizkg C.panwm removal by £ïltration (Patania et al., 19951, may impact the degree o f pathogen passage that occurs during early breakthrough Ntration, resulting in differences between studies, such as the one observed between the present study and that of Patania et al. (1995). Cornmensurate with other studies (Patania et al., 1995; Nieminski and Ongerth, 1995), the elevated filter effluent C.pawum concentrations during end-of-run and breakthrough filtration were loosely correlated with increasing fdter effluent particle counts and turbidity. This study supported previous studies that suggested that oocyst-sized microsphere removals may be good surrogates for C.p a m m removd by filtration (Swertfeger et al,,1998). A good linear fit between oocyst and microsphere removals was presented in Figure 6-17. As previously discussed, the rnicrosphere findings are particularly important because no reliable quantitative surrogates for the removal of C. parvurn during water treatment exist at this time. The microspheres offer several advantages for use over oocysts in treatment process evaluations such as those reported in this study. The microspheres cost substantially less than oocysts, do not require antibody stainùig, do not pose the public health threats of C. parvum (although they could not necessanly be introduced into fiiL1scale plants), are resilient during treatment, and rnay possibly lend themselves to automated enGeration. As was shown above, the microspheres also appear to be removed at levels that are compamble to oocyst removals (or slightly lower in the case of stable operation), suggesting that they are generally conservative surrogates that can be effectively used for investigating C.p a m m removal in treatment evaluations. The severe reduction in C. pawum removal during the end-of-nui and breakthrough experiments relative to the optimized filtration experiments should be considered in the context of the experirnental conditions. Since filter influent C. p a m m concentrations are not typically anywhere near the -los oocystsK influent concentrations used during these experiments, the removal data collected during this investigation should not be used to quantitatively predict differences in oocyst removals at various points in the filter cycle in fiill-scale plants. The end-of-run and early breakthrough filtration data are noteworthy because they clearly indicated a severe deterioration in C. pantunz removal by filtration during operating conditions in compliance with the 0.3 NTU IESUrIR filter effluent turbidity requirement. From an operational perspective, these data might challenge the appropriateness of an upper turbidity Limit of 0.3 NTU for all points in the filter cycle. The data suggest that placing filters out of senice eadier in the filter cycle (perhaps when effluent turbidities are near 0.1 NTU) may be a desirable strategy for maximizing pathogen removal. The high C. p a m m and yeliow microsphere concentrations that were found in the filter effluents after only fifteen minutes of seeding suggested that the piimary mechanism of oocyst passage through the filter dUnng end-of-nrn and breakthrough filtration was either non-attachent or weak, transient attachent and subsequent detacbment. The relatively low filter effluent concentrations of blue microspheres also supported this conclusion. Though far from incontroverhble, this conclusion is in general agreement with other studies that suggested non-attachment was an important mechanism of particie passage through fdters during breakthrough operation (Ginn et al., 1992; Moran et ai., 1993b). 6.2.4 Coagulant Effects Delivering the optimal coagulant dosage during water treatment can be challenguig because the residence tirne for water in treatment plants is relatively short. In locations where influent water quaiity can change rapidly, similar rapid responses in coagulant dosing are not aiways easily attained. Even though the treatment plant may be capable of producing low-turbidity water when high-turbidity water consistently enters the plant, the speed of the raw water quality changes relative to the coagulation response may lead to increased particle and potentially pathogen passage through the treatment process. Adequate chernical pretreatment has been repeatedly shown to be a critical step in maintaining good particle removal during filtration (O 'Melia, 1985; Vaidyanathan and Tien, 1988; Tobiason and O'Melia, 1988). Several studies have suggested the importance of coagulation processes for improving filter removal efficiencies of pathogens (Logsdon et al., 1981a; Ongerth and Pecoraro, 1995; Patania et al., 1995). The importance of coagulation for C. pamrn removal by filters was also demonstrated by the bench-scale investigations (Chapter 5) that indicated >5-log rernoval of oocysts durkg optunized treatment while only approximately 1-log removal of oocysts during coagulation failure. Other studies have yielded sirnilar results (Chartes et al., 1995; Swertfeger et al., 1998). 6.2.4.1 Experimental Design As described in Chapter 3, the coagulation regime applied at Ottawa utilizes a high alun dose with silicate addition for combined TOC and particle removal. To investigate the effect of coagulation on C. p a m removal by filtration, two basic types of experiments were performed: no coagulation and sub-optimal coagulation. Coagulation effects on Cparvum rernoval by filtration were then assessed by comparing these conditions to optimal coagulation conditions (stable operation experiments). The no coagulant experhnents were conducted to estimate the worst-case condition of coagulant failure. Coagulants were removed from the treatment process in several different combinations- These combinations were as follows: 1 . No Coagzilant - Extended Duration No coagulant was used in either the pilot plant or the jar-coagulation apparatus for three fiter nuis prior to the experiment, simulahg a long-term coagulant failure. Coagulant aid (silicate) was maintained at its' optimum dosage. No Coagulant - Short Duration No coagulant was used in the pilot plant or the j ar-coagulation apparatus, sirnulating a short-term COagulant failure. Coagulant aid (silicate) was maintained at its' optimum dosage. During these experiments, the primary coagulant (alun) pump was turned off for a short period of tune prior to seeding the microorganisms. Fihers are typically already conditioned during such short-term failures with substantial amounts of floc present in the fdters that c m act as effective collectors for some period of time. Comparing data from this experiment to those obtained during the no coagulant - extended duration experiments revealed the effects of filter conditioning with optimally coagulated water pnor to coagulation failure. 3. No Coagrlant in the Pilot Plant. These experiments were identical to the no coagulant - short duration experiments with the exception that although no coagulant was fed to the pilot piant, it was added to the jar-coagulation apparatus. Coagulant aid (silicate) was maintained at its' optimum dosage in both the pilot plant and the jar-coagulation apparatus. This experiment investigated coagulant effects associated with the addition of jar-coagulated microorganisms to the fdter influent. 4. No Coagulant in the Jar. This experiment was identical to the stable operation experiments with the exception that no coagulant was added to the jar-coagulation apparatus. Coagulant aid (silicate) was mainrained at its' optimum dosage in both the pilot plant and the jar-coagulation apparatus. Like the previous experiment (no coagulant in the pilot plant), this experiment investigated potential filter aid effects associated with the relatively high flter influent concentration of coagulant that resulted fiom the addition of jarcoagulated microorganisms at the fdter influent location and the relative role of filter conditioning. 5. N ü Silicate (Coagulant Aîd,. In this experiment the activated silica feed was discontinued for two filter cycles prior to the experiment. Normal alum coagulation was maintained in the pilot plant and jar-coagulation apparatus, simulating an extended duration coagulant aid failure. This experiment was conducted to investigate the role of activated silica (as compared to alum) in the coagulation/filtration process at Ottawa. C. parvurn oocysts were seeded into the filter influent for one hour during these experiments. Samples were collected at 15, 30, 45, and 55 minutes after the start of seeding. B. subtilis spores were also seeded into the filter influent during ail of these experiments except the no coagulant-extended duration experiment, during which yellow polystyrene microspheres were seeded. In addition to the no coagulant experirnents, sub-optimal coagulation experiments were conducted to determine the effects of changing coagulation conditions (without a change in raw water quality) on pathogen passage. The experiments sirnulated a 40 to 60 percent underfeed in coagulant dosage fiom the optimum. The silicate concentration did not change fiom the optimum concentration during these experiments. FiIter effluent turbidities of 0.2 to 0.3 NïU, which were stiU in cornpliance with IESWTR requirements, were targeted for the sub-optimal coagulation experiments. It was difficdt to predict how long it would take the filters to respond to the coagulant underfeed; therefore, during some experiments the target filter effluent turbidity was exceeded. Jar-coagulated C. parvum oocysts were seeded into the Nter influent for one hour during the sub-optimal coagulation experiments. Samples were collected at 15, 30, 45, and 55 minutes after the start of seeding. B. subtilis spores and polystyrene microspheres were also seeded during some of these experiments. During the one experiment that included microsphere seeding, both blue and yeilow polystyrene microspheres were seeded into the filter influent. As during the breakthrough experiments, the seeding of blue microspheres was ptanned for a one-hour period of stable operation approximately five hours prior to the penod of deteriorated filter effluent turbidities resulting fiom sub-optimal coagulation. Since it was difficult to predict exactly when the effects of sub-optimal coagulation would be observed in the filters, the seeding of the blue microspheres was actuallly conducted four hours pnor to the detenoration in fdter effluent turbidities that resulted from the coagulant underfeed. As during the breakthrough experiments, the goal of seeding the blue microspheres was to help elucidate mechanistic behavior during sub-optimal coagulation (detachment versus non-attachment) by accumulating the spheres on the filter, just before the filter effluent quality started substantially deterïorating. Yellow microspheres were seeded concurrently with the C.pawum oocysts, during the one-hour seeding penod when filter effluent turbidities were deteriorating. The Cparvurn and yeiIow microsphere concentrations in the filter effluents were considered indicative of either non-attachent or very poor, transient aitachment and subsequent detachment. The coagulant effects experiments and seeding conditions are summarized in Table 6.1 0. Table 6-10 Summary of Coagulant Effects Experiments at the Ottawa Pilot Plant Experiment No Coagulant Extended Duration Date C.parvurn B. subtilis 12/13/99 J J Blue Spheres Yellow Spheres J No Coagulant Short Duration No Coagulant in Plant No Coagulant in Jar No Silicate Sub-Optimal Coagulation (-50% underfeed) 6.2.4.2 Results Filter effluent turbidity and total particle counts, C.parvum, B. subtilis, and polystyrene microsphere removals, and particle reductions through the treatment process during the no coagulant and sub-optimal coagulation experiments at Ottawa are summarized in Table 6.1 1. The general filter performance data and detailed instantaneous turbidity and particle data are available in Appendix D (Tables D.4 and D.5 respectively). The detailed instantaneous C. panrurn and B. szibtilis and microsphere data are also available in Appendix D (Table D.6 and D.7, respectively). Table 6.1 1 Filter Performance During No Coagulant and Sub-Optimal Coagulation Experiments at Ottawa Date Log in Removai Login ~ e d . . Effluent Concentration (mean 5 standard deviation) C, p a m r n B. snbtilis Blue Yellow Particles Particles Turbidity Spheres Spheres >2pm (+/:/mL) (NTU) No Coagulant - Extended Duration 1 2 13/99 0 3 f0.02 0.2 f0.19 - No Coagulant - Short Duration 02/09/99 2.1 2 0.33 0.4 & 0.29 08/04/99 3.1 0.20 0.3 + 022 Overali 2.6 0.56 0 3 f0.28 + No Coagulant in Plant 06129i99 4.8 + 0.02 07/13/99 5.0 2 0.03 OveraU 4.9 f0.12 No Coaguiant in Jar 08118/99 5.6 f0.03 2.5 + O. 12 2.2 + O.18 2.4 _+ 0.22 3.2 f0.1 1 - - - - Sub-Optimal Coagulation (50% underfeed) 03/18/99 1.5 + 0.24 0.9+ O.13 - 03/23/99 3.3 + 0.42 1.7 + 0-25 05/04/99 5.3 0.17 0.9 2 0-03 03/1O/Oû 4.3 + O.13 Overd 3.6 2 1.48 1.2 2 0.43 1.7 f0.61 204 + 233 * ~ o reduction g of particles through rreatment process (plant influent to filter effluent). 0.59 +038 The box-and-whisker plots in Figure 6.22 through Figure 6.24 summarize the respective C p a m m and B. subrilis removals by fütration and the total particle reductions through the treatment process during the various no coagulant and sub-optimal coagulation experiments. As would be expected. the largest detenoration in oocyst, spore, and particle reductions (relative to stable operation) was observed during the no coagulantextended duration experiment when filter effluent hubidities were hi& (-2.2 NTU) and approximately equal to settled water turbidities. Almost no C. p a m m or B. subtilis removal or particle reduction was observed during this experiment, emphasizing the importance of good chernical pretreatment for the removal of pathogens by filtration. - Figure 6.22 7 No Coag. in Jar Sub-Optimd Coag. - No Coag. in No Coag. No Coag. Plant Short Durauon Extended Duration No Silica Box-and-whisker plot of C.parvurn removals by fdtration d u ~ the g no coagulant and sub-optimal coagulation experiments at Ottawa. - 6 -- C 3 & 2 E 3 2 1 Swble Operation -- 7 T n 520 O Figure 6.23 n=X n=8 n=4 n=J -- 6 -- 5 -- 3 -- 3 7 -I I Stable Operation n= 4 n = 12 Sub-Optimal Coag. II II 6 No Coag. in Jar No Silica 1 r I L O - No Coag in No Coag. No Coag. Pht Short Dwauon Extended Duration Box-and-whisker plot of B. subtiZis removals by filtration during the no coagulant and sub-optimal coagulation experiments at Ottawa. Stable Operation Figure 6.24 Sub-Optimal Coag No Coag in Jar No Silica No Cmg. ùi NO Coag. NOCmg. Plant Short Duration Evtended Duntion Box-and-whisker plot of total particle (22 p ) reductions through the plant during the no coagulant and sub-optimal coagulation experiments at Ottawa- Although filter effluent turbidities were also high (-0.6 to 0.8 NTU) during the no coagulant-short duration experiments, the deterioration in oocyst and spore removals by filtration and total particle reductions through the treatment process (relative to stable operation) was considerably less than during the no coagulant-extended duration experiment. More specifically, when coagulant was absent for only a short duration (several hours pnor to and during seeding), the fdter s a managed to achieve -2.7-log removal C. parvum. This level of oocyst removal represented an -3-log reduction relative to stable operation (based on median removals during the experiments). The respective -4-log and -2-log decreases in B. szrbtilis removals and total particle reductions through the treatment process were generally consistent with this trend. The absence of coagulant in the pilot plant and absence of coagulant during jarcoagulation experiments (respectively Iabeled "no coagulant in plant" and "no coagulant in jar") hvestigated the relative role of coagulant in oocyst coagulation and general filter performance. As indicated by Figure 6.22 and Figure 6.23, the C.p a m m and B. subtih removals during these experiments were both in the range obsenred d d g stable operation. The absence of coagulant in the pilot plant had a somewhat more noticeable, though still modest effect on C. parvum and B. subîilis removals by fdtration. Unlike the C.panium and B. sltbtilis removals, particle reductions were considerably lower during the experiments during which coagulant was absent fiom the plant, but not the jar (Figure 6.24). This result appeared consistent with differences in filter effluent turbidity and particles. During the no coagulant in the jar experiments, filter effluent turbidity and particles were consistently 0.03 NTU and 0.4 particles/mL respectively; however, during the no coagulant in the plant experiments filter effluent turbidities ranged fkom 0.65-0.76 N T ü and 403-536 particles/mL respectively (Table CS). The absence of activated silica during both pilot plant and jar coaguiation appeared to have no effect on the removal of C. parvum or B. subtilis by filtration, or the total particle (2 2 p ) reductions through the plant. As demonstrated by Figure 6.22 through Figure 6.24, the microorganism removals and particle reductions during the no silicate experiments were in the range of those observed during stable operation. This result was also consistent with the low filter effluent turbidities and particle counts, which were 0.03 NTU and <5 particles/mL respectively during the no silicate experirnent (Table C S ) . The sub-optimal coagulation experiments (40-60% decrease in coagulant concentration) demonstrated varied effects on C.parvzrrn removals by filtration, which ranged h m 1.3 to 5.5-log. C.p a m m removals d e g sub-optimal operating conditions ranged fiom comparable to or up to %log lower than those achieved during stable operation. It shouId be noted, however, that C. pamrm removals during sub-optimal coagulation when filter effluent turbidities were below 0.3 NTU (May 4, 1999 and March 10, 2000) were within -1-log of the removals achieved during stable operation. B. subtilis removals by filtration and total particle reductions (12 pm) through the treatment process similarly varied, respectively ranging fiom 0.8 to 2.0-log and 0.8 to 2.5-log during sub-optimal coagulation conditions. This range of microorganism removals and particle reductions may be in part explained by the wide range of fdter effluent turbidities (0.13-0.98 NTU) encountered during these experiments. During the one sub-optimal coagulation experiment perfomed with microspheres (March 10, 2000), C.p a m m removals ranged from 4.1 to >4.4log. Similady, the polystyrene microsphere removals ranged fkom 3.8 to 4.4log. Relative to the other sub-optimal coagulation experiments, the filter effluent turbidity was low (-0.17 NTU) at the stmt of this experiment and increased to approximately 0.36 NTU by the end of the seeding period. Consistent with the hirbidity data, total fdter effluent particle concentrations (i2pm) ranged from 109 to 151 particles/ml during the experiment. The filter effluent C-parvtrm, microsphere, and particle concentrations during the March 10, 2000 suboptimal coagulation experiment are presented in Figure 6.25. This fi*ou~e similar removals of C.p a m m oocysts and polystyrene microspheres. 160 100000 Seeding Period as C . p a r v u m ( o o c y ~ ) 100 -- . - 3 E -.-g ;= 80 Maximum Effluent O q s t Concentration Observed (n = 32) During Stable (Optimized)Operation -- O ?z c 60 -- O :-100 : 40 -- 0 IO:00 A M 1 1:00 A M 1200 P M 1:00 PM -:O0 PM 3100 PM The Figure 6.25 Filter effluent particle, C.parvurn, and microsphere concentrations during March 10,2000 sub-optimal coagulation experiment at Ottawa. The relationship between oocyst and microsphere removals by the pilot fïiter during the no coagulant and sub-optimal coagulation experirnents was fairly linear, as indicated in Figure 6.26, with a coefficient of determination (R') of 0.95. This relationship must be regarded as provisional, however, because data are only available for the two extremes. Nonetheless, the data in Figure 6.22 and Figure 6.26 provide additional support for the hypothesis that oocysts-sized polystyrene microsphere removals are reasonable indicators of C. parvurn removal by filtration. Further investigations are necessary to determine if this relationship is consistent during stable operation and other non-optimal operating periods. Potentiai surrogates for C. p a m m removal during filtration will be discussed in further detail in Chapter 7. 0.0 Figure 6.26. 1.O 3-0 3.0 4.0 C.parvum Removal (logIo) 5.0 6 .O Relationship between C. pamirn and microsphere removals by filtration during the no coagulant-extended duration, no silicate, and sub-optimal coagulation experiments at Ottawa. Figure 6.27. Relationship between C. p a m m and B. subtilis removals by filtration and total particle (2 2pm) reductions through the plant during al1 no coagulant and sub-optimal coagulation experiments at Ottawa. B. subtiiis spore removals by filtration and total particle reductions during the no coagulant and sub-optimal coagulation experiments were compared to C. parvum removals in Figure 6.27. niese data indicated some correlation between C.purvzrrn and B. subtiZis removals. Linear regression indicated a coefficient of determination (R') of 0.55. The relationship between C.p a m m removals b y filtration and total particle reductions was considerably weaker, with a coefficient of detemination (R') of only 0.48. Further discussion of these relationships is provided in Chapter 7. To elucidate the mechanisms associated with C. parvurn passage through filters, the Mach 10, 2000 sub-optimal coagulation experiment included the seeding of blue microspheres approximately 5.5 hourç before sub-optimal coagulation conditions affected filter performance. The filter influent concentrations of these microspheres would ideally be near O spheresk during sample collection, so that effluent concentrations could be attributed to detachment; however, these desired concentrations were not achieved due to non-plug-flow conditions in the standing water above the fdter media (Table C.7). 6.2.43 Statistical Analysis Confidence intervals for the individual C.p a m m removais duMg coagulant effects experiments were calculated using the meîhod described in Chapter 4. The endpoints of the 95% confidence intervals for oocyst removal are sumnarized in Table 6.12. The importance of coagulation for filtration is clear from these data. Cornparison of the 95% confidence intervals for stable operation (Table 6.3) and the no coagulant (short and extended duration) experiments indicated that the lower C.p a m m removals observed during the no coagulant experiments were sigificantly different from those observed during stable operation (Table 6.12, cr=0.05). Relative to stable operation, the no coagulant in plant, no coagulant in jar, and no silicate experiments failed to demonstrated significant differences in C.p a m m removal (Table 6.3 and Table 6.12, a=0.05).Sub-optimal coagulation yielded a large range of C.p a m m removais that were likely related to the range of filter effluent turbidities during these experiments. Data from the no coagulant-extended duration, no coagulant in the jar, no coagulant in the plant, and no silicate experiments could be pooled (whereas the other conditions could not be pooled due to continuing changes in settled water turbidities). The resulting overall 95% conf~denceintervais were 0.25- to 0.28-log, 5.1- to 6.0-log, 4.7- to 5.2-log, and 4.6- to 5.2-log respectively. The overall removals the no coagulant in the jar experiment failed to demonstrate significant ciifferences in C. pamrrn removd relative to stab'le operation, whereas the no coagulant in the plant, no silicate, and no coagulantextended duration experiments demonstrated C. pawtrm rernovals significantly different fkom those obtained during stable operation. The overaii 95% confidence intervals for microsphere removals during the no coagulantextended duration and no silicate experiments were 3.7- to 3.8-log and 0.22- to 0.25-log respective1y. The microsphere removals during the no coagulant-extended duration during ) experiment were not significantly different from C. p a m m removals ( ~ ~ 0 . 0 5 that experiment, whereas microsphere removals during the no silicate experiment were significantly different from the C. p a m m removals (a=O.OS). The overall role of microspheres as surrogates for C.psrvzrm rernoval by filtration is addressed in Chapter 7. Table 6.12 95% Confidence Intervals and C.parvum Removal Ranges During the No Coagulant and Sub-Optimal Coagulation Experirnents at Ottawa No CoagulantExtended Duration 12/13/99 No CoagulantShort Duration No Coagulant in Plant No Coagulant in Jar 08/ 18/99 No Silicate 12/17/99 Sub-Optimai Coagulation 6.2.4.4 Discussion As indicated by Figure 6.22 and Figure 6.23, the C. p a m m and B. subtilis remvals during the no coagulant in jar experiment were both in the ranges observed d u ~ stable g operation. This result may suggest that the presence of a l a m hydrolysis species and destabilized particles in the filter played a greater role in pathogen rernoval than coagulation of the pathogens themselves. Consistent with this hypothesis, the absence of coagulant only in the pilot plant only had a somewhat more noticeable, though still modest effect on C. p a m m and B. szrbtilis removals by filtration. The same trend was not observed with total particle reductions through the treatment process. Relative to stable operation, particle reductions did not change considerably during the no coagulant in the jar experiment; however, they decreased dramatically during the no coagulant in the plant experiments (Figure 6.24). The absence of activated silica during both pilot plant and jar coagulation appeared to have no effect on the removal of C. parvuni and B. subtilis by filtration and total particle (22 p ) reductions through the plant. This result demonstrated that the silica did not play an important role in pathogen removal during filtration. Rather, given other performance data, the role of silica was likely as a settliq aid that helped to reduce settled water turbidities and provide for longer filter cycles. During this experiment, total particle reduction through the treatment process was similar to that observed during the stable operation experiments (Figure 6.24). The sub-optimal coagulation experiments (40-60% decrease in coagulant concentration) demonstrated varied effects on C.parvum removals, which ranged from 1.3 to 5.5-log. C. p a m m removals during sub-optimal coagulation conditions were therefore anywhere from comparable to >3-log lower than those achieved during stable operation. The C.p a m m removals during sub-optimal coagulation when filter effluent turbidities were below 0.3 NTU were within 1-log of the removals achieved during stable operation, whereas the poorer removal were associated with considerably higher filter effluent turbidities (- 0.8 - 0.9 NTLJ). The C.parvum removal capacity of the filter decreased perhaps as non-optimally coagulated particles deposited within the filter, subsequently resulting in less favorable attachrnent (removal) conditions and poorer effluent quality. The overall and long term effects of poor coagulation were revealed in the no coagulantshort duration and no coagulation-extended duration experiments, which yielded substantially lower removals of C.parvum relative to stable operation. Other studies have demonstrated that the relative effects of poor coagulation can be source-water andor coagulation regime specific, although the general trends are sùniIar (Huck ef al., 2001). Therefore, care should be taken when extrapolating such results to other source waters or coagulation regimes. The results of the various coagulation investigations cIearly indicated that substanhal deterioration of C.p a m m removals by filtration could occur if chernical pretreatment kvas not adequately optimized and maintained. Moreover, substantial deterioration in C.parvum removals even occurred at filter effluent turbidities below 0.3 NTU, when filters were in cornpliance with the requirements of the ESWTR. The no coagulant in the plant, no coagulant in the jar, and long- and short-term coagulation failure experiments also suggested that filter cooditioning (at leaçt at utilities such as Ottawa where a relatively high alun dose is used) might play a role in maintaining C. parwrrn removal during short-term coagulation failures. 6.2-5 Hydraulic Step Hydraulic conditions can significantly impact the quality of filter effluents. It is generally recognized that filter performance is adversely affected by non-steady flow (Tmssell et al., 1980). Severai studies have concluded that different filtration rates do not necessarily adversely impact protozoan removal by filters (Al-Ani et al., 1986; Hom et al., 1988). Fitzpatrick et al. (1999) demonstrated that large and sudden changes in flow drarnaticaily deteriorated particle removal by fllters wide smaller changes that were implemented gradually did not always increase particle counts; it is possible that simdar relationships could exist between C.pawurn removai by filters and changes in flow. 6.2.5.1 Experimental Design Experiments evaluating the effect of hydraulic changes were performed to assess the effect of fdtration rate changes on the removal of C.parvurn and B. subtilis. Hydraulic steps were imposed during stable (optimized) operating conditions. Each of the hydraulic step experiments consisted of a 25% increase in filtration rate over less than one minute and was achieved by opening the fdter effluent valves. The higher rate was rnaintained throughout the remainder of the filter cycle. The experiments were designed to represent a scenario that results in increased flow to the filters, such as when one filter is put out of service resulting in increased flow rates through the other filters that remain in operation. The experimental design tested the hypothesis that microorganisms accumulate within the filter during stable operation and then are released following a sudden increase in hydraulic loading. During the hydraulic step experiments, the microorganisms were seeded in to the filter influent for five hours, which was an extended period of tirne relative to the other experimental conditions investigated. The seeding occurred during stable filtration conditions with the presurnption that microorganisms would accumulate in the filter during this period. The hydraulic step was imposed immediately after the seeding period. The goal of the hydraulic step experiments was to yield information regarding the detachment of microorganisms; one way of achieting this was by seeding the filter with microorganisms and then initiating the hydraulic step when filter influent rnicroorganism concentrations were low (ideally near O microorganisms/l). Samples were collected prior to, during, and after the hydraulic step. The flow increase occurred at a tirne of 300 minutes (at the end of the five-hour seeding period). Samples were collected at 280 and 295 minutes (prior to the increase in hydraulic loading) to confïrm that the filter was removing microorganisms at levels comparable to those achieved during the stable filter operation experiments. A 5-mhute composite sample was collected at 300 minutes to collect what passed through the filter as the hydraulic step occurred. Samples were also collected at 5 minute intervals after the hydraulic step was imposed (305, 3 10, 3 15 minutes) to assess any subsequent effects on water quality. Additional samples were collected during the second and third experiments (320 and 360 minutes) to detemine if microorganism removals retumed to baseline levels (ie.,those achieved during stable filter operation) after the change in flow rate. The hydraulic step experiments and seeding conditions are summarized in Table 6.13. Table 6.13 Summary of Hydraulic Step Experiments at the Ottawa Pilot Plant Experïment Hydraulic Step Date C.p a m m B. sztbtiZis 06/07/99 06/ 15/99 06/2 1/99 J J J J Blue Spheres Yellow Spheres 4 J Results 6.2-52 The hydraulic step experiments were conducted in Ottawa on June 7, 15, and 22, 1999. Even though the same protocol was foilowed during each experiment, the resulting impact on water quality differed between each of the experiments. Filter effluent turbidity and particle counts, C.pawurn and B. subtilis removals by filtration, and total particle reductions through the treatment process during the 25% iacrease in flow hydradic step experiments are sumrnarized in Table 6.14. The general fiiter performance data and detailed instantaneous tubidity and particle data are available in Appendix D (Tables D.4 and D.5 respectively). The detailed instantaneous C.parvum and B. subtiZis data are also available in Appendix D (Table D.6). Table 6.14 Filter Performance Duing Hydraulic Steps at Ottawa Date C. p a m m + + + LogIoRemoval Loglo~ e d . ' Effluent Concentration (mean istandard deviation) B. subtilis Blue YeIIow Particies Particles Turbidity Spheres Spheres 2 2pm (#/d) (NTZT) 0.7 t 0.5 1 -2.0+1.01 1562117 0.24+0.11 3.4 + 0.32 -- 2.4 0.95 99 t 140 0.05 0.02 ---1.8 + 0.69 1.7 0.87 2.2 & 0.93 125 11 126 0.14 2 0.12 06/07/99 0.2 1.28 + 06/15/99 4.0 4 0.58 06/21/99 2.7 0.78 Overall 2.5 1.74 * L Oreduction ~ of particles through treatment process (plant influent to filter effluent). + - - + Both füter effluent particle counts and turbidity considerably increased for a period of approximately 30 minutes (Figure 6.28) following the fIow increase on June 7. The filter effluent turbidity during stable filter operation portion pnor to the hydraulic step was -0.06 NTU and the filter effluent particle concentration (22 pn) was -0.5 particles/mL. As a result of the hydraulic step, the filter effluent turbidity and particle concentration temporarily increased up to 0.37 NTU and 297 particles/ml respectively. During the course of that filter cycle, neither the fdter effluent turbidity nor effluent particle concentration returned to the baseline levels that had been achieved prior to the hydraulic step (Figure 6.28). GeneraUy consistent with the stable operation experiments, filter effluent concentrations of C.parvum and B. stibtiIis were refatively low (respectively c l 0 oocystsL and <IO0 cfun) during the stable operation portion of the filter cycle prior to the hydraulic step. The increase in filter effluent turbidity and particle concentration (Figure 6.28) was commensurate with an increase in filter effluent microorganisms, with effiuent concentrations of C. parvurn reaching 4412 oocysts/l and B. subtilis reaching (Figure 6.29). Detailed turbidity, particle, and microorganism concentration 2000 and rernoval data are available in Appendix D. - 450 Seeding Pefiod (--- 400 -- F 350 O o -6.5 m h ( - 2 . 7 g p d ) o o 0 , ,-8.1 Particles >= 2pm Sarnpling Times 4 - Turbidity O 0 mBi (-3.3 g p d ) > L a 4:00 PM 5:00 PM Time Figure 6.28 Turbidity and particle response of filter during hydraulic step experbnent on June 7, 1999 at Ottawa pilot plant. - 350 Seeding Period aa 300 -- 350 -- El a ParUcles >= 2 p m 1 A 0 CNr'L A M;~.UmurnEffluent Ooqst Concennation Observed (n = 33) During Stable (Optimized) Opention 3 300 -- --E \ 1 5 250 -d & N II ^, 700 -- -.e: '3 0 C c .*.-------*-...*-...-.*.--*..**.-...* 150 -100 -50 H-ic -- Stcp ù O-L 3 ' O; - Figure 6.29 Particle and microorganism response of filter during hydraulic step experiment on June 7, 1999 at Ottawa pilot plant. 450 Seeding Penod * + 400 - 350 3 -g e * * * Sampling Times 0.3 -- 500 - - -- O 3 B -. 2.50 -- 5 -O A N Il -.-^,a zoo -1 -- 0 1 U r g 150 -LOO C4 -- -- 2 i 0.1 so -y4.-4-O 2 0 0 PM Figure 6.30 3:00 PM Time 0.0 .):O0 PM Turbidity and particle response of filter during hydraulic step experiment on June 15, 1999 at Ottawa pilot plant. The hydraulic step had a very different impact on filter effluent water quality on June 15. The filter effluent turbidity only slightly increased whereas the particle concentration peaked considerably (Figure 6.30). The filter effluent turbidity during stable fdter operation pnor to the hydrauiic step was approximately 0.05 NTU and the filter effluent particle concentration (2 2 pn) was approximately 1.0 paaicle/mL. As a result of the hydnuiic step the filter effluent turbidity and particle concentration temporarily increased to 0.09 NTU and 381 particles/ml respectively. Unlike the June 7 experiment, both the filter effluent turbidity and particle concentration returned to the baseline levels that were being achieved prior to the hydraulic step (Figure 6.30). Despite the considerable increase in Nter effluent particle concentration (over a period of approximately 30 minutes), no appreciable changes in filter effluent C. p a m m and B. subtilis concentrations relative to the stable operation period of the füter cycle were observed (Figure 6.3 1). While relatively low concentrations of B. subrilis spores were found in the fdter effluent during the stable and hydraulic step portions of the experiment (< 500 c m ) , alrnost no C. pawum oocysts were detected. Figure 6.31 Particle and rnicroorganism response of filter during hydraulic step experiment on June 15, 1999 at Ottawa pilot plant. A third hydraulic step experiment was performed on June 22. Filter effluent turbidity and particle data were not available during this experiment due to difficulties with the data acquisition system. These problems made it impossible to exactly pinpoint when the hydraulic step (and associated sampling) occurred. Filter effluent turbidity and particle data collected pnor to the experiment indicated that the fdter was likely operating at less than optimal conditions just prior to the hydraulic step; the effluent particle counts were slightly above what was typically observed at the Ottawa pilot plant during stable filter operation. A slight increase in fdter effluent oocyst concentrations was observed as a result of the hydraulic step, however, the filter effluent spore concentrations did not demonstrate this trend. Although the filter effluent oocyst concentrations were slightly elevated as a result of the hydraulic step, the increase was not as dramatic as that which occurred during the June 7 experiment. Confidence intervals for the removal of C. parvum during the hydraulic step experiments were not calculated. This was because the codidence intervals could be somewhat misleading because the filter influent concentrations of the microorganisms were decreasing as sampling occurred. C. p a m m removals based on filter influent and effluent pairs of data would not account for any oocysts already accumulated on the filters during the seeding period. 6.2.5.3 Discussion One hydraulic step experirnent at Ottawa (June 15) did not appear to yield substantial increases in filter effluent microorganism concentrations (C. parvum and B. subtilis) between the stable operation and hydraulic step portions of the experiments. These data suggest that little detachment of microorganisms occurred as a result of 25% increase in flow hydraulic steps. Considerable changes in filter effluent C.parvum concentrations occurred during the other two of the hydraulic step experiments at Ottawa (June 7 and 22). The presence of oocysts in the filter effluent samples during these experiments was likely due to at least a moderate amount of detachment fiom the filters; this was particularly evident in the Iune 7 data where the filter effluent oocysts were higher than the influent concentrations. Further mechanistic investigations wodd be necessary to appropriately address this speculation however. Overail, the traditional performance data (filter effluent turbidity and particle concentration) proved to be good indicators of treatment performance (Figure 6.28 to Figure 6.3 1). The June 7 and 15 hydradic step experiments underscored the conclusion that filter effluent turbidity and particle counts were not directly indicative of microorganism passage through filters. A hîgher peak particle concentration occured during the .Tune 15 (Figure 6-30) experiment than during the June 7 experiment (Figure 6.28). In contrat, the filter effluent microorganism concentration was dramatically higher on June 7 (Figure 6.29) than on June 15 (Figure 6.31). Moreover, the filter effluent particle spike on June 15 was not accompanied by an increase in effluent oocyst concentration, emphasizulg that increases in performance measures such as particle counts are not necessariiy directly indicative of microorganism passage through fdters. Filter effluent particle concentration, turbidity, oocyst and spore concentrations increased for a short period of time during two hydrauLic step experiments at Ottawa. Even though the increase in effluent microorganism concentrations was temporary, the number of oocysts that passed through the filter was substantial (given that a hi& number of oocysts were seeded into the filter during the experiment). This result was consistent with the findings of Cleasby et al. (1963), Tuepker and Buescher (1968), and Fitzpaaick et al. (1999) who showed that sudden, large flow rate changes cause deterioration of filtered water quality. These fmdings were also consistent with those of Logsdon et al. (198 1) who sinuiarly demonstrated that increases in Giardia passage through filters could be expected as a result of hydraulic changes. The June 15 hydraulic step experiment demonstrated that a 25% increase in flow could leave filter effluent C.p a m m and B. szrbtilis concentrations and turbidity essentially unaffected while moderate spikes in filter effluent particle concentrations (2 2 p) occurred. The difference between these results and those from other hydraulic step experiments that resulted in considerable passage of microorganisms might be explained in part by the work of Cleasby et al. (1963), which demonstrated that particle passage through filters following a disturbance was dependent on the composition of the fiiter influent. These data suggest that the baiance between attachrnent and detachment forces may be variable at a given location and result in performance from no risk to a high risk of microorganism release fiom a filter as a result of hydraulic changes. These results further indicate that it may be possible to optimize or at least identiEy the factors that affect these forces so that the potentiaily severe effects of hydrauiic changes (such as those observed during the June 7 experiment) can be minimized. It is dBïcult to speculate on a cause for the ciifferences in filter effluent water quality between the hydraulic step experiments. The hydraulic step experiments were performed w i t h a week of each other (essentially no variation in water temperature), at approximately 25 to 30 h o u s into the fdter cycle, and with comparable raw water and filter effluent quality and coagulation conditions prior to imposing the hydraulic step. Differences in oocyst and spore removds relative to stable operation were at least partly afTected by the lower (relative to stable operation) filter influent oocyst concenaations that resulted fiom the extended seeding penod during the hydraulic step experiments. It was expected that the microorganism rernovals during the stable operation portion of the hydraulic step experiments might be lower than those observed during the stable operation experiments. This was because of already Iow effluent microorganism concentrations (often near detection limits at Ottawa) during the stable operation experiments and lower influent concentrations (compared to stable operation) associated with the longer seeding period during the hydraulic step experiments. Mthough the filter effluent C. parvurn concentrations were often lower during the hydraulic step experiments than during the stable operation, the decrease in filter effluent oocyst concentrations was not directly proportional to the decrease in influent concentrations. These data emphasized the need to m e r study the relationship between filtration studies using hi& seeded microorganism concentrations as opposed to indigenous microorganism concentrations. A limited number of pilot-scale C. pawurn, B. subtilis, and polystyrene microsphere removal investigations were conducted at the University of Waterloo 0 Pilot Plant in Waterloo, Ontario, Canada. The UW pilot plant was operated in direct filtration mode with in-line fiocculation (contact filtration). It treated synthetic raw water comprised of dechlorinated tap water with kaohnite-induced turbidity (-1 -5 NTU at filter influent). The coagulation regime consisted of a relatively low alum dosage (-5 m f l ) for particle removal. The raw water was coagulated in-line and then filtered by both dual- and trimedia filters. The rationale and specific seeding conditions for each of the experiments were summarized in Chapter 3 and are elaborated upon below. 6.3.1 Stable (Optimized) Operation Stable operation experiments were performed in duplicate to determine the C.pamrm, B. subtik, and po lystyrene microsphere removals achieved by the p ilot-scale filters under optimal operating conditions at W. The stable operation experiments were conducted in duplicate and represented optimized pretreatment and filtration conditions (i-e., no perturbations or upsets). As at Ottawa, the seeding and sampling were conducted after at least four hours of filter operation, during the early to middle portion of the filter cycle, 6.3.1.1 Experimental Design Four stable operation experiments were conducted at UW (two investigated dual-media and two investigated tri-media). The use of both media types allowed for direct cornparison and investigation of potential improvements in pathogen removal that might be associated with tri-media filtration. Jar-coagulated C. pawurn, B. strbtilis, and yeliow, oocyst-sized polystyrene microspheres were seeded into the filter influent for one hour during these experiments. Samples were collected at 20, 40, and 55 minutes after the start of seeding. Filter effluent turbidities of cO.1 NTU were targeted during these experiments. The stable operation experiments and seeding conditions are sumrnarized in Table 6.15. Summary of Stable Operation Experiments at the UW Pilot Plant Experiment Date C.p a m m B. subtilis Stable Operation Duai-Media 11/23/99 1 1/24/99 J J J J Stable Operation Tri-Media 11/28/99 11/28/99 J J 4 J Blue Spheres Yellow Spheres J 4 The filter effluent turbidity and particle counts were consistently low (< 0.1 NTU and < -25 particles/mL, respectively) during the stable operation experiments at the UW pilot plant. Thou& excelient filter effluent quality water was produced during the stable operation experiments at UW, fiiter effluent particle concentrations and turbidities were somewhat less consistent between the duplicate experiments in each fdter type than between the replicate experiments at Ottawa. Filter effluent kxbidity and particle counts and C. parvurn, B. subtiZÏs, and polystyrene microsphere removals by filtration during stable operation at UW are summarized in Table 6.16. The stable operation perfurmance of the UW filters was generally consistent wÏth that observed at Ottawa; therefore, filter effluent turbidity, particle concentrations, and rnicroorganism concentration data are not exhaustively discussed- The generai filter performance data and detailed instantaneous turbidity and particle data are available in AppendUr D (Tables D.10 and D.11 respectively). The detailed instantaneous C.puwurn, B. strbtilis, and microsphere data are also available in Appendix D (Table D. 12). As summarized in Table 6.16, excellent C. pawurn removals were achieved by the dual- and tri-media filters during stable operation. The >5-iog removals were generally consistent between the replicate experiments for a given media type. C. p a m m removals by the dual-media filter during stable (optimized) operation ranged from 4.7 to 5.4-log, with a mean oocyst removal of 5.0-log (6 samples in total). C.parvlrm removals by tri-media filtration were somewhat higher, ranging fi-om 4.9 to 5.7-log, with a mean oocyst removal of 5.3-log (6 samples in total). Given these data, tri-media fdters did sot appear to offer appreciable advantages in C. pumrrn removal relative to dual-media filters. Since these experiments were only performed in duplicate, M e r experiments would better elucidate this relationship. Though more variable than C.pawum passage through fdtration between the replicate tri-media experiments, B. subtilis spore passage through the UW filters appeared to be consistent with the C. pawurn data B. subtilis removal by dual-media filtration during stable operation ranged fiom 2.9 to 3.3-log, with a mean removal of 3.1-log (6 samples in total). The tri-media filter removed 3.3 to 4.8-log of B. stibtilis, with a mean spore removal of 3.9-log (6 samples in total). These data were generally consistent with the C.p a m removal data that indicated slightly better oocyst removals by the tri-media filter. The confidence intervals for these data are in Appendix D (Table D. 15). Table 6.16 Filter Performance During Stable Operation at UW LogloRemoval Date C. parvztrn B. subtih + 11/23/99 5.0 2 0.25 3.2 0.09 11/24/99 5 . 1 2 0 2 6 3.120.13 Dual-Med. OveraIl 5.0 2 0.23 3.1 _+ 0.13 Loglo~ e d . ' (mean & standard deviation) Blue Yellow Particles Spheres Spheres 1 2pm 4.8 2 0.39 - 4.4 + 0.3 1 - - 4.6 2 0.35 - Effluent Concentration Particles (#/rd) Turbidity . 0 19.2 + 4.2 0.07 + 0.00 2.7 2 0.4 0.06 -t 0.03 10.9 & 9.4 0.06 $0.02 4.6 & O. 16 6.9 2 0.9 0.05 + 0.00 11/28/99 5.1 2 0.12 3.4 2 0.09 6.0 2 0.20 1-620.8 0.03 +0.00 11/28/99 5.6 0.03 4.4 0.30 Tri-Media Overaii 5.3 2 0.31 3.9 _+ 0.57 5.3 _t 0.78 4.2 _+ 3.0 0.04 _+ 0.01 * L O reduction ~ of particles through treatment process (plant influent to filter effluent). + - - Tri-media fdtration also appeared to provide siightly better removals of polystyrene microspheres compared to those achieved by dual-media filtration. Microsphere removais by the dual-media filter during stable operation ranged fkom 4.1 to 5.1-log, with a mean microsphere removal of 4.6-log (6 samples in total). Microsphere removals by tri-media filtration were somewhat higher, ranging fkom 4.4 to 6.1-log, with a mean microsphere removal of 5.3-log (6 samples in totd). The confidence intervals for these data are in Appendix D (Table D. 15). B. subtilis and polystyrene microsphere removals by both dual- and tri-media filtration were generaily lower than those of C. parvzrrn. A cornparison of B. subtiZis and microsphere removals as potential surrogates for C.parvum removals by filtration is provided in Figure 6.32. Total particle reductions through the treatment process could not be examined because raw water particle concentrations were not measured due to equipment limitations (oniy one particle counter was available). These data demonstrated that both B. sztbtilis spore and polystyrene microsphere removals were somewhat indicative of C. pawurn removals by dual- and tri-media filtration at UW. Least-squares iinear regression of the B. sztbtilis and microsphere removal data yielded coefficients of determination (R') of 0.75 and 0.80 respectively. The relationship between C.parvurn removals by filtration during stable operation and B. subtiZis and microsphere removals at UW yielded somewhat different conclusions than the Ottawa data. It is possible that differences in filtration regime (contact filtration as compared to conventional treatment) or coagulation regime might affect how well removals of potential surrogates correlate with C. parvum removals. As mentioned previously, potential surrogates for C.panatm removal b y filtration wiil be discussed in greater detail in Chapter 7. 0.0 Figure 6.32 6.3.1.3 1.O 2-0 3 .O 4.0 C. parvum Removal (log,,,) 5.0 6.0 Relationship between C.parvum, B. subrilis, and microsphere removals by the pilot-scaie dual- and tri-media filten during stable operation at UW. Statistical Analysis Siinîmarized in Table 6.17, the endpoints of the 95% confidence intervals for the individual C. parvum log removds by dual- and tri-media füters during stable operation in were calculated using the method described in Chapter 4. All of these confidence intervals overlap, failing to demonstrate statistically significant differences between the data collected durùig stable operation in a filter of a given media type or between media types (Table 6.17, cx=0.05). This result suggests that, for the raw water investigated, trimedia filters did not offer appreciable advantages in C-parvurn removal during stable operating conditions. The endpoints of the range indicate the lowest and highest C.p a m m removals observed during the operating period and incorporate analytical uncertainty. The C.poniurn rernoval ranges for the dual- and tri-media füters during stable operation were 4.2- to 6.7-Log and 4.4- to 7.0-log respectively. The range for the individual experiments is summarized in Table 6.17. Table 6.17 95% Confidence Intervals and C.p a m m Removal Ranges During Stable Operation at UW - Dual-Media 1 1/23/99 Stable - Dual-Media 1 1 1'24199 Stable Stable - Tri-Media Stable - (logd 4.5 m g d 5.6 (logd 4.2 (Ioglo) 20 40 55 43 5 -3 6.7 4.3 6.7 1.6 4.5 5.8 20 1.6 5.9 4.4 5.9 40 55 4.4 4.6 5.5 5.9 20 4.9 7.0 4.9 4.9 7.0 6-9 4.9 40 55 1 1/38/99 Tri-Media 6.3.1.4 The 20 1 1/2 8/99 6.7 7.0 Discussion Sirnilar to the more numerous experiments at Ottawa, the stable filter operation experiments at UW demonstrated that C. parvum removals of >5-log could be achieved by both dual- and tri-media filtration. Commensurate with the bench-scale findings from Chapter 5, tri-media fdtration provided marginal, but not statistically significant advantages in C-parvurn removal relative to dual-media fdtration. These results are in generai agreement with C.parvurn removal data that have been reported in the literature (Patania et al., 1995; Fox et al., 1998). They provided the baselîne removals to which the other operating conditions (hydraulic step) at UW were compared. 6.3.2 Hydraulic Step As described in Chapter 2, hydraulic conditions can significantly impact the quality of fdter effluents (Trussell et al., 1980). The pilot-scale experiments conducted at Ottawa and described in Section 6.2.5.2 investigated the effects of hydraulic steps on C. pawurn and B. subtilis removal by dual-media filtration. They failed to conclusively demonstrate whether or not considerable deterioration in either particle or pathogen removal could be expected as a result of a sudden, 25% increase in flow to the pilot-scale filter. As a result, M e r hydraulic step experiments were conducted at the U W pilot plant. 6.3.2.1 Experimental Design Experiments evaluating the effect of hydraulic changes were performed to assess the effect of filtration rate changes on the removal of C. pamrm, B. subtilis, and polystyrene microspheres. These experiments were conducted with both dual- and tri-media füters to investigate whether tri-media filter designs could potentially mitigate pathogen passage through filters operating during hydraulic step conditions. As during the Ottawa experiments, hydraulic steps were imposed during stable (optimized) operating conditions. Each of the hydraulic step experiments consisted of a 25% increase in the filtration rate over Iess than one minute, which was achieved by opening the filter effluent valves. The higher rate was rnaintained throughout the remainder of the filter cycle. During the hydraulic step experiments, the C. p a m m oocysts and yeliow microspheres were seeded in the filter influent for one hour prior to the initiation of the hydraulic cep. The seeding occurred during stable filtration conditions with the presumption that microorganisms would accumulate in the filter during this period. The hydraulic step was imposed irnrnediately after the seeding penod. B. subtilis spores were seeded during the one hour penod that began fifteen minutes p i o r to the initiation of the hydraulic step. The goal of this experimental design was to yield information regarding the detachment and non-attachment (or weak attachment and subsequent detachment) of microorganisrns . Sarnples were collected pnor to, during, and after the hydraulic step. The flow uicrease occurred at a time labeled O minutes (when seeding of C. parvum and microspheres ended and fifteen minutes into the seeding of B. subtilis). Sarnples were collected at 15 minutes (prior to the change in hydraulic loading) to confirm that the filter was removing microorganisms at levels comparable to those achieved during the stable filter operation experiments. A sample was collected at O minutes to collect what passed through the filters as the h y h u l i c step occurred. Sarnples were also collected at 10-minute intervals after the hydrauiïc step (10 and 30 minutes) to determine any subsequent effects on water qudity. The dual- and tri-media hydraulic step experiments were each conducted in duplicate. The hydraulic step experiments and seeding conditions are srrmmarized in Table 6.18. 6.3-2.2 Results The hydraulic step experiments were conducted at UW on December 1 0 and 1 1, 1999, and January 15 and 16, 2000. FiIter effluent turbidity and particle counts and C. pumrrn and B. subtilis removals during the 25% increase in flow hydraulic step experhents are surnmarxied in Table 6.19. The summarized data include al1 of the time points (-15, 0, 10, and 20 minutes) except for t = -60 minutes, because these sarnples were only collected during the dual-media experiments; fheir inclusion would make it difficult to compare the dud- and tri-media filtration data. The general fdter performance data and detailed instantaneous turbidity and particle data are available in Appenduc D (Tables D. 10 and D. 11 respectively). The detailed instantaneous C-parvzrm, B. subtilis, and microsphere data are also available in Appendix D (Table D.12). Table 6.18 Summary of Hydraulic Step Experirnents at the UW Pilot Plant Date C.pawum B. subtilis Hydraulic Step Dual-Media 12/10/99 12/11/99 J J J J Hydraulic Step Tri-Media 01/15/00 OU1 6/00 J J J J Experiment Blue Spheres Yellow Spheres J J Table 6-19 Filter Performance During Hydraulic Steps at tTW Date C. p a m m 12110/99 4.6 2 0.69 13/11/99 4.3 2 0.82 DuaMMed. 4.4 2 0.71 Overall LogloRernoval Loglo ~ e d . ' Effluent Concentration (mean + standard deviation) B. subtilis Blue Yellow Particies Particles Turbidity Spheres Spheres 12p11 (ff1m.L) (NTU) 4.4 0.62 1 6 i 1 0 0.05i0.00 1.8 2 0.35 3.1 0.79 25 11 0.06 i:0.00 2.5 + 0.41 - 3.8 0.96 21411 0.05&0.01 2.1 _+ 0.51 01/15/00 4.220.57 1.9+0.33 01/16/00 4-4 5 0.73 2.3 2 0.23 Tri-Media 4.3 2 0.62 2.1 _+ 0.3 1 Overal1 - - - + + + - 4.3 2 0.53 4-6 5 0.53 4.5 _+ 0.52 - + 47 + 30 42528 44 f37 0.06 2 0.0 1 0.06+0.01 0.06 4 0.01 - ~ o reduction g of particles through treatment process @lant influent to filter effluent). The hydraulic step experiments at UW yielded results similar to those observed for the June 15, 1999 hydraulic step experiment at Ottawa (Figure 6.30 and Figure 6.3 1). The filter effluent turbidity did not appreciably increase whereas the particle concentration peaked slightly. The fdter effluent turbidity during stable flter operation pnor to, during, and d e r the hydraulic step was approximately 0.05 NTU. The dual- and tri-media filter effluent particle concentrations (22 p)during stable operation @rior to the hydraulic step) were approximately 10-20 particles/ml. As a result of the hydraulic step the filter effluent particle concentrations temporarily increased to 40-75 particles/mL, with slightly higher peaks observed in the tri-media filter effluent. The filter effluent particle concentrations returned to the baseline levels achieved prior to the hydraulic step. Small, but appreciable changes in filter effluent C. parvum concentrations relative to the stable operation period of the filter cycle were observed concurrent to the moderate increase in filter effluent particles. Filter effluent B. subtilis spore concentrations were aiso elevated, relative to those observed during stable operation, as a result of the hydraulic steps; however, it should be recalled that C. parvurn and B. subtilis were not concurrently seeded during these experiments at UW. Although C. p a m removal data can be somewhat misleading since filter innuent concentrations were decreasing during samphg, the relative removal of oocysts and microspheres could still be compared since they were concurrently seeded. The relative removal of C. p a m m oocysts and polystyrene microspheres during the hydraulic step experiments at UW is presented in Figure 6.33. This figure demonstrates some correlation between oocyst and oocyst-sized microsphere removals by filtration. Leastsquares linear regression of the microsphere removal data yielded coefficients of determination (R') of 0.65 and 0.91 for dual- and tri-media filtration respectively. Combination of aU of the microsphere data resulted in a slightly lower coefficient of determination &) of 0.52. These relationships will be m e r examined in Chapter 7. 1- - Tri-Media Dual-Media -- 0.0 Linear (Combined) Linear (Dual-Media) Linear (Tri-Media) 1.O 2.0 3 .O 4.0 5.O 6.0 C. pantum Remval (log,,) Figure 6.33 Relationship between C-p a m m and microsphere removals by the pilotscale dual- and tri-media filters during bydraulic steps at UW. 63.2.3 Discussion Zn generai, it is difficdt to make mechanistic conclusions trom the UW pilot data. The presence of oocysts in the Hter influent during the imposition of the hydraulic step, albeit at low concentrations, made it difficult to clearly determine whether or not the presence of oocysts in the fdter effluent resuited fi-om detachment or non-attachent. Although it was difficult to elucidate mechanistic information fkom the hydraulic step experiments, the frlter influent C.p a w m and microsphere concentrations ci-ere decreasing while filter effluent concentrations were increasing as a result of the imposed hydraulic steps. These data suggested the possibility of moderate amount of detachment from the fdters, however, they were not as convihcing as the June 7, 1999 Ottawa data which included filter effluent oocysts concentrations that were higher than influent concentrations. Conversely, the B. subtilis effluent concentrations were somewhat higher than during stable operation, suggesting some non-attachment or weak attachment and subsequent detachment of spores within the filter. The traditional performance data (filter effluent turbidity and particle concentration) proved to be good indicators of treatment efficiency. While no increase in filter effluent turbidities was .observe& a noticeable increase in total filter effluent particle counts signaled a moderate increase in fiiter effluent C. p a m m , B. subtilis and polystyrene microsphere concentrations. Although the filter effluent microorganism and microsphere concentrations were slightly higher than during stable operation, this deterioration in finished water quality was typically bnef. These experiments did demonstrate that a 25% increase in flow through dual- and trimedia filters would not necessarily result in dramatically increased filter effluent pathogen concentrations. As mentioned earlier, the UW results and their relationship with the variable results observed at Ottawa might be explained in part by the work of Cleasby et al. (1963) which demonstrated that particle passage through filters following a disturbance was dependent on the composition of the filter influent. These results M e r indicate that it may be possible to optimize or at least identiQ the factors that affect these forces so that the potentially severe effects of hydraulic can be minirnized. A limited number of pilot-scale C. p a m m and B. subtilfi removal investigations were conducted at the Windsor Pilot Plant in Windsor, Ontario, Canada, As described in Chapter 3, the Ottawa and Windsor pilot plants are Wtually identical in design and construction. Operated in a conventional mode with dual-media filtration, the Windsor pilot plant also employed a relatively hi& coagulant dose, but for a different raw water quality than Ottawa, At Windsor, the raw water TOC was considerably low and the alkalinity and turbidity (which could spike as hi& as 350 NTU during the spring) were higher than at Ottawa. The rationale and specific seeding conditions for each of the Windsor experiments were summzrized in Chapter 3 and are elaborated upon below. 6.4.1 Stable (Optimized) Operation Stable operation experiments were performed in triplkate to deterrnine the best C.pavvrrm and B. subtiIis removals that codd be achieved by the pilot-scale, dual-media fiiter at optimal operating conditions at Windsor. As at Ottawa and U W , the seeding and sampling were conducted during the early to middle portion of the Nter cycle, afier at least four hours of filter operation. Though the experiments were pianned without preozonation, pre-ozonation did occur during one experiment. 6.4.1.1 Experimental Design Three stable operation experiments were conducted at Windsor. Jar-coagulated C. p a m oocysts and B. subtilis spores were seeded into the filter influent for one h o u during these experiments. Samples were collected at 20, 40, and 55 minutes after the start of seedùig. Filter effluent turbidities of d . 1 NTU were targeted during these experiments. The stable operation experiments and seeding conditions are summarized in Table 6.20. Table 6.20 Siimmary of Stable Operation Experiments at the Windsor Pilot Plant Experiment Stable Operation Date C.p a m m B. subtilis 1O/14/99* 10/21/99 10/22/99 J J J J 4 ' J Blue Spheres Yeiiow Spheres -stable operation with pre-ozonation. 6.4-1-2 Results Filter effluent turbidity and particle counts were consistently low ( ~ 0 . 1NTU and <-20 particledrnL respectively) during the stable operation experiments at Windsor. Udortunately, on-line data Iogging of fdter effluent turbidities was not available d u ~ g these experiments. Filter effluent particle counts, C.parvurn and B. subtilis removals by fiItration, and total particle (1 2pm) reductions through the treatment process during stable operation at Windsor are summarized in Table 6.21. The general fdter petformance data and detailed instantaneous particle data are presented in Appendix D (Tables D.13 and D.14 respectively). The detaited instantaneous C-parvum and B. subtilis data are also available in Appendix D (Table D. 15). Table 6.2 1 Filter Performance During Stable Operation at Windsor LogIoRemoval Logio~ e d . ' Effluent Concentration (mean + standard deviation) C. parvum B. subtilis BIue Yellow Particles Particles Turbidity Spheres Spheres 1 2pm (#/rd) 0 10/14/99** 4.3 10.39 3.9 0.29 --1.6 0.10 13 2 2.4 -Date + + - - 4.2 2 0.28 3.7 _+ 0.36 1.6 0.06 14 _+ 1.6 'Log reduction of parricles through treatment pmcess @lant influent to filter effluent). ** Experiment performed with pre-ozonation, Overall - Traditional performance rneasures (turbidity and total particle counts) and fdter effluent microorganism and microsphere concentrations were relatively consistent during the stable operation investigations and between replicate experiments. Filter effluent particle counts ranged from 13-16 particles/mL during stable operation at Windsor. The filter effluent particle counts were generally consistent between the replicate experiments. C. parvum removals during stable (optimized) operation ranged fiom 3 -8 to 4.7409, with a mean oocyst removal of 4.2-log (9 samples in total). B. subtiks removals ranged fiom 3.3 to 4.4-log, with a mean removd of 3.7-log (9 samples in total). These data were summarized in Table 6.21. Although there was some variation in removals calculated on the basis of individuai influent-effluent sample pairs of microorganisrns, the calculated removals during a given experiment were fairly reproducible as indicated by the relatively low standard deviations (Table 6.21). The overall standard deviations were also relatively low, demonstrating good reproducibility between the replicate experiments. The confidence intervals for these data are in Appendix D (Table D. 16). B. subtilis removals by filtration were generally lower than C, pamcnz rernovals. A comparison of C. parvum and B. subtilis removals during stable operation did not indicate a clear relationship between oocyst and spore removals (Figure 6.34). The best fit line from least squares Linear regression yielded a coefficient of determination (R') of only 0.18. While spore removals were lower than oocysts removals, the lack of a relationship between oocyst and spore removals suggested that B. subtilis removal data were not indicative of the filter's ability to remove C.pauvurn. Similar results were observed with total particle (22 pm) reductions through the plant. A comparison of C.panturn rernovals and total particle reductions through the treatment process during stable operation did not indicate a clear relationship between oocyst removals and particle reductions (Figure 6.34). The best-fit line from least squares linear regression yielded a coefficient of determination (R') of only 0.2 1. As demonstrated and noted during previous experiments, these data suggested that total particle reductions through the treatment process are not adequate surrogates for C. pawurn removals during fdtration. Figure 6.34 6.4.13 Relationship between C.parvum and B. subtilis removals by fdtration and total particle (22 pn) reductions through the plant at Windsor. Statistical Analysis Confidence intervals for the individual C. parvurn removals during stable operation were calculated using the method described in Chapter 4, and therefore incorporated analytical recovery and uncertainty of recovery. The endpoints of the 95% confidence intervals for oocyst removai are sumrnarized in Table 6.22. Al1 of these confidence intervals overlap, thereby fading to demonstrate statistically significant differences between the data collected during the stable operation experiments (Table 6.22, a=0.05). The stable operation with pre-ozonation experiment did demonstrate significant differences in C parvurn removals between the first and third replicate samples (Table 6.22, a=0.05); more data would be necessary to detemine if this difference was due to experimental drift or some other phenornenon. A comparison of the confidence intervals for the stable operation and stable operation with pre-ozonation experiments generally faiIed to demonstrate statistically significant diffierences between C. p a m m removals during these periods (Table 6.22, a=0.05). The exception to this was the third sample (t = 55 minutes). Overail, the C.pawrlm removal range was 3.6- to 5.0-log during stable operation and 3.6- to 5.2-log during stable operation with pre-ozonation. The range for the individual experiments is summarized in Table 6.22. These results generally suggested that, for the operational conditions investigated, pre-ozonation did not substantiaUy affect C. p a m m removals by fiitration. As mentioned previously, more data would be necessary to better support ùiis conclusion. 6.4.2 Discussion The stable filter operation experiments demonstrated that C. p a m m removals of >4-log codd be consistently achieved by the du&-media pilot filter at Windsor. This result was important because it demonstrated that a piIot plant essentially identical in construction to the one at Ottawa provided -1-log lower C. pamrm removd at optimal operating conditions, when treating a different source water. The higher fdter effluent particle counts during stable (optimized) operation at Windsor also underscored the difference in overall performance and particle reduction by filtration. The Windsor data also demonstrated that pre-ozonation appeared to have little impact on C.p a m removal b y filtration. This experiment was only performed once and provided only three data points. No operational explanatïon was found to justiS why the C. parvum removal data were so variable during this operating condition (relative to other experiments at Windsor, Ottawa, and UW). A more thorough investigation with replicate experiments would be necessary to fully investigate the effects of pre-ozonation on C. parvum removal by filtration. Table 6.22 95% Confidence Intervals and C. p a m m Removal Ranges During Stable Operation at Windsor. Date Experiment Sample Time 20 30 55 CI IOW- a WF R -1, (1% IO) 3.6 3 -7 4.3 (loglo) 3 -9 (~OSIO) 3 -6 R w (lof50) 5-2 Stable Operation with Pre-Ozonation 10113/99 Stable Operation 10/21/99 20 40 55 4. 1 3.8 3.6 4.9 4.4 1-1 5 -6 4-9 Stable Operation 10/22/99 20 40 3.7 3.2 4-3 3 -7 4.3 6.4.3 3.8 ~ 4-1 5.2 Rate Effects As mentioned previously, Cleasby et al. (1963) and Tuepker and Buescher (1968) showed that large flow rate changes cause deterioration of filtered water quality by the detachment of previously retained particles. These relationships supported observations that d e c h h g rate filters may provide better performance than constant rate filters (Hudson, 1959; DiBernard0 and Cleasby, 1980); however, subsequent experiments by Hilmoe and Cleasby (1986) found no significant differences between declining rate and constant rate filters. The authors speculated that the previously reported poorer effluent quality achieved by constant rate filtration might have been caused by the constant rate control system used by DiBemard0 and Cleasby (1980), which might have inadvertently resuited in continuous flow rate fluctuations or swges. If declining rate filtration provides filter performance advantages over constant rate fdtration, it may also provide an operational strategy for mitigahg C. parvlrm passage through filtration. 6.4.3.1 Experimental Design Four rate effects experiments were conducted at Windsor (two investigating constant r filtration and two Uivestigating declining rate fdtration). These experiments involved side-by-side cornparisons of two dual-media filters, one operating in a constant rate mode and the other operating in a d e c k g rate mode. The experiments were performed hours into a -40-hour cycle). approxirnately halfway into the filter cycle (-20 Laboratory limitations (sample processing) precluded these experiments fiom being conducted with both oocysts and spores. Only jar-coagulated B. subtilis spores were seeded into the filter influent for one hour during these experiments. Samples were collected at 20,40, and 55 minutes after the start of seeding. Filter effluent turbidities of CO.1 NTU, essentiaily representing stable operation, were targeted during these experiments. The rate effects experirnents and seeding conditions are summarized in Table 6.23. Table 6.23 Summary of Rate Effects Experiments at the Windsor Pilot Plant Experiment Date C.parwrm B. subtilis Stable Operation Constant Rate 03/ 10/99 03/22/99 J J Stable Operation Declining; Rate 03/10/99 03/22/99 d 6.4.3.2 Blue Spheres Yeilow Spheres Results As during the stable operation experiments, on-line data logging of fdter effluent turbidities was not available during these experiments. Filter effluent particle counts, B. subtilis removals, and total particle reductions through the treatrnent process during the rate effects experiments at Windsor are summarized in Table 6.24. The general filter performance data and detailed instantaneous particle data are available in Appendix D (Tables D.13 and D.14 respectively). The detailed instantaneous B. subtilis data are also available in Appendix D (Table ID. 15). The March 10, 1999 data were disregarded because chernical pretreatment and fiter operation were not optimized d u ~ these g experiments. This result was underscored by the elevated fdter effluent particle concentrations (>2000 particles/mL) in both of the dual-media filters (operating at constant and declining rate modes). The fiiter effluent particle concentrations drtring stable operation at Windsor were typically less than -250 particles/ml during the tirne of year during which the rate effects experiments were performed. Stable operation conditions were achieved by the constant and declining rate fdters during the March 22, 1999 experiments. Table 6.24 Filter Performance During Rate Effects Experiments at Windsor Date LogloRemoval Log10 ~ e d . ' Effluent Concenmtion (mean + standard devîation) C. parvurn B. subtilis Blue YeUow Particies ParticIes Turbidity Spheres Spheres 22 p (fi/rnL) Stable Operation - Constant Rate Filtration 03110199~ 0-8 0.08 -O S 0.02 2 159 29 03/22/99 3.2 5 0.42 1.6 2 0.03 244 1+ 6 - - + - Stable Operation - Declining Rate Filtration + -- - -+ + 03110/99** 2.5 0.49 0.420-01 37422 107 03/22/99 3.1 0.54 1.8 4 0.04 136 _+ 5 - L O ~reduction of particles through treatment process (plant influent to filter effluent). Filter operation was not optimized during the 03/10/99experiments and data were not evaluated. + .. 6.4.3.3 - - --- - Discussion Though each experiment was only perfoxmed once, the B. subtilis spore removals by constant and declïning rate filmtion were essentially the same. This result demonstrated that declinhg rate filtration did not offer an obvious advantage over constant rate filtration for B. subtilis removal, and iïkely C. parvurn removal, at the operating conditions investigated. Further experimentation would be necessary to conclusively determine any potential differences in C. parvum rernoval that might result from operating filters in constant or declining rate modes. The observed spore removals were similar to the removals obtained during the stable operation experiments (that were performed in a constant rate filtration mode). The data were sLightIy more variable during the present experiments, however, with slightly Iower rnean removals. Given that the filter performance changed seasocally, as indicated by the filter effluent particle concentrations (Table 6.21 and Table 6-24), it was difficdt to fûrther compare the results between the stable operation and rate effects experiments. The main focus of this research was to ïnvestigate C.parvurn removal during various portions of the filter cycle and during operational periods when turbidity and particle removal processes are challenged, The majority of these investigations were conducted at the Ottawa pilot plant. C.pa?vurn removals during al1 of the operational conditions investigated at Ottawa (except hydraulic steps) are surnmarized in Figure 7.1. -n=32 n=4 P - --..".-- r 'q " n=8 n=14 n=4 n=8 7 n=16 n=l2 n=8 n=12 n=20 n = 4 "-- -- 5 High Level of Rernoval --4r,+ .i -- 4 a-<... Law Level of Removai - -- 6 2 -- I .-..---- -....---- Alrnost No RemovaI -i spMt N0cnq.m Opsum Jar y.b*[knitg llmR DipcaIrm HoWiaR Not3.g.m Plvu kibapinil cm& Ead-or-Rui -. p . . . . . - NoCoyE8dy Lue B a r D u m m ~ Brahlhwgh ~ h Nocw- O bE=kd Dwmion Operating Penod Figure 7.1 Box-and-whisker plot of C. panrum removals b y filtration during al1 operating penods (except hydraulic steps) investigated at Ottawa. The C.p a m m removal data in Figure 7.1 generally indicated three levels of removal: high, low, and a h o s t no removal. High removals were in the range of those observed during stable operation and were generally >4-log- Oocyst removals during the stable operation, no coaadant in the jar, ripening, no silicate, and no coagulant in the plant experiments were high. The end-of-run, no coagulant-short duration, early breakthrough, and late breakthrough conditions resulted in low removals of C. pawum by fï!tration. These low removals were typically -J-log, a marked decrease relative to stable operation during which C. p a m m rernovals were typically >5-log. Sub-optimal coagulation conditions spanned the high and low C. pawum removal range, while the no coagulantextended duration conditions essentially precluded C.parvum removal by filtration. The hydraulic step experiments were not included in Figure 7.1 because the experiments were conducted in a manner that resulted in slowly decreasing filter influent C. parvum concentrations during the implernentation of the hydraulic step conditions. From Figure 7.1, it can be concluded that the granular media filters best remove C.pamrm fiom waters that are weli-coagulated. The Ottawa pilot plant generally maintained good oocyst removals during brief penods of non-optimal coagulation; however, extended periods of poor coaguiation almost completely prevented oocyst removal. Relative to stable fdter operation, C. pavvurn removals were siightly lower during the ripening portion of the filter cycle. End-of-run and breakthrough operation resulted in dramatically lower oocyst removals relative to stable operation, suggesting that the later portion of the filter cycle can be particularly vulnerable in terms of maintainhg pathogen removal. Using the statistical framework developed in Chapter 4 to account for analytical recovery and uncertauity associated with recovery, the range of C.parvurn removals during each experiment was calculated. The replicate data fiom the each of the stable operation, no coagulant in the jar, stable operation during runoff, no silicate, no coagulant in the plant, and no coagulant-extended duration experiments were pooled to calculate overall 95% confidence intervals for each of these conditions. As described in Chapters 4 and 6, this was possible because no obvious changes were occurring in the independent operational variables (as,settled water turbidity, pretreatment conditions, filter loading rates, etc.) durhg these experiments. For the remaining experiments (sub-optimal coagulation, endof-nui operation, no coagulant-short duration, early breakthrough, and late breakthrough) the range of C.pawum removals during each experiment was calculated based on the lowest and highest endpoints of the highest posterior density (HPD) regions calculated from each of the individual probability density functions (pdfs) for C.parvzrrn removal (each pdf was based on one influent-effluent pair). This "adjusted" range of C. pawum removals at Ottawa is depicted in Figure 7.2, with the operating conditions listed in the same order as in Fi-gure 7.1. The C. p a m removal data in Figure 7.2 indicate the same general trends a s the data in Figure 7.1. In the case of the data collected during dynamic (changing) operating conditions which precluded the pooling of data, however, the ranges of removals are generally larger due to the uncertainty associated with the analytical method for concentrating and enumerating C. p a m m oocysts fiom water. While the unadjusted minimum and maximum removals f?om these experiments spanned a relatively s r n d range of approximately 1-log (Figure 7.1), the adjusted data (Figure 7.2) indicated a substantially larger range of removals that couid be expected (approximately 2-log) given the uncertainty associated with the malytical method used to concentrate and enumerate the C. parvum oocysts. It should be noted that the analytical method used during this thesis research and to generate the range of C.parvum removals in Figure 7.2 was consistent and reliable relative to others that have been reported in the literature (Appendix A). For similar filter Ilifluent and eKluent C. p a m m data collected using a less reliable method, these ranges would be larger. This result underscores the importance of assessing the reliability of available C. pamcm data in the context of the analytical methods used during data collection and enurneration, particularly when non-detects or very low oocyst counts are observed. Operatinp Penod Figure 7.2 95% Confidence intervals and adjusted ranges of C.p a m removals by filtration during al1 operating periods (except hydraulic steps) investigated at Ottawa. kl&ough the data in Figure 7.1 and Figure 7.2 present log removals calculated fiom the seeding experiments, these results should be considered in terms of the relative ciifferences in pathogen removals during the different operating conditions. As discussed in previous chapters of this thesis, the actuai log removals of C. p a m m that can be obtained by full-scale filtration processes are limited by the influent oocyst concentrations. Furthemore, C. p a r h m removals c m be source-water and treatment design specific, as demonstrated in Figure 7.3, which depicts C.parvum removals by dual-media filters during stable operation at the Ottawa, Windsor, and UW pilot plants. LTW Ottawa Windsor Ottawa. Wwgindsor. and W Research Pldom Figure 7.3 Box-and-whisker plot of C. p a m m removals by dual-media frlters during stable operation at Ottawa, Windsor, and UW research platforms. Figure 7.4 also depicts the C. parvum removals by dual-media filters during stable operation at the various research platforms, however, the data from each location were pooled and the 95% confidence intervals were calculated. They were 5.4- to 5.6-log at Ottawa, 4.7- to 5.2-log at UW, and 4.0- to 4.1-log at Windsor. The cofidence intervals account for analytical recovery and uncertainty of recovery. Figure 7.4 M e r emphasizes that the reliability of C.panrurn data will considerably increase with higher counts per sarnple or with pooling of replicate data. These data also demonstrate that similar filtration schemes may offer different levels of C. parvrirn removal. None of the 95% confidence intervals in this figure overlap, demonstrating that the C. panzirn removals at the three research platforms were significantly different from one another (a=0.05).A comparison of the Ottawa and Windsor C.parvurn removals is interesting because the pilot plants are essentially identical in construction, but treat different raw waters (described in Chapter 3). The significant differences in oocyst removal during stable operation at these MO plants suggest that the C. p a m m removal capacity of granular media flters may be source water specific; this may be associated with the necessity of different pretreatment strategies for source waters of various quality- 2-" Ottawa Windsor Location Figure 7.4 7.2 95% confidence intervals for C. parvurn removals by dual-media filters during stable operation (pooled data) at Ottawa, Windsor, and UW. OVERALL ASSESSMENT FOR POTENTLAL SURROGATES FOR PARVC'M B. strbtilis and polysty-rene microsphere removals and total particle (>2 p)reductions through the treatment process were investigated during several periods of operationai challenge to investigate potential surrogates for C. parvurn removal by filtration. The experiments were performed concurrently with the C. p a m m investigations, most of which were conducted at the Ottawa pilot plant. The B. srrbtilis and microsphere removals by filtration and the total particle (22 pm) reductions through the treatment process during all of the operational conditions investigated at Ottawa (except hydraulic steps) are summarized in Figure 7.5 through Figure 7.7 respectively. Cornparison of these figures to the C.p u m m removal data in Figure 7.1 suggests that overall, neither the spore removals nor the particle reductions are reasonable, quantitative surrogates for C.panzurn removal. The microsphere removal data corresponded closely to the C. pawum removal trends. These data are limited, however, since they were only obtained during some of the operath g investigated during this research. Opera- Figure 7.5 Period Box-and-whisker plot of B. subtilir removals by filtration during all operating periods (except hydrauiic steps) investigated at Ottawa. Oprnting Penod Figure 7.6 Box-and-whisker plot of polystyrene microsphere removals by filtration during al1 operating periods (except hydraulic steps) investigated at Ottawa. Operating Period Figure 7.7 Box-and-whisker plot of total particle (Z2 p)reductions through the plant during all operating periods (excebt hydraulic steps) investigated at Ottawa. The C. pawzrm, B. subtilis, and polystyrene microsphere removals by filtration and the total particle (2 2p.m) reductions fiom the Ottawa and UW pilot experiments were combined to assess the potential surrogates for C. pawurn removal by filtration. Some variation in the removals calculated on the basis of individual influent-effluent sample pairs was observed for al1 of these parameters. in general, the caicuIated removals durhg a given experiment were fairly reproducible as indicated by the relatively low standard deviations for the repkate samples; the overall standard deviations were also low, dernonstrating good reproducibility betwcen the replicate experiments (Appendix D). B. subtiiis and polystyrene microsphere removals and total particle (2 2 p ) reductions through the treatment process are presented relative to C. parvzrm removais by filtration in Figure 7.8 through Figure 7.12. These figures respectively correspond to the overaii removals at Ottawa (all conditions investigated at Ottawa), overall removals at stable oepration at UW), and combined removals (ail conditions investigated at Ottawa, Windsor, and UW). Least squares linear regression was used to assess the relationships between C.parvum and potential surrogate removals in dl cases. Figure 7.8 and Figure 7.12 demonstrate that B. subtiZis removals by filtration were generally lower than removals of C.parvum. However, these figures also demonstrate that these removals were not always lower than C.parmm removals, as there are some points on each figure that correspond to spore removals that were higher than oocyst removals. The coefficients of determination (R') for the best fit linear regression models were 0.48 and 0.47 for the Ottawa and combined Ottawa, Windsor, and UW data sets respectively, suggesting that spores of B. subli[is were not adequate quantitative surrogates for C.panrurn removal by filtration. This result was not surprishg given that B. strbtilis spores are -1 p in size, which is at the minimum transport efficiency for particle removal by fdtration. It was also consistent with other studies reported in the literaîure ( e g , Lytle et al., 1996;Nieminski and Bellarny, L 998; Sweafeger et al., 1999). Non-linear relationships (e-g., exponential, logarithmic, etc.) between C.parvurn and B. subtilis removals by filtration were not expected and were not suggested by the data. Figure 7.8 Relationship between C. panmm removals by filtration and total particle (22 pm) reductions through the plant during al1 operational periods investigated at Ottawa. 1.O 0.0 2.0 3.O 1.0 C.porvum R c m o d (log,,) 5.0 7.0 6.0 Relationship between C.parvum and B. subtilis removals by filtration during al1 operational periods investigated at Ottawa. Figure 7.9 -.-iine of q u i d e n c e ---,.-" ...*-- -Lin- _/- (Microspheres) .*- y = 0.8% * 0.34 R ' 0.0 1.O 2.0 3.0 4.0 5.0 6.0 = 0.95 7.0 C.p a n m Remaval (log,,) Figure 7.10 Relationship between C.parvum and polystyrene microsphere removals by filtration during ail operational periods investigated at Ottawa. Dual-Media Microspheres - - - Lmear (Dual-Media Microspheres) _ _Linex - j I 1 I (Tri-Media Mioosp hem) ' y=I.llx-1.06 h e of equivalence 1.O 0.0 Figure 7.1 1 - 3-0 4.0 C punvm Remotal (log,,) 5.0 6.0 Relationship between C. p a m r m and polystyrene microsphere removals by filtration during a l l operational penods investigated at UMr. t 1-- 0.0 2.0 Linear (B.subtiiis ) Linear (Microsphses1 1.O 2.0 3.O 3.0 5.0 6.0 7.0 C.p a m m Removd (log,,) Figure 7.12 Relationship between C. parvum, B. subtilis, and polystyrene microsphere removals by filtration during all operational petiods iqvestigated at Ottawa, Windsor, and UW. Although it might be tempting to suggest that B. subtilis removal provided a conservative indication of a filter's ability to remove C. p a m , this statement is not completely accurate because in some instances spore removals were higher than oocyst rernovals. Whde it is m e that spore removals were generdy lower than oocyst removals, the lack of a reproducible or consistent relationship behveen oocyst and spore removals suggested that B. szibtilis removals were not indicative of the filter's ability to remove C.pawum. A more accurate and consenrative concIusion would be that spore removals were generally indicative of treatment performance, but not necessarily parvzun removal. This conclusion is also supported by other studies presented in the literature (e.g.,Lytle et al., 1996; Nieminski and Bellamy, 1998; Swertficger et al., 1999). As mentioned in Chapter 6, total particle (2 2 p)reductions through the treatment process were calctdated based on raw water rather than filter influent values. Filter influent particle concentrations could not be measured due to floc breakage and accumulation on and in the particle counter sensor chamber. The coefficient of determination for the best fit linear regression mode1 was 0.27 for the Ottawa data set, suggesting that particle removals were hadequate nirrogates for C. parvum removal by filtration. This result was also consistent with studies previousIy reported in the Literature KeChevallier &al., 1991c; Niernuiski and Ongerth, 1995; Swaim et al., 1996). The lack of a linear relationship between C.panmm removals by filtration and total particle reductions through the treatment process was not s u m s i n g given that the nature and distribution of particles change as they pass through coagulation, flocculation, sedimentation, and filtration processes. Non-linear relationships between C. panzum removals by filtration and total particle reductions through the treatment process were expected and demonstrated by the data. As with B. subtiZis, the particle removals tended to be lower than C.paniurn removals at Ottawa and were not necessarily indicative of the filter's ability to remove C. pumim. Overall, particle removals were generally indicative of treatment pei-fomance, but were not quantitatively associated with C. parvum removal. The hydraulic step experiments at Ottawa underscored this point by demonstrating a substantial increase in filter effluent particle concentrations that was not accompanied by an increase in effluent oocyst concentrations. As shown in Figure 7.8 and Figure 7.12, of the potential surrogates investig&d, oocyst- sized polystyrene microsphere removals were closest to C. parvum removals by frlûxtion. The coefficients of determination for the best fit linear regression models were 0.95, 0.78, 0.66,and 0.93 for the respective Ottawa, UMr dual-media, UW tri-media, and cornbined Ottawa and UW (dual-media) data sets, suggesting that polystyrene microsphere removals may likeiy be appropriate surrogates for C.pamim removal by filtration. The UW dual- and tri-media filtration data suggested that the relationships are system specific. The poorer relationships between microsphere and oocyst removals at UW relative to those obsewed at Ottawa also suggested that coagulation (e.g., low coagulant dose for particle removal vs. high coagulant dose for combined TOC and particle removal) or filtration regime (e-g., conventional vs. in-line filtration) may a k o impact the reliability of polystyrene rnicrospheres as surrogates for C.parvum removal. Although these relationships must still be determined the excellent correlation coefficients at Ottawa and the slightly lower correlation coefficients based on more limited data fkom UW suggest that polystyrene microsphere removals should continue to be investigated as surrogates for C. pavvurn removal by filtration. L I 8.1 CONCLUSIONS The key conclusions fiom this thesis research are presented below. I. The Gibbs sarnpler may be effectively used to incorporate analytical uncertainty into confidence intervals on rernovals of discrete particles such as C. p a m m by drinking water treatrnent processes such as granula media filtration. 2. Removals of formalin-inactivated C. parvum oocysts were similar to those of viable oocysts durhg stable operation, ripening, and coagulation failure in both dual- and tri-media filters, suggesting that formalin-inactivated oocysts are good surrogates for viable oocysts. 3. C. p a m m removals were not different in dual- and tri-media filters during stable operation, ripening, hydraulic step, and coagulation failue conditions. These results suggest that, for the water matrices studied, tri-media filtmtion does not provide superior C. parvum removal in cornparison to dual-media filtration. 4. During optimal operating conditions (filter effluent turbidities and cumulative particles <0.1 NTU and ( 2 5 particles/mL respectively), >4.5-log removal of C.panrum could be achieved by filtration. At two of the three pilot plants investigated, >5-log removal of C. pawurn was achieved, even at temperatures as low as 1°C and during spring runoff conditions. C. parvum removals by filtration were moderately lower (by -0.5- to 1-log) d d n g ripening than during stable operation. These differences were minor likely due to the shoa duration of the ripening perïod. C.parvum removals deteriorated substantially (by 3- to 4-log) dduring end-of-m and early breakthrough filtration relative to stable operation, even at filter effluent turbidities below 0.1 NTU. This result suggested that filter operation during breakthrough, as measured by turbidity or even particle counts, should be avoided. During the coagulation f a k e conditions investigated at UW (low coagulant dose for particle removai), C. parvum removal b y dual- and tri-media fdters was significantly compromised, with a >4-log decrease in C.parvum removal relative to stable operation. Coagulant failure at Ottawa (hi& coagulant dose for TOC and particle removal) resulted in a >3-log decrease in C. parvum removal relative to stable operation. The lack of coagulant over a short duration of several hours resulted in this deterioration of C. parvurn removal capacity; however, when the coagulation failure conditions persisted for several fdter cycles, the absence of coagulant resulted in almost no C.parvrim removal b y filtration. Sub-optimal coagulation conditions resulted in considerably deteriorated C. parvurn removals b y filtration, even at fiter effluent turbidities below 0.3 NTU. Relatively rapid changes in hydraulic loading (hydraulic steps) demonstrated varied effects on C.parvurn removal by filtration. In rnost cases, little to no detenoration in filter effluent C. p a m m concentrations occurred. Turbidity monitoring proved more usefbl than particle counting in gauging the effects of hydraulic steps on C. p a m m passage through filters. These events should be investigated M e r to better define how and when they impact pathogen passage. 10. Oocyst-sized polystyrene microspheres appeared to be reasonable surrogates for C.parvurn removal b y fdtration during several operating conditions; however, they shodd continue to be evaluated relative to oocysts to better define the limits of their applicability as surrogztes. 11. As expected, based on filtration theory and research literature, B. sztbtilis spore removals and total paaicle reductions through the treatment process were indicative of treatment performance, but were inadequate as quantitative surrogates for C. pawum removal by filtration. 12. Increasing the number of oocysts in a given sample (eg., by increasing the processed sample volume) can considerably increase data reliability when the observed number of counts is low ( e g - ,below 10 oocysts). This research demonstrated that excellent removal of C. pamtrn oocysts can be achieved by granular media filtration processes during optimized treatment conditions. magnitude of C.pawurn removal was somewhat site- and source water-specific. The A particularly important hding was that, in some cases, poor C. p a m m removal was observed when otherwise excellent filtered water qualities of <0.1 NTU were achieved (during the end-of-run experiments). Similady, relatively large spikes in filter effluent particle counts (during the hydraulic step experirnents) were not necessarily indicative of deterioration in C. pawrrrn removal by filtration. Nonetheless, this research has demonstrated the validity of the water treatment industry's general approach of minunizing filter effluent turbiciity and particle concentrations for maximizing C.parvum removal b y granular media filtration. The experimental results clearly support the importance and usefülness of monitoring the effluent turbidity and particle counts of individual filters. It was shown that C. parvurn removal by granular media fdtration varies during different periods of the filter cycle and sub-optimal performance events that can occur during typical water treatment conditions. These differences can be site- and source-water specific, making it difficult to extrapolate data fkom one study to another. Furthemore, the difficdty in identimg vulnerable penods of operation is compounded by this investigation's demonstration that the relative potentiai for pathogen passage during operation during a given event or penod of the Nter cycle may be very different than that during another event or period of the filter cycle, even though the fùter effluent water quality measured by turbidity or particle counts may be exactly the same. Despite sitespecific details affecting the magnitude of C. parvum removals, this research has indicated that general trends in C. p a m m removal by ganular media filtration can be associated with specific design and operating conditions. The use of the developed statistical fkmework for assessing the reliability of C.parvurn removai data made it easier to compare data fkom different operating periods and research platforms. The statistical approach also emphasized that the reliability of C.p a m m concentration and removal data can be considerably increased when higher counts (e-g., more than 10 oocysts in a given sample) are obtained (e.g., by processing larger sample volumes). Finally, this research demonstrated that the use of polystyrene microspheres as a surrogate for C.parvum removal by filtration may provide utilities with a cost-effective and safe tool for p e r f o h g pilot-scale investigations to better gauge the C. p a m removal capacity of their specific systems. This tool may be particularly useful in the context of the USEPA's Long-Term 2 Enhanced Surface Water Treatment Rule (LTZESWTR) and other such regulations aimed at C. parvum removal from drinking water because they utilize a treatment technology approach for cornpliance, rather than the implementation of a stringent filter effluent monitoring requirement. 8.3.1 Water Treatment Plant Operations and Management The foUowing recommendations fcr maximiMg C.parvum removal by water treatment operations and management are based on the conclusions of this thesis research. 1. Utilize fîiter to waste capacity during ripening if possible. 2. Use a turbidimeter or particle counter as a relative indicator of filter performance during ripening. 3. - Muumize the duration of the filter effluent turbidity and particle spikes associated with npening by optimizing backwashing and pretreatment (i. e., coagulation) conditions. 4. Use particle counters to monitor filter effluents and to signal end-of-run operation and eady breakthrough. 5. Consider the relative cost implications of extended filter cycles versus maintainkg water quality by terminating filter cycles when end-of-nui is signaled (e-g., noticeable, steady rise in particle counts). 6. ClearIy specm filter backwash criteria (maximum run tirne, filter effluent turbidity, head loss, etc.). 7. Maintain optimùed chernical pretreatment by performing jar tests and pilot studies to maximize turbidity and particle removal by coagulation and fdtration. 8. Recognize that source water quality changes may result in sub-optimal coagulation and subsequent sub-optimal filtration. 9. Respond to sub-optimal coagulation conditions by adjusting cos-lmt dosing to optimal levels as quickly as possible. 10. Minimize the efEects of changes in hydraulic loading on filter performance by implementing changes gradually or by adjusting plant operations to minimize hydraulic changes. 11. Do not rely on particle removals or B. subtilis removals as surrosates for C.p a m m removal by filtration. 12. Utilize the statistical fkamework developed in this research, or a comparable approach, to assess the reliability of C. parvttrn removal data. 13. Utilize particle counts and turbidity measurements as indicators of treatment efficiency. Recognize that in some cases, C. p a m m passage through filters can increase without ciramatic increase in either of these parameters. Conversely, recognize that turbidity ancilor particle counts c m increase substantially without resulting in comparable increases in C. païvtim passage through filters. Generally, however, increases in filter effluent particles and turbidity tend to signal increases in the potential for C.p a m m to pass through filters. 8.3.2 Water Treatment Research This research demonstrated that various operating periods and conditions during granular media filtration could rssult in very different levels of C.p a m m removal. These results were obtained by performing bench- and pilot-scale seeding studies utilizing very high concentration of formalin-inactivated C. parvum oocysts that were introduced into filter influents. Several critical questions have arisen fiom this research and are s~m~narized in the following list of research recommendations. Detexmine if removals achieved during studies involving seeding at high influent C. parvzinr concentrations are representative of those aciiieved with lower C.pawum concentrations (such as those that occur indigenously). Detexmine if jar-coagulation of C. parvzrm oocysts is representative coagulation at the rapid mix. Detennine the relative impact of the presence of surfactant (to prevent oocyst aggregation) in C.p a m m samples used during seeding studies on coagulation and removal by filtration. Investigate the potential for the coagulation-flocculation-sedimentation process to result in a specific sub-population of oocysts that subsequently enters disinfection a d o r filtration processes. 5. Further investigate the deterioration of C. parvurn removal during end-ofnui 6. Investigate the relative impact of coagulation regime and filter conditionhg on 7. operation to determine when it begins and how consistently it occurs. p a m m removal by fdtration. Elucidate the impact of hydraulic steps on C. parvzcm throua filters, investigating different changes in hydraulic loading and ixnplementation of the changes at various periods during the Nter cycle. 8. Develop universal (non site-specific) rnethods for evaluating the robustness of C. parvum removals by fdtration. The statistical fiamework developed in this thesis, or something comparable, should be used to assess the reliability of C.parvuPr removal data if applicable. 9. Continue to investigate polystyrene microspheres as surrogates for C.parvum removal, potentially integrating their use into combined disinfection-removal studies. 10. Base conclusions on C.p a m m removal data that are present in reliably countable numbers ( e g - , more than 10 oocysts in a given sample or at least 10 observations). OBJECTIVE To date, one of the ch associated with the detection of C. parvum has been poor and highly variable oocyst recovery. To complete the research objectives stated in Chapter 1, it was necessary to reliably concentrate and enurnerate C. parvurn oocysts fiom water. State-of-the-art analytical methods reported in the literature were considered. A method was then selected, implemented, and optimized to consistently yield high recoveries of C.parvzrm from waters representative of those studied during the present research. Method selection was based on the following criteria: The method had to be relatively reliable (high recoveries with low variability). The method had to allow for concurrent identification of C. parvurn and polystyrene microspheres. The method had to be readily implemented, without requinng extensively trained personnel. The method did not necessan'ly have to discem between viable and non-viable C.parvurn. The rnethod had to be relatively inexpensive, without requiring exorbitantly expensive equipment. Environmental detection and enurneration methods for C~ptosporïdiumoocysts were originally developed Çom those developed for Giardia. One of the greatest challenges associated with the detection of C. parvum has typically been poor and highIy variable oocyst recovery (e-g., Clancy et al., 1994). This is in part because oocysts typically require concentration fiom large samples of water pnor to detection. It has generally been recornmended that at least 1OOL of source waters and 21000L of finished waters be concentrated pnor to enurneration of indigenous oocyst concentrations (Jakubowski et al., 1996). The method proposed by the American Society for Testing and Matenals ( A S w (1993) involved passing large water volumes of water through a polypropylene yarn-wound carû-idge filter followed by an immunofluorescence assay (FA)-Percoll-sucrose gradient protocol and epifluorescence microscopy. The literature strongly and fiequently indicates that the ASTM method is inadequate, however (e.g., Hargy et al., 1996; Yesey and SIade, 1991) and often results in highly variable and typically low oocyst recoveries (e-g, Hargy et al., 2996). LeChevallier et al. (1995) exarnined the ASTM method to determine when major losses occurred. The results revealed that the centrifugation and clarification steps could each result in Iosses as high as 30%. Further difficulties with the ASTM method can result fiom the presence of algae because numerous species of algae can fluoresce and result in false positive counts (Rodgers et al., 1995). Although the United States EPA's new Method 1622 has preliminary indications of recoveries >70% (Clancy et al., 1997), more data are necessary for a thorough evaluation of the method. Due to poor analytical recoveries and low indigenous C. panurn concentrations, experimental evaluations of oocyst removal typically involve seeding with concentrations considerably higher than those typically present in raw water. Seeding with high oocyst concentrations ensures reliably countable (Le., non-zero) concentrations in process effluents, thereby allowing for removal calculations. Subtle differences in methodology c m affect recoveries and data interpretation, however. For example, Musial et al. (1987) showed that the overall efficiency of recovery of an ASTM-like protocol could decrease as oocyst number present in the sarnple decreased; they also demonstrated that processing samples at higher flow rates could also decrease recoveries. Ensuring that spiked oocyst concentrations are accurately measured prior to spiking adds to the consistency of recoveries (Sîraub et al., 1996). Regardless of the analytical methodology used, it is important to evaluate a range of possible or expected oocyst concentrations fiorn al1 waters investigated (eg., filter idluents and effluents tiom a11 raw water sources) so that recovery can be appropriately described. As described above, implementing methods consistently will allow for better data interpretation, particulariy when trying to incorporate the uncertainty associated with recovery into descriptions of C. parvzrm or other pathogen removal by treatment processes (as was done in Chapter 4). Similady, cornparisons between studies should be evaluated with caution because oocyst recovery is often very sensitive to seeded concentration (e.g, Straub et al., 1996; Musial et al., 1987), quality of oocysts used (Dawson et al., 1993), sample volume processed (e-g., Nieminski et al., 1995), and water characteristics such as turbidity (e.g, Nieminski et a l , 1995). Several aIternatives to the ASTM protocol have been examined ( e - g , Nieminski et al. 1995; Whittmore and Carrington, 1993; Vesey et ai., 1993a). Table A.1 includes a concise sumrnary of the more common methods used in detecting C. pawurn fiom water. Observations relevant to the present study are included, however, only the methods that can be readily applied are listed. Those methods in the process of being developed (e-g, PCR, flow cytometry, etc.) are discussed in detail elsewhere in the Iitvrature (eg.? Jakubowski et al., 1997; Vesey et al., 1991). Table A. 1 Siiininary of Coininon Metliods Used in Detectiiig û)ptospoikiiirtiz Meiliod Mciliod Descripiion 'i'ypical Recovcry Soiiie Key Kefcreiices 0bset.vniioiis Relcvatit to Preseiit Stiidy (%) EPA Rlethod 1622 Concentration: vortex flow filtratioii >70 Clniicy ct al., 1997 iio fiirtlicr iiiforiiintioii ciirreiitly civnilnblc polysiilfotie capsule 72 Clniicy cl tri,, 1997 rio flirtlier iiiforiiiatioii currciitly wailt~blc PCTE iiieiiibrniie disk 96 Clniicy et cil., 1997 iio fiirtlicr iiifortiintioii ciirreiitly nvnilriblc Purification: iiiiiiiunoiiiûgnctic separatioii Rosso~iioiidoet nl., 1994 67-83 Hiikhiiri et cil., 1998 35 ;t: 15 Claiicy ct ni., 1999 Enunieration: Receiit stiidies rcported iiiiicli lower recoveries origiiially reportedlespcctcd, scinic as for. ASTM soiw as,fi)rASl'M Dowd aiid I)i11ai, 1097 CI~iicyct ai,, 1997 Stnin in soliitioii prcvciit s bnckgroiriid fl~ioresceiice, Coiicurretit eiiiiiiicrntioii riiid viability nicasiirciiierit, 1-GO LcClievdlicr ct al., 1995 7 1,cClievnllier ct cil., 1990; 1991 a,b Rodgers cr al., 1995 1-largyct cd,, 1996 Ceiitrifiign~ionaiid clarification resiilt iii losses as high ns 30%. Reçovery is grcaily iiiipackd by aiiioiiril arid type of debris prcsciit in tlic wrrntcr, Algiic cati result iii falsc-positives, 1 figlily variable, typicitlly low recovcrics, Iiiiiiii~iiofliioresceiiceAssay ( F A ) Concentration: polypropylene yarri-woiiiid cartridge Good recovery rcquires optiiiinl bend coiiceiiiri~tioti, >62% rccovery at t~irbiditiesof 5000 NTIJ. Purification: flotritioii on Ikrcoll-siicrosc gradient The C. parvum analytical protocol described by Yates et al.' (1997) was employed because of its relative ease of implementation and typically consistent recoveries that averaged between 30 and 50% (Yates, 1997). This recovery range is comparable to those presented for ASTM-like methods in Table A S . This method was designed for use during seeding studies (Le., it was not developed for studying indigenous oocyst concentrations) and involved processing fiter influent and effluent samples volumes of l 300 mL. Processing small sample volumes (of less than I L) required seeding the filter S u e n t location with very high oocyst concentrations (-10' oocysts/L); however, it also eliminated the need for a purification step, thereby increasing oocyst recovery. This method was selected as a starting point because it met a11 of the selection criteria discussed above (unlike U.S. EPA Method 1622 at the time the research commenced). The C-parwzrm analytical method of Yates et al. (1997) utilîzed filter housings (Swimex; MiIlipore Canada Ltd., Nepean, ON.) containing pre-wetted, 25 mm, 0.45 pm cellulose acetate filters (VWR Canlab, Missisauga, ON.). The filter housings were comected to disposable syringes. Approximately 2 mI, of 1% bovine senun albumin (BSA) were passed through the filters and then the water samples were passed through the filters. The syringes then were rinsed with a few milliliters of a buffered detergent solution ( l x phosphate buffered saline PBS] with final concentrations of: 0.1% sodium dodecyl sulphate, 0.1% Tween 80, and 0.01% Sigma Antifoam A and final pH of 7.4) to help maximize oocyst recovery. An additional 2 mL of 1% BSA were then passed through the filters. The membrane filters were then removed fkom the Glter housings and were placed on top of 25 mm, 8.Oprn nitroceIlulose support membranes (Millipore Canada Ltd., Nepean, ON) on a manifold (Hoefer Scientific, San Francisco, CA.); weights held the membranes in place. The sample concentration steps are summarized in Figure A. 1. A standard immunofluorescence assay (FA) was then used to stain the oocysts (USEPA, 1996). Presurnptive microscopie analysis for C.parvum enurneration was performed using epifiuorescence microscopy at 4 0 0 ~magnification (Nikon Labophot 2A, Nikon Canada Inc., Toronto). \ filter membrane placed on manifold afler water sampie is syinge-filtered 7 ' --I.& B. .- -,- %&'.&x vacuum adjusment Figure A.1 C.parvurn analytical method of Yates et al. (1997) Samples of C. p a m m were preserved in a penicillin/streptomycin solution because recent research has indicated that oocysts stored in this prese~ativemay more closely represent oocyst behavior in the natural environment (Li et al. 1997). The oocysts were obtained from a commercial laboratory (Waterbome, Inc., New Orleans, LA. or University of Arizona, Department of Veterinary Science, Tucson, AZ.). Vials of -1 0' oocysts were obtained; they were inactivated with 5% formalin (final concentration) for C. parvztrn in 1X PBS with 0.0 1% Tween 20 to prevent oocyst dumping. Al1 microorganism stocks were refrigerated at 4°C in the dark until use. The C. parvum stock suspension was briefly vortexed and a small portion of the suspension (< 100 pL in total) was removed to enurnerate the oocyst concentration. The stock concentration was deterrnined by averaging five replicate counts with a hemocytometer (Petroff-Hausser Bacterial Counting Chamber, Hausser Scientific Corporation, Horsham, PA). The entire grid (1 mm2) was used in the enurneration process (Nikon Labophot 2A, Nikon Canada hc.,Toronto). Several potential modifications to the C. p a m m concentration and enumeration rnethod descnbed by Yates et al., (199ï) were investigated- The investigated modifications included the use of a different type of filter membrane, the use of direct vacuum filtration rather than syringe filtration, the role of the buffered surfactant rime in rnaximizing oocyst recovery, and the importance of sample handling strategies such as vigorous shakmg prior to oocyst concentration. The results and implications of each set of optimization experirnents are discussed below. A.6.1 Membrane Type (Polycarbonate vs, Cellulose Acetate) Polycarbonate membranes were compared to cellulose acetate membranes of similar pore size (-0.4 pm). As indicated in Table A.1, polycarbonate membranes have been investigated in other studies in the context of ASTM-like protocols (e-g.,Nieminski et al., 1995; Kfir et al., 1995). They are desirable because they do not fold and bubble on microscope slides as much as cellulose acetate membranes. The disadvantage of their use is that because they are somewhat thereby precluding confirmation of internal structures with Nomarski differential interference contrast (DIC) rnicroscopy. Given the hi& concentrations of C.parvum oocysts and the epifluorescence rnicroscopy used durhg this research, identification of C. panum oocysts was not difficult. Since only presumptive microscopic analyses were performed during this research (Le., no confirmation of internal structures), the relative opacity of the membranes was not of concem. Recovery studies using Milli Q T M water and cellulose acetate (0.45 pm) or polycarbonate (0.4 pm) membranes were performed using concentrations of formalin-inactivated C. parvum that would be expected at the filter influent and effluent locations during the pilot-scale expenments completed d u ~ this g thesis research. Al1 of the pipettes, syringes, and glassware were ~ s e with d a few milliliters of a buffered detergent sohtion ( l x phosphate buffered saline PBS] with h a 1 concentrations of: 0.1% sodium dodecyl sulphate, 0.1% Tween 80, and 0.01% Sigma Antifoam A and final pH of 7.4) before and after coming into contact with the water samples. This rime was also passed t h r o u a the membrane filters to ensure higher recoveries. The cellulose acetate and polycarbonate membrane data are presented in Table A 2 and Table A.3 respectively. A comparison of the recovery data failed to demonstrate statistical difEerences between cellulose acetate and polycarbonate membranes used in conjunction with either syringe or manifold , n2=15 for each comparison, Table A.4). filtration (two sided t-test, ~ 0 . 0 5nl= A.6.2 Direct Vacuum Filtration (vs. Syringe Filtration) Direct vacuum filtration was compared to syringe filtration to minimize oocyst losses, allow for easier and faster handling, and to ensure constant pressure on the filter membranes. Although passing less than -50 mL through a membrane with a syringe is not diEcuIt, passing larger voIumes requires the syringe to be refillsd multiple times. As the filter membrane clogs, it becomes increasingly difficult to apply the pressure necessary to pass the water through the filter; this c m potentiaLly lead to large surges in pressure that rnight darnage the membrane. A manifold provides constant vacuum consistently for al1 samples and allows for concurrent processing of multiple samples. Recovery studies using MilliQm water and syringe or direct vacuum (manifold) filtration were conducted in conjunction with the ceilulose acetate and polycarbonate membrane investigations described in Section A.6.1. These recovery experirnents were also performed using concentrations of formalin-inactivated C. pawum that wodd be expected at the filter influent and effluent locations during the pilot-scale experirnents completed during this thesis research (Chapter 6). Al1 of the pipettes, syringes, and glassware were rinsed with a few milliliters of a buffered detergent solution (described in Section A.6.1) before and afier coming into contact with the water samples. This rinse was also passed through the membrane filters to ensure higher recoveries. The syringe and direct vacuum filtration data are summanzed in Table A.2 and Table A.3 for cellulose acetate and polycarbonate membranes respectively. A comparison of the recovery data demonstrated statistically dflerent recoveries with both types of membranes when direct vacuum filtration, as compared to syringe filtration, was used (two sided t-test, a=0.05,nl= n2=15 for each comparison, Table AS). This result clearly demonstrated that the higher recoveries achieved with direct vacuum filtration were significantly different £?om those achieved with syringe filtration. Table A-2 C.pamrm Recovery h m Cellulose Acetate Membranes Concentration Method Cellulose Acetate Processed Volume 2.5 Seeded (Oocysts/L) 1.0 E+6 i- S~ringe Filtration # Counted 1444 1665 1352 1559 478 1370 1828 1346 1537 1663 C. parwum OocystsIL Recovery (%) 5.8 E+5 6.7 E+5 5.4 E+5 6.2 E+5 1.9 E+5 5.5 E+5 7.3 E+5 5.4 E+5 6.1 E t 5 6.7 E+5 Average: Sfd dm.: Coeff: Var.: 466 724 418 714 830 Average: Std d m : CoefJ Var.: Cellulose Acetate + 2.5 1.0 E+6 Manifold Filtration 1528 1586 1896 2021 2223 1874 2249 2091 1776 23 58 6.1 E+5 6.3 E-1-5 7.6 E+5 8.1 E+5 8.9 E+5 7.5 E+5 9.0 E+5 8.4 E+5 7.1 E+5 9.4 E+S Avera~e: Std dm.: Coefj Var.: 638 810 814 616 744 Average: Std dm.: Coeff:Var.: 278 13 Table A.3 C.parvum Recovery h m Polycarbonate Membranes Concentration Method Polvcarbonate +- Processed Seeded Conc. Volume (Oocysts/L) 2.5 1.0 E+6 S~ringe Filtration # Counted 2219 1269 1593 1298 1534 1333 1601 1843 1650 1432 C. pawum Oocysts/L Recovery (%) 8.9 E+5 5.1 E+S 6-4 E+C 5.2 E+5 6.1 E+5 5-3 E+5 6.4 E+5 7.4 E+5 6.6 E+5 5.7 E+5 Average: Std d m : Coeff: Var.: 504 448 624 766 770 Average: srd dm.: CoefJ Var.: Polycarbonate + 2.5 1.0 E+6 Manifold Filtration 1898 1902 2117 1744 1620 2214 1971 2213 1860 1772 7.6 E+5 7.6 E+5 8.5 E+5 7.0 E+5 6.5 E+5 8.9 E+5 7.9 E+5 8.9 E+5 7.4 E+5 7.1 E+5 Average: Std dm*: Coeff: Var.: 658 728 576 750 808 Average: Std dm.: Coeff:Var.: - 279 13 Table A.4 StatisticaI Anaiysis of Cellulose Acetate and Polycarbonate Membranes F-Test T m a r n p l e for Variances Variance Observations df F P ( F c = one-tail F Critical one-tail F-Test T-ample Mean Variance Observations df 239.32 15 14 t-Test T a a m p l e Assuming Equal Variances 147.25 15 14 1.63 0.19 2.48 for Variances . F P(Fc=f) one-tail F Critical one-tail Cellulose Acetate. Pt Poiycartmnate. CA ManifoId 76.42 113.89 15 14 1.52 0.22 248 Mean Variance Observations Pooled Variance Hypothesized Mean ilifference df tstat P(TG-t) one-tail t Critical one-tail P(Tc=t) two-tail t Critical two-tait CA Syringe 58-99 PC Syringe 62.81 t-Test Two-Sanple Assurning Equal Variances PC Manifold 74.96 75.13 15 14 Mean Variance Observations Pooled Variance Hypothesized Mean ûiierence df t Stat P(Tc-t) one-tail t Critical one-tail Pflc=t) two-tail CA Manifold 76.42 PC Manifold 74.96 Table A S StatisticaI AnaIysis o f Direct Vacuum (Manifold) and Syringe Filtration t-Test T N a m ple Assuming Equal Variances F-Test Two-Sample for Variances Variance Observations df F P(Fone-tail F Critical one-tail CA Syrïnge CA Manifold 239.32 15 14 210 0.09 2.48 113.89 15 14 PC Cellulose Acetate. Polycarbonate. PC Syringe 6281 147.25 15 14 1.96 0.1 1 248 CA Syringe 58.99 239.32 15 176.60 O 28 -3.59 0.00 1.70 0.00 205 CA ManifOld 76.42 11 3.89 15 t-Test Two-Çample Assuming Equal Variances F-Test T M a m p l e for Variances Mean Variance Observations df F P(Fc-f) one-taii F Critical one-tail Mean Variance Observations Pooled Variance Hypothesized Mean Difference df t Stat P(Tc=t) one-tail t Critical one-tail P(T-) two-tail t Critical twetail PC Manifold 74.96 75.13 15 14 Mean Variance Observations Pooled Variance Hypothesized Mean Difference df t Stat P(Tc=t) one-tail t Critical one-tail P(Tc=t) two-tail t Cfltical W t a i l PC Syringe 62.81 147.25 15 111.19 O 28 -3.1 6 0.00 1.70 0.00 205 PC Manifold 74.96 75.13 15 The findings fiom the method optimization experiments described in the previous section (Section A.6) were applied to the method of Yates et al. (1997) to yield an optimized parvum analytical protocol. The key results that were integrated into the C.pamnz analytical method fiom the rnethod optimization experiments were: 1. the use of 0.40 pm polycarbonate membranes because they do not warp as readily as cellulose acetate membranes (thereby resulting in more easiiy obtained and reliable counts) and 2. the use of direct vacuum filtration rather than syringe filtration because it is easier, faster, and results in Iess sample loss (higher recovery). The optimized C.parvum concentration and enurneration protocol is presented Table A.6. This protocol was employed for al1 of the C. parvum analyses performed during this thesis research. Table A.6 C. pawurn Concentration and Enurneration Protocol Autoclave the support membranes and filter membranes in 1X PBS. Prepare Primary and Secondary stains accordinj to the protocol provided by the Hydrofluorm* 8/10volume 1X PBS). Combo Kit (1110 volume pritnary or secondary stain, 1110 volume goat seStore prepared stains in a dark place due to their Iight sensitivity. Open manifold ports aod connect the manifold to the vacuum pump- Adjust pressure release valve to obtain a vacuum of 5 hches of mercmy. Close manifold ports and turn off vacuum pump. With flamed forceps add the support membranes and then the filter membranes to the manifold, Place weights on the membranes. (Note of where each sample will be filtered-) Filter 2 mL of 1% BSA through each membrane. Rinse graduated cylinder (larger sample volumes) or pipette (smailer sampIe volumes) with buffered surfactant solution and discard excess solution Shake sarnples bottles to mix the contents. Add appropriate amounts of filter influent and Fdter effluent to each pre-rinsed, labeled graduated cylinder or pipette. CarefulIy add each sample to the appropriate manifold and filter. membrane moist when filtering is cornpleted-) (Add IX PBS to keep the Rime each graduated cylinder with buffered surfactant solution, retaining the solution. Filter through the appropriate membrane. Riose a small test tube with eluting solution and discard excess solution. Add S O a of the positive control to the test tube and add -5 mL of 1X PBS; filter through appropriate membrane. Rinse test tube with eluting solution and filter the retained solution. . Filter 2 mL of 1X PBS for the negative control. 283 Table A.6 C. p a m m Concentration and Enurneration Protocol (Continued) 12. After filtering aii the samples, filter 2 mL of 1%BSA through each membrane. J1 13. After al1 of the BSA has been fütered turn off vacuum. Close al1 ports and add 500pL of the prepared Primary Stain to each membrane. Let stand covered (light sensitive), for 25 minutes. (At this time the slides can be Iabeled and placed on the slide warmer at 37°C.) 14. After 25 minutes, tuni on the vacuum and open the ports to alIow the Primary Stain to filter through. 4@ 15- Rime each membrane 5 times with approximately 1 mL. of 1X PBS- JI 16. T m off vacuum. Close al1 ports and add 500pL of the prepared Secondary Stain to each membrane. Let stand. covered. for 25 minutes, (At this t h e , add 2 drops of 2% DABCO/glyceroi to each siide.) 17. M e r 25 minutes, turn on the vacuum and open the ports to allow the Secondary Stain to drain through, 18. Riuse each membrane 5 times with approximately 1 mL of IX PBS. + 19. Add approxhnately 1 mL Ethanol series in the foiiowing order: 10%. 20%, 40% 80%, and 90.2%. Allow each volume to comp1eteIy drain before adding the next volume. J1 20. Remove the filter membranes fiom the manifolds and transfer them to the labeled slides, using flamed forceps behveen tramfers. AIlow the filters to remain on the slide warmer for 5 minutes. 21. Add 1 drop of 2% DAûCO/glycerol on top of each membrane. J1 22. Place a cover slip on each membrane and let stand on the warmer for 5 minutes. JI 23. Remove air bubbles and excess glycerol with cotton swabs. With cIear fingernail polish in a syringe, seal around the cover slip- Store slides at 4OC in the da&, Enurnerate at 400x ma-gnification. Once the optimized protocol for C. parvum concentration and analysis was established, it was necessary to evaluate the recovery of C- pumrrn oocysts fiom the various water matrices encountered during this thesis research. Recovery experirnents were performed on the following water matrices: 1. Treated Ottawa River Water 2. Treated UW Synthetic Water (1.5 NTU) 3. Treated UW Synthetic Water (3 -5 NTU) Recovery of C.pawum fiom filter irduent and efnuent samples fiom each of the water matrices was investigated. The C. parvum recovery data for the Ottawa, UW 1.5 NTU, and UW 3.5 NTU waters are available in Table A.7, Table A.8, and Table A-9 respectively. A single factor analysis of variance (ANOVA) was used (Table A.10) to determine if there were significant differences in C.parvum recovery fiom the various water matrices studied when the optimized C.parvurn analytical method (Section A.7) was used. These results indicate that the ANOVA analysis failed to demonstrate sipifkant differences between C.parvurn recoveries fiom filter influent and effiuent waters fiom any of the water rnatnces investigated during this thesis research ( ~ ~ 0 . 0 5 ) . Ideally, treated water fiom the Windsor Pilot Plant (filter influent and effluent) would also have been included in these recovery investigations. Given the small number of experiments performed at Windsor (relative to Ottawa and UW) and time cornmitment necessary to complete the recovery studies (while experiments were concurrently being performed in Ottawa), recovery experiments were not performed on the water fÏom Windsor. If M e r C. parvurn seeding investigations were plamed for Windsor, recovery experirnents would be warranted so that accurate statistical analyses of oocyst removal by the treatrnent process could be performed. Given the consistent C.p a m m recoveries Çom the other water matices investigated (discussed above), it was assumed that the oocyst recoveries fkom Windsor filter ïnfiuent and effluent would also be consistent with these recoveries. Table A.7 C.pawttm Recovery fkom Ottawa Water Date 5/ 11/99 Sample Location Processed Volume (m=) Filter duen< LOO Seeded Concentration (Oocysts/L) 1.0 E+6 average &om 50 fieIds of view f # Counted (Oocysts) 7.0 7.2 { ::: (- 7.1 C.parwrn Measured Oocysts/L 7.0 E t 5 7.2 E+5 8.5 E+5 8.2 E t 5 7.1 E+5 Average: Srd Dev.: Coeff: Recovery (%) 70 est. 72 est. 85 est. 82 est71 est. 76 est. 7 est. 9 est. Average: Std Dev.: Coeff: 5/18/99 .~&ted Filter effluent 72 84 70 76 74 Average: Std Dm.: 72 84 70 76 74 75 5 (est,) by field counts @ 400x using Equation 3.1. (Nikon Labophot 2% Nikon Canada Inc.. Toronto. ON) Table A-8 Cspamrrn Recovery fiom UW Water (1-5 NTU) Date 4/1/00 Sam~le Location Filter Muent Processed Volume (mL) 2.5 Seeded Concentration (Oocvsts/L) 1 -0E-tfî5 # Counted (Oocvsts) 128 134 177 C.~ a r v u m Measured (Oocvçts/L~ 5-1 E+4 5.4 7.1 52 6.6 6.3 E+4 E+4 Recovmy IO/) 51 E+4 E+4 E+4 7.8 E+4 5.6 73 7.0 8.2 6.7 8.7 8.1 E+S E+4 E+4 E+4 E+4 E+4 E+4 A verape: Sid Der.,: CoeK Var.: 4/6/00 Filter effluent 500 716 788 840 792 810 754 724 682 8 16 724 A vera~e: Std Dm-: CoefjC Var.: 4/7/00 Fil ter effluent 500 64 54 58 70 80 78 88 74 78 78 . Average: Std Da): C o e e Var.: 15 Table A.9 C.p a m m Recovery fiom UW Water (3.5 NTU) Date Sample Location Processed Volume (mu Filter influenc 1O0 Seeded Concentration # Counted (OocyStdL) (Oocysts) 1.0 E+6 f 8.3 average fiom 50 fieIds of view 6-9 C,pamirn Measured (Oocysts/L) -- Filter effluent Filter effluent 5 1.0 E+5 83 est. 69 est. 71 est. 65 est, 9.0 E+5 90 est. 76 est. 10 est. 14 est. 500 8.7 7.9 6.3 6.0 Et4 E+4 E+4 E+4 7.7 E+4 Average: Sfd Dar: Coeff: 732 822 552 s34 708 Average: Std Dm.: Coeff: 500 LOO 60 84 70 62 88 Average: Srd Dm.: Coeffi -~stixnated (est.) by field counts @ 400x d (%) 8.3 E+5 6.9 E+5 7.1 E-1-5 6.5 E+5 Average: Srd Dm.: Coeff: Filter influent Recovery 87 79 63 60 77 73 11 15 73 82 55 83 71 73 11 16 60 84 70 62 88 73 13 17 g Equation 3.1. (Niion Labophot 2A, Nikon Canada hc.,Toronto. ON) Table A. 1O ANOVA Analysis of C.parvum Recovery fkorn Various Water Matrices Anova: Single Factor SUMMARY - - . .- .. .. .. Gmups Ottawa FI Ottawa FE U W 1.5 FI U W 1.5 FE U W 3.5 FI U W 3.5 FE Count Sum 5 10 14 20 10 10 380.00 751 -40 950.40 t 486.60 742.00 742.00 SS df 5 ANOVA Source of Variation Between Groups Within Groups 517.72 6463.52 63 Total 6981-24 68 Average Variance 76.00 75-14 67.89 74.33 74.20 74.20 48.50 32.76 135.57 71 -22 195.07 122.62 MS 103.54 102.60 F 1.Of P-value Fcnt 0.42 2.36 One additional set of recovery experiments was performed using Ottawa filter innuent and effluent water with a dose of 30 mg/L alum (Alr(S04)3-18H20). This experiment was performed to ensure that the presence of hi& doses of coagulant, as used at Ottawa during the jar coagulation of C. parvurn oocysts pnor to seeding (Chapter 3), did not affect the recovery of C. p a m m oocysts from the water (e.g., perhaps by causing additional aggregation). The C. p a m m recovery data for Ottawa water with 30 mgL alum addition are summarized in Table A.11. A single factor analysis of variance (ANOVA) was again used (Table A.12) to determine if there were significant dBerences in C. p a m m recovery fiom the various water matrices studied when the optimized C.parvurn analytical method (Section A.7) was used. These results indicate that the ANOVA analysis failed to demonstrate significant differences between C. p a m m recoveries fiom filter influent and efnuent waters from any of the water matrices includhg the Ottawa water dosed with high concentrations of alun ( ~ 0 . 0 5 ) . Table A. 11 p a m Recovery fiom Ottawa Water with 30 mg/L Alurn (Ai2(S04)3-18H20) Date SampIe Location 6/2/99 Filter influent Processed Volume (mu 6/2/99 Filter effluent 10 500 Seeded Concentration (00cy-w 1.0 E+6 ff Counted (Oocysts) 6992 8019 5888 7583 7961 32 39 38 41 28 C,parvum Measured Recovery (Oocysts/L) (%) 7.0 E+5 8.0 E+5 5.9 E+5 7.6 E+5 8.0 E+5 Average: Std Dm.: coe, 64 78 76 82 56 Average: Std Dm.: CoefJ; 70 80 59 76 80 73 9 12 64 78 76 82 56 71 11 15 Table A. 12 ANOVA Analysis of C.parvurn Recovery fiom Various Water Matrices Lncluding Ottawa Water with a Hïgh Coagulant Dose Anova: Single Factor SUMMARY - - ...... . Gmups Ottawa-Alum FI Ottawa-Alum FE Ottawa FI Ottawa FE U W 1.5 FI - U W 1.5 FE U W 3.5 FI U W 3.5 FE Count 5 Sum 1O 5 1O 14 20 10 1O 364.43 714.40 380.00 751-40 950.40 1486.60 732.00 742.00 Average 72.89 71-4-4 76.00 75.14 67.89 74.33 74.20 74.20 Variance 78.05 109.77 48.50 32-76 135.57 71-22 195.07 122.62 ANOVA - - -. Source of Variation Between Groups SS df MS F P-value 545.55 7 77.94 0.76 0.62 Total 8309.17 83 F crit 2.13 Method blanks (negative controls) were included almost every time samples were processed for C. parwm; they were not processed on the few occasions where space on the manifold was limited. The method blank consisted of a filtered 2-mL sample of phosphate buffered saline (PBS) at pH 7.4. This control came into contact with al1 of the reagents used to process the C.p a m m oocysts fiom the water samples. The inclusion of this control during al1 (or most) of the C.p a m m analyses ensured that the processed water samples were not contaminated with an outside source of oocysts during processing. The method blank data for the Ottawa, Windsor, and UW experiments are available in Table B.1, Table B.2, and Table B.3 respectively. These tables indicate that negative controls were processed for al1 but two experiments at each of the pilot plants; an "x" in the negative control column indicates that no oocysts were found in control sample. As indicated in Table B. 1 through Table 8 . 3 , no oocysts were found in any of the 50 processed negative control samples, thereby suggesting no contaminant source of oocysts associated with sample processing. Positive controls were included almost every time samples were processed for C.parvum; they were not processed on the few occasions where space on the manifold was limited. n i e positive control sample consisted of a fdtered 50-pL sample of formalinized stool containing Cryptospondium spp. oocysts and Giardia spp. cysts preserved with 10% formalin. This control came into contact with al1 of the reagents used to process the C. parvum oocysts from the water samples. The inclusion of this control during all (or most) of the C. parvum analyses ensured that the identification method was reliable for st;iining oocysts (and cysts). Well-stained oocysts (and cysts) were expected h m every positive control sample; if they were not found, the oocyst counts from water samples processed at the same time as the positive control would have to be considered inconclusive due to inadequate staining The positive control data for the Ottawa, Windsor, and UW experiments are available in Table B.l, Table B.2, and Table B.3 respectively. These tables indicate that positive controls were processed for all but 1 experiment at Ottawa, 2 experiments at Windsor, and four experiments at UW. Although positive controls were not associated with 7 experiments, the experiments represented only 4 sample processing occasions. A "J" in the positive control column indicates that oocysts (and cysts) were found in the control samples. As indicated in Table B. 1 through Table B.3, oocysts were found in al1 of the 49 processed positive control samples, suggesting reliable staining of oocysts. These data support the argument that samples in which little or no oocysts were found t d y had low concentrations of oocysts rather than low counts due to poor staining methods. B.3 FILTER FLUENT AND EFFLUENT NEGATIVECONTROLS Filter influent and effluent negative controls were collected during almost every experiment. Filter uifluent negative controls were collected to ensure that no submantial outside source of oocysts af5ected the C.parvurn removal data; such a source was highly unlikely because seeded oocyst concentrations were typically 105-1o6 oocystsL during the pilot-scale experiments. Although filter influent negative controls could also be processed to determine background concentrations of cysts and oocysts in the frlter influent, it would be sornewhat misleadhg to suggest that the controls collected during this research would adequately serve that purpose. The Nter infiuent negative control volumes that were processed were typicaiiy 0.1 to L L; while inappropriate for determining background concentrations of indigenous oocysts (typically X 0 0 - 1000 L are required) these volumes were adequate for ensuring that background levels of oocysts did not substantially impact the seeded oocyst concentrations. Filter effluent negative controls were collected and processed in 1-L volumes to ensure that the C. p a m m removal data were not affected by background concentrations of oocysts (either indigenous or fiom previous experiments) exiting the filters. The samples were processed in 1-L volumes because the experiments were designed with seeded concentrations aimed at ensuring that oocysts woiild be found in reliably countable concentrations in 1 -L samples of filter effluents. It was originally planned that the collection of filter influent and effluent controls would alternate between the influent and effluent locations during successive experiments. After March 1999, and during most of the pilot-scale experiments, negative controis were collected at both the filter influent and effluent locations to better assure minimal effects, if any, fi-om outside sources of oocysts. The flter influent and effluent negative control data for the Ottawa, Windsor, and UW experiments are available in Table B.1, Table B.2, and Table B.3 respectively. These tables indicate that filter influent negative controls were processed for al1 but 1 1 experiments at Ottawa (due to the original alternating sampling scheme) and 2 experiments at UW. Filter effluent coneols were processed for al1 but 12 experiments at Ottawa (again, pnmanly due to the original alternating sampiing scheme) and 2 experiments at UW. An "x" in the filter influent and effiuent negative control columns indicates that no oocysts (or cysts) were found in control samples. As indicated in Table B.1 through Table B.3, no oocysts were found in any of the 43 processed filter influent control samples or 42 processed filter effluent negative control samples, suggesting no substantial outside sources of oocysts (either indigenous or fkom previous experiments). Table B.1 Methodological Positive and Negative Control and Filter Influent and Effluent Control Results for Pilot-Scde Experiments at Ottawa Experiment Date Euperimental Conditions 8/6/98 Stable Filter Operation (Shakedown #1) stable- ilt ter Operation Stable Fiker Operation StabIe Filter Operation Ripening Ripening Ripening Breakthrough Breakthrough Bd%hrough Onset of Breakthrough No Coagulants in Plant, No Coagulants in Jar Sub-optimal CoaguIation Stable Filter Operation Subaptimal Coagulation Stable Filter Operation During Runoff Stable FiIter Opemion During Runoff Sub-optimal Coa,dation Stable Filter Operation Hydrtiulic step Hydradic step Hydraulic step No Coagulants in Plant No Coagulants in Plant Stable - Seeded at Rapid Mix Stable Filter Operation No Coagulants in Plant, No Coagulants in Jar StabIe Filter Operation, No Coagulants in Jar No Coagulants, No Media No Coagulants Since Backwash No Silicate in Plantflar Onset of Breakthrough Onset of Break'through Stable Filter Operation Onset of Breakthrough Onset of Breakthrough Onset of Brdcthmugh Onset of Breakthrough Processing Positive Negative Date ~ o n t r o l * control" x X x X X X X X x X x X x x x x X X X x x x NP x NE' x X X Y x x x X x X x x x X x Sub-optimai Coagulation ' An h check marli. ( J ) indiutes the presence of oocysts in the sample. X indicates that no oocysts were found in the m p l e . "'Nor sar~~pled. NP Not proces~ed x Filter Filter Influent Effluent controlS controlB Table 8.2 Methodological Positive and Negative Conîroi and Filter M u e n t and Effluent Control Results for Pilot-Scale Experiments at Windsor Expenment Date Experimental Conditions 3/ 10199 Stable Constant Rate Filtration Stable - Declinhg Rate Filtration Stable Constant Rate Filtration Stable - Declining Rate Filtration Stable Operation with Pre-ozonaaon Stable Opercation Stable Operation ri 3/22/99 n 10/13/99 10/21/99 10/22/99 - - A check marlI ( 4 )indiutes the presence of oocysts in the sarnple. Processhg Positive Negative Filter Filter Date Controf controlB Influent Enluent ~ o n t r o l ~~~ontrol"' 3/11/99 4 x Y x 3/11/99 J x K x 3/23/99 J x >C x J x x x 3/23/99 J Y x Y 10/15/99 NP NP x x 10/23/99 NP NP Y x L0/23/99 An x indicates that no oocysts were found in the sample. Not processed. NP Table B.3 Methodologicai Positive and Negative Control and Filter Muent and Effluent Control Results for Pilot-Scale Experiments at UW Expenment Date 11/23/99 11/24/99 11/28/99 11/28/99 I3/10/99 1 x 11/99 1/15/00 1/16/00 Experimentai Conditions Stable Operation - Dual-media n Stable Operation - Tri-media l1 Hydraulic Step - Dual-media 81 Hydrauiic Step - Tri-media n - A check mark ( 4j indiutes the presence of oocysts in the synple. * An JC indicates that no oocysts were found in the sample. Not sarnpled. NP Not processed. NS Processing Positive Negative Filter Date ~ontrol* ~ontrol$ Influent controlf 11/25/99 J x NS 1 1/25/99 4 Y NS 11/29/99 i 1/39/99 NP 13/13/99 12/13/99 NP l/ 17/00 1; 17/00 NP NP J J Filter Effluent ~ontrol~ NS NS Y x x Y NP Np L Y x x Y Y Y x x x x x Table C. 1 Calculation of Beta Paraineters n and b for C poiwrii Recovery from Ottawa Filter Iiifliieiit Seedcd Sniiiple Voluiiic Sceded Obscrvcd Concentration Voliiii~e Proccsscd Nuiiibcr o f Oocysls Nuiiiber o f Oocysts (0ocystslL) 1 .OE+OG (L) O.1 x y E@ ) Vsr@ ) E(~I)lVnr(p) E(p )2/Var(p) I /Vnr(,o ) Niiiiibcr o f trials (%) 0.5 500 35 1 35 1 149 0.70 4.176 E-4 1.68 1 Ed-3 poolcd iiienn recovcry 0,7745 pooled vnriniice 0.0045 poolcd stniidiird devintioii 0,0674 1.180Et3 2,395E-t.3 Cnlculation of Betn Parameters n aiid 1> for C.yn~*i)~rin Recovery from Ottnwo Filter Effluent Seeded Srimple Volunie Scedcd Observed Coiiceiitration Volunic Processed Nuiiiber of Oocysts Niiiiiber of Oocysts (%) (oocysts/L) (L) 1O00 OS 1O0 500 357 0,5 100 500 366 1000 43 1 1O00 0.5 100 500 500 1 O00 0.5 1O0 378 O. 5 1 O0 500 345 1O00 O. 5 1O0 50 1 O0 36 O. 5 100 50 1 O0 42 1O0 0,s 1 O0 50 35 38 100 0.5 100 50 x 357 366 43 1 378 345 36 42 35 38 y E(p ) 143 0.7 1 134 0,73 69 0,86 122 0,76 155 0.69 14 0,72 8 0.84 15 0.70 12 0.76 poolect iiiciiii recovery 0.7645 poolcd vririaiicc 0.0048 poolcd standard devintioii 0,0696 Var@ ) 4.076 E-4 3.916 E-4 2,374 E-4 3.682 E-4 4,269 8-4 3,953 8-3 2,635 E-3 4,118 8-3 3,576 E-3 E@ jlVar(p) E(p)2/Vnr(p) l/Vnr(p) Nunibcr of trials 1.752 E t 3 1.1369 E t 3 3.630 E t 3 2,053 E t 3 1,616 Et3 1.821 E t 2 3.188 E t 2 1.700 Et2 2.125 Et2 1.251 E t 3 1.368 E+3 3,129 Et3 1.552 Et3 1.1 15 E t 3 1.31 1 E t 2 2.678 E t 2 1,190 Et2 1,615 Et2 2.453 E+3 2.554 Et3 4,212 E t 3 2.716 Et3 2,342 E t 3 2.530 E+2 3.795 E t 2 2.429 Et2 2,796 E+2 Cclrc.m Cao- r . - * y q y 0 0 0 , 5: :zzsosaaggzsz r $ 7 ~ t t W t E YNF:N'??V! t t W t W T m T ; T u -Tl???'? - - l - - a a m - a - * m m C I ~ ~ ~ m v l m m v - - - - - 0 0 0 0 C 3 ' 0 0 0 0 9 9 9 9 c O O C O C Table D. 1 C parvum Removal Data fiom Bench-Scale Experiments Date TF of Expeait Stable Opmation Inricüvared C.panwn S d Durauon (min) 10 23 40 Tri-,Media Dual-Media C parvm at Sampting FI FE ~oocystsn) ~oc?cysrsn) 6 5 3 2 Et5 3 5.66 Et5 5 4.84 E+5 C.pnum Log Removd FI FE I-) ~oocystsn) f oocysts/L) I-) Log Removal 5.0 5.3 5.0 5.86 E+5 6'73 E+5 5-98 E+5 5 1 2 5.1 5.8 5.5 4.9 4.6 4.0 4.6 4.26 E+5 4.47 E+5 4.61 E+5 457 E+5 3 5 14 8 5.7 5.0 4.5 Stable Opetrition Viable C. p r v w n Ripening Inacüvitred C. parvum Ripening Viable C.parvum 5 10 15 20 3.68 E+5 3.94 E+5 3 5 1 E+5 3 . n E+S 5 11 32 10 J.8 Table D.2 Cillculatioii of Ctlicorclicnl for Bench-Scale Experiments STABLE - Inaclivaled 1 - STABLE lnaclivated2 - STABLE Inaclivaled 3 - STABLE Viable 1 K - STABLE Vlable 2 - STABLE Vlable 3 - RIPENING inaclivaled 1 - RIPENING lnaclivaled 2 (oocys\slL) 7.14Et05 (00cysldL) 5,52Et05 (00cyslçlL) 0.77 7.14Et05 (mL) (00cyslçlL) 5.86Et05 0.82 (oocy sldmL) O. 1 1.07Et08 (mL) 0.1 (0ocyslslmL) 1.07EtO8 (ml) 1500 Table D.2 Calculntioii of Ctlieoreiicel for Beiich-Scnle Experiments (Contiiiued) Expriment - RlPENlNG lnaclivaled 3 - RIPENING Viable 1 - RIPENING Viable 2 Vi - RIPENING Viable 3 CF - lnaciivated 1 - CF lnaciivaied 2 CF - lnactivated 3 Dual-Media (oocysldL) 5.61Et05 (oocysts/L) 4.24Et05 Tri-media 0.76 (oocysldL) (oocysldL) 638Et05 5.53Et05 (mL) 0.79 0.05 (oocystdinL) 8.42Et07 (mL) (oocystdrnL) 0.05 1.05Et08 (mL) 750 Table D.2 Calculation of Ctheoreticiil for Bench-Sc& Experiiiients (Continued) Experiment - CF Viable 1 - CF Viable 2 - C f Viable 3 Tri-media DuabMedia (mL) (oocystdL) 5.38Et05 0.78 (0ocysIdL) 6.27Et05 (oocysls.'L) 3,75E+05 0.60 O, 1 (oocySIdmL) 1.04Et08 6.90Et05 6,90E+05 6.90Et05 5.78Et05 5.91 Et05 5.09E +O5 0.84 0.86 0.74 6.27E.tO5 6.27Et05 6.27Et05 3.71Et05 4.00Et05 4.20Et05 0,59 0.64 0.67 O, 1 0.1 0.1 1.04Et08 1.04E +O8 1.04Et08 0.1 0.1 0.1 9.41Et07 9.41Et07 9.41Et07 1500 1500 1600 6.90E+05 6,90Et05 6,90E+05 4.62Et05 4.55E.t.05 4.94E+05 0.67 0.66 0.72 6.27Et05 6.27E+05 6.27Et05 3.75Et05 4,00E+05 3.96Et05 0,60 O. 1 1,04Et08 0.64 O. 1 0.1 0.1 0.1 9.41Et07 9.41Et07 9.41Et07 1500 1500 1500 0.63 O. 1 1.04E+08 t .04Et08 (m L) (0ocysldmL) 0.1 9.41E+07 (mL) 1500 (oocysldL) 6.90Et05 Y C C C X X X g= a g~ s s~z cc, cc. F, ~ q , Table D-5 Turbidity and Total Particle Data h m Ottawa Date 816198 Ty~e of Experimenc Stable Filter ûpmtion Seed Duntion at Sampling (min) 15 Stable FiIter Q e m u o n 15 30 45 55 Stable Filtu Operation 15 30 45 55 Stable Frlter Opcntion 15 30 -15 55 Stable Filter ûpuation 15 30 45 55 1/19/00 Stable Filter Operation 15 30 45 55 Stable Filter Operation 15 30 45 55 Stable Filter Opention Runoff Runoff Reduction through entire t r e a t n x ~ pluit. ~t PI Turbidity (NTU) FE Log ~eduction* 1.M 0.02 1.73 PI 4%6 Parcicles >=2 ~ t m (#/mL) FE Log ~ e d u a i o n ' 19 3-47 Table D.5 Turbidity and Total Particle Data korn Ottawa (Continued) Date 7/20/99 TF of Experïment Stable Filter Operation seeded at npid mix Seed Dunuon at Sampling Pt Turbidity FE Log ~eduction* PI Panides >=7 p m (#/mi.) E Log ~eduction* (min) 150 165 1.40 0.08 180 195 2 10 3 3 340 355 370 Filter Ripening Filter Ripening Filter Ripening -1-1 * Reducüon through entire mrment plant 3.56 0.06 1-80 - .-. Table D.5 Turbidity and Total Particle Data f?om Ottawa (Continued) 3/1/00 Ty~e of Experiment Early Breakthrough 3/3/00 Early Brd-through 3/4/00 Early Breakthrough Date Szed Duntion at Smpling (min) -335 PI Turbidity NlV) FE Log ~ e d u a i o n ' 3.28 0.04 PI Puticles >=2 pl(#/dl FE Log ~eduaion- 1.88 7842 1-14 6213 h t e Breakrhrough 1220199 h t e Breakthrough * Reducùon rhrough enùre ue;iuncnt plant. 83 1.87 Table D.5 Turbidity and Total Particle Data kom Ottawa (Contïnued) TF Date 12/22/99 of Expairnent Lare Br&-through No Coagulants. No Media No Coûgulant - Extendeci Duration No Cmgulant - Short Duntion No Cmgulant - Short Duration No Coagulant in Plant No Coagulant in Plant No Coagulmt in Jar No Silicate Seed Dumuon at Sarnpling (min) 15 30 45 55 PI Turbidity NiT.ll FE Log Reduction0 437 0.35 432 030 4.33 030 431 030 1.23 1-16 1.16 1.16 PI Particles >=2 prn (#/mi.) FE Log ~eduction' Table D.5 Turbidity and Total Particle Data i?om Ottawa (Continued) Date 2f 18/99 3/22/99 Type of Experiment Sub-OptimaI Coagulrition (52%reduaion to 20 rn-) Sub-OptimalCoagulation (47%reduction to 20 m a ) Seed Dmtion at Sampling (min) 15 30 35 55 15 30 15 55 5/4/99 Sub-Opûmd Coagulation (49%reduction to 22 m&) 15 30 45 55 31 10100 S~bOpumalC~agulation (49%reduaion to 20 mg/Lj -245 -235 -200 -170 - 140 -65 -35 1s 30 45 55 6n/99 Hydraulic Step (5% inaeasr in flow) 6/15/99 Hydrauiic Step (3% increase in flow) 6/22/99 HydmuIic Step (25% increase in flowj Reduction through encire munent plant PI Turbidity t NTU) FE Log ~eduction' 1.05 0.98 033 PL Parcicles >=Z pm (#/mL) FE Log kduction- 3631 589 0.79 Table D.6 Microorganism Data fiom Ottawa Date -- of Experimeot Stable Filter Operation -' at Sampling (min) tlnxysa) FI FE Log Rernoval Fi KFUIL) FE Log Remotal 15 -+ Stable ~ d t eOperation r Stable Filter Operation *' Stable Filer Operation '-0 est. > 4.9 est. > 5.1 est. > 5.0 est -t Stable Filter Opencion 4.6 4-7 4.6 4.6 Stable Film Operation Stable Filter Operation 1-0esr 1.1 esr 2 3 est. 1-0est. ' 4-8est. -15 esr. 4.4 est. 4 2 es. 1/19/00 Stable Filter Operation Stable Filter Opencion During Runoff Stable Filter Openaon During Runoff 42 4.1 4.2 4.2 Table D.6 Microorganism Data fiom Ottawa (Conhued) O~ ar SampIing Expenment No Coagulanr in Jar ' (min) Date loocyStsn'1 FI FE Log Removai FI (cFu~) PL: Log Removai 3.1 32 32 33 3.4 3.6 33 35 Sub-opamal Coagulation ' (52%reducüon to 20 m g n ) 1.1 0.9 1.O 0.8 1.6 est 1.9 est 2.0 est 1 5 est 0.9 1.O 0.9 1.O Sub-opbai Coagulation (49% reduccion to 20 mgK) Hycirmiïc Step ' (159i n c m in flow) 2.8 LXL 2.6 est, 1.4 0.6 0.4 03 61' 15/99 Hychuiîc tep (15% inin flow) 3.1 est 3 3 est. 3.1 est. 23 1.0 1.6 est 2 7 est. 61ZY99 Hydrauiïc Step n m in flow) (25% i 2.9 est 3.L est 2.6 est. 2.4 est 1.4 est. 1.7 est Table D.7 Microsphere Data fkom Ottawa TP Date of ExpeRment S d Dunrion at Smpling t min) 1/19/00 Stable Filter Openrion 15 3/8/00 End-af-Run Yeiïow Microspheres ts p h e r d ) t3 FE Log Removal 4.7 6.1E+5 Blue Microspheres (spheres/L) FI FE l o g Removd 4.1 4-1 da 1.4 I I 0-4 0-7 4-7 4.6 4-1 3.3 37 2- 1 da da da 5.1 5.1 43 3-4 3.7 3.3 da n/a da da 4.8 4.9 3.8 33 23 1.1 09 1-0 0.9 4.7 4.9 4.4 3-3 3.0 n/a 0.3 nia Table D.7 Microsphere Data from Ottawa (Continued) Date 1 W-Of99 3/10100 Ty~e of Experiment h t e Brd-through Seed Duration at S~rimpling (min) 15 Yeliow Microspheres (spheres/L) FI FE toc Remoni 43E+5 6580 Suhptimal Coagulation ( 4 9 8 reduction ro 1C rng/L) Fdtcr iduent microsphexe &a estimateci by enumemting 50 randan fields of view at 400X (Nilcm Labophol -A. Nikon C d a Inc. Toconta Oh?. Blue Microspheres (spheredi.) FE @ Rrmovai FI ocmL=ar* c \ C!I, o o c o g > W. îr. O U'a\OC\ rt cc. cc. j S 1 1 Cc -a Ir. CC. W. W. 3 W. Table D.13 Turbidity and Total Particle Data fiom UW Hydnulic Step 75% Bow incxwse ddmedi? 1 3 1 1199 Hydraulic Step 25% Uow m e duill-medn 1115/00 Hydnulic Step 25% Bow mcrerise tri-media 1/16/00 Hydnulic Step 25% now maease ui-media - * Reducrion chrough entire m r m e n t plant. - 138 0.06 -. + + - +*+F +I oqqq- P ' . W. Ir. GEE"'r-J Cl- - r¶ V: REFERENCES Al-Ani, M.Y., D.W. Hendricks, G.S. Logsdon, and C.P. Hibler. 1986. Removing Giardia Cysts from Low Turbidity Waters by Rapid Rate Filtration. Jour. A m 7 A , 78(5):66. Aldom, J.E. and A.H. Chagla. 1995. Recovery of C p t o s p o n d i ~ ( m Oocysts fkom Water by a Membrane Filter Dissolution Method. Letters in Appl. MicrobioL, 20: 186. Alvarez, M. B. Bellamy, J.B. Rose, C.J. Gibson, and G. Mitskevich. 1999. Cryptospon'dium Removal Using a Pulsating Blanket Clarifier, Microsand Bailasted Clarifier, and Dissolved Air Flotation in Treatment of a H i m y Colored Florida Surface Water: A Pilot Study in Proc. A WKA Water Quality Technology Conference. Denver, CO.: AWWA. Amirtharajah, A. 1985. The Interface Between Filtration and Backwashing. Wat. Res., 19(5):58 1, Amirtharajah, A. 1988. Some Theoretical and Conceptual Views of Filtration. Jour. A M A , 80(12):36. Arnirtharajah, A. 1993. Optimum Backwashing of Filters with Air Scour: A Review. Water Sei. Tech., 27(10): 195. Amirtharajah, A., N. McNelly, G . Page, and J. McLeod. 1991. Optintum BacXwush of Dual Media Filters and GAC Filter Adsurbers with Air Scour. Denver, CO.: AWWARF and AWWA. Amirtharajah, A. and K. M. Mills. 1982. Rapid Mix Design for the Mechanisms of Alum Coagulation. J. A 74(2), 2 10. Amirtharajah, A. and D.P. Wetstein. 1980. Initial Degradation of Effluent Quality During Filtration. Jour. A m A , 72(9)5 18. APHA, A W A , WEF (American Public Health Association, American Water Works Association, and Water Environment Federation). 1995. Standard Methods for the Examination of Water and Wastewater. 19" ed. Washington, D.C.: APHA. Arora, H., M.W. LeChevallier and W.D. Norton. 1995. The Relationship Between Particle Counts and Parasites: Full-Scale Experiences in Proc. A WKA Annual Conference. Denver, CO.: AWWA. ASM News. 1996. Cryptosporidium and Enteric Disease: A Case in Point. ASM News, 62(l): 14. ASTM (Arnerican Society for Testing and Materials). 1993. Standards on Materinls and Environmental Microbiology, 2nd Edition. Philadelphia, PA.: American Society for Testing and Materials. Atherholt, T. B. and L. R. Kom. 1999. K R Protocol - Alternative Treatment of Parasite Sample Data. Jour. A WKA, 9 1(3):95. Barer, M.R. and A.E. Wright. 1990. A Review: Cqptosponndi~trnand Water. Letters Appl. Microbiol., 11(6):271. Baudin, 1. and J.M. Laîné. 1998. Assessment and Optirnization of Clarification Process for Crprosporidium Removal in Proc. A W A Water Qzraliv Technologv Conference. Denver, CO.: AWWA. Bertolucci, G.C., G. Gilli, E. Carraro, D. Giacosa, and M. h p p o . 1998, Influence of Raw Water Storage on Giardia, Cryptospondium, and Nematodes. Wat. Sci. Tech.,37(2):261. Billica, J.A., D. Gregory, and KR. Gertig. 1999. A Pilot Plant Study to Compare the Peifomance of Dual and Mixed Media Declining Rate Filters in Proc. AJVFE-4 Annual Conference. Denver, CO.: AWWA. Black, AP., F.B. Birkner, and J.J. Morgan. 1965. Destabilization of Dilute Clay Suspensions with Labeled Polymers. Jour. A W A , 57(12): 1547-1560. Blom, G. 1989. Probability and Statitics. Theory and Applications. New York, NY.: Sp~ger-Verlag. Box, G. E., W. G. Hunter, and J. S. Hunter. 1978. Statistics for Experimenters. New York, NY.: J. Wiley & Sons Inc. Box, G.E. and G . C. Tiao. 1973. Bayesian Inference in Statistical Analysis. Reading, MA-: Addison Wesley Publishing Company. Bukhari, Z., R. M. McCuin, C.R. Fricker, and J.L. Clancy. 1998. Immunomagnetic Separation of Cryptosponditrm parvum from Source Water Samples of Various Turbidities. Appl. and Envir. Microbiol., 64(11):1495-4499. Bukhari, Z., T. M. Hargy, J. R. Bolton, B. Dussert, and J. L. Clancy. 1999. MediumPressure UV for Oocyst Inactivation. Jour. A W A , 9 1(3):86. Button, N.P. and P.M. Reilly. 2000. Uncertainty in Incident Rates for Trucks Canying Dangerous Goods. Accident Anal. Prev. 32: 797-801. Campbell, A.T., L.J. Robertson, M.R. Snowball and H.V. Smith. 1995. Inactivation of Oocysts of Cqptosporidium parvum by Ultraviolet Irradiation. Wat. Res., 29(1 l):2583. Casella, G. and E.I. George. 1992. Explllining the Gibbs Sarnpler. Amer- Statistician, 46(3): 167-174. Casemore, D.P., S.E. Wright, and R,L.Coop. 1997. Cryptosporidiosis: H w a n and Animal Epidemiology in Fayer, R. (editor) Cqptospon'ditlm and Cryptosporidiosis. Boca Raton, FL: CRC Press, 1-42. Chappeil, C.L., M.M. Marshall, G. Widmer, P.C. Okhuysen, and C.R. Sterling. 1999. Cqptosponditrrn p a m m (Genotype 2 ) Isolates Vary in their Capacity to infect Cultured Enterocytes and Animal ModeIs presented at A W A International Symposiunz on Waterbome Pathogens. Denver, CO.:AWA. Charles, G., J.M. Morgan, M. Macfhee, M. Kim and D.W. Frederickson. 1995. Benchscale Parasite Spiking - An Alternative to Pilot-scale Giardia and Crptosponditinz Spiking Investigations in Proc. A W A Water Quaiity Technology Conference. Denver, CO.: AWWA. Clancy, J.L., Z. Bukhari, T.M. Hargy, J.R. Bolton, B.W. Dussert, and M M . Marshall. 2000. Using UV to Inactivate Cryptosporidium. Jour. A WKA, 92(9):97- 104. Clancy, J.L., Z. Bukhari, R.M. McCuin, 2. Matheson, and C.R. Fricker. 2999. USEPA Method 1622. Jour. A m A , 9 1(9):6O-68. Clancy, J.L., M.M. Marshall, J.E. Dyksen, E. Kawczynski and K. Carns. 1996. Innovative Electrotechnologies for the Inactivation of Crptospoïidium p a m m Oocysts in Water in Proc. A m A Water Qualiy TechnoZogy Conference. Denver, CO.: AWWA. Clancy, J.L., W.D. Gollnitz and Z. Tabib. 1994. Commercial Labs: How Accurate are They? Jour. A WWA, 86(5):89. Cleasby, J.L., M.M. Williamson and E.R. Baumann. 1963. Effect of Filtration Rate Changes on Quality. Jour. A WWA, SS(6): 869. Coallier, J., M. Prévost, B. Barbeau, L. Bodos and R. Desjardins. 1996. Full-Scale Physical and Chernical Rernovals of a Fast Response and Econornic Microbiological Indicator in Proc. A WWA Water QuaZity Technoloby Conference. Denver, CO.: AWWA. Colton, J.F., P. Hillis, and C.S.B. Fitzpatrick. 1996. Filter Backwash and Start-Up Strategïes for Enhanced Particulate Removal. Wat. Res. 30: lO:2SOZ. Conley, W.R., 1972. High Rate Filtration, Jo~rr.A W K A , 64:3:203. Conley, W.R and K.Y. Hsiung. 1969. Test Program for Filter Evaluation at Hanford. Jour. A M A , 52:2:205. Conley, W.R. and R.W. Pitman. 1960. Design and Application of Multimedia Filters. Jour. A FFWA, 6 1:2:97, Current, W.L. 1988. The Biology of Cgptosporidium. ASM N w s , 54(11):605. D'Antonio, R.G., R-E. Winn, J.P. Taylor, T.L. Gustafson, W.L. Current, M.M. Rhodes, G.W. Gary Jr. and R.A. Zajac. 1985. A Waterbome Outbreak of Cryptospondiosis in Normal Hosts. A m . Int. Med.' 103:886-888. Dawson, D.J., M. Maddocks, J. Roberts and J.S. Vider. 1993. Evaluation of Recovery of Cyptosporidiurn parvuin Oocysts Using Membrane Filtration. Letiers in ApplMicrobiol., 17(6):276. DiBemado, L. and J.L. Cleasby. 1980. Declining-Rate Versus Constant-Rate Filtration. Jour. Envir. Engrg., ASCE. 106(EE6):1043. Dowd, S.E. and S.D. Pillai. 1997. A Rapid Viability Assay for Cryptospon'ditcm Oocysts and Giardia Cysts for use in Conjunction with Indirect Fluorescent Antibody Detection. Cm. Jour. Micobiol., 43 :658. Dugan, N.R., KR. Fox, R.J. Milmer, D.A. Lytle, D.J. Williams, C.J. Parrett, C.M. Feld' and J.H. Owens. 1999. Control of Cgvptosporidium Oocysts by Steady-State Conventionai Treatment in Proc. A m A Annual Conference. Denver, CO.: AWWA. Edzwald, J.K. and M.B. Kelley. 1998. Control of Cryptosporidiztnz: From Resenroirs to Clarifiers. Wat, Sci. Technol., 37(2): 1-8. Edzwald, J.K., M.B. Kelley, H.J. Dunn, G.S. Kaminski, and J.P. Malley. 1996. Control of Cryptosporiditm by Coagulation, Flotation, and Filtration in Proc. A W A Water QuaZity TechnoZogy Conference. Denver, CO.: AWWA. Edzwald, J.K., J.E. Tobiason, L.M. Parento, M.B. Kelley, G.S. Kaminski, H.J. Dunn, and P.B.Galant. 1999. Clarification and Filtration Performance for Removals of Giardia and C~ypfosporidiurnin Proc. AWKA Water Qztali- TechnoZogy Conference. Denver, CO.: AWA. Elimelech, M. 1991. Kinetics of Capture of Colloidal Particles in Packed Beds Under Attractive Double Layer Interactions. Jour. Colloid Inter. Sci., 146(2):337. Fayer, R. and B.L.P. Ungar. 1986. Cryptosporidium spp. and Cryptosporidiosis. Microbiol. Rev., 50:458-483. Ferguson, D.W., M.J. McGuire, B. Koch, R.L. Wolfe, and E.M. Aieta 1990. Comparing PEROXONE and Ozone for Controlling Taste and Odor Compounds, Disinfection ByProducts, and Microorganisms. Jour. A W A , 82(4): 18 1. Finch, G.R., E.K. Black, L-L. Gyürek and M. Belosevic. 1993. Ozone Inactivation of Cwptospondit~mp a m m in Demand-Free Phosphate Buffer Deterrnined by In Vitro Excystation and Animal Infectivity. Appl. and Envir. Microbiol-, 59(12):4203. Finch, G.R., E.K. Black, and L.L. Gyürék. 1994. Ozone and Chlorine Inactivation cf Crvptospondiurn in Pvoc. A W A Water Qua& Technologv Conference. Denver, CO.: AWA. Fisher, R. A. 1938. Statistical Methods for Research Workers, 7" Edition. Edinburgh: Oliver and Boyd. Fisher, L.D, and G. Van Belle. 1993. Biostatistics. Sciences. New York: John Wiley and Sons. A Methodology for the Health Fitzpatrick, C.S.B., A.T. Campbell, and C.J. Cable, 1999. The Effect of Filtration Rate Changes on Filtrate Particle Counts in Pmc. AWEA Wuter Qualis, Technoloa Conference. Denver, CO.: A W A . Fox, KR., N.R. Dugan, R.J. MiItner, D.A. Lytle, D.J. Williams, C.J. Parrett, C.M. Feld, J.H. Owens, E.W. Rice, and C.H. Johnson. 1998. Comparative Removal of Cryptosponndiumand Surrogates in a Low Flow Pilot Plant System in Proc. A W A Water Qualis, Technoloa Conference, Denver, CO.: AWWA. Fox, KR., N.R. Dugan, R.J. Mïltner, D.A.Lytle, D.J. Williams, C.J. Parrett, C.M. Feld, J.H. Owens, E.W. Rice, and C.H. Johnson. 1998. Comparative Rernoval of Cwptosponndium and Surrogates in a Low Flow Pilot Plant System in Proc. A @WA Water QuaZip TechnoZogy Conference. Denver?CO.: AWWA. Fox, K.R. and DA. Lytle. 1996. Milwaukee's Crypto Outbreak: Investigation and Recommendations.Jour. A W A , 8 8(9): 8% Frederickson, D.W., Z. Tabib, S. Boutous, W. Culien, R. Dittmer, L. McCarthy, and R. McCuin. 1995. Examination of the Slide Technique vs Membrane Technique to Recover Giardia lamblia Cysts and Cryptospondium pawum Oocysts in Proc. A BWA Water @ml@ TechnoZogy Conference. Denver, CO.: AWWA. Fuller, R.G., J. Steiner, and R. Butcher. 1995. Evaluation of Filter Performance to Remove Giardia and Cryptosporidiurn City of Billings, Montana in Proc. AFVFVA Annual Conference. Denver, CO-: A W A . Geman, S. and D. Gernan. 1984. Stochastic Relaxation, Gibbs Distributions, and the Bayesian Restoration of Images. IEEE Tram. Pam. Anal. Bach. Intell., 6: 72 1. Ginn, T.M.,A. Amirtharajah, and P.R. Karr. 1992. Effects of Particle Detachment in Granular Media Filtration. Jour. A M A , 84(2):66. Graham, N.J.D.1988. Filter Pore Flocculation as a Mechanism in Rapid Filtration. Wat. Res-, 22(10): l229. Gytkék, L.L., L.R.J. Liyanage, M. Belosevic, and G.R. Finch. 1996. Disinfection of Cn7>tospondum parvum Using Single and Sequential Application of Ozone and Chlorine Species in Proc. A W A Water Qzrality Technology Conference. Denver, CO.: AWWA. Haas, C. N. and J. B. Rose. 1996- Distribution of Ctyptosponditm Oocysts in a Water Supply. Wat. Res., 30(10):2251. Habibian, M.T. and C R O'Melia. 1975. Particles, Polymers, and Performance in Filtration. Jour. Envir. Eneoyg., ASCE, 101~ 5 6 7 . Hali, T., J. Pressdee, R. Gregory, and K. Murray. 1995. Cwptosponditim Removal During Water Treatment Using Dissolved Air Flotation. Water Sci Tech. 3 l ( 3 4 ) : 125. Hansen, J.S. and J.E. Ongerth. 1991. Effects of Watershed Characteristics on the Concentration of Cryptospor5dium Oocysts in River Water. Appl. and Envir. Microbiol., 57(10):2790. Hargesheimer, E.E., C.M. Lewis and C.M. Yentsch. 1992. Evalrration of Particle Counting as a Measure of Treament Plant Performance. h e r i c a n Water Works Association Research Foundation and AWWA, Denver, CO. Hargy, T.M., T.P. Clancy and J.L. Clancy. 1996. Field Evaluation of the ICR Protozoan Method in Proc. A W A Water QualiS, Technology Conference. Denver, CO.: A W A . Hayes, E.B., T.D. Matte, T.R. O'Brien, T.W. McKinley, G.S. Logsdon, J.B. Rose, B.L.P. Ungar, D.M. Word, P.F. Pinslq, M.L. Cummlligs, M.A. Wilson, E.G. Long, ES. Hurwitz and D.D. Juranek. 1989. Large Community Outbreak of Cryptosporidiosis Due to Contamination of a Filtered Water Supply. Nav England Jour. Med., 320: 1372. Hilmoe, D.J. and J.L. Cleasby. 1 986. Cornparhg Constant-Rate and Declining-Rate Direct Filtration of a Surface Water. Jour. -4 WtYA, 78: 12:26. Hom, J.B., D.W. Hendricks, J.M. Scanlan, L.T. Rozelle, and W.C. Trnka. 1988. Removing Giardia Cysts and Other Particles from Low Turbidity Waters Using DualStage Filtration. Jozrr. A W A , 80:2:68. Hubbe, M.A. 1984. Theory of Detachment of Colloidal Particles fiom Flat Surfaces Exposed to Flow. Colloids and Surfaces, Amsterdam. The Netherlands, 12(1/2), 151-178. Huck, P.M., B. M. Cofley, C. R. O'Melïa, and M. B. Emelko. 2001. Filter Operation Effects on Pathogen Passage. Denver, CO.: AWWA and AWWARF. Hudson, H.E. 1959. DechineRate Filtration, Jour. A WKA, 5 1:11:1455. Ives, K.J. 1982. Fundamentals of Filtration in Proc. Symposium on Water FiIfrution, European Federation of Chernical Engineers, Antwerp, Belgium. Ives, KJ. and 1. Sholji. 1965- Research on Variables Affecting Filtration. Jour. Envir. Engrg. - ASCE, 9 1:SA4: 1. Jackson, W., T. Straub, S. Monten, E. Vinluan, D. Boening, and M. Feige. 1997. Determination of Analyst and Method Variabilities Associated with Giardia and CgptosponMdittmin Spiked Quality Control Samples in Proc. A WFKA Wuter Qunlity Technology Conference. Denver, CO.: AWWA. Jakubowski, W., S. Boutros, W. Faber, R. Fayer, W. Ghiorse, M. LeChevallier, J. Rose, S. Schaub, A. Singh, and M. Stewart. 1996. Environmental Methods for Cwtosporidium. Jour. A W A , 88(9): 107. Jeffkeys, H. 1961. Theory of Probability, Third Edition. Oxford, U.K.: Oxford University Press. Joseph, C., G. Hamilton, M. O'Connor, S. Nicholas, R. Marshall, R. Stanwell-Smith, E. Ndawula, D. Casemore, P. Gallagher and P. Harnett. 1991. Cryptosporidiosis in the Isle of Thanet; an Outbreak Associated with Local Drinking Water. Epidemiol. Infect. IO7:509-519. Kelley, M.B., P.K. Warrier, J.K. Brokaw, K-L. Barrett and S J . Komisar. 1995. A Study of Two U.S. A.rmy Installation Drinking Water Sources and Treatment Systems for the Removal of Giardia and Cryptosporïdium in Proc. A W A Water Quali'y Technology Conference. Denver, CO.: A W A . Kfir, R., C. Hilner, M. du Preez and B. Bateman. 1995. Studies on the Revalence of Giardia and Crptosporidium Oocysts in South Afican Water. Water Sci. Technol., 3 1(5-6):435, Konch, D.G., J.R. Mead, M.S. Madore, N.A. Sinclair and C.R. Sterling. 1990. Effects of Ozone, Chlorine Dioxide, Chiorine, and Monochloramine on Cryptosporidiurn parvzirn Oocyst Viability. Appl. and Envir. Microbiol., 56(5) :1423. Lawler, D.F., I.L. Darby and R S . Cushing. 1995. Complexities of Filtration Dynamics: Particles, Flocs, and (Perhaps) Cysts in Proc. A WWA Annrial Conference. Denver, CO.: AWWA. LeChevallier, M.W. and W.D. Norton. 1992. Examining Relationships Between ParticIe Counts and Giardia, Cryptosporïdium, and Turbidity. Jour. A W A , 84( l2):54. LeChevallier, M.W, and W.D- Norton. 1995. Giardia and Cwptospon'diztrn in Raw and Finished Water. Jour. A WKA, 87(9):54LeChevallier, M.W., W.D Norion and R.G. Lee. 1991a. Giardia and Cryptospon'dium spp in Filtered Water Supplies. AppZ- and Envir. Microbiol., 5 7(9):26 17. LeChevallier, M.W., W.D. Norton and R.G. Lee. 1991b. Occurrence of Giardia and Cwptospon'dium spp in Surface Water Supplies. Appl. and Envir- Microbiol., 57(9):26 10. LeChevallier, M-W., W.D. Norton znd J-B. Rose. 1991c. Giardia and Cwptosporidium in Water Supplies. Denver, CO.: AWWARF and AWWA. LeChevallier, M.W., W.D. Norton, J.E. Siegel and M. Abbaszadegan. 1995. Evaluation of the Immunofluorescence Procedure for the Detection of Giardia Cysts and Cryptosporidittm Oocysts in Water. Appl. Environ. Micro biol., 6 1:690. LeChevallier, M.W., T.M. Trok, M.O. Burns and R.G. Lee, 1990. Cornparison of the Zinc Sulfate and Immunofluorescence Techniques for Detecting Giardia and Cryptosporidium. Jour. A W A , 82(9):54. Leland, D., J. McAnulty, W. Keene and G.Stevens. 1993. A Cryptosporidiosis Outbreak in a Filtered-Water Supply. Jour. AWKA, 85(6):34. Levich, V.G. 1962. Physico-Chernical Hydrodmamics, Eaglewood Cliffs, NJ.: Prentic- HalL Li, S.Y., J.A. Goodrich, J.H. Owens, G.E. Willeke, F.W. Schaefer and R.M. Clark. 1997. Reliability of Surrogates for Detennining Cryptosponditrm Removal. Jour. A W A , 89(5):90. Lisle, J.T. and J.B. Rose. 1995. Ctyptospondi2m Contamination of Water in the USA and UK: a %ni Review. Jour. Water SRT-Aqua. 44(3): 103. Logsdon, G.S., 1999. Effective Management and Operation of Coagulation and Filtration in Proc. of the Seventh international Conference of the Israel Sociey for EcoZoa and Environmental Quality Sciences, Jerusdem, Israel. Logsdon, G.S., J.M. Syrnons, R.L. Hoye Jr. and M.M. Arozarena. 1981a. Alternate Filtration Methods for Removal of Giardia Cysts and Cyst Models. Jour. A W A , 73(2): 11 1. Logsdon, G.S., J.M. Sxymons, and T.J. Sorg. 1981b. Monitoring Water Filters for Asbestos Removal. Jour. En vir. Enbgrg., ASCE, 1O7(6): 139%1 3 1 5. Lytle, D.A. and KR. Fox. 1991. Particle Counting and Zeta Potential Measurements for Optimizing Filtration Treatment Performance in Proc. A W A Water Quality Technology Conference. Denver, CO.: A W A Lytle, D . k , KR. Fox, E.W. Rice, J. Owens, and C.H. Johnson. 1996. The Use of Aerobic Spore Forming Bacteria for Evaluating Drinking Water Treatment Performance in Proc. A W AAnnztal Conference. Denver, CO.: AWWA. MacKenzie, W.R. N.J. Hoxie, M.E. Proctor, M.S. Gradus, K.A. Blair, D.E. Peterson, I.J. Kazmierczak, D.G. Addiss, KR. Fox, J.B. Rose and J.P. Davis. 1994. A Massive Outbreak in Milwaukee of C~prosporfdiurnInfection Transmitted ïhrough the Public Water Supply. New E&. JOUEMed-, 331:161-167. Maguire, H. C., E. Holmes, J. Hollyer, J.E.M. Strangeways, P. Foster, R.E. Holliman, and R. Stanwell-Smith. 1995. An Outbreak of Cryptospondiosis in South London: What Value the p Value? Epidemioi. In$, 115: 279-287. McDowell-Boyer, L.M., 1992. Chernical Mobiiization of Micron-Sized Particles in Saturated Porous Media Under Steady Flow Conditons. Environ. Sci. Technol., 26:3:586. McLauchlùi, J., S. Pedraza-Diaz, C . Amar-Hoetzeneder, and G.L. Nichols. 1999. Genetic Characterization of Cryptospor+diurn Strains fiom 2 18 Patients with Diarrhea Diagnosed as Having Sporadic Cryptosporidiosis. Jour. Clin. Microbiol., 37: 10:3 153-3 158. James M. Montgomery, Consulting Engineers, Inc. 1985. Watev Treament Principles and Design. New York: John Wiley & Sons. Moore, A.C., B.L. Herwaldt, G.F. Craun, R.L. Calderon, A.K. Highsmith, and D.D. Juranek. 1994. Waterborne Disease in the United States, 1991 and 1992. Jour. AFKKA, 86(2):87. Moran, D.C., M C Moran, R.S. Cushing and D.F. Lawler. 1993a. Particle Behavior in Deep-Bed Filtration: Part 1 - Ripening and Breakthrough. Jour. A W A , 185: 12:69. Moran, MC., D.C. Moran, R.S. Cushing and D.F. Lawler. 1993b. Particle Behavior in Deep-Bed Filtration: Part 2 - Particle Detachment. Jour. A W A , l85(12):82. Musial, C.E., M.J. Arrowood, C.R. Sterling, and C.P. Gerba. 1987. Detection of Cryprosporidium in Water by Using Polypropylene Cartridge Filters. Appl. and Envir. Microbiol., 53(4):687. Nahrstedt, A. and R. Gimbel. 1996. A Statistical Method for Deterrnining the Reliability of the Analytical Results in the Detection of Cryptosporidil~mand Giardia in Water. Jour. Water SRT - Aqua, 45(3): 101. Nieminski, E.C. 1994. Removing Giardia and Cryptosporidiurn Cysts Removd Through Direct Filtration and Conventional Treatment in Proc- AWKA Annual Conference. Denver, CO.: AWWA. Nieminski, E.C. and W.D. Bellamy- 1998. Application of Pathogen Surrogates to h p r o v e Treatment Plant Performance in Proc. A W A Water Quality Technology Conference. Denver, CO.: AWWA. Nieminski, E-C. and J.E. Ongerth. 1995- Removing Giardia and Ctyptospor-iditrm by Conventionai Treaûnent and Direct Filtration. Jour. A W A , 87(9):96. Nieminski, E.C., F.W. Schaefer and J.E. Ongerth. 1995. Comparison of two Methods for Detection of Giardia Cysts and CptospondrUrn Oocysts in Water. Appl. and E n v k Microbiol., 6 L (5):1714. Nocito-Gobel, J. and J.E. Tobiason. 1995. Effects of Ionic Strength on Colloid h Chernical Socieîy National Meeting and Deposition and Release in Proc. 2 ~ Arne~ican Exposition- Washington, û.C .: ACS. O'Melia, C.R. 1985. Particles, Pretreatment, and Performance in Water Filtration. JourEnvir. Engrg., A X E , 111(6):874. OfMelia,C.R. and W. Aii. 1978. The Role of Retained Particles in Deep Filtration. Progr in Water Technol., London, U.K., 10:(5/6): 167-182. O'Melia, C.R. and D.K. Crapps. 1964. Some Chemical Aspects of Rapid Sand Filtration. Jour- A WKA, 56: 10: 1326. OfMelia, C.R. and W. Sturnm. 1967. Theory of Water Filtration. Jour. A W A , 59(11):1393. Ong, C., W. Moorehead, A. Ross and J. Isaac-Renton. 1996. Studies of Giardia Spp. and Cwptosponndium Spp in Two Adjacent Watersheds. Appl- and Envir. Microbiol., 62(8):2798. Ongerth, J.E. and J.P. Pecoraro. 1995. Removing Ctyptosporidiurn Using Multimedia Filters. Jour. AFVFKA, 87(12):83. Ongerth, J. E. and J. P. Pecoraro. 1996. Electrophoretic Mobility of Cryptosporidiurn Oocysts and Giardia Cysts. J. Environ. Eqrg., 122(3), 228. Ongerth, J.E., J. Riggs, and J . Crook. 1989. A Study of Water Treatment Practices for the Removal of Giardia lamblia Cysts. Denver, CO.: AWWARF and AWWA. Ongerth, J-E. and H.H.Stibbs. 1957. Identification of Cryptosporidiurn Oocysts in River Water. Appl. and Envir. Microbiol-, 53(4):672. Owens, J.H., R.J. Miltner, F.W. Schaefer and E.W. Rice. 1994. Pilot Scale Inactivation of CrptospoM'dium and Giardia in Proc. A WFKA Water QzraZity TechnoZogv Conference. Denver, CO.: A W A . Parkhurst, D. F. and D. A. Stem. 1998. Deteminhg Average Concentrations of Cgptospondizrm and Other Pathogens in Water. Environ. Sci. Technol., 323424. Patania, N.L., J.G. Jacangelo, L. Cwnmings, A. Wilczak, K. Riley, and J. Oppenheimer. 1995. Optirnizution of Filtration for Cyst Remuval. Denver, CO.: AWWARF and AWA. Patel, S., S. Pedraza-Diaz, J. McLauchlin, and D.P. Casemore. 1998. Molecular Characterisation of Cptosporidium pawum h m Two Large Suspected Waterborne Outbreaks. Cornnzun. Dis. Public Health, 1:23 1 -23 3. Payatakes, A C , C. Tien, and R.M. Tunan. 1974. Trajectory Calculation of Particle Deposition in Deep Bed Filtration: Part II. Case Study of the Effect of Dimensiodess Groups and Cornparison with Experimental Data. Jour. &CE, 20:900. Peng, MM., L. Xiao, A.R. Freeman, M.J. Arrowood, A.A. Escalante, A.C. Weltman, C.S.L. Ong, W.R. MacKenzie, A.A. Lal, and C.B. Beard. 1997. Genetic Polymorphism Among Cryptosporidium parvurn Isolates: Evidence of Two Distinct Human Transmission Cycles. Emerg. hfect. Dis.,3 (4):567-572. Plummer, J.D., J.K. Edzwald, and M.B. Kelley. 1995- Removing Cwtospondriztrn by Dissolved-Air Flotation. Jour. A W A , 87(9):85. Press, W.H., B.P. Flannery, S.A. Teukolsky, and W.T. Vetterling. 1989. Numerical Recipes - B e Art of Scientific Computing (Fortran Version). Cambridge, UK: Cambridge University Press. Rajagopalan, R. and C. Tien. 1976. Trajectory Analysis of Deep-Bed Filtration with the Sphere-in-Celi Porous Media Model. Jozrr- d C h E , 22:523 -533. Raveendran, P. and A. Amirtharajah. 1995. Role of Short-Range Forces in Particle Detachment During Filter Backwashing. Jour. Envir. Engrg., ASCE, 12l(12): 860. Rice, E.W., KR. Fow, R-J. Miltner, D.A. Lytle and C.H. Johnson. 1996. Evaluating Plant Performance with Endospores. Jour. A WFEA, 88(9): 122. Richardson, A.J., R.A. Frakenberg, A.C. Buck, J.B. Selkon, J.S. Colborne, J.W. Parsons and R.T. Mayon-White. 1991. An Outbreak of Cryptosporidiosis in Swindon and Oxfordshire. Epidemiol. Infect., 1O7:485-495. Rimer, A. E. 1968. Filtration Through a Tri-Media Filter. Jozrr. Envir. Enqg-, ASCE. 94(3):52 1. Robertson, L. J., A. T. Campbell, and H. V. Smith. 1993. In vifro Excystation of Cgptosporidizrmpamrnr. ParasitoZ., 106:13-29. Robertson, L. J., A. T. Campbell, and H. V. Smith. 1992. Sunival of Oocysts of Cryptosporidium parvum Under Various Environmental Pressures. Appl. Environ. Microbiol., 58:3494-3 500Rodgers, M.R-, D.J. Flanigan and W. Jakubowski. 1995. Identification of Algae L5rhich Interfere with the Detection of Giardia Cysts and Cwptospon'diurn Oocysts and a Mehtod for AUeviatÏng This Interference. Appl. and Erzvir. Microbiul., 6 l(lO):3759. Roebeck, G.G., K.A. Dostal and R.L. Woodward. 1964. Studies of Modifications in Water Filtration. Jour. A WWA, 56: 198. Roefer, P., J.T. Monscvitz and D.J. Rexing. 1995. The Las Vegas Cryptosporidizrm Experience in Proc. A W A Water Qualiw Technology Conference. Denver, CO.: AWWA. Roefer, P., J.T. Monscvitz and D.J. Rexing- 1996. The Las Vegas Cryptosporidiosis Outbreak. Jour.A M A , 88(9):95. Rose, J.B. 1988. Occurrence and Significance of Ctptospondizrm in Water. Jour. A W A , 80(2):S. Rossomando, E.F., M.A. Hesla, E. Olson, F. Thomas, D.C. Gilbert and J.J. Herlihy. 1995. "Continuous" Immunomagnetic Capture of Crvptosporidium and Giardia from Surface Water in Proc. A WFKA Water Qua& TechnoZogv Conference. Denver, CO.: AWWA. Ryan, K-J., J.J. Champoux, W.L. Drew, S. Fallcow, P.C. Neidhardt, J.J. Plorde, and C.G. Ray, eds. 1994. Shems Medicd MicrobioZogy. An Introduction to hfectiozis Diseases. 3rd ed. S t d o r d , Corn.: Appleton and Lange. Ryan, J.N. and P.M. Gschwend. 1994. Effects of Ionic Strength and Flow Rate on Colloid Release: Relating Kinetics to Interface Potential Energy. Jour. CoZZoid and In tedace Sci., 164( 1):21. Schaechter, M.,G. Medfoff, and B.I. Eisenstein. 1993. Microbial Disease. Balitomore, MD.: Williams and Wilkins. Scott, K.N., L.S. Palencia, R.P. Merlo, R.S. Yates, S. Liang, and R. DeLeon. 1997. Evaluation of Cryptospondium and Surrogate Removal Throughout a Fuil-Scale Treatment Plant. International Symposium on Waterbome Crptospon-diur in Proc. A W A Intemationsal Symposium on Waterbome Pathogens. Denver, CO. : A W A . S h m a , M.M., H, Chamour,, D S H , Sita Rama Sanna and R.S. Schechter. 1992. Factors Controlling the Hydrodynamic Detachrnent of Particles from Surfaces. Jour. Colloid and Intevface Sci., 149(1): 121. Shepherd, K.M. and A.P. Wyn-Jones. 1996. An Evaluation of Methods for Simultaneous Detection of Cryptospondium Oocysts and Giardia Cysts fiom Water. AppL and Envir. Microbiol., 62(4j: 23 17. Smith, A.F.M. and G. O. Roberts. 1993- Bayesian Computation via the Gibbs Sampler and Related Markov Chain Monte Car10 Methods. J. R. Statistic. Soc., 55(1): 3-23. Smith, H.V., W.J. Patterson, R. Hardie, L.A. Greene, C. Benton, W. Tulloch, R.A. Gilrnour, RW.A. Girdwood, J.C.M. Sharpe, and G.I. Forbes. 1989. An Outbreak of Waterborne Cstptosporidiosis Caused by Post-Treatment Contamination. Epidemiol. Infect. 103(3):703-705. Smith, H.V., A. Gnmason, C . Benton and J. Parker. 1991. The Occurrence of Cryprosporidium Spp. Oocysts in Scottish Waters, and the Development of a Flurogenic Viability Assay for Individual Cqpfosporidium Spp. Oocysts. Water Sci. Tech., 24(2): 172. Solo-Gabriele, H. and S. Neumeister. 1996. US Outbreaks of Cryptosporidiosis. Jotrr. A WKA, 88(9):76. Spielman, L.A. 1978. Hydrodynamic Aspects of Flocculation in n e Scientzifc Basis for Flocculation. Edited by K.J. Ives. Leyden: Sitjhoff-Noordhoff. Straub, T., W. Jackson, D. Boening and J. Clancy. 1996. Evaluation of Five Methods to Determine the Spike Dose of Giardia and C g p t o ~ p o ~ d i u rin n Proc. AKGKA Watev QzraZity TechnoZogy Conference. Denver, CO.: AWWA. Sulaiman, LM., L. Xiao, C- Yang, L. Escalante, A. Moore, C.B. Beard, M.J. Anowood, and A.A. Lal. 1998. Differentiating h a n fiom Animal Isolates of C~yptospo~diztrn pawurn. Emerg. Infect. Dis., 4:4:68 1-685. Swaim, P.D., M.S. Heath, N.L. Patania, W.B. Wells, and R.R- Trussell. 1996. Hi&-Rate Direct Filtration for Giardia and C ~ t o s p o n ' d i u mRemovai in Proc. Annzrul Conference. Denver, CO.: AWWA. Swertfeger, J., D.H. Metz, J. DeMarco, A. Braghetta, and J.G. Jacangelo. 1998. Effect of Filter Media on Cyst and Oocyst Removal. Joztr- A m A 9 1:9:90. Tate, C.H. and R.R. Trusseli. 1978. The Use of Particle Counting in Developing Plant Design Criteria. Jour. A WKA, 70(12):69 1-698. Tien, C. and A.C. Payatakes. 1979. Advances in Deep Bed Filtration. Am. Inst. of Chernical Engrs. Jour., 25(5):737-759. Timms, S., J.S. Slade, and C.R. Fncker. 1995. Removai of Cptospondilm by Slow Sand Filtration. Water Sci Tech., 3 l(5-6): 8 1. Tobiason, LE.and C.R. O'Melia. 1988. Physicochemical Aspects of Particle Removal in Depth Filtration. Jour. A W A , 80(12):54-64. Trusseil, R., A.R. Trussell, J.S. Lang, and C.H. Tate. 1980. Recent Developments in Filtration System Design. Jour. AW'KA, 72: 12:705. Tuepker, J.L. and C.A. Bauescher, Jr. 1968. Operation and Maintenance of Rapid Sand and Mixed Media Filters in a Lime Softening Plant. Jottr. A W A , 60: 1377. U.S. Environmental Protection Agency. 1996. ICR Microbiology Laboratory Manual. Office of Research and Deveiopment, Washington, D.C. US. Environmental Protection Agency. 1989. National Primary Dnnking Water Regulations: Filtration and Disinfection; Turbidity; Giardia lamblia, Viruses, Legioneh, and Heterotrophic Bacteria. Fed. Reg., 54(124):27486-2754 1. U.S . Environmental Protection Agency. 1998. National Primary Drinking Water Regulations: Interirn Enhanced Surface Water Treatment Rule; Final Rule. Fed. Reg-, 63(24 l):69478-6952 1. U S . Environmental Protection Agency. 1991. Optimizing Water Treatment Plant Performance Using the Composite Correction Program. Office of Research and Development: Cincinnati, Ohio. US. Environmental Protection Agency. 1999a. Method 1622: Cwprosponndiumin Water by Filtration/IMS/FA. EPA 82 1-R-99-001. Office of Water: Washington, D.C. U S . Environmental Protection Agency. 1999b. Method 1623: Cqprospondium and Giardia in Water by Filtration/IMS/FA. EPA 821-R-99-006. Office of Water: Washington, D.C. van Breemen, L.W.C., H.A.M. Ketelaars, W. Hoogenboezem, and G. Medema. 1998. Storage Reservoirs-A First B d e r for Pathogenic Micro-Organisms in the Netherlands. Wat. Sci. Tech., 37(2):253. Vesey, G. and J. Slade. 1991. Isolation and Identification of Cryptospondium fiom U'ater. Water Sci. Tech., 24(2): 165. Vesey, G., J. Slade, and C.R. Fricker. 1991. Taking the Eye Strain Out of Environmental Cvptosporidiurn Analysis. Letters Appl. Microbiol. 13(2): 62. Vesey, G., J.S. Slade, M. Byrne, K. Shepherd and C.R.Fricker. 1993a. A New Method for the Concentration of Cryptospondium Oocysts fiom Water. Jour. Appl. Bacteriol., 75(1):82. Vesey, G., J.S. Slade, M. Byme, K. Shepherd, P.J. Dennis and C.R. Fricker. 1993b. Routine Monitoring of Cryptosporidicirn Oocysts in Water Using Flow C ytometry. Jour. Appl. Bacteriol., 75(1):87. Weiker, R., R. Porter, W.B. Pztt, M.R. Provart, M. Schwartz and T. Kauri. 1994. Cryptosporidiosis Outbreak in Kitchener-Waterloo: Identification, Assessrnent and Future Prevention in Proc. of the Sirth National Conference on DrinLing Water, Victoria, B.C. West, T., P. Daniel, P. Meyerhofer, A. DeGraca, S. Leonard, and C. Gerba. 1994. Evaluation of Cnrptosporidiurn Removd Through High-Rate Filtration in Proc. A W A Anmal Conference. Denver, CO.: AWWA. Whitmore, T.N. and E.G. Carrington. 1993. Cornparison of Methods for Recovery of C-ptosporidiurn from Water. Water Scî. Tech., 27(3-4):69. Widmer, G., S. Tzipori, C.I. Fichtenbaum, and J.K. Griffiths. 1998. Genotypic and Phenotypic Characterization of Cryptosporidium parvzim Isolates fiom People with Aids. Jour. Infect. Dis., 178:834-840. Wiesner, M.R., J. Hackney, S. Sandeep, J.G. Jacangelo and J.M. Laine. 1991. Cost Estirnates for Membrane Filtration and Conventional Treatment. Jour. A ItTtt.'A,86(12):33. Yao, K.M. 1968. Influence of Suspended Particle Size on the Transport Aspect of Water Filration. Ph.D. diss., Univ. of North Carolina, Chape1 Hiil, NC. Yao, KIM., M.T. Habibian and C.R. OrMeLia. 1971. Water and Wastewater Filtration: Concepts and Applications. Environ. Sci Technol., 5: 1105-1112. Yates, R.S. 1997. Personal communication. Yates, R.S., J.F. Green, S. Liang, R.P. Merlo, and R. DeLeon. 1997. Optimizing Direct Filtration Processes for Crptosporidium Removal in Proc. AWWA Water Qualiw Technology Conference. Denver, CO.: AWWA. Yohe, T.L., J. Heichel, B. Stromberg, T.M. Getting, L. Zukus, C. Ball. 1999. The Effect of Low Uniformity Coefficient Anthracite on Dual Media Filtration in Proc. AWWA Ann zral Conference. Denver, CO.: AWWA. Zhu, H.,D.W. Smith, H. Zhou, and S.J. Stanley. 1996. Improving Removal of Turbidity Causing Materials Using Polymers as a Filter Aid. Wat. Res., 30: 1:103.