Journal of Hydrology 380 (2010) 180–190 Contents lists available at ScienceDirect Journal of Hydrology journal homepage: www.elsevier.com/locate/jhydrol Hydrological response of a small mediterranean agricultural catchment Joan Estrany a,*, Celso Garcia a, Ramon J. Batalla b,c a Department of Earth Sciences, University of the Balearic Islands, E-07122 Palma de Mallorca, Spain Department of Environmental and Soil Sciences, University of Lleida, E-25918 Lleida, Catalonia, Spain c Forestry and Technology Centre of Catalonia, E-25240 Solsona, Catalonia, Spain b a r t i c l e i n f o Article history: Received 12 May 2009 Received in revised form 27 September 2009 Accepted 31 October 2009 This manuscript was handled by K. Georgakakos, Editor-in-Chief Keywords: Baseflow Recession curves Rainfall–runoff Under-drained systems Mediterranean Mallorca s u m m a r y Field studies have recently shown that quickflow may be produced in Mediterranean environments as the result of saturation mechanisms, considering that the dominant runoff processes may change throughout the year due to the significant seasonality. One of the most representative agriculture elements of Mediterranean-climate regions is rainfed herbaceous crops. The aim of this study is to achieve a better understanding of hydrological responses and the role played by under-drained systems in runoff variations under such climatic and land use conditions. Field measurements were conducted in Can Revull, a small catchment (1.03 km2) in the island of Mallorca. The mean annual precipitation and potential evapotranspiration are 517 mm and 1010 mm, respectively. The hydrological regime is intermittent with a mean annual discharge of 4 l s1. Based on 3 years of field data (July 2004–June 2007), this study presents a soil water balance of the catchment and an evaluation of the variability of runoff and its components on an annual and seasonal time scale. In addition, the response of runoff components at event-scale was analysed. At the annual and seasonal time scales, it was only possible to observe a succession of three different hydrological periods throughout the year conditioned by the evapotranspiration which also plays an important role in the baseflow response. Precisely, baseflow was predominant, presumably enhanced by the under-drained system. Furthermore, quickflow response was dominated by saturation mechanisms, whereas rainfall excess mechanisms were limited during study period to dry seasons when baseflow was not present and discharge values therefore were lower. Ó 2009 Elsevier B.V. All rights reserved. Introduction Processes involved in the runoff generation in Mediterranean areas are many and diverse, depending on the catchment characteristics but also as a function of the antecedent conditions and the characteristics of the rainfall episode. The traditional perception is that the primary generation mechanism of quickflow is the rainfall falling at intensity greater than the local infiltration capacity of the soil (Beven, 2002). However, field studies have recently shown that, in these environments, quickflow may be also produced as the result of saturation mechanisms (Ceballos and Schnabel, 1998; Latron and Gallart, 2008; Gallart et al., 2008). Furthermore, it should also be considered that Mediterranean rivers show significant seasonality, so that the dominant processes, whatever they are, may change throughout the year (Kirkby, 2005). Extensive rainfed herbaceous crops are one of the most representative agricultural elements of Mediterranean-climate regions. In Spain, this type of crop occupies more than 40% of agricultural land (INE, 2003), mainly because climate is formed by irregular * Corresponding author. Tel.: +34 971172793; fax: +34 971172309. E-mail address: joan.estrany@uib.cat (J. Estrany). 0022-1694/$ - see front matter Ó 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.jhydrol.2009.10.035 rainfall that does not sustain intensive agricultural activity. However, very little attention has been paid to the hydrological behaviour of such type of landscapes despite their environmental and socio-economic importance (Latorre et al., 2001). Considering the significant seasonality, in some Mediterranean areas rain mainly falls in winter, followed by a lesser amount during the warmer spring period. Therefore, rainfed herbaceous crops are preferably located in clay soils since they provide significant water storage during the spring period when herbaceous crops have the highest water needs. However, during rainy winters these fine soil textures become an important constricting factor for herbaceous crops, particularly cereals (Porta et al., 1999). Thus, the installation of an under-drained system such as subsurface-tile drainage becomes essential, as it is the only way to drain fine texture soils created in flat areas during winters. It is an agricultural water management practice in regions with typically seasonal perched water tables, fine soil textures and concave slopes (Stone and Krishnappan, 1997). Subsurface or ‘tile’ drainage removes excess water from the soil, which prevents air and oxygen from getting to the plant roots. Since flow hydrographs typically provide the only available information, the basis of the hydrological processes active in a J. Estrany et al. / Journal of Hydrology 380 (2010) 180–190 catchment cannot be identified and properly interpreted, as different combinations of processes may lead to similar hydrographs. Despite these limitations (Hewlett, 1982), the study of hydrograph and associated runoff volumes (i.e. essentially, quickflow volume, resulting from any hydrograph separation technique) has been the classic approach in catchment hydrology for several decades; this allowed finding relationships between the response at the outlet and other hydrological information available in the catchment (Woodruff and Hewlett, 1970). A research project is in progress in the Na Borges basin since 2004, an agricultural lowland river located in the island of Mallorca, aiming at establishing a comprehensive water and sediment budget (Estrany and Garcia, 2005). In this context, Can Revull is a representative area selected to study hydrological and sedimentary dynamics and contribution of headwater catchments to the large Na Borges basin. This paper focuses on the analysis of the hydrological response of the small Can Revull creek, a Mediterranean catchment on rainfed herbaceous crop supported by an artificial underdrained system. First, an analysis of annual and seasonal rainfall, runoff, Evapotranspiration (ET) and Potential Evapotranspiration (PET) was carried out by means of a monthly soil water balance. Second, runoff components and several related indices were derived to evaluate frequency and variability of the hydrological response at the annual and seasonal time-scale. Finally, dynamics of baseflow and quickflow at event-scale were examined by applying match-strip and stepwise multiple regression methods, respectively. 181 Study area The Can Revull drains an area of 1.03 km2 (i.e. at the outlet gauging station, see Fig. 1) flowing into the Torrentó de Boscana (7.9 km2), a headwater tributary of the Na Borges River; this is a lowland agricultural basin (319 km2) located in the north-eastern part of Mallorca, Spain (Fig. 1). The geology of Can Revull is characterised by a structurally gentle alpine relief in the Central Ranges of the island composed of Miocene calcarenites, which rest discordantly over a deformed Mesozoic–Cenozoic substratum. The maximum altitude of the catchment is 144 m a.s.l. The channel length is 2.4 km and the average channel slope is 4.7% (10% in the first 400 m, and 2% downstream). The climate in the catchment area can be classified as sub-dry Mediterranean (Thornthwaite, 1948), with a mean temperature of 16.5 °C and a mean annual rainfall of 517 mm (1974–2006, data from the Boscana Nou rainfall station, located 1.5 km from the Can Revull gauging station. Autumn is the rainiest season, followed by winter, spring and summer with an inter-annual variability of 23%. The main characteristic of the rainfall is its torrential behaviour, especially during late summer and autumn when the daily intensity can reach 100 mm (i.e. 25-year recurrence interval). During the winter, mid-latitude westernlies and associated frontal systems bring general rainfall (Romero et al., 1998). The flow regime is intermittent; flowing normally from November–December to March–April. Average daily discharge was around 4 l s1 during the study period 2004–2007. Mean annual PET was 1010 mm, estimated from the Thornthwaite method (1948). Fig. 1. Location of the Can Revull catchment. Inset maps: (a) Location of Mallorca and the Na Borges basin; (b) the Boscana sub-basin within the Na Borges basin; (c) the Can Revull small catchment within the Boscana basin. The photograph shows an upstream view of the measuring section. 182 J. Estrany et al. / Journal of Hydrology 380 (2010) 180–190 The main bedrock of the catchment is limestone and this fact is reflected in the soil type. Using the Soil Taxonomy System, they can be classified as Entisols at the basin headwater and Alfisols in the downstream flat areas. Therefore, at these flat areas soils are deep and well developed on Quaternary alluvial sediments supported by an impervious lower-middle Miocene (Burdigalian) layer, and are characterized by a ‘‘silty-clay-loam” soil texture (Diaz Palmer et al., 2006). These impervious sediments do not allow the percolation into deep aquifers. Uniquely, a shallow and unconfined aquifer is built in the soils characterized by a high field capacity >40% (Diaz Palmer et al., 2006), where the installation of an underdrained system is the only way to drain these areas. As a result, these perched water-tables since the Roman times enhance streamflow considering that naturally had formed wetlands. The soils together with the concave low plain topography and the very humid winters are the main factors behind the construction of the under-drained network, which occupies 75% of the catchment. Drains are at a depth of approximately 1 m and backfilled with pervious material, installed as a herringbone system, with laterals laid diagonally across the slope, intersecting a main drain running downslope to the outlet in the bank of an artificial stream (Fig. 2). In fact, an orthogonal network of artificial channels was built to conduct subsurface flow to a natural stream (see main map of Fig. 1). The steep and convex topographical areas are terraced by drystone walls, a historical management practice in the island (Grimalt, 1998) and in the Na Borges basin (Estrany et al., submitted for publication). Rainfed herbaceous crops are the main use of land (91%). They are located on the flat and subsurface drained areas, the main crops being cereals such as wheat and oats, which have the highest water needs during April and May (65%). This fact indicates that these species are well adapted to this type of soils and climatic conditions. Rainfed tree crops, i.e. almond tree cover 6.2% and are found in the steepest areas. Rest 2.7% of land is occupied by forest, where holm oak (Quercus ilex) is the climax vegetation, and the pine tree (Pinus halepensis) occupying the clearer and dryer areas. Methods Field instrumentation A gauging section was constructed at the catchment outlet of the Can Revull basin in December 2003, using a sharp crested thin plate weir. The 90° V-notch weir is a galvanized iron sheet with sharp brass edges, and was secured with concrete foundations and cut-off walls fixed in the bed and banks of the stream with bricks and cement (see plate in photograph of Fig. 1). The water stage is continuously measured using a pressure sensor (Druck Fig. 2. Block diagram of the under-drained system. PDCR-1830-3) linked to a data logger (Campbell CR10X) powered by a 12 V battery, which takes readings with a sample interval of one minute and a log interval of 15 min (recording the average value of the samples between log intervals). The stage/discharge rating curve was derived by means of standard formulae (International Organization of Standards, 1980). At the gauging station, rainfall was recorded using a tipping-bucket Davis rain gauge. This was positioned 1 m above the ground and connected to a datalogger that recorded 0.2 mm precipitation increments. Data analysis Water balance After the calibration of instruments in winter 2004, we selected for this study a period from July 1, 2004 to June 30, 2007. This gives a series of three full years that allows building up a monthly soil water balance, starting and finishing when water-storage in the catchment is the lowest in the year. The Thornthwaite–Mather model (T–M, Thornthwaite and Mather, 1955; Steenhuis and Van der Molen, 1986) was used to estimate the monthly water balance and depth of water excess available for runoff generation and aquifer recharge, from the available rainfall and evapotranspiration records. In this way, PET (Thornthwaite, 1948) was calculated from daily temperature obtained at the B346 Porreres station located 7.5 km to the southwest of Can Revull (see Fig. 1). Monthly ET was derived from PET, total precipitation and derived state variables and flows from the T–M model such as soil-moisture storage, and soil-moisture storage withdrawal. Likewise, the following parameters were also derived from the T–M model (i.e. Gallart et al., 2008): WETMONTH (months) is the annual number of months with water excess, EXCFLOW [–] is the correlation coefficient between the monthly values of water balance excess and observed flow, EXCREL [–] is the proportion of annual flow that was simulated through the water balance simulation, and ET/PET [–] is the proportion of the simulated ET in relation to PET. The available water capacity of the T–M model was calibrated between reasonable bounds to maximise EXCFLOW and to obtain EXCREL and ET/PET values close to unity. Monthly rainfall, runoff and ET were grouped by season (i.e. summer comprises July to September; autumn is from October to December; winter is from January to March; and spring is from April to June). The runoff coefficient was estimated by representing runoff as a percent of precipitation at annual, seasonal and monthly time steps. Runoff components and flow duration curves Runoff components were extracted at event-scale by the straight-line method. The techniques used to separate runoff components are numerous (Nathan and McMahon, 1990), although all are fully arbitrary (Latron et al., 2008). In our case, quickflow and baseflow were separated by plotting a semi-logarithmic hydrograph based upon the backward extension of straight line recession curve segments, to distinguish the individual flow component line with a constant upward slope starting at the beginning of the event and finishing at the end of the second recession segment. Fig. 3 shows an example of the hydrograph separation, which was carried out for 20 flow events along the 3-year study period. From the 15-min discharge measurements, for each runoff component (i.e. quickflow, baseflow and total flow) Flow Duration Curves (FDCs) were established at annual and seasonal scales. For each FDC, two standard indices were derived from total flow, baseflow and quickflow when discharge values exceeded 0 m3 s1. These were the Variability index (Vi) and 30/70 ratio. Lane and Lei (1950) defined Vi as the standard deviation of the common logarithms of discharge determined at 10% intervals from 10% to 90% of the cumulative frequency distribution. Catchments with more J. Estrany et al. / Journal of Hydrology 380 (2010) 180–190 183 Fig. 3. Separation of the storm hydrograph for the flood event occurred at the end of January 2006 at the Can Revull small catchment. sustained flow; i.e. catchments with higher water storage, tend to have lower Vi (<0.5) than catchments with a higher percentage of surface runoff and lower storage. Likewise, FDCs with a steep slope are indicative of streams with more variability than those with a flatter slope. The 30/70 ratio is an approach to estimate these slopes, expressing the flow exceeded 30% of the time to that exceeded 70% of the time. In order to standardize the examination of runoff component behaviour, two indices were also applied: first, a Quickflow Response Ratio (QRR) to indicate the percentage of total precipitation resulting in quickflow; second, a Base Flow Index (BFI), defined as a non-dimensional ratio between the average discharge under the separated baseflow hydrograph and the average discharge of the total hydrograph. In catchments with high groundwater or subsurface contribution to streamflow, the BFI may be close to 1, but it may be equal to zero for ephemeral streams (Smakhtin, 2001). Recession periods A study of recession periods was carried out by matching the strip method (Toebes and Strang, 1964; Nathan and McMahon, 1990), involving the plotting of recession curves on a semi-logarithmic plot to allow the visualisation of very low flows. Each recession curve was superimposed and adjusted horizontally to produce an overlapping sequence by means of the software GraÒ pher . Some authors emphasize that a minimum of 10 years is necessary to provide reliable estimates of recession parameters (Tallaksen, 1995). However, the low frequency of days with rainfall in Mediterranean regions allows greater and longer recession periods (Latron, 2003). A mathematical expression was derived to quantify recession curves (e.g. Toebes and Strang, 1964; Tallaksen, 1995). In addition, parameters were optimized by applying a constrained nonlinear regression procedure using the Levenberg–Marquardt method and statistical software package SPSS. The best fitted storage-outflow model for a shallow unconfined aquifer is Depuit-Boussinesq. It is an exact solution for the nonlinear differential flow equation, being a second degree hyperbola Q¼ Q0 ð1 þ X 0  tÞ2 ð1Þ where Q is the rate of flow (l s1), t the time discharge (days) and X0 is a non-constant recession coefficient. Equation yields the outflow from an unconfined aquifer and gives a good fit under conditions that match these well. However, sometimes recession curves do not fit correctly to this equation, as happened in Can Revull. Latron (2003) suggested the following solution to improve recession curve fit: Q¼ n X i¼1 Qi ð1 þ X i  tÞ2 ð2Þ For three curves n = 2 with parameters optimized by applying a constrained nonlinear regression procedure with SPSS. These equations have been integrated in order to measure available reservoir for runoff, which can be estimated using the following equation: R¼ n X i¼1 Ri ð1 þ X i  tÞ ð3Þ where R is available reservoir for runoff (mm). Rainfall–runoff relationships For each flow event, several variables extracted from the hyetograph and hydrograph were derived and placed into two groups (Table 1a): (i) antecedent conditions and (ii) event conditions. These variables were available for 26 events, nine of which corresponded to three multi-peak flow events included in the 20 events previously mentioned. Thereafter, stepwise multiple regressions were applied in order to analyse rainfall–runoff relationships using the statistical software SPSS. These regressions were applied with variables in both linear and logarithmic forms, whilst maximum peak discharge (Qmax) and quickflow runoff (Qrunoff) were the dependent variables. In order to assess the magnitude of the relationship between variables, a Pearson product–moment correlation analysis was applied, analysing both types of variables together (Table 1b). It was found 184 J. Estrany et al. / Journal of Hydrology 380 (2010) 180–190 Table 1 (a) Pre-event and event conditions variables used in stepwise multiple regressions to explain rainfall–runoff relationships and (b) Pearson correlation matrix between selected parameters. Coefficients in bold are significant at the 0.01 level and coefficients in italics at the 0.05 level. Pre-event conditions Event conditions (a) Precipitation as soil moisture index AP1d Antecedent precipitation 1 day before (mm) AP3d Antecedent precipitation 3 day before (mm) AP7d Antecedent precipitation 7 day before (mm) AP15d Antecedent precipitation 15 day before (mm) AP21d Antecedent precipitation 21 day before (mm) Precipitation and derived variables Ptot Total event precipitation (mm) IPavg Average rainfall intensity (mm) IPmax 300 Maximum 300 rainfall intensity (mm) IPmax 50 Maximum 50 rainfall intensity (mm) Runoff variables Maximum peak discharge (l s1) Quickflow runoff (mm) Qmax Qrunoff Time relationships between rainfall and runoff Time storm Rainfall storm event duration (min) TstartQ Time interval between start of the storm and start of hydrograph rise (min) T for Q Time interval between median rainfall storm and peak discharge (min) (b) Log Qmax Log Qrunoff Time storm TstartQ T for Q Ptot IPavg IPmax 300 IPmax 50 AP1d AP3d AP7d AP15d AP21d Log Qmax Log Qrunoff Time storm TstartQ T for Q Ptot IPavg IPmax 300 IPmax 50 AP1d AP3d AP7d AP15d AP21d 1 0.759 1 0.274 0.372 1 0.055 0.183 0.512 1 0.015 0.021 0.774 0.611 1 0.027 0.026 0.704 0.481 0.666 1 0.455 0.466 0.282 0.116 0.199 0.257 1 0.412 0.496 0.142 0.095 0.039 0.431 0.962 1 0.429 0.519 0.197 0.066 0.076 0.396 0.930 0.973 1 0.384 0.391 0.345 0.269 0.430 0.582 0.285 0.372 0.323 1 0.560 0.546 0.242 0.336 0.397 0.528 0.353 0.446 0.363 0.803 1 0.534 0.344 0.317 0.291 0.379 0.607 0.351 0.407 0.342 0.630 0.785 1 0.375 0.247 0.321 0.262 0.377 0.654 0.361 0.421 0.350 0.558 0.655 0.915 1 0.362 0.252 0.184 0.029 0.180 0.572 0.423 0.443 0.396 0.662 0.558 0.740 0.812 1 that the dependent variables were more significant in logarithmic form, whereas independent variables were more significant in linear form. The derived equations were only used as descriptive least squares fit to the catchment data, and should not be extended beyond the limits of the data from which they were derived. Independent variables not already in the equation, with a probability of a smaller F were introduced at each step, if said probability was sufficiently small. An F level of 5% was used. The variables already introduced into the regression equation were eliminated if their F probability was sufficiently high. The method concluded when no other variable was susceptible to inclusion or elimination. The exact influence of each of these variables was measured with b coefficients. These coefficients are dimensionless parameters and measure the effect of a particular independent variable on the variation of the dependent variable, and as they are dimensionless, may be directly compared (Gregory and Walling, 1973). Results and discussion Soil water-balance Fig. 4 shows a general overview of the hydrological behaviour of the catchment. The simulated monthly water balance shows rainfall for a long-term period (1974–2006) and rainfall, runoff, ET and PET for the study period 2004–2007. Considering that the average annual rainfall for the period of record 1974–2006 was 517 mm and the standard deviation just 22.7 mm, the mean annual rainfall for the three study years can be considered near to the average value (496 mm). However, one important difference was observed at the seasonal time scale during study period. Spring showed very low values, with total rainfall depths lower than summer, the season with lowest rainfall for the long-term period. In the monthly time scale, October accounted for the highest long-term rainfall, but presented an important reduction during the study period; the same phenomenon was observed from May to August (both included). In contrast, November, December, February, March and April displays rainfall values well above the long-term mean. In addition, runoff depths illustrated a somewhat different pattern, winter being the season with the highest value (81 mm), followed by autumn, spring and summer with runoff values of 28, 13 and 0.2 mm, respectively. The derived parameters from the T–M model (Table 2) show that only 11 months were accounted with water excess (30% of the study period), whilst at the gauging station flow was measured during 17 months (47% of the study period). This difference may indicate that the model underestimates the role played by the under-drained system in the runoff generation. Nevertheless, at annual and seasonal time-scales during the study period the simulated monthly water balance was well correlated with observed flows (EXCFLOW) and the proportion of flow simulated by the water balance (EXCREL) suggest initially that saturation processes are dominant. Furthermore, PET and ET exemplify a different pattern. Summer is the season with the highest average value of PET (477 mm), followed by spring, autumn and winter with 313, 153 and 68 mm, respectively. However, ET is conditioned by seasonal rainfall distribution and by the decreasing of soil-mois- J. Estrany et al. / Journal of Hydrology 380 (2010) 180–190 185 Fig. 4. Monthly water balance: mean monthly rainfall, runoff and simulated actual evapotranspiration during the study period (2004–2007). Histograms show rainfall distribution (top) and reference evapotranspiration (bottom) for long-term (1974–2006). ture storage because water becomes more difficult to remove from the soil being less available for ET. Hence, summer had the highest seasonal average ET (90 mm), followed by autumn, winter and spring with 88, 68 and 62 mm, respectively. Finally, the ET/ PET parameter explains the seasonal evolution of the hydric stress suffered by the vegetation. Consequently, summer had the lowest value of ET/PET indicating that little water is available for vegetation; whilst in winter the low values of PET involved the maximum water availability for ET. Considering that autumn and spring are transitional seasons, they exhibited different values caused by the scarcity of rainfall in spring during study period. Whilst autumn showed a middle value of ET/PET, the spring’s value was very low indicating a maximum hydric stress for the vegetation in a period when cereals have the highest water needs. Runoff coefficients were 55.5% in winter, 20.1% in autumn, 14.7% in spring and 0.2% in summer, being the annual average 24.8%. These runoff coefficients are very high if compared to those observed in other catchments with similar climatic conditions, where runoff coefficients are usually below 10% and never exceed Table 2 Derived parameters from the water balance analysis during each study year and seasonal averages. WETMONTH EXCFLOW EXCREL ET/PET Years 2004–2005 2005–2006 2006–2007 Average period 4 3 4 – 0.76 0.64 0.57 0.66 1.02 1.55 1.64 1.32 0.54 0.64 0.52 0.57 Seasons Summer Autumn Winter Spring 0 4 6 1 0.00 0.86 0.14 0.97 0.00 2.06 1.10 1.14 0.19 0.58 1.00 0.20 25% (Ceballos and Schnabel, 1998; Tzoraki and Nikolaidis, 2007). These high runoff coefficients can be primarily explained by the presence of the under-drained system constructed in high waterstorage soils located in flat areas and supported by an impervious sediment layer (i.e. lower-middle Miocene). Moreover, this layer reduces the deeper percolation creating perched water-tables which naturally had formed wetlands, but since Roman age enhances streamflow by means of the under-drained system. Using these main elements of the water balance (rainfall, runoff and ET), linear regression relationships were established on a monthly time-scale. Nevertheless, none of them were statistically significant (r2 < 0.2), as has been reported in other Mediterranean catchments (Ceballos and Schnabel, 1998; Latron et al., 2008). The lack of a simple relationship is explained by the high range of evapotranspirative scenarios (Fig. 4), indicating wet periods (late autumn, winter and early spring) during which the water supply exceeds the evapotranspiration demands and dry periods during which water supply does not cover these demands (late spring, summer and early autumn). Despite of the lack of linear relationships, the simple water balance confirmed a general succession of three different periods during the year determined by evapotranspiration. This general succession has already been highlighted by other authors in small mountainous catchments where climate experiences changes related to altitude (e.g. Gallart et al., 2002; Latron et al., 2008). Hence, seasonal PET and rainfall distribution resulted in the following succession: (a) In spring and summer evapotranspiration demand is very high, soils become dry and cracked, and subsurface-tile drainage does not work. Under these conditions, only longer or repeated events such as those of September 2006 are able to generate runoff. 186 J. Estrany et al. / Journal of Hydrology 380 (2010) 180–190 (b) A transition period, in early autumn and in early spring, during which rainfall feeds catchment water reserves, before water flows into the channel. (c) A wet period corresponding to late autumn and winter, when catchment water reserves have been refilled and evapotranspiration losses are low compared to rainfall depths; during wet conditions, rainfall is mostly available for runoff. Runoff components and flow duration curves Soil water balance components and the derived parameters from the T–M model of the catchment provides only a partial interpretation on the effects of the variables derived from hyetograph and hydrograph on varying runoff characteristics. In order to further evaluate the variability of the hydrological response and the influence of catchment characteristics, runoff was separated into Fig. 5. Flow-duration curves for total discharge, baseflow and quickflow during the study period: (a) Annual; (b) Summer; (c) Autumn; (d) Winter and (e) Spring. J. Estrany et al. / Journal of Hydrology 380 (2010) 180–190 quickflow and baseflow for 20 events. Subsequently, annual and seasonal FDCs (Fig. 5) and indexes were established for total runoff and its components to provide further insight into the hydrologic response of the catchment and its underlying causes. At the annual time scale total flow was present for 32.4% of the time. The low value of the Vi (0.201) indicates the catchment’s high dynamic storage, confirmed by the quickflow that was only present 2.1% of the time (Table 3). Meanwhile, significant differences were observed at the seasonal scale: – Winter presented flow 90% of the time and had the lowest Vi of all seasons (0.17), showing that highest dynamic storage along the year. Nevertheless, the 0.84 value from the 30/70 ratio denotes that winter flow variability was high –albeit short term – as high soil water content and rainfall was mostly available for quickflow which was present 4.9% of the time, the highest percentage at the seasonal time scale. – Autumn, with 21.8%, showed the second highest percentage of discharge duration. In this case, the Vi is the highest (0.27); at the beginning of the season, soil moisture deficit was at maximum and rainfall must first refill catchment water storage. For the same reason the 30/70 ratio value is the lowest although quickflow was present for 2.7% of the time. – Spring presented flow 18.9% of the time with a Vi of 0.190, denoting that dynamic storage was still significant as catchment reserves were still high. Accordingly, and despite the low amount of rainfall accumulated during this season, water was mostly available for quickflow as soil water content was still high. Therefore, the 30/70 ratio was higher than for autumn, although quickflow was only present for 1.6% of the time. – Finally, summer was not considered, as flow represented just 0.8% of the time. Table 4 shows the values for rainfall, quickflow, baseflow, QRR and BFI for the annual and seasonal time-scales. Results of applied indices allowed summarizing the runoff component behaviour: Table 3 Relative seasonal and annual total flow and quickflow duration for study period 2004–2007. Variability index and 30/70 ratio for total flow at seasonal and annual time-scale for the study period 2004–2007. Season Total flow (%) Quickflow (%) Total flow Variability index 30/70 ratio Summer Autumn Winter Spring 0.82 21.73 89.98 18.88 0.47 2.65 4.94 1.62 – 0.270 0.167 0.190 – 0.686 0.838 0.773 Annual 32.37 2.13 0.201 0.786 Table 4 Total and relative values for each runoff component, Quickflow Response Ratio (QRR) and Baseflow Index (BFI) during each study year and seasonal averages. Years 2004–2005 2005–2006 2006–2007 Average period Seasons Summer Autumn Winter Spring a b Quickflow (mm) Quickflow (%) Baseflow (mm) Baseflow (%) QRRa 10.5 14.5 9.2 11.4 6.1 14.1 9.9 10.0 162.8 87.9 84.1 111.6 93.9 85.9 90.1 90.0 2.34 2.71 2.03 2.35 0.94 0.86 0.90 0.90 0.2 3.1 6.2 1.9 96.8 10.7 7.7 14.7 0.0 25.5 74.9 11.2 3.2 89.3 92.3 85.3 0.19 1.54 4.52 3.48 0.03 0.89 0.92 0.85 Quickflow Response Ratio. Baseflow Index. 187 (a) QRR showed low values (average = 2.4%) indicating a lack of adequate conditions for quickflow generation. This is due to the fact that under-drained system depletes the water table and decreases the occurrence of conditions for soil saturation. Thus, on the seasonal time scale, winter had the highest QRR (average = 4.5%), reaching values of up to 7.6% in 2005– 2006 due to the main rainfall events occurred when catchment water reserves were filled. Meanwhile, spring presented the second highest QRR (average = 3.5%) because then catchment water reserves are still large. Although autumn had more rain than spring, it presented lower values according to the transition period explained in the section ‘‘Field instrumentation”. (b) BFI for the total study-period with discharge was 0.92. Indeed, most streamflow was baseflow confirming that saturation processes are then dominant and related to the water release through the under-drained system, despite the fact that 67.6% of the time the stream did not flow. Thus, for the seasonal scale, BFI only exhibited a low value (0.03) during summer, being the only season where quickflow was predominant following an ephemeral behaviour, i.e. flash-floods. Baseflow response: recession curves Recession curve variations were analyzed emphasizing the seasonal effects in relation to rainfall and evapotranspiration variability (e.g. Wittenberg, 2003). Analysis was done considering the high proportion of baseflow in streamflow dynamics caused by the soil water-storage and the under-drained network acting in the catchment. From instantaneous discharge measured daily at 6.00 h UTC when rainfall was negligible, 17 recession periods were selected, ranging from 3 to 17 days and discharges from 1.28 to 1.5 l s1. Afterwards, using the strip match method, three master recession curves (MRC) were derived through the set of common lines (Fig. 6). Particular attention was paid to emphasise the particular influence of PET; multiple correlation coefficients were about 0.98. As a result, three PET scenarios can be described: 1. No PET influence: Average daily PET was 0.55 mm. The mean duration was 22 days and it is representative of recessions occurring in late December, January and the beginning of February. 2. PET influence-1: Average daily PET was 0.98 mm. The mean duration was 17 days, and it is representative of recessions occurring in late February and March. 3. PET influence-2: Average daily PET is 1.91 mm. The mean duration was 13 days, and it is representative of recessions occurring in April. BFIb As it is shown by recession coefficient (Table 5), the very steep segments in the upper part of the recession curve for not influenced MRC would appear to reflect the shallow surface storage that quickly responds to storm rainfall but is thereafter exhausted quite rapidly. The other parts of the curve have lower recession coefficients showing high retention capacity caused by low topography, geology and soil texture, maintaining 34% of the hydrological reservoir at the end of the recession period which was released slowly through the under-drained system. Influences 1 and 2 recession coefficients demonstrate PET influence in the lower parts of these curves, retaining just 25% and 21%, respectively, of the water reserves at the end of each recession period. Similarly, initial reservoir volumes are clearly different for each MRC. Non-influence 188 J. Estrany et al. / Journal of Hydrology 380 (2010) 180–190 Fig. 6. (a) Master Recession Curves and (b) Evolution of reservoir volume available for runoff. Table 5 Recession curve equations, discharge (Q), maximum available reservoir (R) and recession coefficient (X) for t = 0, 7 and 13 days. MRC Equation Not influenced Q¼ Influenced 1 Q¼ Influenced 2 Q¼ 106:2 ð1þ0:54tÞ2 25:5 ð1þ0:06tÞ2 22:4 ð1þ0:09tÞ2 þ þ þ 21:8 ð1þ0:03tÞ2 11:1 ð1þ210:92tÞ2 14:5 ð1þ0:53tÞ2 Avg PET per day (mm) Q0 (l s1) Q7 (l s1) Q13 (l s1) R0 (mm) R7 (mm) R13 (mm) X0 X7 0.55 128.0 22.1 14.4 83.1 51.0 39.0 0.125 0.035 0.030 0.98 36.6 13.3 8.2 32.6 17.1 10.5 0.091 0.064 0.064 1.91 36.9 10.3 5.5 22.6 8.8 4.8 0.133 0.096 0.093 initial volume was 83 mm; accordingly, influences 1 and 2 initial volumes were 33 and 23 mm, respectively. Quickflow response: event-scale rainfall–runoff relationships Quickflow was present for just 2.1% of the study period, but represented the response of the catchment to rainfall. Rainfall–runoff relationships were analysed, including variables from the hyetograph and hydrograph derived from 26 events. Variables were grouped as independent (antecedent conditions and precipitation), and dependent (related to discharge). Independent variables can be classified according to theories of runoff generation mechanisms: (a) intensity precipitation variables related to the Horton theory and (b) antecedent conditions and total precipitation related to the Dunne theory. According to Horton (1933), surface runoff can occur when rainfall intensity exceeds the infiltration capacity of the soil. Furthermore, rainfall duration must be greater than ponding time. According to Dunne’s statement (Dunne and Black, 1970), if rainfall intensity is lower than infiltration capacity, surface runoff is produced by precipitation over the area where water table is at the surface. At the event scale, overland flow is likely to occur if the initial water table is shallow. Within this hydrological context, Table 6 shows the multiple regression equations that were derived considering that there is a strong positive correlation between dependent variables and pre-event conditions variables, as well as a negative correlation between these dependent variables and precipitation intensity variables. In each case, the multiple correlation coefficients were about 0.85. These two equations summarize the hydrological processes involved in runoff generation within the catchment: – The first equation explains the maximum peak discharge, which integrates the two runoff components. As seasonal BFI values indicate, when flow was present, baseflow and subsur- X13 face-tile drainage played a significant role along most of the study period. Consequently, antecedent conditions were important for runoff generation, a fact corroborated by the two related variables in the equation: antecedent precipitation for 3 and 7 days before peak occurs. b coefficients (0.33 and 0.53, respectively) indicate that maximum peak discharge was higher with high soil water content. Likewise, if total event precipitation is greater, maximum peak discharge will be higher. Its b coefficient (0.66) is the highest of all variables, indicating that it was the most influential variable confirming that saturation processes were dominant. Finally, average rainfall intensity occurred during study period shows a negative tendency (b coefficient is 0.49) due to the fact that the highest rainfall intensities took place during summer when the soil water content was low and there was no flow into the channel, causing maximum peak discharges to be lower although rainfall intensities are higher. – The second equation was integrated for one of the runoff components, i.e. quickflow runoff. In this case, antecedent conditions are not as important as in the first equation. Therefore, antecedent precipitation over the three previous days is the only preevent condition variable in the equation. Its b coefficient (0.57) is lower than the antecedent condition variables from the first equation. Variables related to precipitation such as total event precipitation and 5 min rainfall intensity are the most significant as they are directly related to quickflow. Thus, total event precipitation showed a b coefficient of 0.97 indicating greater significance than in the first equation. As with the first, rainfall intensity negatively modified the equation. But in this case it was even larger due to the 5 min rainfall intensity variable. The b coefficient is also high (0.72) indicating a significant influence. Finally, a second variable illustrated a negative tendency in the equation: the interval between mean rainfall storm event and peak discharge in minutes. If this time interval is 189 J. Estrany et al. / Journal of Hydrology 380 (2010) 180–190 Table 6 Derived equations for rainfall–runoff relationships applying the stepwise multiple regression technique and subsequent b coefficients. Log Q max ¼ 1:248 þ 0:005X 1 þ 0:006X 2 þ 0:015X 3  0:013X 4 where Qmax; Peak maximum discharge (l s1) b coefficients Qrunoff; Quickflow runoff (mm) Qmax (a) X1; X2; X3; X4; X5; X6; 0.328 0.530 0.658 0.492 – – Antecedent precipitation 3 days before (mm) Antecedent precipitation 7 days before (mm) Total event precipitation (mm) Average rainfall intensity (mm h1) 50 Maximum rainfall intensity (mm h1) Time interval between median rainfall storm and peak discharge (min) longer, quickflow runoff will be lower indicating that the water content of the soil was low, causing quickflow runoff to be lower. The quickflow response has been studied by means of maximum peak discharge and quickflow runoff as dependent variables. According to the previous results, it may be suggested that both were mainly controlled by the Dunnes’ mechanism, although the second one in a lesser extend. Therefore, soil water storage capacity had to be complete before any significant runoff was generated. It is necessary to emphasise that rainfall data series between 1988 and 2001 years recorded at the Porreres rain gauge, located 7 km from the Can Revull gauging station (Fig. 1), allow us to verify that these rainfall intensities for the entire study period 2004–2007 are representative of the long-term record. Thus, the average of the maximum 30-min rainfall intensities at the reference site was 27.1 mm h1 (YACU, 2002), whereas it was 35.5 mm h1 during the study period. Summary and conclusions We have examined the hydrological dynamics in a small Mediterranean catchment for a period of three hydrologically average years. This catchment is characterised by its extensive herbaceous agricultural activity; of particular interest is the under-drained system (constructed in soils with high water-storage which naturally formed wetlands). Several conclusions can be drawn as follows: (1) The monthly soil water balance provided a general overview of the hydrological variables. The derived parameters from the soil water model suggest that saturation processes are dominant; in addition, significant seasonality confirms that evapotranspiration is the main variable controlling the catchment’s hydrological response. As such, the mean annual runoff coefficient was relatively high (24.8%). The lack of a linear relationship between rainfall and runoff on the annual, seasonal and monthly time-scales can be related either to the short study period and the specifically to high range of evapotranspirative scenarios. Consequently, it was only possible to observe a succession of three different hydrological periods throughout the year conditioned by the potential evapotranspiration. (2) The flow duration curves show that discharge was only present for 32.4% of the time, of which 2.1% of the time was quickflow. However, there are some significant differences at the seasonal scale. While flow was present for 90% of the time in winter, in summer it was only 0.8%. Quickflow was present for 4.9% of the time in winter and 0.5% in summer. These seasonal differences are better observed by applying the Variability index and 30/70 ratio, which con- Log Q runoff ¼ 2:644 þ 0:016X 1 þ 0:038X 3  0:015X 5  0:001X 6 Qrunoff 0.572 – 0.974 – 0.716 0.474 firm the succession of three different periods where waterstorage dynamics play an important role in runoff generation. (3) Quickflow Response Ratio (QRR) and Baseflow Index (BFI) are two good indicators of runoff component behaviour. QRRs were low (2.4%) on the annual time scale. Thus, potential evapotranspiraton causes a significant decrease of the water-table during dry periods and the under-drained system depletes the water table during wet periods; as a result, quickflow generation is limited. Previous rainfall is needed to feed catchment water reserves, also considering that the intensity of rainfall to exceed soil infiltration capacity need to increase due to the cracking of clay soils during dry periods. At the seasonal scale, QRR values clearly explain the succession of three hydrological periods associated with water-storage dynamics. These periods are clearly defined by the average annual BFI value of 0.92, indicating a high subsurface contribution to the streamflow. As a result, baseflow is controlled by an unconfined and shallow aquifer formed in deep soils supported on an impervious Miocene layer and artificially under-drained. (4) A study of recession periods was conducted to assess the significance of baseflow response. Three master recession curves were distinguished according to different evapotranspiration scenarios occurring in late autumn and winter, late winter, and the beginning of spring. An equation based on the Depuit-Boussinesq model for shallow unconfined aquifers was fitted and available reservoir volumes were estimated by integration. Some remarkable differences were observed between the three curves under different rates of PET. (5) Finally, stepwise multiple regressions were established to analyse the quickflow response by means of the event-scale rainfall–runoff relationships. Maximum peak discharge and quickflow runoff were dependent variables and variables related to rainfall and antecedent conditions were independent variables. Both dependent variables were mainly controlled by antecedent conditions and event rainfall volume, whereas the role of rainfall intensity was negative, thus suggesting the predominant role of saturation (Dunne’s mechanism) rather than rainfall excess (Horton’s mechanism) during the study period. Results provide one of the first insights into the hydrological response of a catchment characterized by Mediterranean rainfed herbaceous crops artificially under-drained. However, more detailed long-term data is needed to achieve an accurate understanding of the hydrological functioning of the Can Revull and similar catchments. The use of continuous information on the evolution of both saturated and unsaturated soil scenarios and its relation to the 190 J. Estrany et al. / Journal of Hydrology 380 (2010) 180–190 catchment scale runoff response needs further attention. Only then models will be improved considering the complex response caused by the long dry periods as well as the important role of the underdrained systems. Acknowledgments This work has been funded by the Spanish Ministry of Education and Science (Research project REN2001-0281) and by and agreement between the Spanish Ministry of Environment (LUCDEME project) and the University of the Balearic Islands. The authors wish to express their gratitude to Sebastià Barceló Sansó (alias Revull), the landowner who generously permitted the instrumentation of the catchment and allowed access. Thanks must also be given to Joan Miquel Carmona for his technical assistance during the construction process and during field work, to Jérôme Latron for his technical assistance in water balance and recession curves, to Ricardo Alberich for his statistical support. The authors are indebted to Francesc Gallart for the constructive revision that greatly improved the final version of the manuscript. References Beven, K.J., 2002. Runoff generation in semi-arid areas. In: Dryland Rivers. John Wiley, Chichester, pp. 57–105. Ceballos, A., Schnabel, S., 1998. Hydrological behaviour of a small catchment in the dehesa landuse system (Extremadura, SW Spain). Journal of Hydrology 210, 146–160. Diaz Palmer, A., Garcia, C., Servera, J., Úbeda, X., 2006. Spatial variability of total nitrogen, total carbon and organic carbon content in the top-soil of the Na Borges basin, Mallorca, Spain. Zeitschrift für Geomorphologie 143, 83–94. Dunne, T., Black, R.D., 1970. Partial area contributions to storm runoff in a small New England watershed. Water Resources Research 6, 1296–1311. Estrany, J., Garcia, C., 2005. Monitoring suspended sediment fluxes in the Na Borges basin, Mallorca, Spain. In: Batalla, R.J., Garcia, C. (Eds.), Geomorphological Processes and Human Impacts in River Basins. IAHS Publication 299, Wallingford, pp. 119–125. Estrany, J., Garcia, C., Walling, D.E., submitted for publication. Investigation of soil erosion and redistribution in a Mediterranean lowland agricultural catchment using caesium-137. Gallart, F., Llorens, P., Latron, J., Regüés, D., 2002. Hydrological processes and their seasonal controls in a small Mediterranean mountain catchment in the Pyrenees. Hydrology and Earth System Sciences 6, 527–537. Gallart, F., Amaxidis, Y., Botti, P., Canè, G., Castillo, V., Chapman, P., Froebrich, J., García-Pintado, J., Latron, J., Llorens, P., Lo Porto, A., Morais, M., Neves, R., Ninov, P., Perrin, J.L., Ribarova, I., Skoulikidis, N., Tournoud, M.G., 2008. Investigating hydrological regimes and processes in a set of catchments with temporary waters in Mediterranean Europe. Hydrological Sciences Journal 53 (3), 618–628. Gregory, K.J., Walling, D.E., 1973. Drainage Basin Form and Process. A geomorphological approach. Edward Arnold (Publishers) Ltd., London. 458 pp. Grimalt, M., 1998. L’home com a factor geomorfològic a Mallorca. L’enginyeria popular amb finalitat antierosiva. In: Fornós, J.J., (Ed.), Aspectes geològics de les Balears (Mallorca, Menorca i Cabrera). Universitat de les Illes Balears, Servei de Publicacions i Intercanvi Científic, Palma, pp. 423–434. Hewlett, J.D., 1982. Principles of Forest Hydrology. University of Georgia Press, Athens, Greece. Horton, R.E., 1933. The role of infiltration in the hydrologic cycle: EOS, transactions. American Geophysical Union 14, 446–460. INE – Instituto Nacional de Estadística, 2003. Anuario Estadístico de España. Ministerio de Economía y Hacienda, Madrid, 755 pp. International Organization of Standards, 1980. Water flow measurement in open channels using weirs and venturi flumes – Part 1: thin plate weirs. ISO 1438/11980 (E). Kirkby, M.J., 2005. Organisation and process. In: Anderson, M.G. (Ed.), Encyclopedia of Hydrological Sciences, vol. 1. John Wiley, Chichester, pp. 41–58. Lane, E.W., Lei, K., 1950. Streamflow variability. Transactions of the American Society of Civil Engineers 20, 1084–1134. Latorre, J.G., García-Latorre, J., Sanchez-Picón, A., 2001. Dealing with aridity: socioeconomic structures and environmental changes in an arid Mediterranean region. Land Use Policy 18 (1), 53–64. Latron, J., 2003. Estudio del funcionamiento hidrológico de una cuenca mediterránea de montaña (Vallcebre, Pirineos Catalanes). PhD thesis. Universitat de Barcelona, Barcelona, 298 pp. Latron, J., Gallart, F., 2008. Runoff generation processes in a small Mediterranean research catchment (Vallcebre, Eastern Pyrenees). Journal of Hydrology 358 (3– 4), 206–220. Latron, J., Soler, M., Llorens, P., Gallart, F., 2008. Spatial and temporal variability of the hydrological response in a small Mediterranean research catchment (Vallcebre, Eastern Pyrenees). Hydrological Processes 22 (6), 775–787. Nathan, R.J., McMahon, T.A., 1990. Evaluation of automated techniques for base flow and recession analysis. Water Resources Research 26, 1465–1473. Porta, J., López-Acevedo, M., Roquero, C., 1999. Edafología para la agricultura y el medio ambiente. Mundi Prensa, Madrid. 876 pp. Romero, R., Guijarro, J.A., Ramis, C., Alonso, S., 1998. A 30-year (1964–93) daily rainfall data base for the Spanish Mediterranean regions: first exploratory study. International Journal of Climatology 18, 541–560. Smakhtin, V.U., 2001. Low flow hydrology: a review. Journal of Hydrology 240, 147– 186. Steenhuis, T.S., Van der Molen, W.H., 1986. The Thornthwaite-Mather procedure as a simple engineering method to predict recharge. Journal of Hydrology 84, 221– 229. Stone, M., Krishnappan, B.G., 1997. Transport characteristics of tile-drain sediments from an agricultural watershed. Water, Air and Soil Polution 99, 89–103. Tallaksen, L.M., 1995. A review of baseflow recession analysis. Journal of Hydrology 165, 349–370. Thornthwaite, C.W., 1948. An approach toward a rational classification of climate. Geographical Review 38, 55–94. Thornthwaite, C.W., Mather, J.R., 1955. The water balance. Laboratory of Climatology Report No. 8, Drexel Institute of Technology, New Jersey, USA. Toebes, C., Strang, D.D., 1964. On recession curves 1: recession equations. Journal of Hydrology 3, 2–15. Tzoraki, O., Nikolaidis, N.P., 2007. A generalized framework for modeling the hydrologic and biogeochemical response of a Mediterranean temporary river basin. Journal of Hydrology 346 (3–4), 112–121. Wittenberg, H., 2003. Effects of season and man-made changes on baseflow and flow recession: case studies. Hydrological Processes 17 (11), 2113–2123. Woodruff, J.F., Hewlett, J.D., 1970. Predicting and mapping the average hydrologic response for the Eastern United States. Water Resources Research 6 (5), 1312– 1326. YACU, 2002. Estudio de caracterización del régimen extremo de precipitaciones en la isla de Mallorca. Memoria. (Translation to English: Characterization of the extreme rainfall regime on the island of Mallorca). Secció d’Estudis i Projectes, Direcció General de Recursos Hídrics, Conselleria de Medi Ambient, Govern de les Illes Balears, Palma de Mallorca.