Ecological Modelling 371 (2018) 25–36 Contents lists available at ScienceDirect Ecological Modelling journal homepage: www.elsevier.com/locate/ecolmodel Deviations on a theme: Peat patterning in sub-tropical landscapes a,⁎ a a a Christa L. Zweig , Susan Newman , Colin J. Saunders , Fred H. Sklar , Wiley M. Kitchens a b b T Everglades Systems Assessment Section, South Florida Water Management District, 3301 Gun Club Road, MSC 4350, West Palm Beach, FL, 33406, USA Department of Wildlife Ecology and Conservation, Newins-Ziegler Hall, University of Florida, Gainesville, FL, 32611, USA A R T I C L E I N F O A B S T R A C T Keywords: Landscape pattern Positive feedback Patterned peatlands Restoration Everglades Sub-tropical Patterned landscapes have long been a popular setting to test hypotheses about the effects of processes on ecosystem structure. The combination of scale-dependent and positive feedback theory is one of the most wellsupported in current literature, each explaining separate aspects of patterned peatlands. These theories have been developed from dynamics in boreal peatlands and tested in boreal systems, but the mechanisms that control peat patterning in a sub-tropical system have yet to be acknowledged in theory. Statistical evidence for different mechanisms are present in the biophysical features of sub-tropical patterned peatlands, such as the ridge and slough landscape (RSL) within the greater Everglades wetland system. We use data from the RSL to test whether features of a sub-tropical, patterned peatland conform to positive and scale-dependent feedback patterning theories developed in boreal peatlands and use dynamic simulation to explain our results. The analysis of surface elements and nutrient differences within the RSL and our dynamic simulations indicate that positive and scaledependent feedback may not be appropriate theories for sub-tropical peat patterning. Decomposition, rather than production, appears to be more important for abrupt microtopographical elevation differences, and differential nutrient concentrations are due to vegetation types, rather than increased evapotranspiration from greater vascular plant growth. Our model expands on the current theories for RSL maintenance, incorporating vegetation types and life history traits into differential peat deposition, which create the signature microtopographical differences found in the Everglades, and demonstrates that the underlying ecological patterning processes in sub-tropical peatlands are likely very different from boreal peatlands and require further discussion and study. 1. Introduction Patterned landscapes have long been a popular setting to test hypotheses about the effects of processes on ecosystem patterning (Couteron and Lejeune, 2001; Eppinga et al., 2008; Klausmeier, 1999; Rietkerk and Van de Koppel, 2008). Specifically, research in patterned peatlands has introduced a set self-organization theories to explain how regularly patterned microtopography exists on the landscape (Belyea and Lancaster, 2002; Couwenberg and Joosten, 2005; Rietkerk et al., 2004; Swanson and Grigal, 1988). These include scale-dependent feedback (Eppinga et al., 2008; Rietkerk et al., 2004), differences in hydraulic conductivity between hummocks and hollows (Couwenberg and Joosten, 2005; Foster et al., 1983; Glaser, 1992), uplift by frost heaving (Brown, 1970; Eppinga et al., 2008; Moore and Bellamy, 1974), and the role of seasonal flooding (Sakaguchi, 1980; Seppälä and Koutaniemi, 1985). Two different mechanisms within the scale-dependent feedback theory, nutrient accumulation and peat accumulation, are the most ⁎ well-supported in current literature (Eppinga et al., 2009a; Watts et al., 2010), each explaining separate aspects of patterned peatlands. The peat accumulation mechanism assumes that sharp microtopographical differences between ridges and hollows are reinforced by greater plant production on drier ridges, primarily from the increased presence of vascular plants (Belyea and Clymo, 2001; Hilbert et al., 2000; Rietkerk et al., 2004; Swanson and Grigal, 1988). Plant productivity, which determines the input of biomass to peat production (Belyea and Clymo, 2001; Eppinga et al., 2009a; Hilbert et al., 2000), is a function of peat thickness, a proxy for water stress on productivity (Eppinga et al., 2009a). Net peat accumulation in this mechanism is primarily a function of plant production and peat decomposition, both driven by water table depth. Evidence for this mechanism, expressed as sharp microtopographical differences, is a bimodal “frequency distribution of surface elements (vascular plant biomass or acrotelm thickness)” (Eppinga et al., 2008) and is well documented in boreal peatlands (Eppinga et al., 2009a). This feedback has been studied along a climactic gradient (Eppinga et al., 2010) and focused on the evapotranspiration (ET) and Corresponding author. E-mail addresses: czweig@sfwmd.gov (C.L. Zweig), snewman@sfwmd.gov (S. Newman), fsklar@sfwmd.gov (F.H. Sklar), wiley01@ufl.edu (W.M. Kitchens). https://doi.org/10.1016/j.ecolmodel.2018.01.008 Received 22 May 2017; Received in revised form 4 December 2017; Accepted 16 January 2018 0304-3800/ © 2018 Elsevier B.V. All rights reserved. Ecological Modelling 371 (2018) 25–36 C.L. Zweig et al. Fig. 1. Study sites for elevation and biomass data within Water Conservation Area 3A South (WCA3AS), FL USA. Small squares are 1 km2 study plots and large, white rectangles indicate the plots used in each quadrant analysis. decomposition and plant production (Aerts, 1997; Meentemeyer, 1978). Boreal-based models include production and decomposition as a function of water stress, and do not explicitly use temperature or differences in species’ attributes (Eppinga et al., 2009a; Eppinga et al., 2009b; Rietkerk et al., 2004), which are likely important in a sub-tropical climate. Statistical evidence for these differences between boreal and sub-tropical processes would be present in the biophysical features of sub-tropical patterned peatlands, such as the ridge and slough landscape (RSL) within the greater Everglades. The Everglades RSL is an example of a patterned peatland with parallel drainage topography whose patterning has been relatively stable on the landscape scale for thousands of years (Bernhardt and Willard, 2009) until water management for flood control and water supply was initiated in the early 20th century. The historical RSL was a dominant landscape type of the central portion of the Everglades and is a focus of the current restoration strategy. The RSL consists of long, linear ridges of sawgrass (Cladium jamaicense Crantz.); longer hydroperiod sloughs—open water areas with submerged aquatic plants and water lilies; and occasional tree islands oriented parallel to the slowmoving flow of water from northwest to southeast. The RSL also includes a wet prairie community, a short-statured emergent community that is sometimes considered a result of over-draining (McVoy et al., 2011), but is a stable and important community for critically endangered wildlife (Zweig and Kitchens, 2014). This landscape has been fragmented by compartmentalization, impoundment, and reduced flows (Ogden, 2005) and is now in a degraded state (Nungesser, 2011; Wu et al., 2006; Zweig et al., 2011). We use data from the RSL, explicitly considering wet prairie habitat, to test whether features of a sub-tropical, patterned peatland do, in fact, conform to two patterning mechanisms of scale dependent feedback: precipitation ratio in sites dominated by temperate and sub-arctic climates (Scotland, Sweden, and Siberia). The nutrient accumulation mechanism, on the other hand, explains landscape patterning; particularly regular string or maze patterning, from the directed flow of water and nutrients into ridges by increased ET of vascular plants (Rietkerk et al., 2004). As nutrients accumulate within ridges from plant uptake and recycling, vascular plant growth should increase and pull more water and nutrients by ET, reinforcing the nutrient accumulation cycle (Eppinga et al., 2008). This occurs at a local scale within the ridge, creating a negative effect for plants growing further from the ridge and generating a scale-dependent feedback (Eppinga et al., 2008; Rietkerk et al., 2004). These mechanisms have been developed from dynamics in boreal peatlands and tested as mathematical models or empirical models in temperate or boreal systems (Belyea and Clymo, 2001; Couwenberg and Joosten, 2005; Eppinga et al., 2008; Hilbert et al., 2000; Rietkerk et al., 2004). A limited number of studies have discussed non-boreal microtopography (Cheng et al., 2011; Couwenberg and Joosten, 2005; McVoy et al., 2011; Nungesser, 2003) and some aspects of scale-dependent patterning theories have been applied to the sub-tropical Everglades (Heffernan et al., 2013; Larsen and Harvey, 2010; Watts et al., 2010), including nutrient accumulation effects (Larsen et al., 2007; Larsen et al., 2015). These models were mainly theoretical studies, but one data-based analysis suggested that sub-tropical peatlands do conform to boreal-based peat accumulation mechanisms through elevation bimodality (Watts et al., 2010). However, we hypothesize that the mechanisms that control peat patterning in a sub-tropical system should be different because increased temperatures, longer photoperiod, lack of seasonality, and distinct vegetation communities contribute to differences in structuring processes through changes in 26 Ecological Modelling 371 (2018) 25–36 C.L. Zweig et al. Fig. 2. Composite STELLA peat deposition model structure for all vegetation communities in WCA3AS, FL USA. See Table 3 for variable definitions. 2.1. Peat accumulation mechanism peat and nutrient accumulation. Prior studies within the RSL did not explicitly consider the wet prairie community or combined it with slough (Watts et al., 2010). With productivity rates and biomass values between that of ridges and sloughs, consideration of this community is expected to influence modality of biomass surface elements. We tested modality of data-based surface elements from the RSL (plant biomass and elevation) to determine the influence of the peat accumulation mechanism; and differences between data-based nutrient levels in ridge and sloughs as a basis for confirming the nutrient accumulation mechanism in sub-tropical peatlands. We also use a peat accumulation model based on species-specific growth and decomposition rates to put the data-based results into context. To address the applicability of the peat accumulation mechanism to sub-tropical peatlands such as the Everglades, we tested for evidence of bimodality in elevation and biomass. Modality was tested by comparing the fit of our data to a unimodal and bimodal normal distribution (Eppinga et al., 2008; Watts et al., 2010) with the statistical software R (version 3.0.1, package ‘mixsmsn’), except in the case of biomass. Biomass data was heavily right-tailed, so we used a package that could model skewed distributions (a skewed-t distribution), and determined fit by Bayes’ information criteria (BIC) to compare to a similar RSL study (Watts et al., 2010). We analyzed the biomass data two ways: as a complete dataset and spatially by quadrant northeast, southeast, northwest, and southwest (Fig. 1). Quadrants were used as proxies for water depth 9–30+ cm deeper average water depths over the past 20 ears in the south due to impoundment (EDEN) and peat depth (average depth = 105 cm and 78 cm for the east and west, respectively. unpublished data, Zweig) along landscape gradients. The southeast quadrant was the most affected by impoundment and had the highest water depths (Zweig and Kitchens 2008). 2. Methods Biomass and elevation data were taken from a long-term monitoring study (Zweig and Kitchens, 2008; Zweig and Kitchens, 2009; Zweig and Kitchens, 2014) within Water Conservation Area 3A South (WCA3AS, Fig. 1). Briefly, 20–1 km2 plots were placed on the landscape in a stratified random manner—stratified by water depth, elevation, and peat depth (Zweig and Kitchens, 2008), and 2–3 transects were randomly located in each plot. Transects always started in a slough and ran perpendicular to the flow gradient into a ridge. Transects ranged from 30 to 100 m in length. We clipped above ground biomass in 0.25 m2 quadrats at 3 m intervals along transects. Samples were sorted to species, oven dried to a constant weight, and weighed to the nearest 0.1 g (n = 8464). Our annual November sample events began in 2002 and continued to 2011. We also measured water depths at each 3 m interval to derive highresolution elevation data. We acquired water stage surfaces from the Everglades Depth Estimation Network (EDEN, http://sofia.usgs.gov/ eden/) for each day of sampling in 2002 and subtracted the measured water depth from the stage measurement at each site to obtain elevation of each sample point. We only used 2002 water depth data (n = 1131) to avoid any changes in elevation associated with walking transects and disturbing peat. 2.2. Nutrient accumulation mechanism To examine the applicability of the nutrient accumulation mechanism to sub-tropical patterned peatlands, we tested for differences in nutrient concentrations between ridge and sloughs using nutrient data collected for a prior study examining the effects of enrichment throughout all of Water Conservation Area 3A (Bruland et al., 2007). Levene’s statistic for data restricted to our study site (WCA3AS) indicated that variances were not homogenous, so we compared nutrient concentrations in the top 10 cm of peat with a Kruskal-Wallis one way analysis of variance test to determine significant differences between community types: ridge, prairie, and slough (n = 184). We analyzed phosphorus instead of a suite of nutrients because the Everglades is an extremely phosphorus-limited system (McCormick et al., 1996). 27 Ecological Modelling 371 (2018) 25–36 C.L. Zweig et al. We ran the model 200 times to capture random variance within the model for 100 years under three different hydrologic scenarios: historic, 90% of historic depths, and 110% of historic depths. We used the last 20 years of hydrologic data from Site 64 in WCA3AS (EDEN) and repeated it four times to create a 100 year hydrologic dataset, which causes some periodicity in the data. Site 64 represents the hydrology of the best conserved area of the RSL (Watts et al., 2010). Differences between hydrology for vegetation communities were calculated by adding 10 cm to Site 64 depth data (in a slough) for wet prairie and 20 cm for the ridge, so the same rainfall/water management pattern would produce different hydroperiods, minimums, and maximums. Although low, 20 cm is within the range of microtopological differences between ridge and slough in WCA3AS (Watts et al., 2010) and typical of our dataset. Wet prairie was assumed to be in between these elevations at 10 cm below the ridge. Of the environmental scalars for production (I, NS, T, t, Shydro, Smin, and Smax), all but T and t were affected by the different hydrologic regimes. We did not test the sensitivity of the model to temperature, as typical temperatures for our area do not normally fall higher than T or lower than t. We did want to have temperature as part of the model for future climate scenarios. To examine the difference between aboveground and aboveground/ belowground biomass as a surface element in a bimodal analysis for the peat accumulation mechanism, we also ran each peat deposition model (ridge, prairie, slough) 50 times with randomly assigned initial biomasses for each species. The random biomasses were selected from field data values and the initial species ratios were consistent with previous model runs. As with the field data above, we fit uni- or bimodal normal distributions to the aboveground-only data and the aboveground/belowground ratio data to test for evidence of sharp microtopographical differences and to determine if adding belowground biomass would influence modality. 2.3. Peat accumulation model We developed a closed-form peat accretion model for the RSL in Systems Thinking Experimental Learning Laboratory with Animation (STELLA v. 10.0.4, ISEE systems, Lebanon, New Hampshire, USA) for two purposes. One, to understand how different species within vegetation communities contribute to peat deposition and elevation, including belowground biomass. And two, to model the difference between peat deposition (and elevation) when the nutrient accumulation mechanism is applied to determine if this feedback was reflected in the Everglades landscape. Our STELLA models were built from the conceptualization of the mechanics (Fig. 2) of peat deposition in three different vegetation communities typical to the RSL: ridge, wet prairie, and slough. These simulate peat deposition at an annual time step and as a function of above- and belowground biomass for each particular species within each community type within a 30 m × 30 m area. Each model only had a few dominant species available for production and growth. The ridge community contained C. jamaicense and Bacopa caroliniana as an edge species. Wet prairie and slough models contained the same species, but were initialized with different biomasses: Nymphaea odorata, Utricularia spp., B. caroliniana, Eleocharis cellulosa, and Eleocharis elongata. We developed two types of dynamic simulations: one with nutrient accumulation to capture scale-dependent feedback and one without any nutrient differences between vegetation communities to reflect only hydrodynamic control. The main driving state variable was biomass, which was a function of production. Belowground biomass was estimated as a simple ratio of aboveground biomass. Aboveground biomass (B) began as an initial condition defined as the average mass of the species in the community from our field data. Annual production (P) was simulated as a logistic function (Grace, 2001) of biomass and what we suggest are critical environmental scalars (air temperature, hydroperiod, maximum water depth, and minimum water depth). Inundation (I) was only applied to C. jamaicense and Eleocharis spp (Fig. 2), but at varying water depths for the different vegetation types—Eleocharis spp was configured to be less affected by deeper water than C. jamaicense. For simplicity, the environmental scalars for production I, NS, T, t, Shydro, Smin, and Smax (Table 1) were combined into one term, denoted by ‘Scal’ in the equation below. Production for each species was summed for community P, then is added to standing biomass (B): Senescence (S) controled the flow from B to litter (L) and was a function of B, P, and for wet prairie/slough communities only (Fig. 2): dry-down loss (dd). Slough species, such as N. odorata, was simulated to lose significant amounts of aboveground biomass when water depths are shallow. Dry-down loss was an IF-THEN statement dependent on hydroperiod (H), but cannot lose more biomass than B – S. Production, standing biomass, and the ratio of B to maximum biomass (Bmax) controlled senescence(s): ( ( )*0.5*B ).Biomass S = (P *randomnumberbetween0.9and1) + B Bmax 3. Results 3.1. Peat accumulation mechanism 3.1.1. Elevation data The lowest elevations occurred in the SE quadrant, near 190 cm above sea level, with the highest elevations occurring in the north quadrants (Fig. 3). Overall, the pooled landscape elevation data fit a bimodal distribution model better than unimodal distribution and that trend continued for the NW, NE, and SW quadrants (Table 2). In the SE, the most impounded and lowest of the quadrants, the elevation data was distinctly unimodal (Fig. 3). 3.1.2. Biomass data The biomass data distribution was heavily right-tailed, with the highest density of data points occurring between 0 and 10 g/0.25 m2 for all quadrants (Fig. 4), and could not be modeled to fit a normal distribution—bimodal or unimodal. Model fit analyses for a skewed-t distribution indicated that biomass was unimodal in the northern quadrants and bimodal in the SE (Table 3). The ΔBIC for model fit in the SW was less than 1.0 and evidence for one model over the other is not strong (Kass and Raftery, 1995). Separating vegetation communities (Fig. 5) demonstrates that the heavy right tail of the biomass distribution is caused by sawgrass ridges and that they have relatively even distributed weights. from senescence that flowed into litter (L), was then split between 3 processes: 1) gaseous/dissolved transport (decomposition), 2) particulate transport, and 3) accretion. These three processes were dependent on the species in the community, as they differ in their lability and lignin content (Osborne et al., 2007). Particulate transport was a transformation term for flocculent material that was then moved out of the area by flow; a flow of materials would come to bear in a spatial model. Gaseous/dissolved transport was governed by temperature, hydroperiod (anoxic status), and nutrient status (+10% production for ridge center, 8% for ridge edge, and 5% for wet prairie). Peat was deposited from the below- and aboveground accretion processes, but had one outflow: oxidation (O). Accreted peat can be oxidized by exposure to air and by microbial processes, so some proportion of peat was lost to oxidation each year, depending on the hydrologic conditions. 3.2. Nutrient accumulation mechanism There was a significant difference (Kruskal-Wallis test, p < 0.001) in total phosphorus between ridge (average 0.43 mg/kg ± 0.013 SE) and slough (0.35 mg/kg ± 0.015 SE), but not between either ridge or slough and prairie (0.36 mg/kg ± 0.043 SE). There were only 8 prairie samples which could account for the higher standard error in the prairie data. The difference between total phosphorus in ridge and slough 28 Ecological Modelling 371 (2018) 25–36 C.L. Zweig et al. Table 1 Peat deposition model variables and definitions from Water Conservation Area 3A South, FL USA. Symbol Variable Definition Type and Source P B AG Prod AG Biomass Yearly production Total biomass of all species in community I Inundation Slows C. jamaicense production in wet years T Maxtemp scalar prod Threshold to account for maximum air temperature t Mintemp scalar Threshold to account for minimum air temperature Shydro Vegsprodhydro scalar Scalar species production by hydroperiod Smin Vegsprodmin scalar Scales species production by minimum water depth Smax Vegsprodmax scalar Scales species production by maximum water depth S AG senesc Grams of biomass that turns over into litter dd Agdrydownloss A function to control for loss of species biomass from drydowns Min Max H L F Annual min Annual max Hydroperiod AG Litter Particulate transport Annual minimum water depth Annual maximum water depth Number of days in a year where water depth is > 0 Amount of senesced vegetation each year Amount of carbon that is removed from model by flocculent transport A Peat D AG Accret Peat GaseousDissolved Transport Annual aboveground accretion Annual amount of peat deposited Amount of carbon that is removed from model by microbial respiration NS NutrientStatus Only used in scale-dependent models: increases production by position in ridge NSd NutrientStatus2 Only used in scale-dependent models: increases decomposition by position in ridge Maxd Maxtemp scalar decomp H2 Hydroperiod scalar 2 Hloss Hydropd Loss 1 O Oxidation PBG BG Prod Threshold that accounts for increases decomposition with increased air temperatures Decreases decomposition if hydroperiod > 364. Represents anoxic conditions Scalar for oxidation by hydroperiod. When hydrperiod is high, scalar decreases oxidation Represents amount of carbon lost when peat is exposed to oxygen and microbial oxidation Ratio of AG prod. Changes with species BBG BG Biomass SBG BG Senesc Total belowgaround biomass of all species in community Grams of biomass that turns over into litter Logistic growth equation (Grace, 2001). Initial biomass for 900 m2 area, receives flows from production. Field data; Zweig and Kitchens (2008, 2009): 3500–315,000 g varying by species and community IF-THEN statement (Newman et al., 1996; Pezeshki et al., 1996; Childers et al., 2006). Slows production incrementally by minimum and maximum depths. IF-THEN statement, slows down production by 10% if temperatures are over 35.6 °C. Steward and Ornes 1985. Max temperature was chosen to be above the average high (28 °C) for sawgrass distribution and for the Everglades. IF-THEN statement, slows down production by 10% if temperatures are under 4.4 °C. Steward and Ornes 1985. Max temperature was chosen to be below the average low (8–9 °C) for sawgrass distribution and for the Everglades. Graphical function between 0 and 1. Curves were drawn for each species using their tolerance for inundation recorded in an extensive review (Richards and Gann, 2008). Graphical function between 0 and 1. Curves were drawn for each species using maximium conditions recorded in an extensive review (Richards and Gann, 2008). Graphical function between 0 and 1. Curves were drawn for each species using their tolerance for minimum water depths recorded in an extensive review (Richards and Gann, 2008). Random amount of production (between 90 and 100%) + a proportion of standing biomass (Davis et al., 2006) IF-THEN statement as a function of min depth and hydroperiod. This accounts for the tendency of aquatic plants (such as N. odorata) to lose all aboveground biomass when water depths are below the surface for a period of time. User-defined depth. Defined by users from real or modeled hydrology. User-defined depth. Defined by users from real or modeled hydrology. User-defined hydroperiod. Defined by users from real or modeled hydrology. Stock variable. Initialized at 0, receives flows from AG senesc. N/A 20–40% of AG Litter, depending on lability of species. Determined from decompositions rates from literature (Serna et al., 2013; DeBusk and Reddy 1998; Penton and Newman, 2008; Rubio and Childers 2006). AG Litter – Particulate transport – GaseousDissolved Transport. N/A Initialized at 0, receives flows from AG and BG Accret. N/A AG Litter and scaled by, temperature, nutrient status, and hydroperiod. Determined from decompositions rates from literature (Serna et al., 2013; DeBusk and Reddy 1998; Penton and Newman, 2008; Rubio and Childers 2006). IF-THEN statement that increases production for ridge center, less for ridge edge, and even less for prairie. Allows us to test the positive feedback hypothesis for peat patterning (Rietkerk et al., 2004). Ridge can be designated interior or edge and this scalar increases production by a small amount (Chambers and Pederson, 2006; Ewe et al., 2006). IF-THEN statement that increases decomposition as a function of available nutrients for ridge center, less for ridge edge, and even less for prairie. Decomposition is increased by a certain amount depending on spatial position (interior ridge, ridge edge, wet prairie) to account for higher nutrient rates and accelerated decomposition (Davis, 1991; DeBusk and Reddy, 1998) IF-THEN statement. Increases decomposition if the maximum temperature is above 35.6 °C IF-THEN statement. Decomoposition is decreased if the marsh is inundated > 364 days limiting aerobic conditions (DeBusk and Reddy, 1998). Graphical function between 0 and 1. Function was drawn to reflect loss of peat when exposed to air (Stephens, 1956) Function of Peat and Hydropd Loss 1. Represents flow of peat from Peat stock as governed by Hydropd Loss 1 Function of AG Prod by species AB/BG ratio. Root/shoot ratio determined by species from literature: Miao and Zou (2012), Serna et al. (2013), Daoust and Childers (1998), Edwards et al. (2003), Busch et al. (1998) Initial biomass, receives flows from production. N/A LBG ABG DBG BG Litter BG Accret BG Decomp Amount of senesced vegetation each year Annual belowground accretion Same as AG Gaseous Dissolved Transport Random amount of production (between 90 and 100%) + a proportion of standing biomass. Identical to aboveground biomass senescence equation Initialized at 0, receives flows from BG senesc. N/A BG Litter – GaseousDissolved Transport. N/A Function of lability of species, temperature, nutrient status, and hydroperiod. Determined from decompositions rates from literature. (Serna et al., 2013; DeBusk and Reddy, 1998; Penton and Newman, 2008; Rubio and Childers, 2006) 3.3. Peat accumulation model communities was statistically significant, but has not shown to have a significant effect on vegetation production (Chambers and Pederson, 2006; Ewe et al., 2006) between the two communities. This was reflected in the nutrient status runs of the dynamic simulation. Annual biomass estimates, litter fall, and production by species, were within realistic ranges (one standard deviation from the average; Table 4) of our field data and also within the ranges in published literature (Craft et al., 1995; Davis et al., 2006; Miao and Zou, 2012; 29 Ecological Modelling 371 (2018) 25–36 C.L. Zweig et al. Fig. 3. Histograms of elevation data (cm above sea level) in WCA3AS, FL USA by quadrant. The red line indicates the kernel density for the data and blue lines represent the mode(s) for the best fit model. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) different from each other. In the nutrient runs (Fig. 6b), the middle of the ridge had increased production (and increased decomposition) from the non-status runs, the edge had some increased production, wet prairies even less, and sloughs had no increased production. The ridge edge and middle were not significantly different from each other (Kruskal-Wallis, p = 0.16), and the wet prairie (0.0078 mm/ yr ± 0.00045 SE) produced approximately the same as the slough (0.0080 mm/yr ± 0.00041 SE). The 110% hydrology was similar to the hydrology in the impounded SE corner of WCA3AS and wet prairie produced the most biomass and peat deposition. While this seemed counterintuitive, it appears to be due to a species shift. The community composition of this ‘wet prairie’ at the end of 100 years of high water depths was more similar to the definition of slough than any other community (Fig. 7). The model was able to simulate a transformation between communities under certain conditions, but only wet prairie and slough because they shared model species. This also provided validation of the model design, as it captured the prairie species decline in biomass with wet conditions and slough species increase (Zweig and Kitchens, 2008). In the analysis of modeled aboveground versus aboveground/ Table 2 Model fit statistics (Bayes’ information criteria values (BIC)) for bimodality of elevation in Water Conservation Area 3A South, FL USA. * Indicates model with lowest BIC value (unimodal or bimodal). + Indicates no real difference between uni- and bimodal fit. Data set Mean 1 Mean 2 σ1 σ2 BIC All* All SWQuad* SWQuad NWQuad* NWQuad NEQuad* NEQuad SEQuad SEQuad* 228.09 207.50 212.08 197.56 226.05 220.24 228.46 216.73 189.44 188.84 194.61 142.88 415.38 405.94 217.02 146.62 210.50 140.74 421.56 98.81 149.81 154.85 9970.43 10043.65 2402.86 2419.19 2759.04 2770.78 2488.00 2542.63 1720.63 1705.75 194.30 203.80 190.85 188.36 116.85 25.65 68.40 192.45 Δ BIC 73.22 16.33 11.74 54.63 −14.88 Serna et al., 2013), providing model validation. In the historic hydrology model, sawgrass ridges deposited twice the amount of peat (Fig. 6a) than wet prairie or slough (Kruskal-Wallis, p < 0.001), and the wet prairie and slough were not significantly 30 Ecological Modelling 371 (2018) 25–36 C.L. Zweig et al. Fig. 4. Histograms of biomass data (g/0.25 m2) in WCA3AS, FL USA by quadrant. The red line indicates the kernel density for the data and blue lines represent the mode(s) for the best fit model. The green line for the SE quadrant indicates the second mode, but the difference between uni- and bimodal is not distinct. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) belowground biomass did not affect the modality of the data. Table 3 Model fit statistics (Bayes’ information criteria values (BIC)) for bimodality of plant biomass in Water Conservation Area 3A South, FL USA. * Indicates model with lowest BIC value (unimodal or bimodal). + Indicates no real difference between uni- and bimodal fit. Data set Mean 1 Mean 2 σ1 σ2 BIC All* All SWQuad+ SWQuad+ NWQuad NWQuad* NEQuad NEQuad* SEQuad* SEQuad 7.81 6.02 6.42 4.30 9.03 7.42 9.00 7.36 6.92 4.38 115.28 2869.11 4372.62 2719.57 4478.24 3001.34 5214.84 2210.59 3599.61 2809.02 5339.87 23429.68 62969.10 63064.56 17743.53 17742.82 19949.80 19946.13 13062.66 13057.01 12294.19 12313.55 98.63 107.60 95.99 124.48 25486.77 19021.78 17156.48 35440.70 4. Discussion Δ BIC While there was initial evidence that sub-tropical patterned peatlands processes were similar to boreal patterned peatlands (Watts et al., 2010), our ecological understanding of the RSL sub-tropical system (Zweig and Kitchens, 2008; Zweig and Kitchens, 2009; Zweig and Kitchens, 2014) indicated that differences in climate should significantly affect structuring processes and these patterned peatlands would not conform to current patterning theories. Most existing literature focuses on the theory of scale-dependent feedback, specifically the peat and nutrient accumulation mechanisms, to describe both microtopographical differences and distinctive patterns in flat peatlands (Belyea and Clymo, 2001; Eppinga et al., 2009a; Rietkerk et al., 2004; Rietkerk and Van de Koppel, 2008). We used a combination of field data and modeling to ascertain their applicability in sub-tropical patterned peatlands. 95.46 −0.71 −3.67 −5.65 19.36 belowground biomass, the aboveground biomass fit a unimodal normal distribution better than a bimodal (ΔBIC = −15) and the aboveground/belowground biomass ratio, with a skewed t distribution, also fit a unimodal distribution (ΔBIC = −19). The addition of 31 Ecological Modelling 371 (2018) 25–36 C.L. Zweig et al. Fig. 5. Kernel density estimate graphs for biomass by quadrant and vegetation community type in WCA3AS, FL USA. Black = slough, blue = wet prairie, and green = ridge. similar to boreal peatlands (Eppinga et al., 2008). The biomass bimodality tests had to include skewed distributions, not the typical normal distributions (Eppinga et al., 2008), as the data was heavily right-tailed. Even with a skewed-t distribution, which was a better model fit than a normal distribution, the only bimodal quadrant was the southeast. This southeastern, impounded end of the RSL has lost a majority of its wet prairies (Zweig and Kitchens, 2014; Zweig et al., 2011) and it may be the wet prairie biomass, bridging the lower biomass of the sloughs and higher biomass of the ridges, that creates a unimodal distribution in RSL field data (Fig. 5). We investigated the possibility that aboveground biomass, while easier to collect, is not appropriate as a surface characteristic and that we were missing critical information by not including belowground biomass. However, our peat accumulation model analysis determined that it was not a lack of data that created a unimodal distribution and the addition of belowground biomass did not affect its modality. Table 4 Summary statistics for field and model output biomass (g/m2). Real Data Community Ridge Prairie Slough Modeled Data Community Ridge Prairie Slough Average Biomass 605.6 266.2 190.76 Min Biomass Average Biomass 632.84 457.64 448.44 Min Biomass 0.2 2.84 0.04 354.52 142.92 153.76 Max Biomass 3985.04 1692.44 1975.32 Median biomass 466.52 224.68 153.32 Max Biomass 828.04 809.56 666.44 Median biomass 643.28 441.96 430.8 4.1. Peat accumulation mechanism According to the peat accumulation mechanism (Eppinga et al., 2008; Rietkerk et al., 2004), distinct states in ridge and hollow landscapes should be evident from the bimodality of surface characteristics (Eppinga et al., 2008; Eppinga et al., 2009b; van de Koppel et al., 2001). Elevation data (representing peat depth) from across the landscape within WCA3AS indicated that there are sharp microtopographical differences (Fig. 3) in the three less-inundated quadrants (NW, NE, and SW) and no statistically significant bimodality in the quadrant with the highest impoundment (SE). This concurs with other RSL studies (Watts et al., 2010; Zweig et al., 2011) and provides additional evidence of the degradation of the RSL pattern by impoundment and sustained high water. However, aboveground biomass, another surface characteristic, was unimodal, even though it seemed logical to assume that a sawgrassdominated ridge would have more standing biomass than deep sloughs, 4.2. Reasons for deviations 4.2.1. Disturbance There are two likely ecological reasons as to why biomass did not reflect the peat accumulation mechanism through bimodality within the Everglades. First is landscape instability and anthropogenic disturbance. The lack of agreement between bimodality of elevation and biomass in the RSL might be indicative of degradation in the landscape pattern—first seen in the vegetation and, later through peat deposition, in elevation. Like many wetlands, large sections of the Everglades RSL have been both over-drained and inundated for extensive time periods (Davis and Ogden, 1994; Givnish et al., 2008; Sklar and Van Der Valk, 2002; Troxler and Richards, 2009), each having a similar effect on the landscape pattern, namely a reduction of elevation differences between ridge and slough (Givnish et al., 2008; McVoy et al., 2011; Watts et al., 32 Ecological Modelling 371 (2018) 25–36 C.L. Zweig et al. Fig. 6. Peat deposition model results for WCA3AS, FL USA. a) Model with no nutrient (scale-dependent feedback) effects. b) Model with nutrient (scale-dependent feedback) effects. system (McVoy et al., 2011), areas of wet prairie are expected to transition to slough (McVoy et al., 2011), but according to the peat deposition models here, the disappearance of wet prairies could be followed by a microtopographical collapse like the southeastern quadrant of WCA3AS (Fig. 6). Wet prairies may not be a transitional community between two stable states, but a stable state of its own, capable of maintaining its position on the landscape. Drying events that encourage the presence of wet prairies have occurred within WCA3AS, but the most constant disturbance factor for the last 90 years has been inundation from pooling against the southern/eastern levees (Zweig et al., 2011). Pooling is particularly apparent in the southeastern quadrant which has the additive effect, even over the southwest quadrant, of lower elevations. Decreased peat accretion on the ridges from decreased sawgrass productivity under a flooded regime (Childers et al., 2006) and increased peat accretion in the sloughs from constant inundation and lack of flow act to reduce the elevation difference between ridge and sloughs. This trend from the literature is also reflected in the unimodality of elevation in the 2010; Zweig et al., 2011). Literature has pointed to the presence of wet prairies on the landscape as an artifact of hydrologic degradation, not a true community state in the RSL. Prairies are referred to as a transitional community between more stable ridges and sloughs (Heffernan et al., 2013; McVoy et al., 2011; Watts et al., 2010), arising when sloughs are over-drained and short-stature emergent vegetation can dominate. If this hypothesis is correct, then the current, observed unimodal nature of aboveground biomass is a precursor to microtopographical collapse, where a feedback of drier conditions that encourage wet prairie growth and increased peat deposition from wet prairies will reduce the elevational difference between ridge and slough. However, according to the peat deposition model, for either type of historic hydrology (scale-dependent or not), wet prairies deposited approximately the same amount of peat as a slough in 100 years, but much less than a ridge; maintaining their position and stability on the landscape relative to the other communities. If more water were added to the system, as has been suggested to mimic a pre-drainage Everglades 33 Ecological Modelling 371 (2018) 25–36 C.L. Zweig et al. Fig. 7. Change of biomass (g/m2) by species from slough and wet prairie peat deposition models for WCA3AS, FL USA, including two hydrologic histories to demonstrate when changes occur in community composition. a) Wet prairie biomass over 100 years with 110% historic hydrology. b) Slough biomass over 100 years with 110% historic hydrology. c) Wet prairie biomass over 100 years with historic hydrology. d) Slough biomass over 100 years with historic hydrology. peatland. Vegetation communities, seasonality, nutrient dynamics, and structuring process are very different between boreal and sub-tropical wetlands. Processes in boreal peatlands are influenced by colder temperatures that slow decomposition and induce frost heaving (Brown, 1970; Moore and Bellamy, 1974), and they are also acidic systems (Eppinga et al., 2009b; Siegel et al., 2006; van Breemen, 1995) whose vegetation communities are largely dominated by one non-vascular genus, Sphagnum (Moore and Bellamy, 1974; van Breemen, 1995). The presence of varied, vascular vegetation communities in the Everglades (Zweig and Kitchens, 2008; Zweig and Kitchens, 2009) and their fate in the process of decomposition and peat production could likely to be the main factor contributing to the difference between the bimodality of elevation and biomass. Difference in elevation between the ridges and sloughs lies, not in the amount of biomass, but how that biomass (above- and belowground) transforms into peat through decomposition. A recent study (Serna et al., 2013) suggests that decomposition rates of RSL vascular plants are influenced more by species than habitat or water depth, a sharp contrast to models from the boreal peatlands where decomposition is a function of water levels (Eppinga et al., 2009a; Eppinga et al., 2010; Rietkerk et al., 2004). Greater net peat production on ridges is not being reinforced by greater plant production, as is suggested by the lack of bimodality of biomass, but by the characteristics of the species that grow on ridges. In wet and dry conditions, C. jamaicense was the slowest to decompose, over E. cellulosa and N. odorata, due to more recalcitrant material in live and standingdead biomass (Serna et al., 2013). This is supported in the paleo record by greater peat accumulation during drier eras of the Holocene that were a C. jamaicense-dominated environment (Jones et al., 2014). Slough vegetation with less recalcitrant biomass, such as N. odorata, provide high quality substrate for microbial communities and increases inundated southeastern quadrant (Fig. 3) and peat deposition model results (Fig. 6). In the 110% historic hydrology—a hydrology more similar, but slightly shallower, than what the southeastern quadrant of WCA3AS has experienced in the last 20 years—the wet prairie transitions to a slough and the slough community is depositing more peat than the ridge, reducing the microtopography of the system. Results from the dynamic simulations also suggest that the peat accumulation mechanism—drier ridges produce more vascular peat biomass, which increases the elevation of the ridge and encourages more plant growth that translates into sharp microtopographical differences—may not apply to the RSL. Increased vascular plant growth on ridges does not appear to be the mechanism for increased ridge elevation according to the bimodality analysis. In the peat deposition model, during the historic hydrologic regime—a hydrology from the middle and northern ‘conserved’ (Watts et al., 2010) sections of WCA3AS—the ridge deposits more peat than wet prairies or sloughs (enough to create a sharp, microtopographical difference in ridges and slough elevation), however it does not contain a sharp increase in biomass. The biomass of ridges was evenly distributed (Fig. 5), likely because C. jamaicense is ubiquitous and can persist in areas where conditions are not ideal (Zweig and Kitchens, 2008). Field data also shows that wet prairies contain a moderate amount of biomass (Fig. 5). This suggests that the mechanism for sharp microtopographical differences in the RSL may be the differential decomposition of vegetation community species not differences in plant production. 4.2.2. Climactic differences This leads to the second, underlying ecological possibility for a unimodal distribution in a RSL surface characteristic: a fundamental difference between the vegetation of a sub-tropical versus a boreal 34 Ecological Modelling 371 (2018) 25–36 C.L. Zweig et al. theory to explain patterning in the RSL or other sub-tropical peatlands, but we have contributed an increased understanding of species influence on net peat accretion and present a deviation from current borealbased mechanisms. decomposition rates (Serna et al., 2013). Even without sharp distinctions in the amount of biomass between ridge and slough, species harvested in slough/transition transects (N. odorata, Utricularia sp., Eleocharis sp.) would degrade more completely than C. jamaicense (DeBusk and Reddy, 1998; Larsen et al., 2011; Penton and Newman, 2008), contributing to less net peat deposition and maintaining the sharp microtopographical difference. In this manner, not all surface elements of the RSL conform to the peat accumulation mechanism but still express a sharp difference in peat accretion between ridge and slough states. Acknowledgements We would like to acknowledge M. Nungesser anad S. Gray for their careful review. References 4.3. Nutrient accumulation mechanism Aerts, R., 1997. Climate, leaf litter chemistry and leaf litter decomposition in terrestrial ecosystems: a triangular relationship. Oikos 439–449. Belyea, L.R., Clymo, R., 2001. Feedback control of the rate of peat formation. Proc. R. Soc. Lond. B: Biol. Sci. 268, 1315–1321. Belyea, L.R., Lancaster, J., 2002. Inferring landscape dynamics of bog pools from scaling relationships and spatial patterns. J. Ecol. 90, 223–234. Bernhardt, C.E., Willard, D.A., 2009. Response of the Everglades ridge and slough landscape to climate variability and 20th-century water management. Ecol. Appl. 19, 1723–1738. Brown, R.J.E., 1970. Occurrence of Permafrost in Canadian Peatlands. National Research Council of Canada, Division of Building Research, Ottawa, Canada. Bruland, G.L., Osborne, T.Z., Reddy, K., Grunwald, S., Newman, S., DeBusk, W.F., 2007. Recent changes in soil total phosphorus in the Everglades: Water Conservation Area 3. Environ. Monit. Assess. 129, 379–395. Busch, D.E., Loftus, W.F., Oron, L., Bass, J., 1998. Long-term hydrologic effects on marsh plant community structure in the southern Everglades. Wetlands 18, 230–241. Chambers, R.M., Pederson, K.A., 2006. Variation in soil phosphorus, sulfur, and iron pools among south Florida wetlands. Hydrobiologia 569, 63–70. Cheng, Y., Stieglitz, M., Turk, G., Engel, V., 2011. Effects of anisotropy on pattern formation in wetland ecosystems. Geophys. Res. Lett. 38, L04402. Childers, D.L., Iwaniec, D., Rondeau, D., Rubio, G., Verdon, E., Madden, C.J., 2006. Responses of sawgrass and spikerush to variation in hydrologic drivers and salinity in Southern Everglades marshes. Hydrobiologia 569, 273–292. Couteron, P., Lejeune, O., 2001. Periodic spotted patterns in semi-arid vegetation explained by a propagation-inhibition model. J. Ecol. 89, 616–628. Couwenberg, J., Joosten, H., 2005. Self-organization in raised bog patterning: the origin of microtope zonation and mesotope diversity. J. Ecol. 93, 1238–1248. Craft, C.B., Vymazal, J., Richardson, C.J., 1995. Response of Everglades plant communities to nitrogen and phosphorus additions. Wetlands 15, 258–271. Davis, S.M., 1991. Growth, decomposition, and nutrient retention of Cladium jamaicense Crantz and Typha domingensis Pers. in the Florida Everglades. Aquat. Bot. 40, 203–224. Davis, S.M., Ogden, J.C., 1994. Everglades: the Ecosystem and Its Restoration. CRC Press, Boca Raton, FL, USA. Davis III, S.E., Childers, D.L., Noe, G.B., 2006. The contribution of leaching to the rapid release of nutrients and carbon in the early decay of wetland vegetation. Hydrobiologia 569, 87–97. DeBusk, W., Reddy, K., 1998. Turnover of detrital organic carbon in a nutrient-impacted Everglades marsh. Soil Sci. Soc. Am. J. 62, 1460–1468. Edwards, A.L., Lee, D.W., Richards, J.H., 2003. Responses to a fluctuating environment: effects of water depth on growth and biomass allocation in Eleocharis cellulosa Torr. (Cyperaceae). Can. J. Bot. 81, 964–975. Eppinga, M.B., Rietkerk, M., Borren, W., Lapshina, E.D., Bleuten, W., Wassen, M.J., 2008. Regular surface patterning of peatlands: confronting theory with field data. Ecosystems 11, 520–536. Eppinga, M.B., De Ruiter, P.C., Wassen, M.J., Rietkerk, M., 2009a. Nutrients and hydrology indicate the driving mechanisms of peatland surface patterning. Am. Nat. 173, 803–818. Eppinga, M.B., Rietkerk, M., Wassen, M.J., De Ruiter, P.C., 2009b. Linking habitat modification to catastrophic shifts and vegetation patterns in bogs. Plant Ecol. 200, 53–68. Eppinga, M.B., Rietkerk, M., Belyea, L.R., Nilsson, M.B., Ruiter, P.C.D., Wassen, M.J., 2010. Resource contrast in patterned peatlands increases along a climatic gradient. Ecology 91, 2344–2355. Ewe, S.M., Gaiser, E.E., Childers, D.L., Iwaniec, D., Rivera-Monroy, V.H., Twilley, R.R., 2006. Spatial and temporal patterns of aboveground net primary productivity (ANPP) along two freshwater-estuarine transects in the Florida Coastal Everglades. Hydrobiologia 569, 459–474. Foster, D., King, G., Glaser, P., Wright, H., 1983. Origin of string patterns in boreal peatlands. Nature 306, 256–258. Givnish, T.J., Volin, J.C., Owen, V.D., Volin, V.C., Muss, J.D., Glaser, P.H., 2008. Vegetation differentiation in the patterned landscape of the central Everglades: importance of local and landscape drivers. Global Ecol. Biogeogr. 17, 384–402. Glaser, P.H., 1992. Ecological development of patterned peatlands. In: Wright, H.E., Coffin, B.A., Aaseng, N.E. (Eds.), The Patterned Peatlands of Minnesota. University of Minnesota Press, Minneapolis, MN, USA, pp. 27–43. Grace, J.B., 2001. The roles of community biomass and species pools in the regulation of plant diversity. Oikos 92, 193–207. Heffernan, J.B., Watts, D.L., Cohen, M.J., 2013. Discharge competence and pattern formation in peatlands: a meta-ecosystem model of the everglades ridge-slough We also tested the application of the nutrient accumulation mechanism to explain pattern regularity in the sub-tropical RSL. Evidence for this mechanism included a concentration of nutrients under ridges, as higher ET rates on ridges pulls in more water/nutrients and concentrates them in the ridge (Rietkerk et al., 2004). There was a significant difference between total soil phosphorus in the ridge and slough in WCA3AS, but as with biomass, the wet prairie acted as an intermediate value between the two. This concurs with the nutrient accumulation mechanism, as the nutrient concentrations are higher in the ridge and decrease as you move out of the ridge into the wet prairie and lowest in the sloughs. This has previously been demonstrated in the Everglades, but for tree islands (Ross et al., 2006; Wetzel et al., 2005). However, in other parts of the Everglades, ridge and slough samples were not significantly different (Ross et al., 2006). While the difference between total phosphorus in ridge and slough communities in WCA3AS is statistically significant in some areas, it does not appear to have a significant effect on vegetation production (Chambers and Pederson, 2006; Ewe et al., 2006) between the two communities. We hypothesize that nutrient differences are not necessarily due to higher ET rates and accumulation, but again, the difference of species within the vegetation communities. Both C. jamaicense and E. cellulosa reabsorb phosphorus and nitrogen from dead leaves (40–80%) before they fall, efficiently retaining nutrients within the plant (Serna et al., 2013). Even without differential ET concentrating extra nutrients into the ridge, a community dominated by C. jamaicense and/or E. cellulosa would have higher soil nutrients from this process. What little nutrients that are lost in the litter by these species are consolidated into peat. In sloughs, N. odorata litter contained over 50% more phosphorus than live leaves (Serna et al., 2013). This high nutrient content encourages microbial decomposition and shortens the turnover time for phosphorus into the water column, not into the soil (non-floc layer). So, while the nutrient accumulation mechanism has been demonstrated on tree islands in the RSL, differences in species’ life histories, instead of increased ET and nutrient accumulation, may be a better explanation for ridges, wet prairies, and sloughs in the RSL. Larsen et al. (2015) found the mechanism for nutrient accumulation in the RSL to be a function of differential hydrologic exchange between ridges and sloughs during dry down periods which could compliment our theory of differential decomposition and nutrient cycling by species. 5. Conclusion While sub-tropical patterned peatlands initially appeared to concur with current peat patterning mechanisms, we have demonstrated that their underlying ecological patterning processes may be very different from boreal peatlands and requires further discussion. There has been a suite of literature modeling the origin and maintenance of the RSL by hydrology and flow (Givnish et al., 2008; Heffernan et al., 2013; Larsen et al., 2011; Larsen et al., 2007). Our model expands on the current theories for RSL maintenance, and is the first to incorporate vegetation types and life history traits into differential peat deposition that create the signature microtopographical differences. There is likely no single 35 Ecological Modelling 371 (2018) 25–36 C.L. Zweig et al. Richards J.H. and Gann D., Greater Everglades Sub-Team Consulting Services to Determine Plant Community Depth and Hydroperiod Optima and Tolerances. Rietkerk, M., Van de Koppel, J., 2008. Regular pattern formation in real ecosystems. Trends Ecol. Evol. 23, 169–175. Rietkerk, M., Dekker, S., Wassen, M., Verkroost, A., Bierkens, M., 2004. A putative mechanism for bog patterning. Am. Nat. 163, 699–708. Ross, M.S., Mitchell-Bruker, S., Sah, J.P., Stothoff, S., Ruiz, P.L., Reed, D.L., Jayachandran, K., Coultas, C.L., 2006. Interaction of hydrology and nutrient limitation in the Ridge and Slough landscape of the southern Everglades. Hydrobiologia 569, 37–59. Rubio, G., Childers, D.L., 2006. Controls on herbaceous litter decomposition in the estuarine ecotones of the Florida Everglades. Estuaries Coasts 29, 257–268. Sakaguchi, Y., 1980. On the genesis of banks and hollows in peat bogs: an explanation by a thatch line theory. Bull. Dep. Geogr. Univ. Tokyo 12, 35–58. Seppälä, M., Koutaniemi, L., 1985. Formation of a string and pool topography as expressed by morphology stratigraphy and current processes on a mire in Kuusamo, Finland. Boreas 14, 287–309. Serna, A., Richards, J.H., Scinto, L.J., 2013. Plant decomposition in wetlands: effects of hydrologic variation in a re-created Everglades. J. Environ. Qual. 42, 562–572. Siegel, D., Glaser, P., So, J., Janecky, D., 2006. The dynamic balance between organic acids and circumneutral groundwater in a large boreal peat basin. J. Hydrol. 320, 421–431. Sklar, F.H., Van Der Valk, A., 2002. Tree Islands of the Everglades. Kluwer Academic Publishers, Norwell, MA, USA. Stephens, J.C., 1956. Subsidence of organic soils in the Florida Everglades. Soil Sci. Soc. Am. J. 20, 77–80. Swanson, D.K., Grigal, D.F., 1988. A simulation model of mire patterning. Oikos 309–314. Troxler, T.G., Richards, J.H., 2009. δ 13C, δ 15N, carbon, nitrogen and phosphorus as indicators of plant ecophysiology and organic matter pathways in Everglades deep slough, Florida, USA. Aquat. Bot. 91, 157–165. Watts, D.L., Cohen, M.J., Heffernan, J.B., Osborne, T.Z., 2010. Hydrologic modification and the loss of self-organized patterning in the ridge–slough mosaic of the Everglades. Ecosystems 13, 813–827. Wetzel, P.R., van der Valk, A.G., Newman, S., Gawlik, D.E., Troxler Gann, T., CoronadoMolina, C.A., Childers, D.L., Sklar, F.H., 2005. Maintaining tree islands in the Florida Everglades: nutrient redistribution is the key. Front. Ecol. Environ. 3, 370–376. Wu, Y., Wang, N., Rutchey, K., 2006. An analysis of spatial complexity of ridge and slough patterns in the Everglades ecosystem. Ecol. Complexity 3, 183–192. Zweig, C.L., Kitchens, W.M., 2008. Effects of landscape gradients on wetland vegetation communities: information for large-scale restoration. Wetlands 28, 1086–1096. Zweig, C.L., Kitchens, W.M., 2009. Multi-state succession in wetlands: a novel use of state and transition models. Ecology 90, 1900–1909. Zweig, C.L., Kitchens, W.M., 2014. Reconstructing historical habitat data with predictive models. Ecol. Appl. 24, 196–203. Zweig, C.L., Reichert, B.E., Kitchens, W.M., 2011. Implications of discontinuous elevation gradients on fragmentation and restoration in patterned wetlands. Ecosphere 2, art98. van Breemen, N., 1995. How Sphagnum bogs down other plants. Trends Ecol. Evol. 10, 270–275. van de Koppel, J., Herman, P.M., Thoolen, P., Heip, C.H., 2001. Do alternate stable states occur in natural ecosystems? Evidence from a tidal flat. Ecology 82, 3449–3461. landscape. PLoS One 8, e64174. Hilbert, D.W., Roulet, N., Moore, T., 2000. Modelling and analysis of peatlands as dynamical systems. J. Ecol. 88, 230–242. Jones, M.C., Bernhardt, C.E., Willard, D.A., 2014. Late Holocene vegetation, climate, and land-use impacts on carbon dynamics in the Florida Everglades. Quat. Sci. Rev. 90, 90–105. Kass, R.E., Raftery, A.E., 1995. Bayes factors. J. Am. Stat. Assoc. 90, 773–795. Klausmeier, C.A., 1999. Regular and irregular patterns in semiarid vegetation. Science 284, 1826–1828. Larsen, L.G., Harvey, J.W., 2010. How vegetation and sediment transport feedbacks drive landscape change in the Everglades and wetlands worldwide. Am. Nat. 176, E66–E79. Larsen, L.G., Harvey, J.W., Crimaldi, J.P., 2007. A delicate balance: ecohydrological feedbacks governing landscape morphology in a lotic peatland. Ecol. Monogr. 77, 591–614. Larsen, L., Aumen, N., Bernhardt, C., Engel, V., Givnish, T., Hagerthey, S., Harvey, J., Leonard, L., McCormick, P., McVoy, C., Noe, G., Nungesser, M., Rutchey, K., Sklar, F., Troxler, T., Volin, J., Willard, D., 2011. Recent and historic drivers of landscape change in the Everglades ridge, slough, and tree island mosaic. Crit. Rev. Environ. Sci. Technol. 41, 344–381. Larsen, L.G., Harvey, J.W., Maglio, M.M., 2015. Mechanisms of nutrient retention and its relation to flow connectivity in river–floodplain corridors. Freshwater Sci. 34, 187–205. McCormick, P.V., Rawlik, P.S., Lurding, K., Smith, E.P., Sklar, F.H., 1996. Periphytonwater quality relationships along a nutrient gradient in the northern Florida Everglades. J. North Am. Benthol. Soc. 15, 433–449. McVoy, C.W., Said, W.P., Obeysekera, J., VanArman, J.A., Dreschel, T.W., 2011. Landscapes and Hydrology of the Predrainage Everglades. University Press of Florida, Gainesville. Meentemeyer, V., 1978. Macroclimate and lignin control of litter decomposition rates. Ecology 59, 465–472. Miao, S., Zou, C.B., 2012. Effects of inundation on growth and nutrient allocation of six major macrophytes in the Florida Everglades. Ecol. Eng. 42, 10–18. Moore, P.D., Bellamy, D.J., 1974. Peatlands. Elek Science, London, England UK. Newman, S., Grace, J., Koebel, J., 1996. Effects of nutrients and hydroperiod on Typha, Cladium, and Eleocharis: implications for Everglades restoration. Ecol. Appl. 6, 774–783. Nungesser, M.K., 2003. Modelling microtopography in boreal peatlands: hummocks and hollows. Ecol. Modell. 165, 175–207. Nungesser, M.K., 2011. Reading the landscape: temporal and spatial changes in a patterned peatland. Wetlands Ecol. Manage. 19, 475–493. Ogden, J., 2005. Everglades ridge and slough conceptual ecological model. Wetlands 25, 810–820. Osborne, T.Z., Inglett, P.W., Reddy, K.R., 2007. The use of senescent plant biomass to investigate relationships between potential particulate and dissolved organic matter in a wetland ecosystem. Aquat. Bot. 86, 53–61. Penton, C.R., Newman, S., 2008. Enzyme-based resource allocated decomposition and landscape heterogeneity in the Florida Everglades. J. Environ. Qual. 37, 972–976. Pezeshki, S.R., DeLaune, R.D., Kludze, H.K., Choi, H.S., 1996. Photosynthetic and growth responses of cattail (Typha domingensis) and sawgrass (Cladium jamaicense) to soil redox conditions. Aquat. Bot. 54, 25–35. 36