Flooding affects food web structure and basal sources supporting sh guilds in a subtropical wetland and shallow lake


 Aquatic ecosystems exchange nutrients and organic matter with surrounding terrestrial ecosystems, and floods import allochthonous material from riparian areas into fluvial systems. We surveyed food web components of a wetland and shallow lake in a subtropical coastal region of Brazil to examine how community trophic structure and the entrance of allochthonous material into the food web were affected by floods. Stable isotope analysis was performed for samples of terrestrial and aquatic basal production sources and aquatic animals to trace the origin of organic matter assimilated by aquatic animals and estimate vertical trophic positions and food chain length. Lake and wetland trophic structures were compared for cool/wet and warm/dry seasons. Food web structure was hypothesized to differ based on hydrology, with the more stable lake having greater food web complexity, and seasonal flooding resulting in greater allochthonous inputs to the aquatic food web. We compared spatial and temporal variation in assemblage trophic structure using an adapted isotopic ellipse approach that plots assemblage elements according to δ13C on the x-axis and estimated TP on the y-axis. Lake trophic structure was more complex with longer food chains compared to that of the wetland. A greater contribution from terrestrial resources to animal biomass was observed in the wetland during the cool/wet period, and food chains in both habitats tended to be longer during the cool/wet period. Findings supported the hypothesis of greater assimilation of allochthonous sources during floods and greater trophic complexity in the more hydrologically stable system.


Introduction
A major issue in aquatic ecology is energy dynamics and the production and transfer of organic materials that sustains the communities and drives ecosystem functioning. Transfer of organic material between terrestrial and aquatic habitats is bidirectional, especially in uvial systems subject to hydrologic variation. Hydrologic variation affects geomorphology, sediment dynamics, habitat availability and connectivity, and resource and population dynamics. Several conceptual models predict the origin of organic matter supporting aquatic communities along lateral and longitudinal gradients of rivers (Vannote et al 1980;Junk et al 1989; Thorp and Delong 1994;Junk and Wantzen 2008), however, similar models are generally lacking for lakes and wetlands. Lakes vary in size, shape, depth, hydrology, thermal regime, productivity, as well as in the degree to which they exchange organic material, including organisms, with adjacent terrestrial systems.
Variation among these factors yields between-lake variation in trophic state (Carlson 1977) and food web structure (Wetzel 2001). Aquatic communities of small, shallow lakes tend to be supported by aquatic macrophytes, whereas in larger, deeper lakes, food webs are mostly supported by phytoplankton, with macrophytes contributing to variable degrees within littoral zones (Vander Zanden and Vadeboncoeur 2002; Kruger et al 2015). The major production sources supporting aquatic food webs may vary spatially within a lake. For example, Rodrigues et al. (2015) found different trophic states on northern and southern limits of a long and shallow lake in southern Brazil. They attributed those differences to wind-driven water currents affecting water quality and the dominant types of primary producers (i.e., macrophytes vs. phytoplankton) (Rodrigues et al 2015).
Terrestrial organic matter enters aquatic food webs via direct consumption of terrestrial animals (Nakano et al 1999) and plant material (Correa and Winemiller 2014) or as detritus (Vannote et al 1980). Allochthonous inputs usually are greater when ooding transports material or allows access for aquatic organisms between terrestrial and aquatic habitats (Junk et  By increasing physical connections between habitats, oods also affect vertical trophic structure, or the food chain length (Sabo et al 2018). Two major hypotheses have been proposed regarding the vertical trophic structure: the productivity hypothesis (Pimm and Lawton 1977;Briand and Cohen 1987) and the ecosystem size hypothesis (Cohen and Newman 1991;Post et al 2000). A relationship between average food chain length and community stability also has been proposed (Pimm 1982;Sabo et al 2010). There is some evidence that average food chain length is positively correlated with ecosystem productivity, size, and stability (Pimm and Lawton 1977;Post et al 2000;Sabo et al 2010).
Based on analysis of data from diverse ecosystems, Takimoto and Post (2013) found a strong but variable effect of ecosystem size on food chain length, no consistent effect of disturbance (inverse of stability), and a weak but fairly consistent effect of resource availability (productivity).
Ratios of stable isotopes of carbon ( 13 C/ 12 C, reported as δ 13 C) and nitrogen ( 15 N/ 14 N, reported as δ 15 N) have been used to describe food web structure (DeNiro and Epstein 1978; DeNiro and Epstein 1981). Consistent carbon isotopic differences have been shown between primary producers in pelagic and littoral zones of lakes and between hydrological conditions in wetlands (Wantzen et al 2002). The nitrogen isotopic ratio undergoes a fairly consistent shift between adjacent trophic levels and, therefore, provides a means to estimate trophic positions of consumers and the length of food chains (Post et al 2000;Post 2002). Therefore, the positions of organisms in biplots of isotopic space (e.g., δ 13 C vs. δ 15 N) have been used to compare trophic niche, similarity and breadth (Newsome et al 2007).
To test if seasonal oods are associated with assimilation of organic matter of terrestrial origin by aquatic animals and if maximum food chain length and community trophic structure are related to ecosystem size, we surveyed aquatic and riparian food web components during cool/wet and warm/dry seasons in a shallow lake and nearby wetland in a coastal region of southern Brazil. The lake is more stable hydrologically with much less temporal variation in area and perimeter/area ratio) compared to the wetland that can sometimes lack surface water entirely during periods of drought. We hypothesized that (1) trophic structure is more complex in the lake compared to the wetland because of its greater area and greater hydrologic stability, (2) terrestrial input to the aquatic food web is greater during ood period, and (3) the in uence of ooding is greater in the wetland, where the aquatic/terrestrial interface (i.e. wetted perimeter/area ratio) is greater.

Study area and sampling design
To investigate potential effects of the ood pulse on aquatic food web structure, we conducted eld sampling during cool/wet (Winter -W) and warm/dry (Summer -S) periods at two locations of Taim Hydrological System: Caçapava wetland (hereafter 'wetland') and Nicola Lake (hereafter 'lake') ( Figure 1). The Taim Hydrological System lies within a coastal landscape of forests, grasslands, dunes, lakes and wetlands that supports high biological diversity (Motta-Marques et al 2013). The region is subtropical, and patterns of precipitation, temperature and wind produce strongly seasonal variation the hydrological cycle. During austral winter, high rainfall and low evaporation result in accumulation of water within the wetland, which in turn promotes connectivity with riparian habitats. The opposite tends to occur during austral summer, when lower precipitation and higher evaporation reduce water levels and exchanges of material and organisms with terrestrial habitats (Bastos et al 2014). The magnitude of the seasonal ood pulse varies interannually due to El Niño and La Niña climatic events, which respectively produce positive and negative rainfall anomalies in southern Brazil (Grimm et al 1998). In addition to oods, wind can promote exchange of organic matter, nutrients and organisms between adjacent habitats ( The wetland site is a system of ephemeral water bodies covering an area of approximately of 2 km 2 . Average water depth is 0.3 m with a maximum depth of 1 m during the cool/wet period. The lake is a perennial with an area of approximately 2 km 2 and average water depth of 0.6 m, but depth can reach 1.3 m during the cool/wet period. Both aquatic systems are surrounded by grasslands dominated by species of Poaceae (C4 grasses). Aquatic macrophytes in both habitats are mostly plants that use the C3 photosynthesis pathway. This distinction is important, because δ 13 C values are usually distinct between plants using the C3 or C4 pathway. Aquatic macrophytes (C3) in these systems are characterized by relatively low δ 13  Benthic particulate organic matter (BPOM) was collected by taking the top layer (~1 cm) of sediment using a spatula. Samples of an additional basal resource, capybara feces (Hidrochoerus hidrochaeris, Rodentia), were also collected. Fecal pellets usually were common along the water margins at both study sites and were often found oating in the water, especially during the cool/wet period.
Fish and macroinvertebrates were sampled using a rapid assessment method (Price and Harris 2009) involving several methods as described in (Bastos et al 2014). To collect sh, three sets of gillnets (8 x 2 m each, each set with adjacent meshes sizes of 30 mm) were placed in the water overnight. Insects, snails and clams were collected by hand from littoral and surrounding areas at both sites. A plankton net (250 µm) was used to obtain zooplankton samples at both sites during the cool/wet period. This method could not be employed during the warm/dry period when there was little water in the wetland and suspended detritus in the lake quickly lled the plankton net.

Data analysis
The average δ 15 N of aquatic invertebrates identi ed as primary consumers (λ = 2) was used as baseline to estimate consumer trophic positions using the following the equation: where λ is the baseline trophic level; δ 15 N consumer is the δ 15 N valued of each consumer species of category; δ 15 N baseline is the average δ 15 N value of the second trophic level baseline collected in the same location and period as the consumer; and TDF is the trophic discrimination factor for δ 15  Trophic position (TP) estimates were computed for each macroinvertebrate and sh specimen and then averaged for functional categories as described above. Average δ 15 N values of the primary consumers used as baseline for TP estimates were compared among sites and periods using analysis of variance (ANOVA) after evaluating normality and homocedasticity. The same analysis was performed for the average values for TP of consumer functional categories.
Only baseline and consumer samples obtained at all sites and seasons and in su cient number (n > 5 per site and period) were included in this analysis. When data assumptions were not achieved, non-parametric ANOVA (Kruskal-Wallis) was performed (Zar 2010). When the main effect was signi cant, multiple comparisons (Tukey post-hoc test) were performed for parametric data.
To estimate relative contributions of terrestrial (allochthonous) vs. aquatic (autochthonous) primary producers to consumer biomass, Bayesian stable isotope mixing models were applied using the package SIAR (Parnell et al 2008) in R. To achieve higher resolution and better ecological inferences from mixing models, primary producers with similar isotopic composition were grouped a priori as suggested by Phillips et al. (2005). Hence, algae and macrophytes were considered as aquatic primary producer sources and terrestrial C 4 grasses as the principal terrestrial primary producer source in all models.
Because it is di cult to obtain pure samples of phytoplankton, samples of suspended particulate organic matter

Results
Among all studied food web components, basal food sources revealed greatest variation on δ 15 N values. Basal sources from the wetland had δ 15 N values ranging from -4.1 to 6.4‰ during the cool/wet period, and from -4.4 to 6.6‰ during the warm/dry period (Table S1). A similar pattern was observed for basal sources from the lake, with a δ 15 N ranging from -1.9 to 7.2‰ during the cool/wet period, and from -1.96 to 5.82‰ during the warm/dry period (Table S1, Figure 2). Fishes tended to reveal greater variation in δ 15 N values during the cool/wet period at both sites (Table S2, Figure 2).
Omnivore and piscivore were the most common sh guilds at both sites during both seasons, and this allowed for spatiotemporal comparisons of trophic positions, estimates of basal source contributions to sh biomass, and the size and similarity of isotopic ellipses. The Kruskal-Wallis test indicated that δ 15 N of primary consumers (invertebrates at the second trophic level) was not signi cantly different between sites, but signi cantly higher during the warm/dry period (lake: 4.5 ±1.6‰; wetland: 3.9 ±1.9‰) than the cool/wet period (lake: 3.14 ±0.9‰; wetland: 2.64 ±1.4‰) (p = 0.023). Average TP of sh guilds showed the opposite pattern ( Figure 3, Table 1). ANOVA revealed no interaction between site and season; therefore, TP of omnivores and piscivores was tested separately by site and season.
Omnivores had higher average TP during the cool/wet period (3.9) than warm/dry period (3.5) (p = 0.001). The piscivore guild revealed the same pattern, with higher average TP during the cool/wet (4.8) than warm/dry period (3.9) (p = 0.001) (Figure 3). TP was not signi cantly different between sites for piscivores (p = 0.145), but omnivores had higher values in the lake (3.9) than wetland (3.6) (p = 0.001). Because trophic discrimination factors are assumed to be different for omnivores and piscivores, source contributions to consumer biomass were estimated separately for these groups using the isotopic mixing model (Table S3). The models revealed spatial and temporal differences in the proportional assimilation of sources among sh trophic guilds ( Figure 4, Table 2). Overall, aquatic basal sources (C 3 plants + algae) rather than terrestrial sources (C 4 plants) were assimilated by omnivorous and piscivorous shes in greatest proportions (ranging from 3 to 77%, for the 95th percentile con dence interval of model estimates). Estimated contribution of C 3 plants to these sh guilds ranged from 29% in the wetland during the cool/wet period to 77% in the wetland during the warm/dry period. Algae also were important contributors to these sh guilds, with greater contributions in the wetland during the warm/wet period (16 -62%) compared the warm/dry period (3 -41%). C 4 plants seemed to contribute little to sh biomass in the lake, with estimates ranging from 0% during the cool/wet period to 45% during the warm/dry period. In general, the contribution of C 4 plants to piscivorous sh in the wetland was slightly greater during the warm/dry period compared with the cool/wet period; the opposite pattern was observed for omnivorous sh from the lake ( Figure 4, Table 2). C 3 plants were estimated to be the dominant source supporting both omnivorous and piscivorous shes in the wetland during the warm/dry period ( Figure 4, Table 2). Standard ellipses (SEA C ) of sh assemblages varied in size, shape, position and overlap within δ 13 C x TP biplots ( Figure 5) in relation to sites and periods. Ellipse area of the local sh assemblage was greater in the lake than wetland ( Figure 5A), and greater during the warm/dry period (lake = 4.07; wetland = 2.24) than the cool/wet period (lake = 2.99, wetland = 1.69). Isotopic areas occupied by the lake sh assemblage was larger in both dimensions (δ 13 C and TP), and sh from the lake tended to have lower δ 13 C values than those from the wetland ( Figure 5A). Between-site overlap of isotopic ellipses was higher during the warm/dry period, with 93% of the wetland area overlapping with 52% of the lake area. During the cool/wet period, 19% of the lake ellipse area overlapped with 25% of the wetland ellipse area.
Inter-season ellipse overlap was higher for the wetland, with 30% of the warm/dry area overlapping with 40% of the cool/wet area; in the lake, 19% of the warm/dry area overlapped with 25% of the cool/wet area ( Figure 5A).
Although the standard ellipse area occupied by omnivores (lake warm/dry = 2.23; lake cool/wet = 1.93; wetland warm/dry = 1.71; wetland cool/wet = 1.10) was smaller than the area occupied by the entire sh assemblage, both revealed similar patterns of spatiotemporal variation ( Figure 5B). Within a given season, TP and δ 13 C dimensions of ellipses for sh assemblages spanned similar ranges at the two sites. The exception was the cool/wet sample from the wetland with relatively narrow δ 13 C range. For assemblages at both sites, the TP dimension had a greater range during the cool/wet period. Conversely, the δ 13 C dimension of the lake sh assemblage had a smaller range during the cool/wet period, and that of wetland sh assemblage was greater during that same period. Within-site, inter-season ellipse overlap was higher for the wetland sh assemblage (19% of warm/dry, 30% of cool/wet) than the lake assemblage (6% of warm/dry, 7% of cool/wet) ( Figure 5B). Ellipse overlap between sites was higher during the warm/dry period (62% of wetland, 47% lake) than cool/wet period (17% of wetland, 10% of lake).
Standard ellipse areas differed between seasons to greater extent for wetland piscivores (warm/dry = 1.08, wetland cool/wet = 0.65) when compared to piscivores from the lake (warm/dry = 1.44, cool/wet = 1.14) ( Figure 5C). The range of δ 13 C was relatively narrow for piscivores from the lake during the cool/wet period and for the wetland during the warm/dry period. The TP dimension was broadest for piscivores from the lake and narrow for piscivores from the wetland during the cool/wet period. Piscivore SEA C had no overlap between the warm/dry and cool/wet periods at both sites, and this was largely due to separation in the vertical dimension (TP axis) ( Figure 5C). Piscivore SEA C had higher overlap during the warm/dry period, when overlap was 84% for the wetland sample and 66% for the lake sample.
During the cool/wet period, overlap was 27% of the ellipse of wetland piscivores and 15% of the ellipse for lake piscivores.

Discussion
Between-habitats differences in food-web structure Contrary to our initial expectation, the overall isotopic niche of local sh assemblages did not enlarge during the cool/wet period, but instead was smaller when compared with the summer/dry period. A possible explanation could be related to the greater habitat size and connectivity when water levels were higher during the cool/wet period, especially in the wetland. Under these conditions, sh may have access to seasonally abundant resources, which might result in exploitation of the most pro table and available resources by diverse sh species, resulting in a reduction in interspeci c trophic niche differences (Winemiller 1989). An increase in spatial similarity of aquatic community structure in tropical oodplains during the annual ood pulse was previous reported (Thomaz et al 2007), and it is known that complexity of trophic structure in streams is related to environmental heterogeneity (Zeni and Casatti 2014). Although the isotopic space of local sh assemblages decreased during the ood period, we cannot necessarily assume that trophic niches of individual species had shifted. As pointed out earlier, isotopic ellipses are only proxies for trophic niches and should not be interpreted as highly accurate and reliable indicators (Newsome et al 2007). In some cases, communities assimilating material from a greater diversity of sources can have smaller isotopic spaces than those using fewer but more isotopically distinct sources. In other scenarios, spatiotemporal shifts in isotopic baselines may hinder comparisons of isotopic spaces (Hoeinghaus and Zeug 2008). Although we could not control for the in uence of seasonal changes in water level on isotopic signatures of basal sources, our approach using trophic position (instead of δ 15 N values) should have reduced potential in uences of temporal and between-habitats differences in isotopic baselines on sh isotopic ellipses, at least in the vertical position. Future studies employing complementary approaches (e.g., stomach contents analysis) with isotopic analysis based on surveys across relevant spatial and temporal scales and sources with su cient isotopic discrimination will be in need to better understand the effects of hydrology on the trophic niche of aquatic organisms.
Basal sources and assimilation of allochthonous matter under high water conditions Our nding corroborated the initial hypothesis that terrestrial input to the aquatic food web is greater during the cool/wet period when water levels are higher, especially in the wetland. An increase in the interface between aquatic and terrestrial habitats promotes great contribution of terrestrial sources to the aquatic food web ( Our isotope mixing models estimated that aquatic C3 plants were the most important carbon source supporting shes in both lake and wetland. Algae also were estimated to be an important source supporting the biomass of omnivorous and piscivorous shes, and some sh species could been strongly supported by food chains originating from algae. In at least one other study, algae (periphyton) was estimated to be the most important carbon source for benthivorous shes in a tropical oodplain (Lopes et al 2015).
Vertical trophic structure changed seasonally in sh assemblages at both sites. Fish δ 15 N tended to be higher and more variable during the cool/wet period. Most sh guilds had higher average TP during this period. These ndings suggest an overall expansion of the vertical trophic structure during the cool/wet period, possibly in response to greater foraging options within expanded aquatic habitats. For omnivores, a greater variety of available food resources can lead to an expansion in horizontal structure at the same time of a contraction in the vertical structure of food webs (Akin and Winemiller 2006).

Conclusions
Seasonal hydrological pulsing in a shallow lake and wetland in coastal southern Brazil was associated with shifts in both vertical and horizontal aspects of sh assemblage trophic structure. Findings corroborated our hypotheses of higher complexity in the trophic structure of the lake compared with the wetland, and this was largely due to elevated trophic positions of omnivorous shes and an overall expansion of trophic niches. We also found an increase the contribution of terrestrial basal sources to sh biomass during the cool/wet season, especially in the wetland. By plotting consumer TP instead of δ 15 N in relation to δ 13 C, we obtained more reliable estimates of assemblage trophic niche spaces (ellipses). This approach should strengthen inferences from comparisons of community metrics using stable isotope data. Figure 1 Map showing sites surveyed, Nicola Lake and Caçapava wetland inside the TAIM Hydrological System, which is located inside the Patos-Mirim hydrological system, southern Brazil, South America.  Relative contributions of basal production sources to omnivorous and piscivorous shes in the wetland and lake during cool/wet and warm/dry periods. Bayesian credible intervals of the feasible contributions of each basal production source to the sh guilds: 50 (darkest gray), 75 (medium gray) and 95% (lightest gray). Cool/Wet -red ellipses; Lake Warm/Dry -green ellipses; Lake Cool/Wet -blue ellipses). Right column: Density plots showing the credibility intervals (50 (darkest gray), 75 (medium gray) and 95% (lightest gray)) and modes (black dots) of the estimated standard ellipse areas (SEAB) and the mode (red squares) of the estimated standard ellipse area corrected for small samples (SEAC) by site and season.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. TableS1S3.docx