Winter coexistence in herbivorous waterbirds: Niche differentiation in a floodplain, Poyang Lake, China

Abstract The classical niche theory supports the idea that stable coexistence requires ecological differences between closely related species. However, information on waterbirds coexistence in the entirely landlocked freshwater system of Poyang Lake is not well understood, especially when the available biomass of their food in the area decreases. In this study, we tested the ecological segregation mechanisms in the 2015/2016 and 2016/2017 wintering periods among eight herbivorous waterbirds (including the Siberian crane Grus leucogeranus, hooded crane Grus monacha, white‐naped crane Grus vipio, common crane Grus grus, greater white‐fronted goose Anser albifrons, bean goose Anser fabalis, swan goose Anser cygnoides, and tundra swan Cygnus columbianus) at Poyang Lake. Using field observations and species niche and foraging habitat selection models, we investigated the abundance, distribution, and food sources of these eight waterbird species to quantify and compare their habitat use and ecological niches. Our results showed that niche segregation among the waterbirds, with respect to food types, time, and spatial location, allow them to coexist and use similar resources. The water level gradually receded in the sub‐lakes of the Poyang Lake, which could provide food sources and various habitats for wintering herbivorous waterbirds to coexist. We demonstrated that the differences in habitat use could mitigate interspecific competition, which may explain the mechanism whereby waterbirds of Poyang Lake coexist during the wintering period, despite considerable overlap in the dietary niches of herbivorous waterbirds.


| INTRODUC TI ON
Wetlands are critical foraging areas for waterbird species, and these areas sustain a high level of biodiversity. It is often believed that their relatively high productivity is the principal factor that determines the coexistence of sympatric species using similar food resources during breeding and non-breeding seasons (Chatterjee et al., 2020;Wiens, 1989). Based on the classical niche theory, stable coexistence of species requires segregation of their niches to mitigate competition for limited food resources, thereby allowing multiple species to forage simultaneously within the same region (MacArthur, 1958;Pianka, 1974). However, the complex dynamics of wetland structures in response to hydrological regime shifts or climate change could limit the food sources for waterbirds, thereby enhancing dietary competition between species (Cumming et al., 2012;Lorenzon et al., 2017). Specifically, many waterbird species show large overlaps in their diets and foraging sites (Henry & Cumming, 2017).
Habitat use is another important niche dimension for the coexistence of waterbird species, and the overlap of their niches can be reduced through resource partitioning and habitat differentiation (Schoener, 1974;Xu et al., 2021). The habitat availability of waterbirds in seasonally inundated wetlands is strongly affected by water level changes (Baschuk et al., 2012;Holm & Clausen, 2006). As habitats vary in structures and resources seasonally, waterbird species with different feeding guilds migrate to favorable areas. The relationship between water level and habitat availability strongly affects the waterbird species that occupy ecological niches at different water depths (Mei et al., 2016;Polla et al., 2018). For example, geese generally forage in exposed riparian grasslands, whereas cranes favorably forage in mudflats or shallow water areas (Jia et al., 2019). Tundra swans prefer to feed in deeper water, and diving birds forage at high water depths (Jiang et al., 2015). When food resources are scarce, waterbirds adapt various strategies for coexistence that enable them to optimally utilize the available food resources. Field evidence showed that Siberian cranes (Grus leucogeranus) shifted their diet from tubers of submerged macrophyte to a different plant (Potentilla limprichtii) when their preferred tuber was extremely scarce . Many waterbird species have evolved unique anatomical specializations, such as special beak shapes, long necks and tarsi, and behavioral specializations that enable them to forage in different areas with variable water depths (Elphick, 2008;Ntiamoa-Baidu et al., 1998).
Poyang Lake is a large floodplain wetland and an important Ramsar site, to which hundreds of thousands of wintering waterbirds migrate every year, particularly herbivorous waterbird species, such as geese, cranes, and swans (Barter et al., 2005;Ji et al., 2007;Ruan et al., 2018). Poyang Lake provides various habitats and abundant food resources for the waterbirds to coexist while foraging and wintering as the water level recede gradually during the low water period (Aharon-Rotman et al., 2017;Yang et al., 2020). As a result, this wetland provides an excellent opportunity to examine the coexistence of herbivorous wintering waterbirds that use different foraging habitats in response to lake water level. In this study, we explored the potential mechanism through which herbivorous waterbirds coexist for wintering at Poyang Lake in order to develop appropriate protection measures. Moreover, the following two hypotheses were tested: (1) niche partitioning occurs among overwintering waterbirds for food, space, and time, which facilitates coexistence and (2) gradual falling water levels in dry season may affect the availability of food and suitable habitat area, affecting waterbirds' abundance and distribution at Poyang Lake.

| Study area
Poyang Lake (28°11ʹ-29°51ʹN, 115°49ʹ-116°46ʹE) is the largest freshwater lake in China, and it is situated in the middle reaches of the Yangtze River and the northern part of Jiangxi Province ( Figure 1). This lake is one of two large lakes that are freely connected to the Yangtze River. Owing to its monsoonal climate, Poyang Lake exhibits considerable seasonal and interannual variation in water levels (Min, 2007;Zhang et al., 2014).
The considerable differences in water levels shape the distinct landscapes between the wet summer and dry winter seasons in Poyang Lake. During the wet period, landscape in the wetland mainly consists of open water, with the water surface area reaching 4000 km 2 . However, besides permanent open water, the wetland landscape mainly consists of herbaceous meadows, mudflats, and isolated water bodies during the dry period. It is a favorable habitat for migratory waterbirds overwintering in this period (Barter et al., 2005;Ji et al., 2007).
During the low water period, the three sub-lakes consist of water bodies, mudflats, and vast meadows with herbaceous vegetation that are mainly occupied by Carex spp. and sedge communities (Jian et al., 2001;Zhang, Li, et al., 2012;Zhang, Yin, et al., 2012

| Water level data collection
The water levels in the study sites were monitored using the nearest hydrographic stations during the flood period. As the water bodies were isolated from open water during the dry period, we monitored the water levels of the three sub-lakes at the sluices (Yellow Sea Datum) at the beginning and middle of each month. Additionally, the water level of Huangjinzui was recorded by the Duchang hydrological station throughout the year ( Figure S1).

| Vegetation surveys
Considering the overwintering period of the waterbirds at Poyang We collected tuber samples from the submerged vegetation using a stainless steel mud collector. Five replicate samples were randomly collected near each sampling point and brought back to the laboratory. Tuber samples were carefully washed with water and weighed using electronic scales (accuracy 0.001 g). These samples were then placed in an oven for 30 min at 105℃ and then dried at 55℃ for 2 days to obtain the dry weights. The biomass of the submerged vegetation tubers was calculated for both dry and fresh weights.

| Waterbird surveys
Waterbird data were obtained from surveys conducted at the study sites between 2015/2016 and 2016/2017 over two overwintering periods. Generally, migratory waterbirds winter at Poyang Lake from October to March of the following year. Therefore, we surveyed the waterbirds at the study sites twice per month from early October to late March and recorded their species and abundance F I G U R E 1 Location of four study areas (Meixihu, Changhuchi, Baishahu, and Huangjinzui) in Poyang Lake, China. Black dots represent the sampling sites of vegetation survey as well as their foraging sites in each 300 × 300 m quadrat (N = 120) comprising the vegetation sampling point on a printed paper map of the area. We did not conduct the surveys on days with extreme weather (i.e., foggy, windy, and heavy rain days) to avoid biased measurements. Before counting, the observation of foraging behavior was typically conducted for 10-20 min. Surveys were conducted simultaneously at fixed locations along fixed routes to avoid repeated counting and to improve counting accuracy. In order to ensure that the monitoring field covered the entire area and did not interfere with the behavior of the waterbirds, one or two permanent observation locations (fixed points) were established at each sub-lake. We adopted the look-see counting method (Barter et al., 2005;Cao et al., 2008), and waterbird species were distinguished based on the methods suggested by Mackinnon and Phillipps (2000) and Barter et al. (2005).

| Habitat variables
We collected data related to geographical characteristics, habitat features, and human disturbance factors for each studied site (Table S1). The elevation data were provided by the Department of Water Resources, Jiangxi Province (Yellow Sea Datum). The geographic locations of the vegetation quadrats were determined using a GPS device. Given the difficulty to accurately identify waterbirds' locations in the field. We recorded species identity and their numbers referring to the grid on a map with some obvious references, such as micro-topology, vegetation. As a result, location data of waterbirds in this study were relatively accurate.
The World Geodetic System 1984 was used while recording all the locations. The difference between elevation and water level was represented as the water table (WT). The biomass of the submerged macrophyte tubers before overwintering (TBI) is a potential food for tuber eaters, whereas the biomass after overwintering (TBII) is unused food. The availability of food (TBD) is the difference between tuber biomass before and after overwintering.
The height (CHC) and coverage (CCD) of Carex spp. before overwintering are represented as CHI (cm) and CCI (%), respectively, and those after overwintering are represented as and CHII (cm) and CCII (%), respectively. The height and coverage of Carex spp. before and after overwintering can be used to develop indices of food availability for short grass foragers. We also measured the distance (m) to the nearest road (DR), village (DV), and lake center (DC) from each foraging location of the waterbirds. All distances were calculated using ArcGIS (version 10.1).
All data were imported into the ArcGIS software, including the latitudes and longitudes of the waterbird foraging positions, geographical characteristics, habitat, food, and hydrological and human disturbance factors. All foraging habitat characteristics are listed in Table S1. Several 300 × 300 m quadrats (N = 120) comprising vegetation sampling points were interpolated using the inverse distance weighted method and the number of species investigated in each quadrat was counted separately (Bancroft et al., 2002).
Subsequently, the factors influencing the feeding habitats of waterbirds were extracted according to their feeding sites in each quadrat.

| Statistical methods
The values are reported as mean ± SD. The normality of the distributions was assessed using the Kolmogorov-Smirnov test, with the significance level set at 0.05 (Lilliefors, 1967). When the data were not normally distributed (Table S2), Spearman's rank correlation test was performed to assess the correlation among the habitat parameters (Table S3; Maritz, 1995). One-way analysis of variance and Duncan's new multiple range test (Hsu, 1996) (Barzen et al., 2009;Wang, Fox, et al., 2013;Wang, Jia, et al., 2013).
Principal component analysis was used to analyze the variation within and among these variables for different bird species (Conner & Adkisson, 1977). We selected the first (PC1), second (PC2), and third (PC3) principal components whose eigenvalues were greater than 1 and whose contribution values were greater than 10.00%. We employed R package nicheROVER (Swanson et al., 2015) to calculate the pairwise of niche overlap probability for eight waterbird species and niche width of each bird in three dimensions (PC1, PC2, and PC3). We calculated corresponding items using the equations below.

Niche overlap:
where X A and X B was correspond to randomly selected principal components in three dimensions from different bird species, and let N R (A) and N R (B) denote their respective niche regions. W. There was a probability of an individual from bird species A to be found in the modeled niche region of bird species B and vice versa (i.e., overlap probability).
The difference between these two values facilitates the assessment of asymmetric niche overlap between groups. We modeled 1000 samples and ran 1000 iterations to compute niche regions with a probability level of alpha 0.95 (i.e., 95.00% probability) and 95.00% confidence intervals, and used 10 Monte Carlo draws to plot the elliptical projection (i.e., default priors).
where N R was a given niche region in three dimensions ( (Burnham & Anderson, 2004). Models with ΔAICc < 2 have support for being the best model in a set. Additionally, we used Akaike weights (ω i ) as an indicator to support the best models.
All statistical analyses were performed using the software R 3.6.1 (R Core Team, 2019).

| Abundance and distribution of herbivorous waterbirds
The number of herbivorous waterbirds in the four study sites varied throughout the survey period ( Figure 2 and Table S4). The distribution of the white-naped crane ranged from a WT of −60 cm to 60 cm, whereas for the Siberian crane the distribution ranged from a WT of −60 cm to 40 cm. Although the hooded cranes were distributed in a similar WT region as the Siberian cranes, the hooded cranes waded shallower than the Siberian cranes. In addition, the common cranes were distributed in the WT region ranging from −140 cm to 0 cm and were mainly concentrated at −20 cm.
Goose species distributed in the WT region ranged from −240 cm to 40 cm. The distribution of the tundra swans was wider than that of other herbivores.

| Habitat characteristics
The potential food resources for the waterbirds were distributed in a step-like manner with peaks in the study sites, and their distribution was spatially uneven ( Figure S2). In general, Carex spp. was distributed in the higher elevation areas. In contrast, the tubers of Our results revealed substantially differences among the favorable habitat characteristics of the eight waterbird species (Table 1).
With regard to food source factors, the TBI, TBII, and TBD were the

| Species niches and foraging habitat selection models
The principal component analysis performed on the habitat environment characteristic variables yielded three principal components (PCs) that explained 62.53% of the total variation in the analyzed samples ( Table 2) Overall, all eight waterbird species showed obvious niche differentiation. The highest niche overlap probability was obtained for WNC-SC (91.78%, Figure 5 and Table S5), whereas the lowest value was obtained from TS-SC (0.07%). In addition, the niche overlap of TS and other birds were clearly differentiated (value < 50.00%). Our results also indicated that BG had the narrowest niche width and HC the widest niche among the niche width of the eight waterbird species (Figure 6). Niche width was significantly different among the eight waterbird species (F = 2508.27, df = 7, p < .001), and LSD test results showed all species, niches were significantly different (LSD: ALL P values <.001, Table S6) for eight waterbird species. We identified the best foraging habitat selection model (top-rank model, ΔAICc = 0) for the eight waterbird species (Table 3) (Table S7).

| Species coexistence maintained by niche segregation
Niche segregation among waterbirds with respect to food types, and spatial location allow them to coexist and use varied food resources.
The niche of a species as a hyper-volume proposed by Hutchinson (1957), allows us to understand how closely related species coexist by exploiting different environmental resources. The stable coexistence of sympatric species can be attributed to ecological niche differentiation, which suggests that species may specialize in different food resources. In our study, niche overlap was observed for the eight species of wintering herbivore waterbirds at Poyang Lake ( Figure 5). Based on the results of previous dietary studies, these wintering waterbird species can be divided into the following two feeding groups: Carex spp. foragers and tuber eaters (Barzen et al., 2009;Wang, Fox, et al., 2013;Wang, Jia, et al., 2013); this is consistent with our results from the Poisson regression models (Table 3,  Schoener, 1974).
When competing for the same food resource, two (or more) species can coexist in overlapping distribution areas via using the same resource at different time or spatial locations (Han et al., 2019;Jean-Baptiste et al., 2012;Tokeshi, 1998). Thus, species that use the similar food resources are susceptible to competitive interactions over sharing limited resources, and to coexist, they should exert a mechanism to reduce, but not necessarily eliminate, negative competitive interactions (Chatterjee et al., 2020;Simberloff & Dayan, 1991). In this study, there was large temporal variation in waterbird abundance in surveyed areas (Figure 2). For Carex spp. foragers, the individual number peaked at Huangjinzui in mid-November and around January for the other three sub-lakes.

F I G U R E 3
Distribution of individuals of eight waterbird species according to the water table. BG, bean goose; CC, common crane; GWG, greater white-fronted goose; HC, hooded crane; SC, Siberian crane; SG, swan goose; TS, tundra swan; WNC, white-naped crane Moreover, a small peak in waterbird numbers occurred at the three sub-lakes between February and March. The overall trend for tuber eaters was similar to that of Carex spp. foragers, but the overwintering peak in Meixihu and Changhuchi was earlier than that in Baishahu. Additionally, the results of the present study suggest that direct competition among wintering waterbirds was avoided via the occupation of different WTs on a spatial scale, which helped to allocate resources more efficiently (Figure 3). For example, with regard to tuber-eating waterbirds, the white-naped crane, Siberian crane, and hooded crane forage in mudflats and shallow water, mainly in shallow water habitats, whereas the common crane and swan goose in grasslands and mudflats, mainly in mudflat habitats. The distribution of tundra swans was wider than that of other herbivores. Although the tuber-eating waterbirds were distributed in a similar region of habitat, they occupied different WTs (Table 1). Thus, birds segregated their niches by occupying different WTs, thereby minimizing niche overlap. Moreover, we found that the niche width among eight waterbird species exhibited significant variation in this study ( Figure 6). Based on results from PC1, PC2, and PC3, the niche width and overlap of the tundra swans exhibited marked differences with other waterbird species (Figures 5 and 6). Similarly, other foragers of the same group showed different degrees of niche separation. For example, the bean goose owned the narrowest niche, while the hooded crane had the widest niche in wintering habitats ( Figure 6). Niche differences allow species to complement each other and make better use of existing resources (Büchi & Vuilleumier, 2014;Carroll et al., 2011;Northfield et al., 2010).

| Abundance of wintering birds correlated with habitat
Habitat quality, including availability of food resources and suitable habitat area, influences the abundance and distribution of waterbird species and determine coexistence. In floodplain systems, water   (Clausen, 2000;Paracuellos, 2006;Wang, Fox, et al., 2013;Wang, Jia, et al., 2013). Because the water level is controlled by a sluice, the sub-lakes of Poyang Lake have unique topographic characteristics and form special hydrological fluctuations, resulting in the spatial distribution pattern observed for the wetland vegetation. Herbivorous waterbirds used different wintering sites within one or more sub-lakes of Poyang Lake and forage various on different wetland vegetation as food sources. Our results showed that Carex spp. were distributed at high elevations in the study sites, which was closely related to the water recession observed in the early overwintering period ( Figure S2). In the middle and low elevation areas, the growth of Carex spp. was largely in response to lake water level (Feng et al., 2020;Li, Qian, et al., 2019;Li, Yu, et al., 2019). As the water level decreased, areas at high elevations were firstly exposed; this could provide a moderately environment for the growth of Carex spp. (Aharon-Rotman et al., 2017;Barzen, 2008).
However, it is noteworthy that the abundance of geese inhibited the growth of meadows with short Carex spp. (Table S3), indicating that geese preferred to forage for Carex spp. at the early growth stage wherein Carex spp. contains high protein and low structural carbohydrate content (Cadieux et al., 2005;Zhang et al., 2016).
Therefore, the resource quality of vegetation suitable for Carex spp.
foragers was strongly affected by the water level of the sub-lakes (Hassall et al., 2001). Previous studies have suggested that floods that receded rapidly or much earlier could accelerate the exposure and growth of sedge meadows, making them unsuitable for wintering geese use (Hassall et al., 2001;Wang, Fox, et al., 2013;Wang, Jia, et al., 2013). The gradual recession in the water level of the sublakes ensured sustainable access to food resources for overwintering waterbirds.
Our results suggested that water depth and the inundation duration of the wet season had significant adverse effects on the tuber biomass of submerged vegetation (Figure 4). Water depth and inundation duration can influence the distribution and growth of submerged vegetation by limiting light intensity, thereby affecting their photosynthetic ability (Li et al., 2020). Few submerged vegetation tubers could survive in Huangjinzui because the water depth was relatively high in the wet season and rapidly receded during the dry season. Therefore, the unique hydrological variation in the sub-lakes of Poyang Lake could provide diverse foraging habitats and abundant food sources for wintering herbivore waterbirds.
Recent studies suggested that herbivorous waterbirds were restricted to winter in several better connected lakes since more lakes had been hydrological isolated from the middle and lower areas of the Yangtze River, particularly for Poyang Lake (Xia et al., 2017).
Poyang Lake is a favorable wintering ground for herbivorous water- Tai Lake has been lost because of eutrophication and persistent high water levels (Qin et al., 2007;Zhao et al., 2017). Similar observations were made at Shengjin Lake, precluding the availability of favorable habitats for the early arrival of wintering herbivorous waterbirds (Fox et al., 2011;. However, Poyang Lake has high wetland vegetation productivity and has become one of the most important wintering grounds for waterbirds in the East Asian-Australasian Flyway (Barter et al., 2005;Wang, Fox, et al., 2013;Wang, Jia, et al., 2013).

| Implication for conservation
The findings of this study demonstrated that differences in habitat use could mitigate interspecific competition, which may explain the mechanism through which waterbirds of Poyang Lake can coexist during the wintering period, despite considerable overlap in the dietary niches of herbivorous waterbirds. Therefore, we should adopt some strategies to protect waterbirds for wintering, such as protecting existing resources, and recovering of submerged vegetation and benthic invertebrates. In doing so, it could be possible to mitigate competition promoting the abundance of food resources. During the non-breeding period, investigation of the coexistence of species and the utilization of existing resources may provide significant insights into the conservation and management of wintering waterbirds and F I G U R E 5 Ten random elliptical projections of trophic niche region for each bird species in Poyang lakes (elliptical plots). Also displayed are one-dimensional density plots (lines) and two-dimensional scatterplots. The eight waterbird species displayed are BG, bean goose; CC, common crane; GWG, greater white-fronted goose; HC, hooded crane; SC, Siberian crane; SG, swan goose; TS, tundra swan; WNC, whitenaped crane  Note: K is number of parameters. AICc is Akaike's information criterion corrected for small sample size. ΔAICc is difference in AICc relative to minimum AICc. ω i is the Akaike model weight. Variable descriptions are found in Table 2.

DATA AVA I L A B I L I T Y S TAT E M E N T
TA B L E 3 Results from foraging habitat selection models of eight bird species based on corrected Akaike information criterion (AICc)