Species Diversity and Community Assembly of Cladocera in the Sand Ponds of the Ulan Buh Desert, Inner Mongolia of China

: In deserts, pond cladocerans suffer harsh conditions like low and erratic rainfall, high evaporation, and highly variable salinity, and they have limited species richness. The limited species can take advantage of ephippia or resting eggs for being dispersed with winds in such habitats. Thus, environmental selection is assumed to play a major role in community assembly, especially at a ﬁne spatial scale. Located in Inner Mongolia, the Ulan Buh desert has plenty of temporary water bodies and a few permanent lakes ﬁlled by groundwater. To determine species diversity and the role of environmental selection in community assembly in such a harsh environment, we sampled 37 sand ponds in June 2012. Fourteen species of Cladocera were found in total, including six pelagic species, eight littoral species, and two benthic species. These cladocerans were mainly temperate and cosmopolitan fauna. Our classiﬁcation and regression tree model showed that conductivity, dissolved oxygen, and pH were the main factors correlated with species richness in the sand ponds. Spatial analysis using a PCNM model demonstrated a broad-scale spatial structure in the cladoceran communities. Conductivity was the most signiﬁcant environmental variable explaining cladoceran community variation. Two species, Moina cf. brachiata and Ceriodaphnia reticulata occurred commonly, with an overlap at intermediate conductivity. Our results, therefore, support that environmental selection plays a major role in structuring cladoceran communities in deserts.


Introduction
A metacommunity is defined as a set of local communities in which species may exchange through dispersal [1]. Concept and theory of metacommunities not only help to explain patterns of distribution and interaction of local communities, but also provide a realistic framework for understanding ecological processes at a regional scale.
Cladocerans disperse passively over long distances with wind, water birds, and similar media [2,3]. With their abundant propagules and large population size, cladocerans easily colonize new habitats [4]. In a local community, species richness and abundance are regulated by environmental conditions, biotic interactions, and dispersal [4,5]. Environmental processes usually dominate at the finest spatial scale (e.g., a single wetland), whereas dispersal and neutral processes contribute to beta diversity at broader scales [6]. The assembly of local cladoceran communities in a well-connected landscape is determined more by environmental or ecological than by spatial processes. Predation, conductivity, primary productivity, and pH are ecological factors directly affecting cladoceran community structure [5,7,8]. As cladocerans rapidly respond to environmental change, they are used as indicators of ecological change [4,9]. The concept of indicator species has been widely applied in niche analysis and assessment of habitat quality [10,11]. It has been developed to explore complex associations between species and habitats [12][13][14][15][16].
Desert ponds are ecologically distinct. In such ponds, cladocerans suffer harsh conditions, low and erratic rainfall, high evaporation, and high and variable salinity. A few well-adapted species thrive there. For example, in a saline lake (salinity 5.5%) studied by He [17], a single species of Moina dominated the cladoceran community. As sand ponds are filled by rainwater within a very short period of the year, desert cladocerans have rarely been investigated. As a result, spatial distribution and species diversity of desert cladocerans remain unclear. Systematic surveys in such habitats are required to explore regional endemicity and understand species diversity. For example, a new ctenopod, Diaphanosoma bopingi, was identified from Sahara [18]. Given the slow evolution of this group, this species was assumed to be the result of local speciation [18,19].
Ulan Buh desert is located near Bayan Nur city, Inner Mongolia, China, and has a mean annual precipitation of around 100 mm. In this desert, a number of temporary dune ponds occur, many of which have formed artificially in the last decade. Although individual dune ponds are physically isolated, strong winds, cattle, and birds visiting the area should promote cladoceran dispersal and colonization. Although the ponds are at slightly different altitudes within the desert, they can be influenced significantly by winds with very different densities. In the perspective of metacommunities, environmental selection could explain variation in the cladoceran community structure.
In the present study, we aimed to identify cladoceran species diversity and explored processes structuring the community mainly with the detailed investigation in 2012. Our study will greatly improve our understanding of species diversity and community assembly of Cladocera in the desert, and may have important guidance for species investigation and conservation of desert aquatic systems.

Study Object
Ulan Buh desert is located in the east of Inner Mongolia. It harbors a series of permanent lakes fed by groundwater and a large number of temporary ponds surrounding the permanent lakes. More ponds were artificially created beside roads connecting local villages. We took surveys from 2011 to 2013 but had enough samples only in 2012. We performed our sampling after a rainfall when many temporal ponds were filled. Cladocera in a few permanent lakes, more or less, could form a species pool for the temporary ponds in some cases. Winds, water birds, and local livestock are the potential medium for dispersal.
In 2012, we investigated 37 ponds and 4 permanent lakes in an area of 15 km (west to east) × 40 km (north to south) ( Figure 1). Physical coordinates ranged from 40.4019 • N to 40.5264 • N latitude and 106.4817 • E to 106.9269 • E longitude. Altitude ranged from 1025 m to 1055 m.

Field Sampling and Laboratory Work
We sampled cladocerans with a dip plankton net (diameter of the net edge was 40 cm) with a mesh size of 120 µm, one sample was taken for each pond by horizontal tows 5 m over the pond surface area. Geographical coordinates and altitude data were recorded by a hand-held GPS device. Water chemistry at 0.5-1.0 m below the water surface was measured by a portable analyzer (YSI: PRO-PLUS), including water temperature (T), pH, dissolved oxygen (DO), and conductivity (Cond). Pond surface size (S), maximum depth (D), and vegetation coverage rate (V) were measured, fish and tadpole presence were recorded visually. Two states of water quality (N) were estimated de visu: "1" represents clean or high transparency, "2" represents turbid or low transparency. Samples were preserved in formaldehyde with a final concentration of 5%. All samples were examined under a dissecting microscope (Olympus: SZXZ-ILLB) to determine the species diversity per sampling site. Species were identified using an optical microscope (Olympus U-LH100-3). Relative abundance was counted under an optical microscope using a Sedgewick Rafter counting chamber. At least 500 individuals or at least 50% volume of each sample were counted depending on the cladoceran abundance.

Field Sampling and Laboratory Work
We sampled cladocerans with a dip plankton net (diameter of the net edge was 40 cm) with a mesh size of 120 μm, one sample was taken for each pond by horizontal tows 5 m over the pond surface area. Geographical coordinates and altitude data were recorded by a hand-held GPS device. Water chemistry at 0.5-1.0 m below the water surface was measured by a portable analyzer (YSI: PRO-PLUS), including water temperature (T), pH, dissolved oxygen (DO), and conductivity (Cond). Pond surface size (S), maximum depth (D), and vegetation coverage rate (V) were measured, fish and tadpole presence were recorded visually. Two states of water quality (N) were estimated de visu: "1" represents clean or high transparency, "2" represents turbid or low transparency.
Samples were preserved in formaldehyde with a final concentration of 5%. All samples were examined under a dissecting microscope (Olympus: SZXZ-ILLB) to determine the species diversity per sampling site. Species were identified using an optical microscope (Olympus U-LH100-3). Relative abundance was counted under an optical microscope using a Sedgewick Rafter counting chamber. At least 500 individuals or at least 50% volume of each sample were counted depending on the cladoceran abundance.

Statistical Analysis
Environmental variables included lake surface size (S), depth (D), conductivity (Cond), DO, pH, tadpole presence, fish presence, vegetation coverage (V in %), and water quality (N); geographic variables including latitude, longitude, and altitude. Salinity was

Statistical Analysis
Environmental variables included lake surface size (S), depth (D), conductivity (Cond), DO, pH, tadpole presence, fish presence, vegetation coverage (V in %), and water quality (N); geographic variables including latitude, longitude, and altitude. Salinity was represented as conductivity [20]. Two types of species matrices were built based on species presence/absence and relative abundance.
We applied a classification and regression tree model (CART) to determine which environmental variables correlated with species richness [21]. The species occurrence data were further used for species richness estimation. Chao 2 and Jackknife estimators were used, as these models are particularly suitable for small sample sizes [22]. Other estimators are presented as alternative results. Log transformation was used for all estimators so that the lower bound of the resulting interval is at least the number of observed species [23]. The lowest ratio of variance/estimator was used as the best estimation [24].
Redundancy analysis (RDA) was used to examine the variation of cladoceran community structure along environmental gradients. lake or pond size (S) was log-transformed (S ranges from 10 m 2 to 4000 m 2 ). All environmental variables were used in the RDA model to fit the species occurrence matrix and relative abundance matrix. Unadjusted R2 was calculated through "Ezekiel's equation" to create an adjusted R2 (Peres-Neto et al., 2006).
Variables with VIF (variance inflation factors) <20 were used as these variables are not significantly collinear with other variables. Based on the permutation test's p-value and AIC value, the forward selection was performed to select significant variables (p < 0.05).
A PCNM method (principal coordinate of neighbor matrices) [25,26] was used to construct spatial structures that can represent all scales. The first PCNM variables (e.g., V1) describe broad-scale processes covering the whole research area, like strong dispersal. The last variables describe fine-scale processes, which cannot be related to environmental descriptors but community dynamics, such as neutral processes. All positive eigenvectors were selected to build a PCNM variable matrix. These PCNM variables were used as spatial explanatory variables in RDA analysis with species occurrence matrix and relative abundance matrix. Variance partitioning was performed to determine how much variation was explained by each of the explanatory variables.
Classification of habitat types was performed by k-means clustering for each variable independently, including pond size, depth, conductivity, DO, pH, and vegetation coverage. The number of habitat groups was decided by using the maximal "ssi" value. The habitat types for predation were grouped into two types: presence and absence. As more species combinations create too many possibilities, only one species and two species combinations were used as species indicators. "Indval" index was calculated for the association between single species or two species combination and associated habitat types. The significance of maximum "Indval" index was tested using a permutation test (per = 999). "Indval" index contains two components: specificity and fidelity. High specificity indicates species associated with specific habitats, and high fidelity reflects a high probability of seeing a species in a specific kind of habitat.

Environmental Variables and Spatial Landscape
The size of the ponds varied from 10 to 4000 m 2 , but only two ponds had an area of more than 1000 m 2 , and half of the ponds (56%) were smaller than 100 m 2 . Pond depth ranged from 0.3 to 3 m, with a mean of 1.01 m. Conductivity was high, from 1404 µS/cm to 12,619 µS/cm, with mean of 4243 µS/cm. pH values ranged from 7.98 and 10.06. DO was between 2.92 to 15.0 mg/L, with an average 9.76 µg/L. Tadpoles occurred in 15 ponds, and small fishes were only observed in 2 ponds. Except for one pond, aquatic plants were abundant, with an average cover 71%. Among 37 investigated ponds, 15 ponds were high-transparency or clean-water.
The two groups significantly differed in altitude and conductivity. Eastern ponds had higher conductivity. Even though a smaller number of ponds are located in the western region, it contained more species. Moina cf. brachiata was the most common species in the ponds, but it was only found in the western group.
Eight estimators of species richness were applied to the incidence data ( Table 2). The homogeneous model was the lowest, with 13 species (14 species were actually observed). The second-order Jackknife model produced a high estimator of 19 species, similar to Chao2 (Chao, 1987) (18 species). The Chao2 model (Chao, 1987) offered the largest 95% confidence interval: 14.480 to 53.842 species. Most estimators recovered a larger prediction than the actual observation. The 1st order Jackknife estimator showed the lowest variance/estimator: 17 species at least and 26 species at most should live in the study area. Cladoceran species richness for each pond varied from 0 to 7. One of the ponds did not have any cladocerans. On average, a pond had four species. The classification and regression tree model (CART) showed that pH, DO, and conductivity were the main variables affecting cladoceran richness (Figure 3). Three ponds with the highest species richness had a high pH value (≥9.73). When pH < 9.725, ponds were classified into two groups according to DO ≥ 9.345 mg/L. Eight estimators of species richness were applied to the incidence data ( Table 2). The homogeneous model was the lowest, with 13 species (14 species were actually observed). The second-order Jackknife model produced a high estimator of 19 species, similar to Chao2 (Chao, 1987) (18 species). The Chao2 model (Chao, 1987) offered the largest 95% confidence interval: 14.480 to 53.842 species. Most estimators recovered a larger prediction than the actual observation. The 1st order Jackknife estimator showed the lowest variance/estimator: 17 species at least and 26 species at most should live in the study area.

Species Association and Coexistence
Among all 105 combinations of species, 16 showed significant associations with corresponding pond groups. These contained 6 species and 10 species combinations (Table 3)

Variation of Cladoceran Communities along Environmental Gradients
For the incidence data, all environmental variables together explained 12.8% of the total variance (Figure 4). The first two constrained axes explained 5.5% and 2.4% of total variance. The first constrained axis was mainly contributed to by conductivity, fish, and tadpole presence, pond size, pH, DO, and vegetation coverage. Occurrence of M. cf brachiata positively correlated with conductivity, while D. longispina negatively correlated with it; the occurrence of S. smirnovi and C. reticulata positively correlated to DO and vegetation coverage. Chydorus sphaericus positively correlated with pond size. Forward selection showed that conductivity and DO were the most significant explanatory variables, which together explained 12.4% of the variance. Conductivity explained 9% of the variance and DO explained 4.5% of the variance.  Redundancy analysis (RDA) for relative abundance data showed that the environmental variables together explained 29.2% of the variance; the first two constrained axes explained 26.8% (Figure 4). The first constrained axis is mainly related to conductivity, fish and tadpole presence, pond size, pH, DO, and vegetation coverage. Ponds dominated Redundancy analysis (RDA) for relative abundance data showed that the environmental variables together explained 29.2% of the variance; the first two constrained axes explained 26.8% (Figure 4). The first constrained axis is mainly related to conductivity, fish and tadpole presence, pond size, pH, DO, and vegetation coverage. Ponds dominated by M. cf brachiata were mainly situated along the first constrained axis, while ponds dominated by C. reticulata and D. longispina were situated along the second constrained axis. Relative abundance of M. cf brachiata positively correlated with conductivity, fish and tadpole presence. Relative abundance of C. reticulata were positively correlated with DO and vegetation coverage, while relative abundance of D. longispina was affected by water pH and depth. Forward selection of variables also showed conductivity and DO as significant explanatory variables. Conductivity and DO together explained 25.8% of total variance; conductivity explained 17.5%, DO explained 10.3%.

Variation of the Cladoceran Community along Spatial Gradients
The PCNM model produced 36 PCNM variables. When applied to species presence/absence data, 17 spatial eigenvectors were positive while only the first two spatial eigenvectors (V1 and V8) were significant. This suggests that broad-scale spatial processes affect cladoceran occurrence; variance partitioning shows that PCNM variables (V1 and V8) explained 5.8% of the total variance and 4.3 % was purely explained by PCNM variables (V1 and V8). Three significant PCNM variables (V1, V4, and V5) were selected for cladoceran relative abundance, indicating only broad-scale spatial processes. They explained 15.9% of the total variance (See Table 4). Table 4. Variance partitioning of cladoceran community structure. (E, selected environmental variables base on forward selection in redundancy analysis including conductivity and DO; S1, significant positive PCNM variables, including V1 and V8; S2, significant positive PCNM variables, including "V1", "V4", and "V5").

Cladocera Diversity
Our investigation suggests limited species richness of Cladocera in the dune ponds of the Ulan Buh desert. Temporary ponds contained more species (16) than large permanent lakes (14) (see species list in Appendix A). Generally, cladoceran species changed at both temporal and spatial scales, thus, further sampling is needed to better describe species diversity in permanent lakes in the region. However, studying adjoined temporary ponds collectively provided species information to understand species diversity and distribution in the permanent lakes. Both temporary ponds and permanent ponds play important roles in maintaining total zooplankton species diversity [31].
As 37 dune ponds were investigated, estimators could be calculated. Chao2, first-, and second-order Jackknife suggested that more species can be expected. Similar estimators were used in extrapolating species numbers for 212 samples with 72 observed species in Thailand. Some 76-82 species were expected to be present even for such a large sample size [32]. ATBI (All Taxa Biological Inventory) suggested that any lowland lake is expected to have around 50 species, and species lists should be based on a minimum of three samples through seasons and years to minimize the under/overestimation [24,33]. In the present study, we sampled all ponds we could find, yet curiously, all Bosminidae and Ilyocryptidae, widespread all over the world, were absent here.
There are a few cladoceran studies in desert waters that showed high variation in species richness. Green (1976) described a sand-dune lake from New Zealand with only 1 species, Bosmina meridionalis, but the species of Bosmina were absent in our study [34]. A desert study from Xinjiang (China) also showed low Cladoceran diversity, with 7 species from 5 lakes between Tarim floodplain and Konchedarya river [35]. In contrast, high richness (34 species) was observed in 12 wet and 2 mud samples in Lençóis Maranhenses, a tropical desert in northeast Brazil. Six species were only found by hatching resting eggs/ephippia from collected mud samples resulting in a total of 15 species [36]. The Lençóis Maranhenses is a coastal strip of 155,000 ha with thousands of temporary freshwater pools fed by seasonal rains (November-May). The case study in the Brazilian desert revealed a great diversity in sediment, partly supported by hatching methods. Our study employed plankton sampling but no sediment work. Further surveys using more comprehensive sampling methods might therefore yield additional species.
In 2005, Alonso and colleagues started a lake survey of Mongolia. Until 2017, 1123 lakes have been cataloged and investigated in fourteen expeditions, almost covering the whole country. In total, 73 cladoceran species have been found, including 1 haplopod, 3 ctenopod, 1 onychopod, 16 members of Daphniidae, 5 Moinidae, 6 Macrothricidae, 5 Bosminidae, 36 Chydoridae (http://www.geodata.es/mongolian_lakes accessed on 23 April 2018). The present study area is about 250 km away from Mongolia, and it has geomorphic features similar to that country. Its cladoceran fauna is expected to be more or less similar too. Three of the six cladoceran species found in a study of northeast Mongolia appeared in our study. Even though their sampling period was quite similar to ours, the low cladoceran diversity may be due to the high salt concentrations of lakes [37]. Long-term studies in drainless saline lakes of the Uldza-Torey basin reported 13 cladoceran species, and frequently occurring three species, D. magna, M. brachiate, and D. mongolianum all occurred in our study [38]. The study in Badain Jaran desert (located about 400 km northwest of our study area) showed diverse littoral cladocerans [39], among which four of these cladoceran species were present in our samples. A study in Jili lake located in north Xinjiang, China, reported 12 cladoceran species [40], and only three of them appeared in our samples.
Environmental variables such as conductivity, water pH, and DO were the main variables correlated with species richness in our study. Conductivity functions as one of the most important environmental filters, reflecting local environmental conditions. Environmental selection, as well as spatial processes affected cladoceran communities of the ponds. Three pelagic species, viz. C. reticulata, D. longispina, and M. cf brachiate, dominated, but with a clear niche partitioning. M. cf brachiata appears to be good indicator of high salinity and is adapted to tadpole predation, while D. longispina is absent under tadpole predation. Interspecific differences in colonization and competitive abilities may also determine the species composition, which was revealed in rock pools on the Baltic islands off the south coast of Finland [41].
Deserts are mobile, and dunes move with the wind. The maximum between-pond altitude difference was 30 m within our area (40 × 15 km 2 ). Significant correlations between altitude and longitude as well as latitude tell that eastern ponds are at higher altitudes than western ponds. This implies that the wind effect is strong in our study area. Furthermore, Cladocera produce ephippia that can be dispersed more or less easily by wind [3,42]. A study on the rock pools of the Appledore islands showed differences between ephippia of Daphnia and Moina in the ease with which they were dispersed [3]. Large numbers of Daphnia ephippia occurred in besides rock pools, against a few of Moina. Compared to Daphnia, ephippia of Moina have lower buoyancy, resulting in low dispersal. The M. cf brachiata in the present study was limited to the western ponds by environmental selection (high conductivity), even under the prevailing northwest winds. However, M. cf brachiata was still able to colonize the east ponds with high conductivity, given enough time. Louette and De Meester (2005) measured colonization by Cladocera of 25 newly dug pools in an area in Belgium of 200 km length and 50 km width [4]. In the built-up communities, a total of 20 Cladocera species were identified after 15 months; C. sphaericus, Daphnia obtusa, and Simocephalus vetulus were in more than half of the pools, but Moinidae and Sididae were absent. Daphniids represented almost 50% of the colonization events, with Daphnia always the pioneer cladoceran species. D. mongolianum, a sidid, was only present in two permanent lakes and one linked temporary pond, implying that this species is limited to permanent lakes. Diaphanosoma tolerates fluctuations in water salinity and is found in up to 15 g/L salinity [43]. The explanation of its restricted distribution in our study might therefore be low dispersal ability.
C. rectangula, C. sphaericus, C. reticulata, and S. smirnovi were the common species in the present study, and their combinations associated with habitats with different pH values, suggesting they all can be found in a wide range of water types. Walseng (2003) identified littoral microcrustaceans as indicators of acidification in Canadian ponds, and species richness increased when aquatic ecosystems recovered from acidic to nonacidic [7]. Among 64 identified Cladoceran species, Sinobosmina sp. and Sida crystallina were good indicators of acidification. This may be the reason why these two genera were absent from our study area. Chen et al. (2010) examined the impact of lake trophic state on the Cladoceran communities in 33 Irish ponds, highlighting that cladoceran functional structure strongly shifted with nutrient enrichment and should be considered as multiple biotic indicators [8]. Even though the nutrient level was not a significant variable to Cladoceran communities in the current study, it may potentially impact Cladoceran functional structure.

Community Assembly
Forward selection in RDA analysis suggested that conductivity and DO significantly explained community variation. Ponds in the present study were all shallow water bodies, thus their DO level was controlled by phytoplankton and aquatic vegetation, and varies throughout the day. In this kind of habitat, cladoceran communities are not significantly altered by DO level. Water conductivity is usually related to salinity, which strongly affects zooplankton communities [44,45]. Zooplankton species richness decreased with lake salinization in the lakes of Tibet, and the zooplankton community shifted from the dominance of copepods and small cladoceran species to large saline filter-feeding cladocerans and phyllopod species [46] A similar pattern emerged in our study, which dominant cladoceran species in our sand ponds shifted from smaller C. reticulata to larger M. cf brachiata. Salinity is a serious stress factor for freshwater cladocerans, thus, species must either adapt to it or perish. M. cf brachiata is able to survive at high salinity, while C. reticulata only appeared in lower salinity ponds. Several species of Moina have a good tolerance to salinity. Their tolerance can even reach 15‰ and they still dominate in lakes with 5.5‰ salinity [17].
In general, mineralization, salinization, and eutrophication all had a significant effect on the cladoceran community [40]. In our study, significant environmental variables, conductivity, and DO together explained the majority (12.4%) of the variations in the cladoceran community. The other environmental variables such as pH, temperature, and water depth are important, but did not significantly contribute to the variation of the cladoceran community.

Conclusions
Our study presented a brief outlook on species richness, fine spatial distribution, and community structure of cladocerans in a desert setting. C. reticulata and C. rectangula were the most common cladocera in the studied region, while C. reticulata, D. longispina, and M. cf brachiata all formed dominant populations. M. cf brachiata and C. reticulata were suggested as good indicators of ponds with high conductivity and lower conductivity, respectively. The combination of these two species coexisted in ponds with moderate conductivity. Species richness is limited by harsh habitats and the community structure was shaped by environmental selection as expected. Conductivity was the most important environmental variable, and a broad-scale spatial structure was significant. Large permanent lakes function as a species pool for temporal ponds nearby. The species diversity and local community in temporal ponds demonstrate environmental filtering or species sorting of local habitats for species dispersed from the permanent lakes.

Conflicts of Interest:
The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper. Table A1. Checklist of species from 37 sand ponds, species from 4 neighboring permanent lakes was also listed. Lake: cladoceran species present in ponds, Pond: cladoceran species present in ponds; + means present, -means absent.