Elevational patterns of fish functional and phylogenetic community structure in a monsoon climate river basin

Understanding the patterns and drivers of biodiversity across space and time is commonly based on species diversity, which may ignore species' functional role and evolutionary history and result in an incomplete understanding of community assembly. It is suggested that integrating species, functional, and phylogenetic diversity could provide a more holistic assessment of community assembly in natural ecosystems. This study aimed to explore the elevational patterns and environmental drivers of multiple facets of fish diversity and community structure in a subtropical river during the wet and dry seasons.


| INTRODUC TI ON
Understanding the patterns and environmental drivers of biodiversity across space and time is central to community ecology, biogeography, and conservation biology (Davies, 2021;Heino, 2011).
Typically, studies addressing this subject have been based on species richness (SR).However, focusing on SR alone largely ignores the fact that species within communities are non-equivalent and characterized by distinct functional traits and evolutionary relationships (Mason et al., 2007;Webb et al., 2002), compromising our ability to fully understand biodiversity dynamics.Recently, ecologists have recognized that functional (FD) and phylogenetic diversity (PD) provide complementary metrics quantifying the diversity of ecology, life history, and evolutionary histories within ecosystems.FD measures the range of species traits (i.e., ecological, morphological, and physiological strategies) that determine their performance and ultimately ecosystem functioning (Petchey & Gaston, 2006).In comparison with SR, FD is more closely related to environmental variation (Mason et al., 2007).PD captures variation in evolutionary relationships among species within a community and thereby reflects the degree of relatedness (Faith, 1992).It is suggested that PD could help elucidate the influence of biotic interactions and biogeographical history on biodiversity distribution (Webb et al., 2002).Further, assuming a strong phylogenetic signal, closely related lineages tend to show more similar suites of traits than distantly related lineages (Webb et al., 2002); PD could thus act as a good surrogate for FD.
By contrast, when studied traits are labile, FD and PD may be decoupled and exhibit divergent trends along environmental gradients (Poff et al., 2006).Overall, it is now widely accepted that integrating multiple facets of diversity could provide important information on spatiotemporal dynamics of community composition (Cadotte et al., 2013;Heino & Tolonen, 2017).
Functional and phylogenetic data has proven useful in inferring the underlying processes shaping community structure (Heino & Tolonen, 2017;Swenson et al., 2012).According to the classic theory of community assembly, biotic communities are the results of deterministic (i.e., environmental filtering and limiting similarity) or stochastic factors (e.g.random dispersal and ecological drift) (Chase & Myers, 2011;Weiher & Keddy, 1995).Specifically, observed values of FD and PD are compared with those generated from null models, and the magnitude of deviation between them could be applied to reveal the strength of assembly processes acting on communities (Mason et al., 2007).For example, if co-occurring species within an assemblage are on average more functionally or phylogenetically similar than expected by chance (i.e., clustering pattern), environmental filtering is assumed to be dominant, where the environment acts as a filter selecting species with certain similar characteristics (Webb et al., 2002;Weiher & Keddy, 1995).Conversely, if cooccurring species are on average functionally or phylogenetically more dissimilar compared to null expectations (i.e., overdispersion pattern), limiting similarity is expected to play a central role by either competitively excluding ecologically similar species or driving the evolution of divergent traits through character displacement (Webb et al., 2002;Weiher & Keddy, 1995).A third alternative is that if coexisting species are neither more similar nor dissimilar than expected (i.e., random pattern), stochastic processes are expected to outweigh deterministic factors in regulating community structure (Heino & Tolonen, 2017;Li et al., 2023;Webb et al., 2002).Yet, this apparent randomness could be a consequence of the combined effects of environmental filtering and limiting similarity on ecological communities (Heino & Tolonen, 2017;Weiher & Keddy, 1995).After a long-term search for a single assembly process in natural ecosystems, ecologists have now recognized that these processes can vary across communities and need not be necessarily mutually exclusive.
Focusing on the relative importance of deterministic and stochastic processes may contribute to a better understanding of community assembly (Graham et al., 2009;Mouillot et al., 2007).
Mountain regions are unusually biodiverse, harbouring more than 85% of the world's species for several animal groups (Rahbek et al., 2019).Environmental conditions change rapidly along elevational gradients on mountains, leading to various habitats and climate zones over short geographic distances (Graham et al., 2009;Montaño-Centellas et al., 2020).Therefore, elevational gradients provide a unique opportunity to explore diversity distribution and the underlying assembly mechanisms.Although elevational patterns of SR (e.g.linearly decreasing) are among the most well-documented patterns in ecology (Rahbek, 1995), it is not yet fully understood how community structures in terms of FD and PD vary with elevation and their underlying drivers (Graham et al., 2009;Qian et al., 2023).In general, high elevations are considered to be more stressful environments (e.g.cold, low productivity, and intense solar radiation) than low elevations (Graham et al., 2009;Zhao et al., 2022).Thus, based on the stress-dominance hypothesis (Weiher & Keddy, 1995), environmental filtering should influence communities at high elevations where environmental conditions are harsh, while limiting similarity should prevail in low-elevation communities where more abundant resources and higher population growth rates can result in more intense competition and niche partitioning.Although this prediction the underlying forces at two ends of the elevational gradient became more prominent in the dry season.

K E Y W O R D S
elevation, environmental filtering, fishes, functional diversity, limiting similarity, phylogenetic structure is intuitively sound and there is empirical evidence supporting it (Kuczynski & Grenouillet, 2018;Zhou et al., 2022), some counterexamples exist (Ding et al., 2021;Wang et al., 2022).For instance, phylogenetic structure of amphibians on Mountain Emei, China showed a positive relationship with elevation, implying a potential shift of assembly mechanisms from environmental filtering to limiting similarity (Wang et al., 2022).Using bird as a model system, Montaño-Centellas et al. (2020) found no global increase or decrease of FD and PD across 46 elevational gradients, highlighting the uniqueness of each mountain.These results suggest that community assembly processes along elevational gradients are complex and thus merit more research across montane systems.
Nevertheless, most previous works on fish community assembly are based on a single-season survey, which may limit our understanding as to how communities are assembled over time (Fitzgerald et al., 2017;He et al., 2022).This is especially true for large rivers, exhibiting significant seasonal variations in hydrology and nutrients (Li et al., 2020;Xia et al., 2023).Environmental filtering is expected to be more prominent during the wet season since fishes can more freely track suitable habitats (Heino et al., 2021).In contrast, the role of biotic interactions (e.g.competition) is anticipated to be strengthened in the dry season.This is because reduced habitats and resource scarcity during the low-water period probably lead to increased fish abundance and subsequently strong competition (Fitzgerald et al., 2017;Montaña et al., 2014).Conversely, stable assembly mechanisms across time have also been reported in dynamic ecosystems (Li et al., 2019;Mouchet et al., 2013).However, to our knowledge, no study has evaluated the patterns and determinants of fish functional and phylogenetic structure along an elevational gradient, while simultaneously incorporating seasonal dynamics.
Here, we explored the elevational patterns of multiple facets of fish diversity and community structure, as well as their environmental drivers in the Chishui River basin, China, during the wet and dry seasons.Specifically, we: (1) investigated how fish SR, FD, and PD vary with elevation in the wet and dry seasons separately; (2) evaluated seasonal changes in the relative importance of different assembly processes in shaping fish functional and phylogenetic structure along an elevational gradient; and (3) determined the extent to which chemical, temperature, and physical variables explain fish diversity and community structure across seasons.Based on previous findings on community assembly (Montaño-Centellas et al., 2020;Zhou et al., 2022), we first hypothesized that fish SR, FD, and PD should exhibit a linearly negative relationship with elevation across seasons (H1, Figure 1).Second, we predicted deterministic processes, instead of stochastic processes, should be dominant in controlling fish functional and phylogenetic structure (H2, Figure 1) (Graham et al., 2009).Furthermore, we expected a potential change of underlying processes from limiting similarity to environmental filtering as elevation increased (H2a) (Weiher & Keddy, 1995).In addition, the influence of limiting similarity and environmental filtering should be strengthened in the dry and wet seasons, respectively (H2b, Figure 1) (Montaña et al., 2014).Finally, we anticipated that each environmental variable should explain variations in fish diversity differently (Xia et al., 2023).

| Study area
The present study was conducted in the Chishui River basin (27°20′-28°50′ N, 104°45′-106°51′ E), a free-flowing tributary of the Upper Yangtze River, China.This river originates from Mountain Wumeng in the Zhenxiong County, Yunnan Province and flows for nearly 437 km before joining the Yangtze River in the Hejiang County, Sichuan Province.The drainage area of this river basin is approximately 20,440 km 2 .Soil types in the Chishui River basin mainly comprise yellow loam in the upstream and midstream reaches, as well as purple soil in the downstream (Jiang et al., 2017).Under the influence of subtropical monsoon climate, this river has a clear separation of the wet and dry seasons, with 60% precipitation (annual average: 1000 mm) occurring from June to September (Wu et al., 2011).
The flora in the Chishui River belongs to the boundary zone between the pan-arctic flora and the paleotropical flora, supporting multiple ancient and endemic species (e.g.Alsophila spinulosa and Ampelocalamus scandens) (Xia, Heino, Liu, et al., 2022).

| Biological sampling
During October 2019 and June 2020, fish assemblages were investigated at 40 sites distributed in the mainstem of the Chishui River basin, as well as its main tributaries (Figure 2).Nevertheless, several tributary locations were inaccessible in 2020 because of COVID-19 lockdowns, and their nearby sites were chosen.Considering the high heterogeneity of topography and hydrological regimes in this basin, we employed either active or passive sampling approaches at each site (Liu et al., 2020).Specifically, for wadable streams (i.e., upstream and tributary sites), a backpack electrofishing protocol (CWB-200P, China; 12 V import, 250 V export) was applied, with two hired fishermen slowly sampling a 200 m stretch (~30-40 min) encompassing all available mesohabitats.In non-wadable riverine sites (i.e., midstream and downstream sites), passive fishing gears, including gillnets (mesh size: 4, 5, 10, and 12 cm between alternate knots), shrimp cages (mesh size: 1.1-1.2cm; length: 13 m), and trotline (800 hooks baited with earthworms), were deployed for 12 h.Following our fishermen's experience, these gears were only inspected and emptied at the dawn of the next day to avoid disturbing fishes.By fitting a samplesized-based sampling curve, we have illustrated that sampling coverage reached more than 97% across seasons, indicating adequate sampling (Xia et al., 2023).All collected fish individuals were identified to species level, and then measured (1 mm) and weighed (0.1 g) before releasing from the sites.Voucher specimens were deposited at Institute of Hydrobiology, Chinese Academy of Sciences.
Several environmental factors were recorded in situ before fish sampling (Table S1).Altimeter and flowmeter devices were employed to measure elevation (m) and current velocity (m s −1 ).Wetted width (m) and water depth (m) were quantified using a Leica Rangemaster CRF900 and a long straight ruler, respectively.Water temperature (°C), conductivity (μS cm −1 ), dissolved oxygen (DO, mg L −1 ), and pH were measured by a Handheld YSI Model Professional Plus Instrument.Moreover, we collected water samples to determine total phosphorus, total nitrogen, total dissolved phosphorus, total dissolved nitrogen, soluble reactive phosphorus (all in mg L −1 ), and chlorophyll a (μg L −1 ).In addition, we calculated land use variables by first delineating the upstream buffer boundary (0.5 km in width and 2 km in length) of each site based on a 30 m digital elevation model.
Then, available remote sensing images of Landsat images, Sentinel 2, and ASTER were used to extract land use variables (agriculture, forest, grassland, water, other, and urbanization, all in %) using ArcGIS software (v.10.6.1).

| Quantifying species richness
Fish SR was estimated in each site during the wet and dry seasons.

| Quantifying functional alpha diversity
We compiled six continuous functional traits (age/length at maturation, growth rate, lifespan, maximum body length, and trophic level) and three categorical traits (body shape, trophic guild, and vertical position) for each species (Table S2).Body shape was coded with six levels (anguilliform, compressed, oval, cylindrical, dorso-ventrally F I G U R E 1 Hypothesized effects of elevation on multiple facets of fish diversity and community structure.(a) Individual fish species are indicated by symbols at the tips of a hypothetical phylogeny, with the shapes and colours of the symbols representing different evolutionary and functional roles.(b) Three different scenarios portray the role of deterministic (limiting similarity and environmental filtering) processes or random processes in structuring riverine fish communities.(c) We hypothesized that fish species richness (SR), functional richness (FRic), and Faith's phylogenetic diversity (PD) would monotonically decrease with increasing elevation (H1).Similarly, functional (standardized effect size of functional dispersion, sesFDis) and phylogenetic community structure (standardized effect size of mean pairwise distance, sesMPD) were expected to be negatively related to elevation (H2a).Additionally, we expected that the influence of limiting similarity and environmental filtering should be strengthened in the dry and wet seasons, respectively (H2b).Different coloured solid lines indicate expected diversity patterns in the wet and dry seasons, whereas the dotted lines show expected sesFDis and sesMPD patterns under the influence of stochastic processes.It should be noted that the seasonal colour schemes are consistent across Figures 1 and 3-5.
flattened, and fusiform), trophic guild with six levels (detritivore, herbivore, invertivore, omnivore, piscivore, and planktivore), and vertical position with two levels (benthopelagic and demersal).These traits were selected to represent multiple functions of fishes associated with food acquisition, mobility, nutrient budget, reproduction, and defence against predation (Villéger et al., 2017).For instance, trophic guild/level can mediate predator-prey interactions and trophic cascades, as well as species' responses to resource fluctuations (Kirk et al., 2022).Vertical position reflects fish mobility, occurrence, and habitat preference (Winemiller et al., 2015).We chose these traits because of their sensitivity to environmental variations (Villéger et al., 2017), and widespread usage in studies of fish FD (Dai et al., 2020;Kirk et al., 2022;Xia et al., 2023).Fish functional trait for collected species was obtained from FishBase (Froese & Pauly, 2014).
We computed 'gawdis' distance of fish traits between species pairs with the 'gawdis' package (de Bello et al., 2021).This distance type was used as it can create a balanced contribution of continuous and categorical traits when computing trait distance.
Subsequently, a principal coordinate analysis was undertaken on functional trait distances to derive a multivariate functional space, which provides a spatial summary of interspecific similarities in n dimensions, with functionally similar species grouping closer together than more dissimilar species.Here, the first two dimensions were retained for downstream analyses as these summarize a large fraction (71.8%) of interspecific variability.For FD, we computed functional richness (FRic, i.e., the volume of the convex hull shaping species in the functional space) via dbFD function from the "FD" package (Laliberté et al., 2014) in R 3.6.1 (R Core Team, 2019).

| Quantifying phylogenetic alpha diversity
To obtain a fish phylogenetic tree, we used the most recent and largest phylogeny reported by Rabosky et al. (2018).Their timecalibrated phylogeny was based on 27-gene multi-locus alignment for 11,638 species.Further, using Rabosky et al.'s (2018) tree as a backbone, 19,888 fish species with no available genetic data were placed according to their taxonomic rank (e.g.genus, family, order).Subsequently, to determine divergence times for these species, they sampled from a distribution of waiting times conditioned on rankspecific estimates of the speciation rate and sampling fraction using a custom procedure, which was repeated 100 times to generate 100 fully sampled fish megatrees (31,526 fishes).Sixty-six genera (98.507%) and 79 species (83.158%) of fish in the Chishui River basin  2018) tree, we added them to their respective genera and family using the package "FishPhyloMaker" (Nakamura et al., 2021).
Thereafter, Rabosky et al.'s (2018) tree was pruned to retain the 95 fish species captured by us (Figure S1).
We computed Faith's phylogenetic diversity (PD), which is the total length of branches connecting all species within a community (Faith, 1992).Using the derived phylogeny and fish community data, we calculated PD through pd function in the 'picante' package (Kembel et al., 2010).

| Community structure and null models
We calculated functional dispersion (FDis) and mean pairwise phylogenetic distance (MPD) to represent functional distance and phylogenetic relatedness of fishes in a community, respectively.
FDis is the species' mean distance in functional space to the centroid of all species (Laliberté & Legendre, 2010), while MPD is the average phylogenetic distance among all pairs of species in a community (Webb et al., 2002).FDis and MPD were calculated using the "FD" and "picante" packages, respectively.The two observed diversity indices were compared to those of 999 random communities to test if functional and phylogenetic structure differed from random expectations.Specifically, random communities were generated by using a "taxa.labels" null model approach, which randomly shuffles species names across the trait distance matrix or the tips of the phylogenetic tree 999 times.Then, the standardized effect size (ses) of FDis (sesFDis) and MPD (sesMPD) were calculated to represent fish functional and phylogenetic structure, respectively, using the code provided by Swenson (2014) and the following formula: where obs is the observed value of FDis or MPD, and mean null and SD null represent the mean and standard deviation of FDis or MPD for 999 random communities.A positive ses value indicates functional or phylogenetical overdispersion, whereas a negative value indicates functional or phylogenetical clustering (Webb et al., 2002).

| Phylogenetic uncertainty
Using the 100 phylogenetic trees proposed by Rabosky et al. (2018), a supplementary analysis was conducted to assess the effects of phylogenetic uncertainty on diversity patterns.We found that PD, MPD, and sesMPD calculated from our created phylogeny were all strongly correlated with the mean values of those from 100 possible phylogenies in the wet (Pearson correlation, r > .946)and dry seasons (r > .963).Therefore, the elevation patterns of fish PD and community structure detected here were not influenced by the degree to which relationships remain unresolved or variably inferred across phylogenetic hypotheses (Qian et al., 2020).

| Phylogenetic signal
Phylogenetic signal reflects the tendency of closely related species to resemble each other more than species randomly drawn from the phylogeny.The D (Fritz & Purvis, 2010) and Pagel's λ metrics (Pagel, 1999)  tively.Phylogenetic signal tests were performed with the "phytools" (Revell, 2012) and "caper" packages (David et al., 2018).

| Statistical analysis
We assessed how FD and PD indices were affected by seasonality and environmental variables.Before statistical analyses, we made log (x + 1) or square root (for proportional data) transformations on non-normally distributed environmental variables to improve normality (Liu et al., 2022).Further, we standardized (mean = 0, SD = 1) environmental variables, fish diversity (SR, FRic, FDis, PD, and MPD), and community structure metrics (sesFDis and sesMPD).First, we ran a one-sample Wilcoxon test or t-test to compare fish diversity between the wet and dry seasons (α = .05).Second, a one-sample t-test was used to determine whether sesFDis was significantly different from zero in the wet and dry seasons separately, whereas for sesMPD one-sample Wilcoxon test was applied (Heino & Tolonen, 2017).Third, to investigate the elevational patterns of multiple facets of fish diversity and community structure, linear regression models relating diversity indices and elevation were conducted.
To avoid overfitting, single-variable linear regression analyses were also run to examine the associations between fish diversity and each environmental variable.The relative importance of environmental variables was determined by the R 2 and p values.
We performed a supplementary analysis to explore if fish functional structure is consistent among different trait groups (e.g.α and β traits) as previous studies have suggested that this approach might shed additional insights into community assembly along ecological gradients (Côte et al., 2019;Kirk et al., 2022).
Specifically, vertical eye position, relative eye size, oral gape position, and relative maxillary length were employed as alpha traits (i.e., mainly reflect habitat preference or resource use), while body elongation, pectoral fin vertical position, pectoral fin size, and caudal peduncle throttling were defined as beta traits (i.e., related to locomotion).These traits were obtained from the FISHMORPH ses = obs − mean null SD null database (Brosse et al., 2021).Then, using linear regression, we related sesFDis values calculated from alpha and beta traits to elevation across seasons.

| RE SULTS
During field surveys, we recorded 95 fish species in the Chishui River basin, belonging to 67 genera, 20 families, and seven orders.
The total number of fish species was 73 (mean: 10) and 84 (mean: 10) in the wet and dry seasons, respectively.Fish assemblages across sites were dominated by Cypriniformes and Siluriformes, which accounted for 76.8% and 14.7% of fish richness, respectively.The most widely distributed fish species were Zacco platypus, Hemibarbus labeo, Acrossocheilus yunnanensis, and Sinogastromyzon sichangensis across seasons.
Based on the results of Pagel's λ (0.879-0.971), body length, growth rate, lifespan, age/length at first maturation, and trophic level all exhibited a phylogenetic signal.As for categorical traits, most of them were conserved phylogenetically, as evidenced by the findings that the D statistic approached zero (−2.736 to 1.561, Table S3).The single-variable linear models of fish diversity varied slightly concerning significant predictors and explained variation.Specifically, we found positive effects of depth, temperature, width, and % cover by water on fish SR, FRic, and PD across seasons but negative influence of DO, velocity, and pH on the same response variables.Furthermore, FDis was negatively shaped by velocity and pH in the wet and dry seasons, respectively, but positively affected by depth, width, and % cover by water in the dry season (Tables S4 and S5, Figure 5).Fish MPD was positively correlated with depth, temperature, and % cover by water but negatively to soluble reactive phosphorus, total dissolved phosphorus, and velocity in the wet season.In the dry season, depth, width, % cover by water, and % cover by other were positive predictors of fish MPD, yet DO, pH, velocity, and % cover by forest exerted a negative influence (Tables S4 and S5, Figure 5).Additionally, fish sesFDis was significantly and negatively associated with velocity and pH in the wet and dry seasons, respectively, but positively to width in the dry season (Tables S4 and S5, Figure 5).Fish sesMPD was best accounted for by soluble reactive phosphorus, temperature, total dissolved phosphorus, and velocity in the wet season and by depth, DO, pH, velocity, width, % cover by forest, and % cover by water in the dry season (Tables S4 and S5, Figure 5).

| DISCUSS ION
Examining the patterns and ecological drivers of multiple facets of alpha diversity along an elevational gradient may shed light on underlying assembly processes (Montaño-Centellas et al., 2020).Here, using the Chishui River basin as a natural laboratory, we revealed some clear patterns of fish assembly processes.As expected from our first hypothesis (H1) (Montaño-Centellas et al., 2020;Zhou et al., 2022), fish diversity metrics (i.e., SR, FRic, and PD) were all negatively correlated with elevation.This result suggested that sites at low elevations generally support more functionally and phylogenetically dissimilar species than those in the highlands.Although the SR-elevation relationship has been widely reported for various taxonomic groups (Rahbek, 1995), FRic-elevation and PD-elevation relationships have been examined less well (Sun et al., 2023;Zhao et al., 2022), especially for fishes.Nonetheless, considering the strong and positive associations among the three facets of diversity, such a negative elevational pattern is expected.
Elevational shifts in physical and chemical conditions can explain the negative association between diversity and elevation.Based on the results of linear models, six environmental factors (i.e., depth, velocity, width, % cover by water, temperature, and pH) were identified to be the most important determinants of fish alpha diversity.Thus, the negative diversity-elevation correlations could be attributed to some environment-related hypotheses, including temperature limitation (Jackson et al., 2001) and or the metabolic theory of ecology (Clarke & Gaston, 2006), and species-area or species-volume hypotheses (Connor & McCoy, 1979;Xia, Heino, Yu, et al., 2022).characteristics (e.g.small body size and herbivore).Apart from these environmental variables that covary with elevation, pH was also a negative correlate of fish SR, suggesting that sites with circumneutral conditions should be more speciose than highly alkaline sites.
Similar patterns have been recorded in previous studies on plant communities in European streams (Heino et al., 2005) and ponds (García-Girón et al., 2019).
Existing fish communities were more phylogenetically clustered than randomly generated communities, implying the predominant role of environmental filtering in affecting fish communities.Situated in the transitional zone of the Yunnan-Guizhou Plateau and Szechwan Basin, the Chishui River basin shows an extensive elevational gradient and complex river morphology (Wu et al., 2011), which could act as a filter selecting specific lineages to occur in certain sites.The important role of environmental filtering for fish biodiversity dynamics has been recognized in this system (Xia et al., 2023) and other rivers (Erős et al., 2012;Roa-Fuentes et al., 2019;Wu et al., 2022).
Because most selected functional traits demonstrated a significant phylogenetic signal, fish communities should also be clustered functionally (Webb et al., 2002).However, a random pattern rather than functional clustering was observed, indicating that stochasticity and ecological drift are the primary processes regulating fish trait composition.The effects of stochastic processes on community structure have been found in previous studies on macroinvertebrates and diatoms in the Chishui River basin (Wang et al., 2020), fishes, macroinvertebrates, and macrophytes in the Yangtze River (Jia et al., 2021;Li et al., 2023;Liu & Wang, 2018) and other taxa in terrestrial landscapes (Cadotte et al., 2019;Si et al., 2017).We, therefore, upheld the idea that deterministic processes and stochasticity work together in governing fish community assembly, which only partially supported our second hypothesis (H2).However, it should be noted that random patterns may result from the antagonistic effects of contrasting deterministic processes, and the significant role of stochasticity may thus be overestimated in our study (Heino & Tolonen, 2017).
What are possible explanations for incongruent patterns of functional and phylogenetic structure?First, measured functional traits and phylogeny may capture different ecological and biological processes (Cadotte et al., 2019).For instance, measured fish traits tend to reflect species' capacity for competition, predator avoidance, and reproduction output, while phylogenetic relatedness is assumed to represent unmeasured multivariate traits that better capture the signal of environmental filtering (Zhao et al., 2020).Indeed, the functional approach may ignore those 'difficult to measure' traits (e.g.metabolic rates, physiological traits related to cold and drought tolerance) that could have resulted in a more detectable shift of community assembly processes.Second, phylogenetic structure may reflect the biogeographic history of an organismal group (Graham et al., 2012), which possibly enable the persistence of a few closely related lineages at high elevations with diverse functional traits.
In our case, fish communities at high elevations (~700 to 1000 m) demonstrated positive sesFDis but negative sesMPD values.Third, whether functional and phylogenetic structure show consistent patterns depends on the extent to which traits are conserved phylogenetically (e.g.Blomberg's K greater than two) (Swenson, 2011).
Fourth, the number of measured traits might determine functional and phylogenetic congruence, with FD based on more traits could capture more of the subtle ways in which functional traits are filtered along ecological gradients (Cadotte et al., 2019;Davies, 2021).
Fifth the responses of functional and phylogenetic structure to ecological gradients may vary with spatial scales (Zhao et al., 2020).For example, variation in functional structure might be more easily captured by environmental factors at the mesohabitat scale, while phylogenetic structure appears to be more detectable at the site scale.
Finally, sampling and other methodological issues (e.g.intraspecific trait variation) may contribute to a certain degree of incongruencies between FD and PD.
Fish phylogenetic structure was negatively correlated with elevation, with communities being overdispersed and clustered in the low and high elevations, respectively.The results supported our hypothesis (H2a) and indicated a potential change of assembly processes from limiting similarity to environmental filtering along the elevational gradient.Thus, this study is consistent with the stressdominance hypothesis (Weiher & Keddy, 1995) and studies conducted on various organismal groups along environmental gradients (Graham et al., 2009;Kuczynski & Grenouillet, 2018;Liu et al., 2023;Monteiro et al., 2023;Qian et al., 2020;Zhao et al., 2020).Generally, high-elevation sites tend to show harsh environments (e.g.fast flow, cold, narrow channels, and homogenous food resources), which select species and lineages with certain features to survive and persist.For instance, high velocity typically supports fish species (e.g.

Acrossocheilus yunnanensis, Garra imberba, and Onychostoma simum)
with the mouth positioned under the body and sharp cuticle on the lower jaw (herbivores and detritivores) in the Chishui River basin.
This is because these fishes could cope with hydrological drag and scrape the algae on the stones.Furthermore, the substantial energy costs of inhabiting a fast flow and cold site would lead to reduced FD is and MPD (Bower & Winemiller, 2019) ecological niches (Liu et al., 2019;Xia et al., 2023).Overall, as only a few studies have tested the elevational patterns of fish community structure, more empirical tests are needed to evaluate the generality of our inferences.
We found fish functional and phylogenetic structure in the Chishui River basin were consistently shaped by stochasticity and environmental filtering across seasons, respectively.This observation is somewhat surprising for a large river showing substantial intra-annual variations in environmental conditions and hydrology and thus disagrees with our hypothesis (H2b).Conversely, Montaña et al. ( 2014) revealed that, due to shrinking water areas in the dry season, limiting similarity gained relevance for fish community assembly in tropical rivers.As our fish surveys during the dry season were conducted at a period when summer floods have already occurred several times, it seems that dispersal is not limited for most fishes, and they can thus freely track suitable habitats to avoid fierce competition (Xia et al., 2023).In addition, it has been indicated that strong environmental gradients as in our case could also create stable diversity patterns and community structure over time (Mouchet et al., 2013).Despite no significant seasonal changes of assembly patterns at the basin scale, at low elevations, fish phylogenetic structure became more overdispersed during the dry season, which suggested an increased importance of limiting similarity.Such variation of community structure between seasons may be explained by the seasonal differences of MPD among species.
At low elevations, the gain of some distantly related species (e.g.et al., 2022;Wang et al., 2020).

Leiocassis longirostris, Odontobutis potamophila, and
Although not explicitly assessed in this study, other assembly mechanisms (e.g.predation, climate, and biological invasion) could also produce nonrandom patterns of fish community structure.
First, predation has long been shown to directly influence the coexistence of fish species within a community (Jackson et al., 2001).
For example, prey species (e.g.some small fishes from Gobionidae) may select habitats where predators (e.g.Siniperca chuatsi and Culter alburnus) have difficulties in accessing them, which may lead to different community structures from those when predators are absent.Indeed, Giam and Olden (2016) have demonstrated that fish communities in most US streams had nonrandom patterns and were governed by environmental filtering and predation-prey interactions.Second, large-scale climate variables (e.g.temperature seasonality) may drive fish community structure through indirect regulation of local environmental conditions (Poff et al., 2006).
Furthermore, it was found that temporal changes in climate (e.g. increase of temperature seasonality) may increase the role of environmental filtering in shaping stream fish community structure (Kuczynski & Grenouillet, 2018).Thus, fish communities at high elevations in this system should be given due attention as these communities are strongly controlled by environmental filtering and may be vulnerable to future climate changes (Freeman et al., 2018).Third, biological invasions may also affect the patterns and underlying mechanisms of community assembly (Kuczynski & Grenouillet, 2018;Sanders et al., 2003).Specifically, exotic species could be either functionally similar or dissimilar to native species (Buckwalter et al., 2020;Darwin, 1859), increasing the force of environmental filtering or limiting similarity.Due to the needs of aquaculture, Rhynchocypris lagowskii has been artificially introduced into a tributary of the Chishui River basin, which in turn eliminated other native fish and irreversibly altered community structure.Although fish fauna in the Chishui River basin have not been seriously impacted by biotic invasions, long-term and highly frequent monitoring programs should be implemented.

| CON CLUS ION
In conclusion, we found a decreasing elevational pattern of multiple facets of fish diversity in the Chishui River basin, which were determined by chemical and physical environmental variables.We revealed that fish communities tend to be functionally random but phylogenetically clustered, indicating the joint role of deterministic and stochastic processes in community assembly.Furthermore, environmental filtering was stronger in controlling fish communities at high elevations, whereas limiting similarity prevailed at low elevations.Moreover, the two deterministic processes were more pronounced in the dry season.Overall, this study provides novel insights into the elevational patterns of community assembly in montane ecosystems.).We thank Qingyu Li, Dianming Chen, Zhiheng Wu, Yao Huang, and Yuqi He, as well as many of our colleagues for their assistance in field sampling.We also thank the editor and two anonymous reviewers for their constructive comments that helped improved the manuscript.

CO N FLI C T O F I NTE R E S T S TATE M E NT
All authors declare that they have no conflicts of interest.
were present inRabosky et al.'s (2018)  tree.For the 16 species and one genus that were present in our dataset but absent from Rabosky F I G U R E 2 The 40 sampling sites distributed in the Chishui River basin, China.Figure adapted from Xia et al. (2023).et al.'s ( were used to test phylogenetic signals of categorical and continuous traits, respectively.If D = 0, the trait is phylogenetically conserved as expected under the Brownian motion model of evolution, and D < 0 indicates a highly conserved trait.The value of D = 1 or >1 suggests a random or overdispersed trait on the phylogeny.For Pagel's λ, a value of 0 and 1 indicate no and strong phylogenetic signal according to Brownian motion model, respec-

F
Elevational patterns of fish (a) functional dispersion (FDis), (b) mean pairwise phylogenetic diversity (MPD), (c) the standardized effect size of FDis (sesFDis), and (d) the standardized effect size of MPD (sesMPD) across the wet and dry seasons in the Chishui River basin, China.The solid lines represent significant relationships (p < .05) between diversity indices and elevation in linear models, while the dashed lines imply nonsignificant relationships between diversity indices and elevation.
Specifically, low-elevation sites located in the Szechwan Basin are typically characterized by a warm climate, wide and deep channels, comparatively low velocity, and ample food resources, which may enable different kinds of fishes to complete their life history processes, thereby promoting high SR.By contrast, high-elevation regions in the Yunnan-Guizhou Plateau with stressful environmental conditions (e.g.cold temperature, shallow water, high velocity, and low productivity) only allow the coexistence of fishes with certain F I G U R E 5 Variation in multiple facets of fish alpha diversity (panel a: SR, species richness; panel b: FRic, functional richness; panel c: FDis, functional dispersion; panel e: PD, Faith's phylogenetic diversity; panel f: MPD, mean pairwise phylogenetic diversity) and community structure (panel d: sesFDis, the standardized effect size of FDis; panel g: sesMPD, the standardized effect size of MPD) explained by individual environmental variables across the wet and dry seasons in the Chishui River basin, China.The y-axis represents the R 2 of singlevariable linear models, and only significant models were shown.SRP, soluble reactive phosphorus; TDP, total dissolved phosphorus; Temp, temperature.
Monopterus albus), along with intensified interspecific competition in the dry season because of reduced habitats, may increase MPD and create an overdispersed pattern.Contrastingly, at middle and high elevations, due to harsher environmental conditions in the dry season, the loss of several distinct relatives (e.g.Pseudobagrus truncates, Rhinogobius cliffordpopei, and Misgurnus anguillicaudatus) may lead to decreased MPD and more clustered structure.Our results emphasized seasonal variation of the patterns and mechanisms of fish community assembly at individual sites, and future research on this topic should therefore take seasonality into account (He