Discussion

Despite decades of study, quantifying the relative importance of deterministic versus stochastic processes in natural community assembly remains a key challenge to ecologists (Dini-Andreote et al. 2015; Tucker et al. 2016), especially in dynamic species-rich landscapes (Ruhí et al. 2017). In this study, we investigated the temporal dynamics of the benthic fish communities in two habitat types (i.e. natural vs modified) within a large river-floodplain system over a water level rising-and-drawdown cycle for three consecutive years. We found distinct assembly processes operating between habitat types, contributing to the variation in observed community dynamics, such as temporal β-diversity community stability, and species synchrony. Moreover, our results suggested that the nutrient gradients strongly mediate the community assembly processes (Donohue et al. 2009), exhibiting significant effects on temporal community dynamics.

Distinct normalized ecological stochasticity in natural and modified habitats

Null models have become a valuable tool to infer assembly processes from spatial variations in community structure (Chase et al. 2011; Mori et al. 2015). However, few studies have applied null models to explore the temporal dimension of community turnover. Here, we used the NST based on a null modelling approach (Ning et al. 2019) to quantify the ecological stochasticity using monthly samples of benthic fish community. NST reflects the contribution of stochastic assembly relative to deterministic assembly, providing a better quantitative measure of stochasticity than other randomization-based measurements such as standardized effect size (Ning et al. 2019). While the mean NST was much less than 0.50 in the modified sites, it was greater than 0.50 at the natural sites. These results suggested that deterministic processes, either environmental filtering or biotic interactions or both (Dini-Andreote et al. 2015; Ning et al. 2019) were the main mechanisms underlying the temporal variation in community structure in the modified habitat. In contrast, in the natural sites especially in open waters, stochastic processes, i.e. demographic (e.g. probabilistic dispersal and random birth - death events) and environmental stochasticity (e.g. changes in water level, pH and water clearance) (Chase & Myers 2011;Shoemaker et al. 2020) could be the prevailing processes driving the temporal structural changes in fish communities.
The lower stochasticity (thus, higher deterministic forcing) in modified habitats could be attributed to many interconnected abiotic and biotic factors. First, modification reduces hydrological connectivity (Li et al. 2020), which depresses temporal variation and creates relatively stable environmental conditions. In turn, the stable conditions might trigger environmental filtering and biotic interaction, such as competition. Second, habitat modification cleared submerged macrophytes through excavation of sediment (field observation by authors), leading to further environmental homogenization in terms of structural complexity. Third, modification could impose dispersal limitation through habitat isolation. Dispersal (e.g., random chance for colonization) can be considered as a more stochastic process that induces deviations from expectations based solely on niche theory (Chase and Myers 2011). Studies have shown that ecological connectivity and dispersal play a central role in structuring communities (Cadotte et al. 2006; Vellend et al. 2014). Our findings demonstrated the importance of considering the ensemble of local communities as an integrated metacommunity, and approach which improves our understanding of the processes maintaining biodiversity in complex dynamic landscapes.

Stochasticity is positively related with temporal β-diversity, synchrony and positive covariation

Previous studies have demonstrated that human disruption, such as habitat modification, could cause biotic homogenization (Iacarella et al. 2018), leading to reduced spatial β-diversity in fish communities in floodplain environments (Quintero et al. 2010; Li et al. 2020). In this study, we found significantly lower temporal β-diversity of benthic fish communities in the modified plantations than in the natural Carexsedges and open waters. Moreover, a significantly positive relationship between ecological stochasticity and total temporal β-diversity was confirmed.
Although the difference in species variance rate between habitat types was not significant, species synchrony in communities was generally higher in natural habitats than in the modified sites. In addition, both species synchrony and variance rate were significantly positively related to NST. The higher synchrony in natural habitats can be partially explained by life history theory (Mims & Olden 2012). Native fish have developed life history strategies adapted to the natural inter- and intra- variations in hydrology (McManamay & Frimpong 2015). Therefore, the life cycles of native fish, such as reproduction, larval and juvenile survival, migration and movement of young and mature fish, are synchronized with long-term hydrological regime (King et al. 2003). In the plantation, to ensure the survival of young trees, the natural hydrological regime was greatly modified, especially during low water periods (Li et al. 2020), resulting an interruption to how native biota respond to environmental cues. There is general agreement that dispersal can synchronise spatially distinct subpopulations (Goldwyn & Hastings 2008; Vogwill et al. 2009). Therefore, the lower species synchrony in modified habitats could also be attributed to the dispersal limitation.
The community stability, measured by the time series of a univariate index (the sum of fish abundance), was significantly higher in the modified habitat than in natural habitats. It had significantly negative relationships with NST, species synchrony and variance rate. As we have shown that ecological stochasticity could lead to high synchrony and positive covarying (i.e. abundance changes in the same direction) among fish species in river-floodplain system, these negative relationships suggested that temporal asynchrony (i.e. compensatory dynamics, as in Gonzalez and Loreau 2009) helps to stabilize abundance fluctuations in the modified habitats. Thus, the persistence of the generalists such asCarassius auratus was higher at the modified plantation sites, where environmental conditions are relative stable. As the use of a single univariate index can potential give misleading impression of the overall picture of community stability (Death & Winterbourn 1994), time lag analysis using community distance (Collins et al. 2000 & 2008; Hallett et al. 2016) can provide insights into community stability (Jones et al. 2017). However, the duration of our dataset is not long enough to have a meaningful analysis using this approach.

Nutrient gradients mediate community temporal dynamics

Nutrient enrichment affects temporal dynamics of biotic communities (Cook et al. 2018; Avolio et al. 2014), and many studies documented that elevated nutrient levels cause biotic homogenization, reducing the temporal β-diversity of biotic communities, such as plants (Zhou et al. 2020) and macroinvertebrates (Huttunen et al. 2020). We found significant relationships between nutrient level (TN or TP) and community temporal dynamics. Species synchrony and variance rate had negative relationship with TN whereas the temporal β-diversity decreased significantly with TP. Note that TP and TN were highly correlated (Pearson’s r = 0.73), and we used the variable with the closer relationship in all models. Johnson and Angeler (2014) also found the negative relationship between TP and temporal β-diversity in a meta-analyses of European fish community diversity. The low phylogenetic diversity and loss of functional diversity in high nutrient conditions might be mediated by periodically hypoxic conditions in floodplain settings (King et al. 2012) or the responses of food sources such as benthic diatoms (Allen 2004; Johnson and Angeler 2014).