Our study area is the catchment of the Thousand Islands Lake (TIL), which is located in the subtropical monsoon climate zone of Eastern China ( Figure 1 ). The area has a catchment area of approximately 10,080 km², with an average annual precipitation of 1430 mm. Within the catchment, the sampling sites have an elevation gradient of over 500 m. Studies across such large environmental gradients are extremely helpful in revealing the influence of the environment on community assembly mechanisms (Thrush et al., 2006).

We investigated 147 sites in the TIL catchment in April and May 2021. Sampling sites cover as evenly as possible the wadable section of the river from the headwaters to large rivers. Meanwhile, sampling sites were selected to avoid stream sections directly below the confluence or with severe impacts from anthropogenic disturbances (e.g., wastewater outlets, bridges and villages).

At each sampling sites, macroinvertebrates were collected with a Surber net (30 × 30 cm², 500 μm mesh) from habitats with different substrate compositions (e.g., cobble, rock, fine sediment) and flow conditions (e.g., water depth and current). The collection was repeated three times at each sampling site and the collected samples were stored in a portable holding tank and then sorted and then preserved in 70% alcohol. Macroinvertebrates were identified in the laboratory to the lowest possible taxonomic level (mainly to genus) using both books (Morse et al., 1994; Holzenthal, 2009; Merritt et al., 2017) and online identification resources (https://www.macroinvertebrates.org/).

In our study, metacommunity deconstruction, selection of explanatory variables (Local, Geo and Spatial only, while Bio data were enforced into all models) and variation partitioning analysis were major processes ( Figure 2 ).

In total, five trait groups were collected for each species ( Table 1 , Appendix 2 ). Two groups of traits (i.e., DAG and FFG) were deconstructed and included in the next step of the analysis.

All data processing was performed in R (Version 4.2.1, R Core Team, 2020). First, we prepared explanatory factors datasets including local environmental (Local), geo-climatic (Geo), spatial (Spatial) and biotic interactions (Bio) factors ( Table 2 , Appendix 3 ). In the next step, the abundance matrix was subjected to a Hellinger transformation. Prior to the analysis, we removed variables with significant multicollinearity (with variance inflation factor ≥ 3; vifstep function in the R package usdm) for all explanatory factors (Naimi et al., 2014). The forward.sel function in the R package adespatial was used for forward selection (Blanchet et al., 2008a), which contains two stopping criteria: significance level and adjusted coefficient of determination (adjusted R²). Subsequently, we obtained a final group of Local, Geo and Spatial factors and tested the significance of each component using the anova function in the R package vegan (Oksanen et al., 2019). Details of forward selection were described in Appendices 3 and 4 . Because we wanted to know the effects of biotic interactions on macroinvertebrates communities, biotic interactions factors were forced into all of the variation partitioning models without the above procedures. Finally, the varpart function in the R package vegan was used in the variance partitioning analysis to quantify the pure and shared effects of the four types of explanatory factors on the communities of macroinvertebrates.

Abiotic and biotic variables showed a large variation across the sampling sites ( Table 2 ). For example, conductivity (Cond) ranged from 14.23 to 363.00 µs/cm (mean: 167.07 µs/cm) while the concentrations of nitrate nitrogen (NO3-N) ranged from 0.21 to 3.40 mg/L (mean: 1.02 mg/L).

In total, 199 taxa were observed, belonging to 4 phyla, 9 classes, 22 orders and 89 families. Aquatic insects accounted for 75% of the total richness, with Ephemeroptera (36 taxa) and Diptera (31 taxa) being the most abundant taxa. In terms of traits composition, the dominant functional feeding group and habit trait group were predator (64 taxa) and clinger (88 taxa).

The results of forward selection procedures were in Appendix 4 and 5 . Results of the variance partitioning showed that local environmental, geo-climatic, spatial and biotic interactions all had considerable impacts on taxonomic and trait compositions of macroinvertebrates ( Figure 3 ), supporting our first hypothesis (H1). The total explained variation by the four explanatory factors for taxonomic composition and trait compositions was 15% and 28%, respectively. The pure effect of environmental filtering (i.e., local environmental and geo-climatic) was smaller than that of spatial processes for both taxonomic and trait compositions, which rejected part of our second hypothesis (H2). In addition, biotic interactions contributed more to the explained variation in taxonomic composition than local environmental and geo-climatic variables.

For the dispersal ability groups, the total explained variance ranged from 9% to 61%. The highest pure effect of environmental filtering (local environmental and geo-climatic) was for DAG04 (2% and 1%), followed by DAG03 (1% and 1%), and the lowest was for DAG02 (only geo-climatic 1%) and DAG01 (only geo-climatic 1%), which supported our third hypothesis (H3a). Meanwhile, the highest pure effect explained by spatial was for DAG04 (52%), followed by DAG01 (15%) and DAG02 (11%) and the lowest was for DAG03 (4%) ( Figure 4 ).

For the functional feeding groups, the pure effect explained by biotic interactions was 2% for scrapers and non-scrapers, which did not prove our third hypothesis (H3b). In addition, the pure effect of spatial on scrapers (7%) was much higher than other functional feeding groups (2%) ( Figure 4 ).

Our results showed that local (i.e., environmental filtering and biotic interactions) and regional (i.e., dispersal limitation) processes jointly shaped the metacommunity structure of stream macroinvertebrates in the Thousand Islands Lake Catchment, which supports our first hypothesis (H1). This stands in line with observations in previous studies that biodiversity patterns in stream ecosystems are influenced by a multitude of ecological processes at local and regional scales (Poff, 1997; Heino et al., 2007; Chiu et al., 2020). We have investigated the potential role of biotic interactions in community assembly of stream macroinvertebrates, which has received limited attention in previous research (Barbee, 2005; Schneck et al., 2013). Despite of this, the proportion of explained variation in our study remained quite low (15–28%). Although low explained variation was also observed in previous studies that investigated the factors shaping metacommunity of stream macroinvertebrates (Cai et al., 2017; Perez Rocha et al., 2018; Branco et al., 2020), such phenomena indicates that some underlying drivers of macroinvertebrates metacommunity structure may have been overlooked. Indeed, many environmental variables (e.g., detailed information on hydromorphology of sampling sites), environmental regimes (e.g., temporal regimes of flow and nutrient concentrations), spatial processes (e.g., different dispersal pathways between sampling sites) and stochastic events (e.g., random local extinctions), albeit challenging to measure, have not been included in the analyses but might have substantial influence on metacommunity structure of stream organisms (Heino et al., 2015; Wu et al., 2022b; Wu et al., 2023).

The effects of environmental filterings on taxonomic and trait compositions were largely consistent, while the spatial factors had a greater effect on both taxonomic and trait compositions than the environmental variables, which partially rejects our second hypothesis (H2). The unexpected large influence of spatial processes is consistent with the findings of recent studies (Mugnai et al., 2022; Wu et al., 2022a). We speculate that this may be related to the following reasons. First, previous studies had generally concluded that spatial processes had a limited impact on small organisms, with environmental filtering playing a dominant role (Leibold, 1998; Cottenie et al., 2003). However, over the last two decades an increasing number of studies demonstrated that spatial processes also have a significant impact on small organisms (Liu et al., 2021), and this was similarly the case in streams (Wan et al., 2015; Wu et al., 2022a). Second, the heterogeneous mountain landscape provided a variety of unique ecological niche opportunities and microhabitats for macrobenthos, as well as relatively large dispersal barriers for species with poor dispersal abilities (Li et al., 2021). As a result, these factors might contribute to the strong influence of spatial factors on taxonomic and trait compositions. Surprisingly, although biotic interactions did not affect trait composition, they explained a non-negligible percentage (2%) of the taxonomic composition. Macroinvertebrates play an important role in stream ecosystems, some of the macroinvertebrates diet also includes coarse or fine particles and zooplankton, and macroinvertebrates may also be food for some fish (Guo et al., 2016; Schmera et al., 2017). Therefore, if future studies consider quantifying biotic interactions across multiple trophic levels, the explanatory rate of this process may be further improved.

These three ecological processes (i.e., environmental filtering, spatial processes and biotic interactions) have varying effects on different macroinvertebrate traits groups (supporting H2), as found in our findings and other related studies (Souffreau et al., 2015; Wan et al., 2015; Liu et al., 2016). Trait-based deconstruction approach can improve our understanding of the mechanisms of community assembly in different traits groups. This demonstrates the importance of using traits in bioassessment. For example, certain species that are little affected by spatial processes (mainly mass effects) and are closely linked to their environment may be excellent candidates for biological evaluation (Leboucher et al., 2020; Bried and Vilmi, 2022). However, only part of the priori expectations has been verified (H3).

We expected that the importance of environmental filtering was positively associated with the dispersal ability of macroinvertebrates. For taxa with strong dispersal ability, it is assumed that they could overcome the barriers within the catchment and find suitable habitats. Therefore, dissimilarity of their compositions could be used to track environmental gradients within the catchment (Heino, 2013). The predicted importance of environmental filtering for different groups of dispersal ability was broadly in line with our expectations. At the same time, the effect of the spatial processes gradually decreased from DAG01 to DAG03, which was in line with our expectations. However, the effect of the spatial factors on DAG04 was surprisingly high. This might be due to the strong mass effects on DAG04 (i.e., strong aerial dispersers) in this particular basin.

Mass effects allow species to persist by dispersing to unfavorable environments (Mouquet and Loreau, 2003). Species in DAG04 have a strong dispersal capacity, which may lead to such species being subject to mass effects or even severing their links with their environment (Heino et al., 2017; Corte et al., 2018; Wu et al., 2022a).

For FFG, we hypothesized that the effect of biotic interactions on the scraper would be stronger than on the other groups. However, we did not detect an obvious difference between the two groups. The use of diatom traits alone to indicate complex biotic interactions associated with macroinvertebrates is likely insufficient. Moreover, we only considered diatom traits while scrapers also feed on other components of periphyton (i.e., fungi, microbes, bacteria, plant detritus, and animals) on substrates (Stelzer and Lamberti, 2002; Heino, 2005). A more comprehensive representation of the effect of biotic interactions on scrapers would be possible if the above factors were considered in the future studies. We also observed that the spatial processes had a greater effect on the scrapers than other functional feeding groups. In the Thousand Islands Lake Catchment, many scrapers are aquatic obligate species (e.g., Gastropoda) and have poor dispersal ability. Dispersal limitation is likely to play an important role in shaping their spatial distributions (Firmiano et al., 2021).