DISCUSSION
In the analyses of richness and beta diversity, we found that environmental conditions captured most of the variation in fish species diversity, suggesting that environmental conditions were satisfactorily represented using average values and standard deviations, whereas the spatial component contributed little to structuring fish communities. Interpretation of spatial processes and its association with dispersion, made using spatial eigenvalue maps, has to be done carefully because they bring not only information on spatial structuring but also on environmental conditions (Diniz-Filho et al. 2003; Hawkins et al. 2007).
The beta diversity of different guilds, especially insectivores and omnivores species, displayed little correlation with environment conditions and a greater one with spatial processes. This result may be explained by the spatial component representing collinear unmeasured environmental conditions rather than effects of space itself, or that insectivorous and omnivorous fish have greater dispersion abilities. A higher dispersal capacity can hide effects of environmental gradients, as species can rapidly colonize sites that have unfavorable conditions, and suppress effects of local extinction (Grönroos et al. 2013). If this was the case for insectivorous and omnivorous species, Mass Effect may be the predominant mechanism structuring their communities. In contrast, detritivores fish and the total community were structured by Species Sorting effects, or an interaction between Mass Effect and Species Sorting (Cottenie 2005), as these communities showed a strong relationship with environmental conditions and a weak one with space.
Although the Local W model was the best representation of space for the entire ichthyofauna community, we observed a weak relationship between fish communities (richness and total beta diversity) and spatial components. This weak association may have occurred due to two factors: (i) the ability of fish to actively select habitats would be more important than dispersion, indicating the SpeciesSorting mechanism of metacommunity structure (Leibold 2004), which is considered the main mechanism responsible for structuring natural meta communities (Cottenie 2005; Van der Gucht et al. 2007; Vanschoenwinkel et al. 2007; Landeiro 2011), (ii) the ”path” along which fish species have dispersed is environmentally unsuitable, restricting their dispersion (Grönroos et al. 2013). Although the characteristics of streams were controlled in order to be representative of natural conditions, the connection between these points was not controlled. Thus, it may be that sites, although preserved, are connected by non-suitable drainage that act as a barrier and limit the dispersion of ichthyofauna (Grönroos et al. 2013).
Linear distances without considering hydrographic units fail to represent spatial process in stream fishes communities. The use of the linear distance is a simplistic and insufficient way to measure spatial processes in aquatic systems (Landeiro 2011). When including the geographic barriers into the linear distance procedure (called Local W here), the linear distance was able to provide a good representation of the geographic patterns of the fish community. The performance of Local W is better than the dendritic distance (W Water), shown previously as the best means to represent spatial processes in aquatic systems. Using Local W, we were able to identify a simple and robust way of representing spatial processes considering the physical barriers separating communities. This result was consistent across all components of the community (i.e. overall community or discrete feeding guilds) and ecological descriptors (i.e. species richness or beta diversity).
A second feature of our results is related to the dispersion of detritivores fish, which cannot be represented using simple Euclidian distances between points, considering or not geographic barriers or hydrological distances. The best way to represent space process for this guild was applying the Hydalt W model. It considers the dendritic distance between points (and barriers), and the direction of the flow. The application of connectivity models associated with flow direction is advocated by some authors as being the most suitable for aquatic organisms (Pearson et al. 2007; Peterson and Ver Hoef 2010). However, with the analysis of our data, we observed that this approach is more complex than the Euclidian distance (considering barriers - Local W), and was only necessary for detritivores fish.
The low dispersion of stream fish is evidenced by the negative relationship between fish communities and spatial filters. Considering that small eigenvalues ​​are related to local structuring and higher eigenvalues ​​to larger scale structuring (Grffith et al. 2006; Blanchet et al. 2008), we interpret that fishes from stream are dispersing across small spatial scales, as richness and beta diversity were negatively related to the filters. This result supports the idea of ​​preserved spots connected by altered ”paths”. Thus, fish species with low dispersal ability tend to express spatial structure more clearly than ones with higher dispersal ability (Thompson and Townsend 2006; Astorga et al. 2011).
Species Sorting predicts a correlation of metacommunity structure only with environmental gradients (Leibold et al. 2004). The interaction of space (dispersion) and environment in the structuring of metacommunities is attributed solely to the Mass Effect model2. However, is it not possible to separate these two mechanisms, and metacommunities that are related to space and environment are understood as interaction between Species Sorting and Mass Effect (Cottenie 2005; Vieira and Tejerina-Garro 2020; Brasil et al. 2019). This interaction was found in 29% (46) of the communities analyzed by Cottenie (2005) and also found for fish (Landeiro et al. 2011; Vieira and Tejerina-Garro 2020; Vieira et al. 2020) and other groups (Brasil et al. 2019). However, the ichthyofauna show a greater correlation with environmental conditions than spatial processes, suggesting that Species Sorting is the key mechanism in structuring fish communities (37% of the metacommunity analyzed by Cottenie (2005) show this result). This result does not rule out the occurrence of dispersion (Cottenie 2005; Landeiro et al. 2011; Vieira and Tejerina-Garro 2020; Brasil 2019), but reinforces the idea that dispersion occurs locally and weakly (negative relation with the spatial eigenvectors maps).
Species Sorting is related to many freshwater organisms, such as macroinvertebrates (Grönroos et al. 2013; Heino et al. 2013), snails (Hoverman et al. 2011) and bacteria (Van der Gucht et al. 2007). The influence of the Species Sorting mechanism in communities at large scales (three basins) was found indicating that there was no variation in the metacommunities among basins, but a continuous variation according to an environmental gradient (Heino et al. 2014). Dispersion has little influence on metacommunity patterns at large scales (Grönroos et al. 2013; Heino et al. 2013). In fact, dispersion decrease at large scales as the same way that environmental gradient increase the correlation with metacommunity structure (Grönroos et al. 2013; Astorga et al. 2011). In addition, fish can only migrate using drainage and some species tend not to migrate or are prevented from migrating due to physical barriers such waterfalls or dams. Furthermore, species that actively disperse tend to select the environment in which they will settle, further reducing effects of spatial structure (Grönroos et al. 2013).
The variation partitioning of fish communities between environmental conditions and geographic space tend to associate environmental conditions with niche (i.e. environmental conditions) or neutral (i.e. geographic distances) theory (Smith and Lundholm 2010). When communities are more correlated with environmental conditions than geographic distance, the communities are structured by niche theory. On the other hand, when the geographic distance are more correlated than environment, the discussion is focused on neutral theory. This dichotomous view is overly simplistic and does not reflect the complexities of the multiple mechanisms that concurrently structure natural communities. Communities are not the effect of only one of these two theories, but the interaction between them (Juen and De Marco 2012). More recent analyses (Cottenie 2005; Van der Gucht et al. 2007; Grönroos et al. 2013; Astorga et al. 2011; Vieira and Tejerina-Garro 2020; Vieira et al. 2020; Brasil et al. 2019; Heino 2011) demonstrate the association of the community with the mechanisms proposed by Leibold et al. (2004). In this case, the relationship between communities and environmental conditions are related to the Species Sorting mechanism, the relationship between community and space to the neutral or Patch Dynamic mechanisms, and the interaction between space and conditions related to the Mass Effect mechanism (Cottenie 2005). However, the dissociation of Mass Effect and Species Sorting is not trivial, since the dispersion limitation in communities within Species Sorting can produce a pattern with relationship between space and environmental conditions, a Mass Effect dynamic (Cottenie 2005).
Therefore, we conclude that the ichthyofauna of Cerrado streams are structured by an interaction between Mass Effect and Species Sorting mechanisms. Among the trophic guilds considered, only the beta diversity of insectivores and omnivores is influenced by the geographical space, suggesting an effect of Neutral or Patch Dynamic models. Finally, we found that the linear distance measures that take into account the physical barriers (Local W in this study) are the best representations of spatial patterns for the fish communities, except for the detritivores that are influenced by the flow direction variable in the connectivity models (Hydalt W). Regarding the environmental conditions, it is necessary that these were represented by the average and some metric (e.g. heterogeneity of substrates, beta diversity and metrics variance) that measures their variation to best evaluate the degree of environmental heterogeneity.