Studies applying species distribution models (SDMs) have multiplied exponentially over the last years, however only few deal with freshwater ecosystems. As streams and rivers have environmental conditions that diverge strongly from terrestrial and marine habitats, adaptations have been explored on how SDMs can be applied in freshwater ecosystems in order to improve model performance and prediction accuracy[1-3]. SDMs are mostly applied to large areas, from the regional to the continental scale (> 10.000 km2) at rather coarse spatial resolutions (> 1 km2). These models support research for large taxonomic groups across wide areas. The main outcomes of such models are continuous distribution predictions for single species and the relative importance of the environmental predictors that drive the distribution model. Frequently, predictions for the future are also made, based on future climate scenarios. These distribution predictions give hints on possible developments in the occurrence ranges of single species. In most cases, warm-adapted species expand their predicted distributions, while endemic and cold-adapted species see their suitable habitats reduced5. Freshwater SDMs have also been applied to much smaller scales (< 10.000 km2 ), at finer spatial resolutions (< 1 km2 ) (Fig. 1). This approach has focused on the catchment, using the stream network as a modeling extent, rather than continuous surfaces. In this way, environmental predictors unique to freshwater ecosystems can be included, such as discharge and flood events6. In addition, common predictors such as land use are adapted as to portray the hierarchical framework of stream networks. The combination of these properties has yielded improved predictions and model performance. Small scale freshwater SDMs have many potential applications as a regionally focused decision support tool in scientific monitoring, environmental conservation and landscape planning. Models developed have mainly focused on Europe and particularly Germany, but there are also experiences in China (Fig. 1). Large and small scale SDMs are built as ensemble models and consist of several replicates using different algorithms and allocation of pseudo-absences (random non-occurrences required to calibrate the model). Thus, for each species modeled, replicates can easily exceed 100 and many species are frequently modeled in parallel, which makes remote computing power absolutely necessary. Ongoing research further develops small scale (high resolution) freshwater SDMs applied to the Kinzig catchment (Hessen) using data from Senckenbergs long term ecological research (LTER) site Rhein-Main-Observatorium. Further applications are also ongoing or planned at regional and continental scales.
Freshwater Species Distribution Models
Senckenberg Forschungsinstitut und Naturmuseum Frankfurt