Freshwater ecosystems need efficient protection which requires detailed information regarding the spatial distribution of its environmental characteristics, which allows simple habitat suitability assessments for freshwater species. Such characteristics can be assessed with regionalisation analyses, where environmental characteristics are spatially clustered to highlight similarities or disparities across a given study area. While large drainage basins are useful for large-scale estimates, it is equally important to address small streams which contribute most to the stream network length. The question however remains, what is the relative impact of the spatial scale and the choice of variables on regionalisation analyses? We tested for scale- and variable-contingent effects in freshwater habitat clusters using three analysis designs. We used the Hydrography90m high-resolution stream network dataset and aggregated land cover, hydro-geomorphological and climatic variables across the sub-catchments of six drainage basins distributed across continents and climatic zones. We then employed k-means cluster analyses and tested the effect of (i) spatial scale, (ii) the choice of environmental variables and (iii) the combination of scale and variables on the resulting habitat regionalisation. Our results show that similar broad habitat cluster patterns emerged regardless of the analysis design, whereas basin-specific analyses uncovered new smaller habitat clusters. Land cover stood out as the most influential variable regardless of the analysis design. Our findings highlight the importance of addressing the spatial scale in freshwater regionalisation analyses for assessing environmental characteristics that are unique to a given drainage basin, which could provide guidance for an improved mapping of high-resolution freshwater habitat patterns globally.
