Migratory fish are among the most threatened freshwater taxa globally. Diadromous species depend on connected freshwater, estuarine and marine habitats to complete their life cycle, yet these environments are increasingly fragmented and degraded. Species distribution models (SDMs) are a widely used tool to identify priority habitat, guide restoration actions and to assess vulnerability to environmental change. However, because SDMs relate species occurrences to the environmental variables at observed locations, they may only reflect a spatial subset of the realised niche, overlooking ecological processes that operate at different spatial scales. For migratory fish species, conventional SDMs may therefore inadequately represent their complex life history and thus lead to inaccurate predictions of potential distribution.
We present a hierarchical framework that represents ecological processes operating at multiple spatial scales. We evaluated this framework using simulated occurrences of a virtual diadromous species, for which the true distribution is known, allowing differences in model performance to be attributed to model structure rather than uncertainty in occurrence data. We tested whether (i) hierarchical SDMs more accurately represent the potential rather than the realised niche under future environmental conditions, (ii) hierarchical SDMs provide more robust predictions than conventional SDMs and finally, (iii) incorporating estuarine variables improved predictions for diadromous species.
We found that SDMs in general represent the potential niche rather than the realised niche when transferred to novel environmental conditions. Compared with conventional SDMs, hierarchical SDMs more accurately captured basin-scale and river reach-scale ecological processes, improving predictions of which river basins support populations and which reaches provided suitable habitat. Incorporating estuarine variables further improved predictions, highlighting the importance of representing recruitment processes that occur outside of the freshwater environment. Our results demonstrate that hierarchical SDMs provide a practical framework for representing ecological processes across multiple spatial scales and improving predictions of distribution for migratory fish.