Better representation of linear features in species distribution models: Mapping the distribution of the Scaly-sided...

Accurate species distribution modelling is essential for conservation planning, particularly for river-dependent species whose habitats are poorly captured by conventional distance-based predictors. The endangered Scaly-sided Merganser (Mergus squamatus) relies on linear freshwater habitats during winter, yet riverine habitat structure is often oversimplified in broad-scale species distribution models. Here, we developed a multi-algorithm modelling framework to map wintering habitat suitability for the species in the Dongting Lake Basin, China. Occurrence records were compiled primarily from systematic surveys and spatially thinned to reduce sampling bias. We compared MaxEnt, random forest, generalized additive models, and boosted regression trees under five-fold spatial block cross-validation, and generated an ensemble prediction across algorithms. The final predictor set emphasized hydrological and topographic attributes, including maximum water width, reservoir capacity, density of the river network, valley depth, landform type, and channel-network base level. Predictor importance and response curves indicated that wintering habitat suitability was shaped mainly by open-water scale, terrain position, valley structure, and regulated water storage. The ensemble prediction identified priority high-suitability areas mainly along the middle and lower Yuan River and the upper Xiang River, with smaller local patches along the Li and Zi Rivers and selected reservoir systems. These findings suggest that hydrological and topographic predictors can enhance the ecological representation of river-dependent waterbird habitats in SDMs and provide spatial evidence for targeted river-reach conservation.