Developing a Robust SDM Pipeline for Predicting Climate-Driven Shifts in Freshwater Fish Distributions

Understanding climate-induced shifts in freshwater biodiversity requires rigorous modeling approaches. This study applies a spatially explicit species distribution modeling (SDM) framework, incorporating high-resolution (1 km) environmental predictors and 3,307 field survey sites collected between 2016 and 2022 in South Korea. Using Maxent modeling, we assessed the potential distribution shifts of 130 freshwater fish species under three climate scenarios (SSP1-2.6, SSP3-7.0 and SSP5-8.5). To enhance model accuracy and applicability across regions, we implemented a robust preprocessing pipeline, including spatial filtering to correct sampling bias, multicollinearity reduction (Pearson correlation and variance inflation factor), pseudo-absence generation, and ensemble modeling to minimize uncertainty. Model performance was validated using AUC and TSS metrics, ensuring reliable predictions for conservation planning. Our results indicate that 48 species are at risk of significant range loss by 2080, with climate-driven contractions disproportionately affecting endemic and endangered species, in addition to cold-water taxa. Specifically, 17 endemic species and 6 nationally endangered species, including Coreoleuciscus splendidus and Cottus koreanus , were identified as highly vulnerable, highlighting the urgent need for conservation interventions. Regional projections pinpointed biodiversity loss hotspots, particularly in the Yeongsan and Nakdong River basins. This study demonstrates how a reproducible SDM pipeline, combining rigorous preprocessing and climate scenario analysis, can provide actionable insights for freshwater conservation planning. Our approach is adaptable to other regions, offering a transferable framework for assessing climate-driven biodiversity shifts in freshwater ecosystems.