Validating a Past Local Extinction Model Using Site‐Level Disappearances in a Cold‐Adapted Salmonid Under Global Warming

ABSTRACT Global warming poses an increasing threat to biodiversity by elevating the risk of local extinctions. Species distribution models are widely used to forecast extinction risk under future climate scenarios; however, empirical validation of these models remains rare. Here, we validated a past local extinction model using site‐level disappearances—defined as resurvey‐based absence at sites where the focal species was recorded 20–60 years ago—which can serve as a practical surrogate for local extinction. We focused on a cold‐adapted salmonid fish, the southern Asian Dolly Varden ( Salvelinus curilus ), in Hokkaido, Japan, and further improved extinction risk models by incorporating spatial climate variation, seasonal thermal regimes, subsurface water effects, and biotic interactions. We resurveyed 138 historical sites between 2018 and 2024 by electrofishing to assess the presence of southern Asian Dolly Varden. Long‐term air temperature data and modelled groundwater temperatures were analysed to estimate warming trends. Multiple disappearance risk models were developed incorporating summer temperature trends and interspecific competition with white‐spotted charr ( S. leucomaenis ), masu salmon ( Oncorhynchus masou ), and rainbow trout ( O. mykiss ), and evaluated using generalised linear mixed models and predictive performance metrics. Local disappearance was detected at 12 sites (8.70%), double the number predicted by the previous model based on mean annual air temperature (6 sites, 4.35%). No significant correlation was detected between local disappearance and either latitude or elevation. Models incorporating increases in summer air temperature and interspecific competition effects provided the best predictions, whereas those based on annual mean or subsurface temperatures performed poorly. Our results indicate that seasonal thermal dynamics and biotic interactions, rather than annual mean, exert a critical influence on local disappearance. Limited dispersal ability and habitat specificity likely constrain range shifts that could mitigate the impacts of climate change. Incorporating these complex factors substantially improves the accuracy of disappearance‐based extinction risk models. By empirically validating a climate‐driven extinction model using local disappearances, this study highlights the importance of ecological realism in extinction risk modelling. The findings offer valuable guidance for international conservation strategies aimed at freshwater species facing global warming.