Nested species distribution models highlight scale and climate-dataset effects on adaptation planning for sweet chestnut (Castanea sativa) forests in Türkiye

Background Projections of forest habitat suitability under climate change are highly sensitive to modeling scale and climate dataset selection, yet these sources of uncertainty are rarely quantified explicitly. Sweet chestnut ( Castanea sativa Mill.) forests in Türkiye occupy climatically marginal zones where topographic heterogeneity strongly mediates climate impacts, providing an ideal system for evaluating how scale-dependent processes alter conservation priorities and refuge identification.. Methods We developed a hierarchical SDM framework integrating national-scale ensemble models (~ 1 km resolution) with regional-scale models (~ 120 m resolution) incorporating downscaled climate, topography, and hydrological variables. Using WorldClim 2.1 and CHELSA 2.1 in parallel across three CMIP6 global circulation models under SSP1-2.6 and SSP5-8.5 scenarios, we explicitly quantified both scale- and dataset-driven uncertainty in current and future projections. Results Despite high predictive performance (AUC > 0.95; TSS > 0.75), national-scale models exhibited pronounced dataset-driven spatial divergence, with future projections revealing contrasting conservation narratives: WorldClim predicted severe habitat contraction (50–80% loss), while CHELSA indicated substantial persistence in high-elevation micro-refugia (10–20% loss under SSP1-2.6). Regional-scale integration of topographic and hydrological constraints reduced predicted suitable habitat by 60% relative to national projections, primarily by excluding climatically permissive but ecologically unsuitable lowland areas. Conclusions Hierarchical SDMs integrating macroclimatic projections with local environmental constraints substantially reduce prediction uncertainty and enhance ecological realism in topographically complex landscapes. Our results demonstrate that coarse-scale climatic models systematically misidentify conservation priorities by overlooking topographically-buffered micro-refugia, with direct implications for adaptive forest management and climate adaptation planning in mountain ecosystems globally.