Capturing microclimates: 30-m downscaled climate dataset reveals fine-scale suitability patterns of invasive pests in karst landscapes

Abstract

Karst landscapes are characterized by extreme topographic heterogeneity, which limits the capacity of conventional coarse-resolution climate data to resolve local microclimates, thereby introducing uncertainty into biological invasion risk assessments. This study developed a high-resolution climate downscaling framework integrating dynamic temperature lapse rates and error correction to reconstruct a 30-m climate dataset for the Southwest China Karst. Using both the downscaled 30-m dataset and the original 1-km dataset, optimized MaxEnt models were applied to predict the potential distributions of two invasive pests, Bactrocera dorsalis and Tuta absoluta . The downscaling framework outperformed conventional interpolation, improving accuracy for maximum temperature, minimum temperature, and precipitation by 12.5% (R 2 = 0.984), 15.1% (R 2 = 0.989), and 4.4% (R 2 = 0.918), respectively. The optimized MaxEnt models showed stable discriminatory performance for both species, with final training AUC values of 0.803–0.835 and ENMeval cross-validated validation AUC values of 0.778–0.809. Relative to the 30-m model, the 1-km model showed broader suitability patterns on steep valley walls for B. dorsalis and expansive basin floors for T. absoluta due to spatial averaging. At the 30-m scale, B. dorsalis exhibited dendritic distributions along dry–hot karst river valleys, driven primarily by Annual Mean Temperature (37.2% contribution) and NDVI (23.9%). Conversely, T. absoluta displayed patchy distributions concentrated in flat plateau basins and depressions, associated with Annual Mean Temperature (27.9%) and Tree Canopy Density (23.5%). Differentiated management strategies are proposed, including linear interception at valley gateways for B. dorsalis and grid-based monitoring in flat agricultural zones for T. absoluta . These findings indicate that high-resolution climate data can refine the spatial interpretation of invasion risk in complex terrains and provide a more detailed spatial basis for precision pest management in karst ecosystems.