Key points are not available for this paper at this time.
This study investigates spatiotemporal dynamics of Aerosol Optical Depth (AOD) in the Sichuan Basin from 2007–2022 using MODIS 3 km AOD products. Integrating Geodetector and Random Forest (RF), we identify dominant drivers of AOD heterogeneity, evaluate their interactions, and develop a predictive model. Geodetector revealed elevation, temperature, and PGDP as factors with statistically significant differences both between drivers and across internal strata. Elevation ranked first in explanatory power (q = 0.522), followed by temperature (q = 0.511), wind speed, and PGDP. All driver interactions exhibited enhancement effects. The RF model achieved strong performance (internal R² = 0.901; 2022 validation R² = 0.822), with elevation contributing >55% of importance. NDVI and wind speed were secondary predictors, while anthropogenic factors showed spatial association but limited predictive contribution. This framework combines spatial attribution and machine learning to diagnose and predict AOD heterogeneity, offering mechanistic insights and a practical tool for air quality management in topographically constrained regions.
Hu et al. (Sun,) studied this question.