Generating High-Resolution and Long-Term SPEI Dataset over Southwest China through Downscaling EEAD Product by Machine Learning

Drought is an event of shortages in the water supply, whether atmospheric, surface or ground water. Prolonged droughts have negative impacts on ecosystems, agriculture, society, and economy. Although existing drought index products are widely utilized monitoring, coarse spatial resolution greatly limits their applications regional local scales. Machine learning driven by remote sensing observations offers opportunity to monitor scale droughts. However, limited time range such as vegetation (VI) resulted a substantial gap generating high before 2000. This study generated spatiotemporally continuous Standardized Precipitation Evapotranspiration Index (SPEI) data spanning from 1901–2018 southwestern China machine learning. It indicated that four Classification Regression Tree (CART) approaches, decision trees (DT), random forest (RF), gradient boosted regression (GBRT) extra (ET), can provide valid information downscaling Estación Experimental de Aula Dei (EEAD) data. The in-situ SPEI dataset produced Penman–Monteith method was used benchmark evaluate temporal performance downscaled SPEI. In addition, necessity VI also assessed. results showed that: (1) ET-based product has best (R2 = 0.889, MAE 0.232, RMSE 0.432); (2) provides no significant improvement for re-construction; (3) topography exerts obvious influence process, (4) shows more consistency with compared EEAD proposed be easily extended other areas without enhance ability long-term monitoring.