Drought Monitoring and Performance Evaluation Based on Machine Learning Fusion of Multi-Source Remote Sensing Drought Factors

Drought is an extremely dangerous natural hazard that causes water crises, crop yield reduction, and ecosystem fires. Researchers have developed many drought indices based on ground-based climate data various remote sensing data. Ground-based are more accurate but limited in coverage; while the cover larger areas poor accuracy. Applying data-driven models to fuse multi-source for reproducing composite index may help fill this gap better monitor terms of spatial resolution. Machine learning methods can effectively analyze hierarchical non-linear relationships between independent dependent variables, resulting performance compared with traditional linear regression models. In study, seven impact factors from Moderate Resolution Imaging Spectroradiometer (MODIS) satellite sensor, Global Precipitation Measurement Mission (GPM), Land Data Assimilation System (GLDAS) were used reproduce standard precipitation evapotranspiration (SPEI) Shandong province, China, 2002 2020. Three machine methods, namely bias-corrected random forest (BRF), extreme gradient boosting (XGBoost), support vector machines (SVM) applied as Then, best model was construct distribution SPEI. The results show BRF outperforms XGBoost SVM SPEI estimation. conditions without ground observation provides comprehensive information by producing a SPEI, which reliability be monitoring.