Remote sensing-based monitoring of agricultural drought in mountainous regions by crop type and yield estimation

Agricultural drought remote sensing monitoring is hindered by coarse precipitation data, unconsidered spatial heterogeneity and simplified drought indicators. This study applied dual-constraint clustering for multi-level agricultural-ecological grid division, integrated high-spatiotemporal resolution precipitation and 30 m-resolution soil moisture indices, and constructed parcel-level drought indices with crop water demand. Based on disaster system theory and a water stress yield estimation model, this study realized integrated drought risk identification, quantitative assessment, and irrigation scheme decision analysis. The validation results indicate that the crop growth stage-specific soil moisture indices effectively support parcel-level drought identification (overall accuracy = 0.911, Kappa = 0.704), the water stress-based assessment model achieves high accuracy (normalized root mean square error = 1.97%), and the irrigation scheme analysis provides practical strategies with corresponding effectiveness evaluations. This study offers a reliable technical approach for precise agricultural drought monitoring and targeted drought mitigation, which is of great significance for improving agricultural drought resilience.