Crop Dynamic Root-Depth-Matched High-Resolution Multilayer Soil Moisture Dataset Generated by In Situ-Remote Sensing Data Fusion

Frequent agricultural droughts pose a serious threat to crop production. Soil moisture (SM), as a key monitoring indicator, is essential for developing effective agricultural drought indices. Most existing SM-based agricultural drought indices overlook variations in crops’ water requirements across developmental stages and root-zone depths, instead focusing on a single SM layer (surface or root zone). A primary challenge for field-scale agricultural drought monitoring is the lack of high-spatial-resolution, multilayer SM datasets. To address this gap, we propose a data-fusion framework that integrates remote sensing products and pointwise in situ observations to generate daily, 30-m resolution, multilayer SM across the root zone. The newly developed multilayer SM dataset demonstrated excellent performance, with an average Pearson correlation coefficient (PCC) of 0.989 and root-mean-square error (RMSE) of 0.034 $\mathrm{m}^{3}{\,}$ . We also tested spatiotemporal generalization and confirmed that the fusion approach robustly captures field-scale SM when downscaling from 9 km to 30 m. Using this multilayer dataset, we developed the dynamic root-depth-matched agricultural drought index (DRDI), which dynamically adapts to crop root-depth distributions across phenological stages. Spatiotemporal analysis using DRDI revealed drought conditions across areas and seasons. Compared with existing indices, DRDI better reflected drought impacts on crop yield and provided more accurate field-scale agricultural drought assessments. Through data sensitivity analysis, DRDI demonstrated resilience to input perturbations. This work has important implications for Geo-AI-based generation of high-quality SM data. More importantly, the proposed drought index offers new perspectives for the geoscience community and agricultural drought applications.