Integrated Drought Index Combining GNSS/GRACE/ RootZone Soil Moisture Data and Its Application

Objectives: Conventional single drought indices fail to capture compound drought characteristics spanning meteorology, hydrology and agriculture, while classic integrated indices such as the Palmer Drought Severity Index (PDSI) rely on sparse in-situ observations with limited spatial resolution. Existing satellite-based monitoring rarely jointly integrates GNSS-derived precipitable water vapor (PWV), GRACE terrestrial water storage anomalies (TWSA, with root-zone soil moisture deducted to avoid redundancy) and root-zone soil moisture (RZSM), leaving a gap in targeted tools for complex island monsoon regions like Kyushu Island, Japan. Methods: A novel Fusion Drought Index (FDI) is developed via the Clayton Copula function, which optimally fits the joint distribution of three standardized single indices: the GNSS-based Standardized Precipitation Conversion Index (GNSS-SPCI), the GRACE-based Drought Severity Index (GRACE-DSI, derived from XGBoost-downscaled 0.1° TWSA), and the Standardized Soil Moisture Index (SSMI). High-resolution PDSI from the TerraClimate dataset serves as the validation benchmark, and the Mann-Kendall trend test plus run theory are applied to quantify drought spatiotemporal characteristics across Kyushu from January 2014 to September 2024. Results: FDI shows a strong positive correlation (0.59–0.69) with PDSI, and significant correlations (0.536–0.617) with all three single indices. Shapley Additive Explanations (SHAP) analysis reveals their respective contributions to FDI are 27.19%, 26.07% and 46.74%. Variance inflation factors of all input indices are below 2.35, confirming no severe multicollinearity. The downscaled GRACE-DSI achieves high stability verified by ten-fold cross-validation. Temporally, FDI successfully identifies two compound drought events in late 2015 and early 2017 that were missed by single indices due to signal mismatches across spheres. Spatially, FDI integrates localized extreme signals from single indices and clearly depicts a north-south drought pattern: the south features high-frequency (16–18 events), short-duration (<2 months) droughts driven by frequent typhoon impacts, while the north presents low-frequency (9–13 events), longduration (up to 3 months) droughts regulated by large-scale climatic signals. No statistically significant widespread drought intensification trend is detected across Kyushu during the study period. Conclusions: The proposed FDI effectively overcomes the limitations of single indices in characterizing multidimensional compound droughts, providing a reliable tool for drought monitoring and risk assessment in island monsoon regions and other complex climatic zones. Future work will incorporate GNSS vertical deformation data to impose physical constraints on GRACE TWSA, further improving monitoring accuracy.