Deriving a clear-sky soil moisture index from ECOSTRESS land surface temperature
Authors
Jia A., Mallick K., Upadhyaya D., Hu T., Szantoi Z., Bhattacharya B., Sekhar M., Skoković D., Sobrino J.A., Ruiz L., Boulet G.
Reference
Remote Sensing of Environment, vol. 329, art. no. 114945, 2025
Description
Agricultural drought threatens food and water security in rapidly growing regions like India and Sub-Saharan Africa, underscoring the importance of remote sensing (RS) for monitoring. However, existing land surface temperature (LST)-based water stress indices often lack sensitivity to soil moisture (SM) deficits in vegetated areas, and high-resolution thermal infrared (TIR) water stress products remain scarce. Additionally, TIR-based indices are rarely validated with ground measurements in Sub-Saharan Africa, limiting their reliability. To address these challenges, we propose a high-resolution (70 m) soil moisture index using ECOSTRESS data, termed Radiative Thermal Inertia (RTI). RTI integrates near real-time noon and midnight ECOSTRESS LSTs with accumulated radiative fluxes, representing the energy required to raise LST by 1 K per unit area. A correction factor (β) accounts for vegetation cover and relative humidity, enhancing RTI's sensitivity to SM variabilities, especially in vegetated regions. First, we employ an innovative climatology-based LST reconstruction method to fill ECOSTRESS data gaps on missed clear-sky days using VIIRS LSTs, achieving accuracies comparable to official clear-sky retrievals (RMSE = 2.31 K at 13:30, 1.91 K at 01:30). These reconstructed LSTs are subsequently used to calculate RTI across 21 soil moisture in-situ sites in Sub-Saharan Africa and India, demonstrating a strong correlation [r = 0.62 for RTI-β] with seasonal SM variability compared to other indicators (Keetch-Byram Drought Index, KBDI; Normalized Difference Water Index, NDWI_ ρ1.24; NDWI_ ρ2.13; and Apparent Thermal Inertia, ATI). While the majority of the drought indices tend to saturate at high fractional vegetation cover (FVC), RTI-β remains stable across a range of vegetation densities. Sensitivity analysis with normalized SM anomalies shows a higher correlation with seasonality-detrended RTI-β (r = 0.70), marking a significant improvement in vegetated areas over the initial RTI and the Scaled Drought Condition Index (SDCI) in sparsely vegetated regions. Spatial and temporal analyses demonstrate the ability of this ECOSTRESS-based SM index to track drought periods and irrigation events. This study addresses a critical gap in high-resolution spatiotemporal surface water stress mapping for agriculture using thermal remote sensing theory. The findings highlight the RTI's potential for future high-resolution TIR missions, supporting agricultural management and drought early warning systems in Sub-Saharan Africa, India, and beyond.
