A new vegetation–thermal integrated approach (NNL-SM) for high-resolution soil moisture estimation

High-resolution soil moisture (SM) information is essential for hydrological, agricultural and climate studies. This paper presents an NDMI–NDVI–LST-based downscaling technique, termed the NNL-SM model, to downscale SMAP SM from 9 to 1 km resolution. The model was validated in Varanasi, India, using ground SM measurements collected on fifteen dates and compared with the VMSMI and triangle optical-based models. By integrating the normalized NDMI, NDVI and LST while minimizing vegetation effects, the NNL-SM model achieved the best performance, with the lowest RMSE (0.028 m3/m3), ubRMSE (0.023 m3/m3), smallest bias (0.001 m3/m3) and highest correlation (R = 0.700). VMSMI showed moderate accuracy, whereas the Triangle model had weaker correlation. Seasonal analysis indicated better winter performance and reduced pre-monsoon accuracy due to vegetation and temperature variability. Overall, the NNL-SM approach showed promising performance for the selected heterogeneous agricultural region and can be further evaluated using hybrid indices and multi-region validation.