High-Resolution Soil Moisture Estimation: A Case Study in Coastal India

Abstract

The high-resolution soil moisture (SM) is significant for agricultural production, hydrological modelling, weather prediction, and climate studies at the regional scale. Current satellite and derived SM products span a wide range of coarse spatial resolutions. This work predicts high-resolution and precise SM estimates for the USA and the western coastal regions of India by utilizing remote sensing data. The study implements and compares five models for SM prediction in the study regions, including machine learning and deep learning models. The extra trees regressor model outperformed all other models in the USA and India datasets. It achieved the best performance metrics, with an MAE of 1.080 mm, RMSE of 1.715 mm, and R2 of 0.958 for the USA, and an MAE of 0.847 mm, RMSE of 1.554 mm, and R2 of 0.955 for India. The study validates the predicted data using 20 stations from the International Soil Moisture Network in the USA. It aims to support decision-making for agricultural practices, hydrology, and climate adaptation by providing valuable insights into the area’s spatio-temporal dynamics of SM. © Indian Society of Remote Sensing 2025.