Machine Learning Driven Enhancement of Soil Moisture using CYGNSS GNSS-R Data and Meteorological Observations over Indian Agriculture Site

This work presents a novel approach to enhance soil moisture estimation by leveraging CYGNSS (Cyclone Global Navigation Satellite System) derived soil moisture and meteorological observations using machine learning models. The study focuses on improving CYGNSS Level 3 Soil Moisture V1.0 data by integrating with key environmental variables like temperature, precipitation, and specific humidity from the MERRA-2 reanalysis dataset for monitoring soil moisture in an agricultural cropland field in Dehradun, India. The research adopted a Support Vector Regression (SVR) model, to significantly enhance the soil moisture. This SVR model demonstrates strong estimated accuracy (RMSE) of 0.05, with correlation of an R-value of 0.84 which indicate a reliable correlation between estimated and in-situ observed soil moisture. The findings suggest that the combination of CYGNSS derived soil moisture and meteorological parameters using SVR can significantly enhance soil moisture estimation, offering valuable insights for agricultural water resource management and water resource planning.