Integrating EOS-04 SAR SMAP Radiometer and CYGNSS GNSS-R Data with Machine Learning for Improved Soil Moisture Retrieval Accuracy

This paper presents an novel methodology for enhancing soil moisture estimation by integrating data from the Cyclone Global Navigation Satellite System (CYGNSS) with Synthetic Aperture Radar (SAR) soil moisture measurements. CYGNSS Level 3 Soil Moisture products offer pre-processed and calibrated estimates for the top 0-5 cm soil layer, leveraging SMAP (Soil Moisture Active Passive) data for calibration. Despite CYGNSS’s high temporal resolution, its spatial accuracy often falls short, necessitating enhancement through high-resolution SAR data from the ISRO’s EOS-04(RISAT-1A) C-band SAR mission. However, the temporal resolution of EOS-04 SAR data is limited, leading to the application of Gaussian Process Regression for data interpolation. Furthermore, machine learning techniques are employed, utilizing additional meteorological variables to refine the soil moisture estimates. The results underscore the potential of combining SAR and GNSS-R remote sensing data with machine learning to improved soil moisture retrieval, making it a valuable resource for applications in agriculture and water resource management.