Time Series Soil Moisture Retrieval from Synthetic Aperture Radar (SAR) Images

Effective farm management for optimal crop yield requires continuous monitoring of key indicators of crop and soil health, with soil moisture being one of the most critical parameters. While in-situ soil moisture sensors provide accurate point measurements, their spatial coverage is often insufficient to capture the heterogeneity across agricultural fields. Synthetic Aperture Radar (SAR) offers a promising alternative due to its sensitivity to surface dielectric properties, enabling indirect estimation of soil moisture through analysis of radar backscatter. In this study, we developed a time series model for soil moisture retrieval using high-resolution S-band UAV-based SAR imagery. Data collection occurred biweekly from March to October during the 2024 and 2025 growing seasons. Ground truth soil moisture data were acquired through buried sensors, handheld probes, and soil sampling. The SAR backscatter was calibrated, converted to decibel values, and analyzed over time. An empirical model linking SAR backscatter to soil moisture was developed using linear regression techniques. This model was validated against in-situ measurements across varying soil textures, roughness conditions, and vegetation cover types. It was successful in validating the linear relationship between SAR backscatter and soil moisture, demonstrating the potential of the model for scalable high-resolution soil moisture monitoring in precision agriculture. Future work will investigate its generalizability to broader land covers and different viewing geometries.