Sensitivity of multi-frequency and multi-polarization SAR to soil moisture at different depths in agricultural regions

• Quantitatively assess penetration and retrieval depths for C- and L-band SAR • Compare penetration depths between compact and full polarizations of SAR. • Compare simulated penetration depth with experiment-derived retrieval depth. • Analyze the effects of incidence angle and vegetation on soil moisture retrieval. • Test six types of polarimetric features for each polarization mode. Soil moisture is critical for various agricultural applications such as crop monitoring and irrigation management. Synthetic Aperture Radar (SAR), especially at L-band and C-band frequencies, has been widely utilized for soil moisture estimation across different polarizations. However, the ambiguity in the depth of soil moisture retrieval remains a challenge due to the complex interactions between SAR signals and ground targets. In this study, the sensitivity of multi-frequency and multi-polarization SAR to soil moisture retrieval at various depths was quantitatively analysed. First, using the Dobson semi-empirical dielectric mixing model, the penetration depth of C-band SAR was found to be limited to the top surface soil (0–5 cm), while L-band SAR could penetrate from 3 cm to more than 20 cm depending on soil properties. Next, the accuracy of soil moisture retrieved at five depths (0–5 cm, 5 cm, 20 cm, 50 cm, and 100 cm) was evaluated using in-situ soil moisture data. Using Random Forest (RF) regression and polarimetric features, the volume scattering of C-band SAR was demonstrated as the dominant scattering mechanism, leading to reduced accuracy across all depths. In contrast, L-band SAR achieved the highest accuracy at the 5 cm depth, constrained by the limited intervals of moisture measurements. Furthermore, the analysis of incidence angle and vegetation coverage revealed that lower incidence angles (15–45 degrees) and lower vegetation conditions improved the accuracy for C-band SAR. L-band SAR, however, exhibited less sensitivity to vegetation coverage when retrieving soil moisture from shallower depths. These findings provide valuable insights for selecting SAR data suitable for soil moisture retrieval in agricultural regions.