Assessment of the Potential Use of Sentinel-1 C-Band SAR Sigma-Naught and Gamma-Naught Features to Support Rice Monitoring Activities

The availability of the cloud-free and publicly accessible Sentinel-1 C-band SAR data allows the development of large-scale and continuous remote sensing (RS)-based rice monitoring activities. This study examines backscatter values derived using ground-range radar cross-section (sigma-naught) and slant-range perpendicular radar cross-section (gamma-naught) on both polarization channels as features for classifying the rice transplanting period on rice fields in Subang Regency, Indonesia, using Classification and Regression Trees, Support Vector Machine, Random Forest, and Gradient Boosting classifiers. Overall, the gamma-naught features produced higher backscatter values over a rice-growing cycle, i.e., ranging from 5.8 to 8.8 percent and from 9.2 to 15.8 percent higher for VH and VV channels, respectively, than that generated by sigma-naught. Furthermore, overall accuracy (OA) and kappa coefficient (K) of gamma-naught features were superior to those derived by sigma-naught in all observed classifiers. Subsequently, the accuracy increment is higher in K than in OA, ranging from 3.2 to 18.6 percent for OA and from 5.1 to 41.1 percent for K. To conclude, the gamma-naught features have much higher potential than sigma-naught in classifying the rice transplanting period, further aiding SAR-supported RS-based rice monitoring activities.