Crop Classification Using Dual-Pol Scattering Powers From Simulated NISAR L-Band SAR Data

Dual-polarimetric (dual-pol) synthetic aperture radar (SAR) data are becoming increasingly available through current and upcoming satellite missions. However, the limited polarimetric information in such data restricts their ability to fully characterize target scattering mechanisms. Recent advancements have introduced methodologies for approximating scattering characteristics from dual-pol observations, opening new opportunities for diverse applications, particularly in agricultural monitoring. This study employs a dual-pol decomposition technique for the HH–HV mode using uninhabited aerial vehicle synthetic aperture radar (UAVSAR) simulated NASA-ISRO SAR (NISAR) L-band data. The method derives three physically interpretable scattering power components: dihedral-like (P_{d-l}), surface-like (P_{s-l}), and unpolarized (P_{u}). We first analyze the behavior of these components across three agricultural sites to assess their sensitivity to variations in canopy structure. Next, we evaluate the interclass separability provided by these scattering powers and compare it with conventional dual-pol backscatter coefficients. Finally, we perform a single-date random forest-based crop classification using the derived scattering powers and benchmark the results against those obtained using backscatter coefficients.