A XGBoost-based geophysical model function relating ocean wind and wave states for C‑band VV‑polarized SAR backscattering

Nowadays, C-band VV (vertical–vertical)-polarized Synthetic Aperture Radar (SAR) ocean wind retrievals commonly rely on CMOD (C-band MODel) series geophysical model functions (GMFs). However, most existing GMFs were built based on low-to-medium resolution scatterometer data and ignore the sea‑state influence on normalized radar cross-section (NRCS). In this letter, two novel GMFs (XGB_CSARMOD_w and XGB_CSARMOD_ww) based on the machine learning technique of eXtreme Gradient Boosting are proposed to model C-band SAR backscattering in VV polarization. Here, XGB_CSARMOD_w uses wind vector and incidence angle to predict NRCS, while XGB_CSARMOD_ww additionally includes significant wave height (SWH). These two GMFs are directly constructed based on collocation between 6392 Sentinel-1A/B SAR interferometric wide images and 70 buoys from 2014 to 2023. The validation results demonstrate that both new GMFs perform better in simulating NRCS compared to traditional GMF of C-SARMOD2. After incorporating SWH, XGB_CSARMOD_ww yields the best performance, achieving a root mean square error of 1.21 dB and a correlation coefficient of 95.17% in NRCS prediction. Our new GMF that simultaneously considers wind and wave conditions provides a new pathway for synergetic retrieval accurate ocean wind and SWH from SAR images under complex sea states.