A Refined Spaceborne GNSS-R Wind Speed Retrieval Framework Considering Swell Dynamics

Spaceborne Global Navigation Satellite System-Reflectometry (GNSS-R) has emerged as a transformative remote sensing technique for ocean geophysical parameter monitoring. However, existing wind speed retrieval algorithms often encounter accuracy bottlenecks due to an incomplete accounting of non-local swell modulations in different sea state conditions. This study identifies that, beyond Significant Wave Height (SWH), the swell period is a critical driver of GNSS-R signal variations, exerting an inherently non-linear influence on sea surface roughness across different wind speed regimes. To address this, we propose a refined retrieval framework. First, the swell period is introduced as a corrective parameter to calibrate scattering cross-section deviations induced by non-local waves, addressing the mismatch between local wind stress and surface roughness. Second, the instantaneous change rate of the Normalized Bistatic Radar Cross Section (NBRCS) is incorporated as a dynamic sensitivity indicator. This feature effectively enhances the model’s capability to distinguish between developing and fully developed seas, providing a temporal dimension to the traditionally static retrieval approach. A Multi-Layer Perceptron (MLP) architecture was optimized using CYGNSS mission data for end-to-end estimation. Validation against ERA5 reanalysis data demonstrates that the integration of swell dynamics and NBRCS temporal variations significantly enhances retrieval robustness across all wind speed ranges. While the model achieved a correlation coefficient (R) of 0.86 and an RMSE of 1.37 m/s across all ranges, the proposed model improves these metrics to R = 0.87 and RMSE = 1.32 m/s. Specifically, the model exhibits improved performance in both the high wind speed range (> 15 m/s), where the RMSE decreased from 2.79 m/s to 2.37 m/s (a 15% improvement), and the low wind speed range (< 10 m/s), with an RMSE of 1.21 m/s. These results underscore the necessity of accounting for swell-related wave age and temporal signal variations to achieve high-precision GNSS-R sensing under diverse and complex global sea conditions.