Neural Network-Based Retrieval Optimization of GNSS-IR for Sea Level Monitoring Under Different Elevation Angles

The Global Navigation Satellite System Interferometric Reflectometry (GNSS-IR) technique has been widely employed to retrieve sea levels by utilizing signal-to-noise ratio (SNR) data, yet its accuracy is often hindered by limitations imposed by satellite elevation angles. In this study, we present a novel methodology that specifically addresses the optimization of sea level retrieval across different elevation angle ranges by incorporating a neural network model. This approach uses SNR residual sequences as key feature inputs to adjust and enhance the retrieval process, thereby mitigating the adverse effects of varying elevation angles on measurement accuracy. Experimental evaluations at the SC02 and PBAY stations demonstrate significant improvements over conventional methods. At the SC02 station, with elevation angle ranges of $5^{\circ }--10^{\circ }$, $5^{\circ }--15^{\circ }$, and $5^{\circ }--20^{\circ }$, the root mean square error (RMSE) is reduced by 3.51%, 18.62%, and 32.03%, respectively, while at the PBAY station, using ranges of $5^{\circ }--20^{\circ }$, $5^{\circ }--25^{\circ }$, and $5^{\circ }--30^{\circ }$, the RMSE improvements are 9.23%, 16.13%, and 21.11%. These results highlight the potential of the proposed neural network-based optimization to not only overcome the intrinsic limitations of GNSS-IR but also significantly enhance retrieval performance, particularly at higher elevation angles. The findings of this study pave the way for more accurate sea level monitoring and improved nearshore sea surface wave height retrieval, contributing to more robust coastal environmental assessments.