• Proposes an adaptive CDF-correction hybrid model for precise SWH estimation. • Develops a physics-guided ensemble algorithm for robust peak wave period retrieval. • Establishes a complete BDS-R to wave energy mapping framework. This study attempts, for the first time, to use BeiDou Satellite Reflected (BDS-R) Navigation System signals for remote sensing of ocean wave energy, preliminarily verifying the feasibility and potential of this technology in capturing the spatiotemporal changes of wave energy. As a novel remote sensing technology, Global Navigation Satellite System Reflectometry (GNSS-R) offers low cost, short revisit periods, and all-weather observation. Furthermore, artificial intelligence (AI) provides new methods for wave prediction by excelling at complex nonlinear problems. This study uses the BDS-R signals received by Fengyun-3E (FY-3E) satellite from June 1, 2023, to June 1, 2025 (732 days) to estimate important ocean wave parameters, including significant wave height (SWH), peak wave period, and wave energy flux. A comprehensive wave energy estimation network is proposed to improve estimation accuracy. SWH is estimated using a deep learning hybrid model (i.e., DilatedResBiLSTM-AttnNet) with bias correction using the adaptive cumulative distribution function (CDF) matching method, while the peak wave period is derived through a physics-guided Bagging Tree (BT) ensemble learning algorithm. The corrected SWH and peak wave period are used as parameters for calculating the wave energy formula, enabling accurate wave energy estimation. The experimental results on the FY-3E BDS-R and WaveWatch III (WW3) test sets show that the DilatedRsBiLSTM-AttnNet model has an retrieval root mean square errors (RMSE) of 0.63 m and a correlation coefficients (CC) of 0.92 in the SWH range of 0-12 m. Compared to existing models, when using WW3 and Jason-3 data as references, their RMSE decreased by 38.17% -48.15% and 13.14% -36.56%, respectively, and their advantages were further validated on National Data Buoy Center (NDBC) buoy data. Moreover, the physical-guided BT method estimates the RMSE of peak wave period to be 1.35 s and CC to be 0.88, which is superior to empirical models; Meanwhile, BDS-R-estimated wave energy shows good agreement with multiple reference datasets (ERA5, WW3, and Jason-3 ku-band), as evidenced by CC of 0.77–0.82 and RMSE of 13.63–20.75 kW/m, demonstrating the potential application of spaceborne GNSS-R technology with BDS-R as an important data source in wave energy estimation.
Bu et al. (Mon,) studied this question.