Novel wave height extraction method using high-frequency radar under low sea states

High-frequency surface wave radar (HFSWR) significant wave height (SWH) extraction is one of the critical steps in developing joint prediction of ocean surface dynamics elements. This study aims to address the limitations of the traditional SWH extraction methods by developing a novel hybrid algorithm. This paper proposes a novel HFSWR SWH extraction algorithm based on the enhanced wavelet neural network (WNN), which combines the radar feature parameters and machine learning method to integrate various types of information to improve wave height extraction accuracy. Barrick's algorithm and different machine learning approaches are compared with the proposed method using the measured data from the Tianjin coast. This study explores the superiority of extracting significant wave height using combined first- and second-order spectral information from HFSWR. The root-mean-square error (RMSE) of the proposed method for 30 consecutive experiments is 0.1070m on average, which has 73% more gain than Barrick's algorithm. The vast majority of RMSE (83%) is lower than 0.1080m for 30 consecutive runs of the proposed method. Experimental results show that the proposed method performs better than the conventional methods under low sea states.