The Empirical Orthogonal Function Theory and Simulation Research for Spaceborne GNSS-R Sea Surface High Wind Speed Retrieval

At present, there are three GNSS-R sea surface wind speed retrieval modeling algorithms, which are least-squares fitting, cumulative distribution function (CDF) and the neural network algorithm, but high wind speed retrieval modeling is still remains a challenge.In this paper, the empirical orthogonal function (EOF) analysis algorithm is proposed for high wind speed retrieval modeling, which obtains the principal components by eigenvalue decomposition of the GNSS-R observation data, and then adopts a linear combination of a principal component and an empirical orthogonal function to establish wind speed retrieval model. The modeling process of this algorithm is simple and fast compared with the former method.In order to verify the feasibility of the algorithm for high wind speed retrieval, we simulated the sea surface delay Doppler maps (DDM) at the interval of 1 m/s and 1° in the range of 20–50 m/s wind speed and 0–70° incident angle due to the deficiency of global high wind speed observation, and meanwhile calculated the normalized bistatic radar cross sections (NBRC). After that, an EOF analysis algorithm is applied to establish the GNSS-R wind speed retrieval model based on NBRCS and incident angle. At the same time, the relationships between the NBRCS and wind speed, NBRCS and incident angle are explained from the perspective of the spectrum. Verified by test, the model demonstrates good performance, with a mean bias of −0.05 m/s and a root mean square error (RMSE) of 0.26 m/s. In the future, more measured GNSS-R observation data and meteorological data in storm will be collected in order to verify performance of EOF modeling algorithm.