RETRIEVAL OF SEA SURFACE WIND SPEED BY SPACEBORNE SAR BASED ON MACHINE LEARNING

In this paper, we presented a preliminary study of retrieving sea surface wind speed (SSWS) from the Senitnel-1 SAR data acquired in Horizontal-Horizontal (HH) polarization based on a machine learning algorithm, to acquire wind field over Arctic ocean with high spatial resolution and accuracy. The interesting is that we found the simple backward propagation (BP) neutral network performs well under guidance of the widely applied Geophysical Model Functions (GMF) CMODs. The network is trained using collocations of S1 data and ASCAT measurements. Compared to the buoy measurements, the bias, root mean square error (RMSE) and scatter index (SI) of wind speed retrieved by the BP neural network model are 0.1 m/s, 1.38 m/s and 19.85%, respectively.