Deriving sea surface wind from synthetic aperture radar based on Fourier transform and neural network

The acquisition of sea surface wind information is of great value for marine activities, such as navigation, fishing and sea operations. In this paper, the ERS-2 SAR data, covering the Yangtze River estuary and its adjacent waters at around 10:30 local time on March 15, 2007, was applied to retrieve wind fields. We improved the sea surface wind direction inversion algorithm based on two dimensions Fast Fourier Transform (2DFFT) in the spectral domain. These sea surface wind directions were then applied as input parameters to derive wind speeds employing Back Propagation Neural Network (BPNN) method. To present and testify the applicability of these algorithms, the wind fields derived from SAR were validated with QuikSCAT SeaWinds-1 measurements. The results show that the SAR inversion results are reliable, and within the range of error tolerance. Specifically, the sea surface wind direction has a root mean square (RMS) error of 9.21°, maximum error of 24.2°. The inversion accuracy is higher in the open sea area. Whereas, it is not really good in the area affected by land, ship, and other artifacts. The wind velocity has a RMS error of 1.74 m/s, maximum error of 3.8 m/s. The effectiveness and reliability of the improved wind direction inversion algorithm and BPNN method for wind speed are demonstrated.