Wind Direction Extraction from Coastal SAR Images Using Cross-spectral Method

A sensitive processing method for extracting the wind direction from SAR (synthetic aperture radar) images is presented. We transfer SAR images to the image spectra using two-dimensional FFT (fast Fourier transform) and take the integral of the image spectra between the wave length 400 m and 800 m, which is high end containing wind direction signatures in Lband SAR. From this and a reference signal (half or full wave rectified sinusoidal wave) we obtain the wind direction by using cross-spectral analysis. To examine the potential of our method, a set of simulated images with various SNR (signal to noise ratio) are used. It is shown that, even if the SNR is −30 dB, the agreement is to within ±3◦. Therefore the processing technique we present allows us the precise estimation of wind direction. We also use 23 JERS-1/SAR scenes from the Tsugaru Strait located between Aomori and Hokkaido prefectures of Japan for the extraction of wind direction. The wind directions retrieved from these images have been compared with in situ wind data from the Kikonai, the Tappi and the Ooma lighthouse. Although the lighthouse estimates are lacking in directional resolution (16 direction, 22.5◦ interval), the SAR estimates obtained from spectra of 6.4-km-square images are shown to agree with the lighthouse estimates to within a mean absolute error of 21◦. The estimates difference may be related to the directional resolution of in situ wind data, the SNR of JERS-1/SAR images, the effect of terrain around the lighthouse, and the ocean current.