Progress in Determination of Wind Vectors from SAR Images

During the past decade, the ability to determine detailed information of ocean surface wind from Synthetic Aperture Radar (SAR) has been generally accepted. It is expected that SAR will provide all-day and all-weather wind parameters with high spatial resolution and high accuracy, which would provide valuable data to marine weather forecasters and SAR model designers. However, the methodology for retrieving wind fields from SAR is far from complete. Present procedures require that a prior wind direction must be available in order to get wind speed information from SAR images. Errors in wind direction can result in significant biases in the SAR-retrieved wind speed. The prior wind direction can be acquired from measurements of other instruments (scatterometer, buoys etc.), from meteorological model outputs, or from wind-induced slicks if evident in the SAR images. The latter method may be limited by the image; many SAR images don't have any obvious wind-induced slicks. He et al. [2005] developed a new gradient method model (GM), which can retrieve wind directions from SAR images without trying to infer slicks. However, thus far, this method can only be used in relatively uniform wind field situations. In this paper, two modifications to the GM method have been developed by introducing an optimal analysis method. Wind speed and direction (with ambiguity) can be retrieved directly. In this method, sigma⁰ from two overlapped sub-blocks of the SAR image are used to retrieve wind parameters, which would violate the original GM method, because it cannot be used in highly varied winds, such as hurricanes. The chief advantage of this method is that wind speed can be retrieved without additional reference wind directional information. The wind direction ambiguity can be removed by background geophysical information of the SAR image or by in-situ measurements. Two cases are considered to illustrate this new method. One is a simulated image with relatively uniform wind field, and the other is a Radarsat-1 image which captures the eye of hurricane Isabel before it made landfall in North Carolina in 2003 as a category 2 storm. In both cases, the retrieved wind result is shown to compare well with Quickscat measurements and in-situ data. However, error exists for highly-varied wind field, especially for the retrieved wind direction. Future investigation will focus on this limitation, and also conduct further comparisons.