Estimation of Surface Winds From SAR Using a Projection Algorithm

Synthetic Aperture Radar (SAR) has been shown to be a useful tool for estimating coastal wind vectors over the ocean. Wind direction is estimated from linear signatures seen in the imagery that are aligned with the wind, and wind speed is estimated by inverting models that predict radar cross section values given wind speed and wind direction. Current work is ongoing in a number of locations to make these approaches operational, mainly by automating the process of estimating wind direction and be developing more robust forward models for radar cross section.

A new approach for automated wind direction estimation will be presented that is based on projections of the SAR image values over various directions. The algorithm has the ability to only estimate directions for image locations that contain sufficient feature information, and to interpolate over smooth image regions without any features aligned with the wind. In addition, a forward radar cross section prediction model that is semi-empirical and based on the two-scale model will be presented and compared to other existing models for predicting C-band and X-band, HH and VV radar cross section values.

The combination of these two algorithms has been automated and validated on 137 comparisons to in situ buoy observations of the local wind. Wind direction was estimated with a RMSE of 39 degrees and wind speed was estimated with a RMSE of 2.2 m/s. If features that are fooling the wind direction estimator (because they are not aligned with the local wind but appear very similar to features that are aligned) are manually removed, the direction RMSE becomes 31 degrees and the wind speed RMSE becomes 2.1 m/s.