Neural network wind retrieval from ERS-1/2 scatterometer data

A neural network methodology is presented to retrieve wind vectors from ERS - 1/2 scatterometer data. The wind directional ambiguities are eliminated by a circular median filter algorithm. All data come from ERS - 1/2 scatterometer data collocated pairs with CMCD4 vector. Comparing the results with CMCD4 and ECMWF wind vector,they agree well, which indicates that it is possible to extract wind vector from the ERS-1/2 scatterometer with the neural network method.