using convolutional neural network in geophysics: wind field evaluation from satellite sar images

This contribution is aimed to show one of the several rising applications of the artificial intelligence in geophysics, i.e. the computation of surface wind speed over the sea from satellite synthetic aperture radar (sar) images. the area of interest is on the svalbard archipelago in the arctic sea. in this area, the deep learning methodology based on a residual neural network (resnet), developed to retrieve wind directions from sar at grid size $le$ 1 km without external information, has been associated with a robust methodology of texture analysis to detect the presence of sea ice over the sea, as the wind can be estimated only over the water. avoiding to illustrate technical details concerning the resnet and texture analysis methodologies, this contribution is focused on showing, as example, a sar derived wind field in the area of interest obtained using resnet and texture analysis.