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 theArtificial Intelligence in geophysics, i.e. the computation of surface wind speed over the seafrom satellite Synthetic Aperture Radar (SAR) images. The area of interest is on the SvalbardArchipelago in the Arctic Sea. In this area, the deep learning methodology based on a residualneural network (ResNet), developed to retrieve wind directions from SAR at grid size ≤ 1km without external information, has been associated with a robust methodology of textureanalysis to detect the presence of sea ice over the sea, as the wind can be estimated only overthe water. Avoiding to illustrate technical details concerning the ResNet and texture analysismethodologies, this contribution is focused on showing, as example, a SAR derived wind field inthe area of interest obtained using ResNet and texture analysis.

کلیدواژه های Using Convolutional Neural Network in geophysics: wind field evaluation from satellite SAR images:

Synthetic Aperture Radar ، Convolutional Neural Network ، texture analysis ، geophysics ، sea wind field