High wind speed retrieval from SAR images using random forest algorithm

This study provides a random forest (RF) based method to retrieve high wind speeds, which takes advantage of RF's powerful nonlinear fitting ability and prevents overfitting. The high wind speed data from the Stepping Frequency Microwave Radiometer (SFMR) are used as the reference data. Compared with two traditional GMF models, the C-band cross-polarized ocean model (C-2PO) and the C-band cross-polarized coupled parameter ocean model (C-3PO), the wind speed inversion results using the RF method have been greatly improved.