Wind Retrievalfor Cfoscat Edge and Nadir Observations Based on Neural Networks And Improved Principle Component Analysis

CFOSCAT is the first rotating fan-beam scatterometer in space. It is on board the China France Oceanography Satellite (CFOSAT) launched on 29 th October, 2018. It can provide wind field products in the unit of Wind Vector Cells. One of the advantages for this novel antenna geometry is being able to obtain more observations of the scene, though for edge and nadir WVCs, sampling issues as well as larger noise level hinder wind retrievals. This is first analyzed in this paper and a new procedure is proposed, tailored for those observations of CFOSCAT. In the proposed method, wind speeds are retrieved from Neural Networks trained for specific edge and nadir WVCs, and based on the obtained speed, a modified principle component analysis (PCA) is applied for CFOSCAT wind direction retrieval. The new process chain is verified by simulated data, and further research is soon planned modifying it for real data after they are released.