Recalibration and Wind Retrieval of GF-3 SAR Wave Mode Measurements

Abstract

In this study, the Gaofen-3 synthetic aperture radar (GF-3SAR) wave mode data were recalibrated, and were further used to retrieve the ocean surface wind. The collocated data used include the European Center for Medium-Range Weather Forecast (ECMWF) and the Advanced Scatterometer (ASCAT) wind data. First, the GF-3 data were recalibrated using a numerical weather prediction (NWP) method. Here the GF-3 measured normalized radar cross section (NRCS) were compared to the simulated NRCS (by collocated ECMWF winds and CMOD5.N model) to determine the recalibration coefficients at different polarizations and incidence angles. By the recalibration, the bias between the GF-3 NRCS and the simulated NRCS was significantly reduced (&0.1dB). Then, the wind speed and wind direction were retrieved from the recalibrated GF-3 data by the CMOD5.N model using the optimal scheme (co-polarized data, VV and HH data) and linear model (cross-polarized data, VH and HV data). Note that the cross-polarized data were only used to retrieved wind speed due to the weak sensitivity on wind direction. Finally, the wind retrievals were validated by the collocated ASCAT winds. The bias and root mean square error (RMSE) of the wind retrievals were obtained at different polarizations and incidence angles. From the validations, the retrieved wind speed and direction have better accuracies than the ones before recalibration. Take the retrieval accuracy at incidence angles from 40° to 40.7° as an example, the RMSEs are 1.89 m/s, 1.92m/s, 2.07m/s and 2.12 m/s for VV, HH, VH, HV data respectively. Moreover, for wind speed retrievals at all incidence angles, the RMSEs are around 2 m/s, while the RMSEs from cross-polarized data are slightly larger than the ones from co-polarized data. For retrieved wind direction, the RMSEs from co-polarized data are lower than 20° Results indicate that the NWP-based recalibration method can significantly improve the accuracy of GF-3 wind retrievals.