Research On Key Technologies Of Sea Surface Wind Speed Retrieval From Spaceborne C-Band Quad-Polarized SAR Measurements

With the rapid development of high spatial resolution remote sensing technology,SAR has become a new technical mean to observe ocean surface wind fields because of its high resolution and capacity to operate in almost all-weather conditions,day or night.It can provide high spatial resolution wind field data for global marine wind resources utilization and development.However,there still exists the following problems in C-band spaceborne SAR wind speed retrieval research: global applicability of available wind speed retrieval models;the effect of resampling spatial resolution on the retrieval accuracy of wind speed;the difference for SAR wind retrieval from different polarization channel data;limitation in application for SAR-retrieve wind speed from quad-polarization system.Based on a large body of C-band RADARSAT-2 quad-polarization fine SAR images together with the buoy measurements wind field and reanalysis wind data,some key problems have been proposed and solved in this paper.The contents are organized as follows:(1)The applicability analysis about wind speed retrieve models of quad-polarization SAR images in difference sea area.This paper compared the performances of commonly used SAR-retrieved wind speed models in difference area of the world's oceans.Firstly,we retrieve ocean surface wind speed from 346 SAR iamges over in the east and west coasts in the United States,Gulf of Mexico and the Bering Sea.The conclusion shows that wind speed retrieve using CMOD5.N function for VV-polarized data are most appropriate and the root mean square error(RMSE)between SAR-retrieved wind speed and buoy-measured is 1.94m/s;CMOD5.N+Z2011 model is the optimal model for wind speed retrieveal from HH-polarized data and the RMSE is 2.37m/s;C-2PO model is the best model for wind speed retrieval from VH-and HV-polarized data,where the RMSE is 1.98m/s and 2.01m/s,respectively.Next,we retrieve wind speed from 58 SAR images over in the East and South China Sea.The results show that CMOD5 function is the optimal model for wind speed retrieveal from VV-polarized data and the RMSE between SAR-retrieved wind speed and ERA-Interim measured is 2.06m/s.However,the SAR-retrieved wind speed errors from HH-polarized and cross-polarized SAR data are relatively large.In addition,wind speed retrieve from cross-polarized SAR images large than conference data.(2)A new polarization ration(PR)model and VH-polarized wind speed retrieval model was developed.Due to the bad performance for wind speed retrieveal from HH-and crosspolarized SAR data in China waters,we proposed two models,called CHN-PR and CHN-CP,baesd on 58 SAR images and ERA-Interim reanalysis wind data.When compared with ERAInterim wind speed,the RMSE of wind speed retreievd from CHN-PR combined with CMOD5 function for HH-polarized data is 1.92m/s,thereby outperforming previous PR models.The RMSE of wind speed retreievd from CHN-CP model for VH-and HV-polarized data is 1.50m/s and 1.53m/s,which has outperfomed the C-2PO and Vachon-CP model.(3)The resampling spatial resolution effec on wind speed retrieveal from quad-polarized SAR images was investigated.This paper developed a comparison strategy for investigation the effect of resampling spatial resolutions on ocean surface wind speed retrieval accuracy with VH-?HH-?VH-and HV-polarized SAR images.Results show that wind speed retrieve accuracy increases with spatial resolution decreasing.The spatial resolution of 850 m,1050m,1350 and1450 m can be considered as the “optimal spatial resolution” for VV-?HH-?VH-and HVpolarized SAR images.(4)The difference between wind speed retrieval from co-and cross-polarized SAR data was investigated.This paper compared the result for SAR-retrieve wind speed from co-and cross-polarized data in difference wind speed section.Using statistical analysi method,an accurate wind speed threshold value algorithm,called WST(Wind Speed Threshold Model)model,is proposed.Subsequently,a hybrid wind speed retrieve model(HWSR)has developed based on WST.The REMS between SAR-retrieved wind speed from HWSR model and buoymeasured is 1.47m/s,which has outperfomed the C-2PO and CMOD.5model.Finnaly,from the point of view of modelling data,radar incidence angle,signal to noise ratio,resampling spatial resolution,ocean current field and SAR imaging mechanism,we explain why the C-2PO models based on the cross-polarized retrieved wind speeds has a rather poor performance at low wind speeds.(5)The applicability analysis about wind speed retrieval from compact polarization SAR data was investigated.Based on 256 quad-polarized SAR images,the RV-,RH-,RL-and RRpolarized SAR images are generated by an RCM simulator.Next,we applied these data to existing wind speed retrieveal models.The result show that wind speed retrieve using Co VePol function for RV-polarized data are most appropriate and the RMSE between SAR-retrieved wind speed and buoy-measured is 1.94m/s;wind speed retrieve using Co Ho-Pol function for RH-polarized data are most suitable and the RMSE is 2.19m/s;SAR-retrieved wind speed from CMOD5.N+Z2011 model are most appropriate for RL-polarized data and the RMSE is 2.31m/s.Based on the relationship between RR-polarized data and buoy wind speed,a new wind speed model,called RCM-RR,has developed.Then,we applied the RCM-RR model RR-polarized SAR data,the results show that SAR-retrieved wind speeds have a good agreement compared with buoy observed data and the RMSE is 2.16m/s.