The Advanced Geosynchronous Radiation Imager(AGRI)onboard China’s Fengyun(FY)-4 satellites,which provides observational data across various wavelengths from visible to infrared(IR),holds great potential for diverse a...The Advanced Geosynchronous Radiation Imager(AGRI)onboard China’s Fengyun(FY)-4 satellites,which provides observational data across various wavelengths from visible to infrared(IR),holds great potential for diverse applications.However,the FY-4A AGRI mid-wave IR(MWIR)band(3.75µm)is often contaminated by stray light in the midnight hours during the 1-2 months before and after the vernal or autumnal equinoxes.In this study,a U-Net-based deep learning model was employed to generate an expedient MWIR band from the FY-4A AGRI long-wave IR band.Validation using normal radiance measurements revealed that MWIR brightness temperatures generated by the deep learning model are very close to those observed by the FY-4A AGRI,with mean absolute error of 1.48 K,root mean square error of 2.39 K,and a correlation coefficient of 0.99.When applying the model to periods of stray light contamination,the brightness temperature anomalies found in the FY-4A AGRI MWIR band are effectively eliminated.The findings of this study could support various scientific applications that necessitate use of the MWIR band during midnight hours,such as identification of fog/low stratus cloud.展开更多
基金Supported by the National Natural Science Foundation of China(42305177 and 42175006)Beijige Foundation(BJG202210)Open Research Program of the State Key Laboratory of Severe Weather(2023LASW-B16).
文摘The Advanced Geosynchronous Radiation Imager(AGRI)onboard China’s Fengyun(FY)-4 satellites,which provides observational data across various wavelengths from visible to infrared(IR),holds great potential for diverse applications.However,the FY-4A AGRI mid-wave IR(MWIR)band(3.75µm)is often contaminated by stray light in the midnight hours during the 1-2 months before and after the vernal or autumnal equinoxes.In this study,a U-Net-based deep learning model was employed to generate an expedient MWIR band from the FY-4A AGRI long-wave IR band.Validation using normal radiance measurements revealed that MWIR brightness temperatures generated by the deep learning model are very close to those observed by the FY-4A AGRI,with mean absolute error of 1.48 K,root mean square error of 2.39 K,and a correlation coefficient of 0.99.When applying the model to periods of stray light contamination,the brightness temperature anomalies found in the FY-4A AGRI MWIR band are effectively eliminated.The findings of this study could support various scientific applications that necessitate use of the MWIR band during midnight hours,such as identification of fog/low stratus cloud.
