A Study On Cloud Detection Algorithm Of FengYun-3D Spectral Imager Over Land

Cloud mask is the basis of quantitative inversion of cloud,aerosol by satellite data.Physical parameters such as cloud,surface properties can be achieved only after cloud detection is carried out.Cloud detection over land is more difficult due to the complexity of the surface type."FengYun-3" is Chinese second-generation polar-orbit meteorological satellite.The fourth satellite called "FY-3D" was successfully launched in November 2017.Researches on the second generation of Medium Resolution Spectral Imager(MERSI-Ⅱ),which is on board FY-3D,are insufficient.This study develops a cloud mask algorithm based on the classic MODIS algorithm for FY-3D/MERSI-Ⅱ over land.The new algorithm makes some adjustments according to MERSI-Ⅱ.MERSI-ii contains 25 spectral channels from 0.47μm to 12μm.However,the lack of the 13.9μm,6.7μm and 3.9μm channels makes some difficulties in the application of the MODIS cloud mask algorithm.The IMAPP software provided by the University of Wisconsin can generate cloud mask products as needed.It is to be noted that the first-level radiation data and the geo-information data of MERSI-Ⅱ have not yet provided by National Satellite Meteorological Center(NSMC),satellite data in this paper is obtained by extracting values of MODIS data in corresponding channels with the adjacent observational time.This study analyzes the effect of the Iack of corresponding spectral channels in MERSI-Ⅱ.Based on the MODIS database in year 2005,the results of 3.8μm and 4.05μm channel performing Brightness Temperature Difference test are studied,respectively.The 3.8μm channel is finally chosen for the new algorithm.Meanwhile,new thresholds are determined by statistical analysis of a large number of satellite data,combined with the original MODIS thresholds.FY-3D/MERSI-μ cloud mask retrievals are evaluated by MODIS and CALIPSO cloud products.Results show that,the lack of high cloud tests performed by 13.9μm and 6.7μm channels does not make significant influence to the final cloud category.It just improves the value of final confidence to some degree.When 3.8μm channel is chosen to perform the infrared Brightness Temperature Difference test,the final confidence value is reduced and pixels of confident cloud are increased.New thresholds can improve results obviously.The validation results show that the consistence between our algorithm and the MODIS products are above 88%,results of confident clear and confident cloud pixels are superior to that of probably clear and probably cloud pixels.Various cloud types,including transparent cirrus,can be detected by our algorithm.The consistence between our algorithm and the CALIPSO products in cloud area are above 90%.Meanwhile,the results are of high accuracy,with category of confident cloud.