Comparison Of Atmospheric Correction Algorithms For MWI Onboard The Tiangong-2 Space Lab Over The Case ? Waters

Atmospheric correction(AC)for coastal waters is one of the key processes in ocean color remote sensing.The accuracy of AC determines the precision of water-leaving radiance and ocean color parameters extraction.The Moderate-resolution Wide-wavelengths Imager(TG2/MWI)is the ocean color sensor onboard the Chinese Tiangong-2 Space Lab,which was launched in 2016,has the characteristics of the high spatial resolution,multiple spectral bands and high signal to noise ratio.TG2/MWI provides the new data for ocean color remote sensing,especially provides the high quality products support to the algorithms research of complex coastal waters and observation of inshore waters phenomenon.In recent years,Hainan Island and its surrounding coastal waters have exploited by humans obviously,the water optical properties have gradually transformed from clear waters to turbid waters.Meanwhile,the obvious water variation is easily measured from coastal turbid waters to clear waters based on the TG2/MWI images.Therefore,taking the damand of AC into account in Hainan Island coastal waters,this paper accomplished the following two main works:1.Comparison of atmospheric correction algorithms for the whole region of interest:Taking the advantages of wide wavelengths and high signal-noise ratio into account for TG2/MWI,four published atmospheric correction algorithms were used in this paper: NIR standard algorithm,UV,NIR-corrected and SWIR algorithm.And these algorithms were evaluated based on the match-up exercise.The results showed that NIR standard algorithm was not fit for turbid waters,some nagetive values were appeared in 413nm;Our quantitative results supported that the NIR-corrected algorithm was the most appropriate algorithm to derive normalized water-leaving radiance with the iterative water optical model,NIR standard algorithm was similar to UV algorithm,but SWIR was the worst.All algorithms showed the better performance at 490 and 555 nm,and low accuracies at the blue(413nm)and NIR bands(750-865nm).2.Comparison of atmospheric correction algorithms based on the classification of water types:Because of the high diversity and spatio-temporal dynamics of the coastal domain,in this paper,the decision tree approach are considered to classify different water types(Class 1 to Class 4),and evaluation of four AC algorithms over these different types.Our results supported that NIR standard algorithm shows the best performance over Class 1(clear)water,NIR-corrected algorithm had the best result over Class 2(relative clear)and Class 4 waters(turbid),and UV algorithm was much more fit for Class 3 water.This research provides the good experience for AC over the coastal waters in ocean color remote sensing,giving the direction of choice of AC algorithms and the reference to the modification of atmospheric correction algorithms based on the validation of AC for the whole region and for each water types,respectively.Finally,some beneficial suggestions and perspectives for further improvements in AC is proposed based on the work summary,such as more efforts are made to get continuous and broad water-leaving radiance data near the coastal waters.In addition,a systematic vicarious calibration will be performed for the TG2/MWI sensor in the future.