Remote Sensing Information Extraction In Bohai Turbid Water From Geostationary Ocean Color Imager Data

Bohai Sea and the Yellow River estuary are typical turbid waters. Their optical propertiesare complex and changeable. The time interval of ocean color observation is increased tohour with the launch of Geostationary Ocean Color Imager (GOCI) in2010. It providesimportant data support to the algorithm research of complex water and observation ofinshore water phenomenon. This paper accomplished the following works:(1) Towards the atmospheric correction difficulty of Yellow River estuary, this paperevaluates the applicability of four atmospheric correction methods: the neighbor cleanpixel method, the UV method, the near-infrared empirical method and the KOSC methodwith GOCI and in-situ data. First, the hypothesis of the four methods is assessed. Theempirical relationship of near-infrared and red bands in the KOSC method is poor. The412nm band is considered dark reasonable in the UV method. The near-infraredspectroscopy of the Bohai Channel is similar to that of real clean pixel, so it is used asclean pixel. The ratios of water’s reflectance and aerosol’s reflectance (k=1.71, ε=1) ofthe near-infrared empirical method can be used in this area.In low turbid water, the retrieved results of four atmospheric correction methods areconsistent. In high turbid water, the temporal and spatial characteristics are reasonablebased on the UV method and the near-infrared empirical method. The result shows thatthe near-infrared empirical atmospheric correction method is the best with the quantitativeanalysis of the Temporal-spatial matching data（18.6~295.22mg/m3）. The root of meansquare error (RMSE) is between0.01~0.03, the average percentage difference is between6%~48%. Especially the APDs within555-680nm are less than15%. The result of the UVmethod is obviously overestimated and that of the neighbor clean pixel method issignificantly underestimated.(2) Due to no inverse algorithm of Particulate Organic Concentration (POC) in theYellow River estuary, three POC retrieval model with745nm,680nm and their compositebands are developed. The POC retrieval model with745nm is the best. Its APD is20.4%and RMSE is18.4. It is also better than the traditional model with blue-green bands. ThePOC temporal and spatial variation from745nm retrieval model is reasonable and theAPD is below30%with GOCI data. It is shown that the POC in winter is the highest among the four seasons and the POC in summer is the lowest. The Bohai bay has thehighest POC and the Bohai Channel has the low POC during the whole year. The POCvariation within8hours has the the same order of magnitude with the seasonal POCvariation. The monthly average POC product with8images per day improves thecoverage of POC distribution, Statistical validity and the reliability of the results.(3) Absorption coefficient is an important water optical property. Its inversealgorithm in turbid water has obvious regional characteristics. Absorption retrievalmodels for412nm,443nm,490nm and555nm with Rrs(660)/Rrs(490) are developed Inthe Bohai Sea based on in-situ measured data. The correlation coefficients are about0.9and the APDs are all below20%. The new model is better than QAA and it is correctedalgorithm. It is found that the absorption coefficient is high in Liaodong Bay, Bohai Bayand Laizhou Bay with GOCI data. Significant difference of absorption coefficients are inthe hourly variation in different dates.Finally, based on the work summary, the next research work is proposed, such asatmospheric correction research plan in Yellow River estuary. In addition, it is alsopointed out that the time series characteristics of GOCI data should be considered fully inthe future.