Retrieval Of Suspended Matter Concentration And Reconstruction Of Missing Data Based On GOCI In Bohai And Yellow Sea

Concentration and distribution of suspended matter is an important research formarine ecosystems, biogeochemical cycles studies, coastal engineering, and harborconstruction. Ocean color remote sensing with its wide observation range, economic,efficient advantage has become an important means for the effective monitoring ofmarine suspended matter distribution. However it can only get the ocean colorinformation in the local and clear zone areas due to the coverage of the clouds overthe ocean and changes of the scanning track. There exist more invalid data in theremote sensing products. So it is necessary to fill missing data in the remote sensingproducts.Retrieval model of suspended matter for geostationary satellite ocean colorimages(GOCI) is developed using the in-situ data and evolutionary modeling method,which observes marine variations of China Bohai Sea and the Yellow Sea with hightime resolution. The evolutionary modeling method based on genetic programmingcould generate automatically several fitting models with the given accuracy, differentstructure and explicit forms. This method can provide a variety of choices for retrievalmodels, from which to determine the optimal suspended matter concentration retrievalmodel.Compared with the existing retrieval models of suspended matter concentration,the evolutionary models have better characteristics with high accuracy and simplestructures. The research shows that the evolutionary modeling method is applicablefor suspended matter concentration and other water constituents from ocean colorremote sensed data.Missing data due to the occlusion of the cloud has brought some difficulties forresearch in the optical remote sensing image. Decomposed intrinsic mode functions are utilized to interpolate and superimpose in order to reconstruct the missing data.The relative errors are small between data reconstruction values and the originalsuspended matter concentration values through the example of suspended matterconcentration. Compared with data interpolation method in different conditions ofdiscrete and continuous missing data, accuracy of suspended matter concentrationbased on empirical mode decomposition was higher than reconstruction of datainterpolation.