Study On Time Series Ocean Color Data Merging And Red Tide Information Extraction

With high temporal and spatial resolution, wider coverage and other advantages, oceancolor data have been widely used in ocean productivity evaluation, ocean pollution, andrelated areas. But in cloud-prone and rainy area，the biggest application limitation of oceancolor remote sensing is data coverage, and the dynamic temporal information has not beenfully exploited yet. Aiming at the red tide detection, techniques of time-series ocean colordata merging were studied, as well methods for red tide extraction and change forecast wereinvestigated. The main contents are listed as following:(1) Processing and spectral analysis for multi-sensor ocean color images: multipleremote sensing images were preprocessed by geocorrecion, atmospheric correction,parameter inversion and mosaicking, then dataset for time-series ocean color datamerging and red tide extraction was established.(2) Merging of ocean color remote sensing images: comparisons among different chlaconcentration images, fluorescence line height images, remote sensing reflectance at443nm and667nm were analyzed, which demonstrate that different source data onthe same day were comparable and consistent, and can complement each other inspatial coverage. The adaptive weighted averaging method was applied on the chlaconcentration images, fluorescence line height images, which can reduce the noiseand the difference between merged and neighborhood pixels, and also can improvespatial coverage and sampling frequency. The bio-optical model of GSM01wasemployed to merge the remote sensing reflectance images and its capacity wasproved in expoiting the redundancy of spectral information and improving productconfidence,but it is sensitive to the noise.(3) Red tide information extraction based on time series ocean color images: accordingto the spectral analysis of red tide, typical red tide cases were analyzed by statisticaland visual analysis methods. Method of integrating chla concentration anddifference, fluorescence line height, modified red tide index, remote sensingreflectance and sea surface temperature was developed to extract the comprehensivered tide information, then simple and feasible threshold method was suggested.Spatial and temporal changes of red tide parameters can be achieved rapidly throughthe visual analysis method, which can highly enhance the detection accuracy andefficiency.(4) Applicaton of relevant technology in red tide forecast: using merged time series of ocean color images in Fujian coastal areas and red tide information extractiontechnology, the ocean color information especially the abnormal chla concentrationand fluorescence line height before the red tide incidents can be used as thescientific basis for prediction.(5) Development of prototype system: based on the key techniques of red tide extraction,information processing system for red tide was constructed, including imagemerging, spectral analysis, time-series data analysis and other modules, whichprovide a great platform for red tide information extraction by using time-series andmulti-sensor ocean color data.