Simultaneously Retrieving Three Water Constituents: TSM, Chlorophyll and CDOM of Case 2 Water

摘要

Retrieving water constituents in case 2 waters by remote sensing is a crucial problem in evaluating ocean first productivity and monitoring various disasters. But it is difficult to accurately and universally develop both bio-optical models and remote-sensing reflectance model because independent temporal and spatial variation of dissolved organic matter (CDOM), chlorophyll and total suspended matter (TSM), high concentration of TSM, as well as the local characters of different regions. Current retrieval algorithms can be classified into two main types: one is linear retrieval algorithm including principal component analysis (PCA) and semi-analytical algorithm, the other is nonlinear optimization algorithm including genetic algorithm (GA) and artificial neural network (ANN). Remote-sensing reflectance model is derived from the radiative transfer equation. In our study, the chlorophyll absorption model and some other paragraphs of bio-optical models are adjusted. The adjustment is based on the water constituents concentration measured simultaneously with remote sensing data in the Yellow Sea and the East Sea of China. Then the equation of remote-sensing reflectance model can be changed into linear matrix of water constituents and coefficients, we find the spectrum curves of total suspended matter coefficient and chlorophyll coefficient turn out significant negative correlation. As a result, when performing matrix retrieval algorithm, chlorophyll concentration and CDOM concentration are out of required accuracy except some special conditions. According to our research, remote-sensing reflectance model is significantly featured by non-linearity and negative feedback. And the experiments prove the linear algorithm is not suitable for retrieving chlorophyll and CDOM of most case 2 waters in China, so the alternative choice of nonlinear optimization algorithm may be a possible way to solve the problem but not linear retrieval algorithm.