Algorithm To Retrieve IOPs For Turbid Inland Water

In this paper, algorithms to retrieve absorption and backscattering coefficients and to decompose aexp(λ), the absorption subtracting absorption of pure water from total absorption, were established.(1) Based on reflection spectrum characters of water, turbid eutrophic Case II water in Lake Taihu was divided into two types with different optical characteristics at Slope=0.32. The Slope refers to the spectrum slope of reflectance between 677nm reflection valley and 701nm reflection peak. Retrieval models with different parameters setting for different water types were established to two divided turbid eutrophic waters. Segmented simulation of backscattering coefficient was adopted.1) 400-685nm backscattering coefficient was simulated through power function. Based on QAA algorithm,550nm and 675nm were choosed as reference wavelength to establish two retrieval models. Based on data assimilation, estimations of these two retrieval models were optimized through the multi-model collaborative retrieval algorithm, so that the optimal backscattering coefficient was determined.2)>685nm backscattering coefficient was simulated as a constant. Finally, absorption coefficient was calculated from a biological optical model. Verification dataset demonstrated that the proposed IOPs retrieval algorithm(QAA-DB) achieves satisfying retrieval accuracy on absorption coefficient of complicated Case II water, showing 19.71% MAPE and 1.3933RMSE.(2) Two algorithms were improved in this study, the algorithms based on band ratios and parameterization method using Gaussian functions. And the aexp(λ) were then decomposed into component functions representing absorption by phytoplankton pigments aph(λ) and sum absorption of non-algal particles and chromophoric dissolved organic matters adm(λ) using the two improved algorithms.1) In the algorithm based on band ratio, the band ratios of components absorption at 450 and 480 nm and the slope of adm(λ) were empirically determined, then adm(λ) and aph(λ) were determined through simultaneous equations and a decreasing exponential function model for adm(λ).2) In the algorithm based on parameterization method using Gaussian functions,12 Gaussian functions representing aph(λ) and a decreasing exponential function model representing adm(λ) were used in parameterizing aexp(λ). The adm(440) was empirically determined firstly and then a least squares optimization algorithm was used to determine the other unknown parameters.Testing using verification dataset, the algorithm based on band ratio showed 24.72% MAPE and 0.81 RMSE; the algorithm based on parameterization method using Gaussian functions showed 27.73% MAPE and 1.02 RMSE. Both algorithms achieve satisfying decomposition accuracy for Case Ⅱ water.