Abstract
For more accurate estimation of multi-substance concentration in the water based on a radiative transfer theory within shorter CPU time, a method based on a four flux model has been developed, and has been studied about it's practicability. The fundamental performance of the method was examined by applying it to simulated image data, in which various cases of three substances to be identified and visible-light bands corresponding to LANDSAT TM data were supposed for the data simulation. The inverse estimation for these cases showed that its accuracy depends slightly on the initial concentrations, indicating that with an adequate initial concentration the inverse estimation can provide the concentration of each substance within 0.07% error. And the influence of the remained atmospheric noise after atmospheric noise calibration had only minor effects on the accuracy in the inverse estimation. Furthermore, the results of the estimation used real Landsat TM data had good agreement with sea truth data.
