Study of retrieving models for chlorophyll-a concentration based on classification of above-water remote sensing reflectance

In this study, according to the characteristics of above-water remote sensing reflectance (Rrs) curves, we put our samples into different classes, and for each class, find a best retrieving model among the first-derivative model, band ratio model and three-band model to show the correlation between Rrs data and the concentration of chlorophyll-a. As a result, it is proved that Rrs curves can be classified into 6 classes by differences in peaks around the wavelength of 560nm or 705nm, as well as valleys around the wavelength of 630nm or 680nm. Compared with non-classified retrieving models based on the whole data set, classified models reduce the absolute value of average relative error (RE) range from 7.39% to 39.85%, effectively improve the retrieving precision.