RESUMO
A hierarchical cluster analysis is applied to the data set of R(rs) spectra. Combing the inherent optical and measured water color characteristics, we determine factors of the variability for each class of R(rs) spectra. The R(rs) measured from Lake Taihu, Chaohu, Dianchi and Three Gorges Reservoir can be divided into three spectrally distinct classes: class 1, class 2 and class 3. Class 1 is associated with water optical property dominated by total suspended matter and phytoplankton. Class 2 is associated with water optical property dominated by phytoplankton. And Class 3 is associated with water optical property dominated by total suspended matter. The three wavelengths factor Chlorophyll a concentration inversion models were developed according to different class water optical characteristics, which means relative error and RMSE are 23.8% and 8.5 mg/m3, and have higher accuracy than the model developed without classification.