The remote sensing of chlorophyll-a in case II water has been far less successful than that in case I water, due mainly to the complex interactions among optically active substances (i.e., phytoplankton, tripton, colored dissolved organic matter, and water) in the former. To address this problem, Gitelson et al. (2008) suggested a three-band model, which can minimize the effects of tripton, colored dissolved organic matter (CDOM), and pure water, and thus promised an accurate estimation of chlorophyll-a. In this study, we used three datasets with different phytoplankton species to test the performance of the three-band model developed by Gitelson et al. The major findings of our study were as follows : (1) the mechanism of the three-band model could work very well for each phytoplankton species (R2>0.84, rRMSE<23%) ; (2) the slope and intercept of the three-band model strongly depended on variation of phytoplankton species ; (3) chlorophyll-specific absorption coefficients at 440nm (a*ph(440)) could be used to predict the slope and intercept of the three-band model for different species of phytoplankton. Compared with the previous three-band model, the RMSEs of the improved three-band model were reduced from 37.2mgm-3 to 7.3mgm-3, and from 34.3mgm-3 to 15.9mgm-3, for Lake Kasumigaura, and Lake Dianchi, respectively.