In the water color remote sensing, hyperspectral data always provide lots of information. However, retrieval of water constituents based on conventional statistical methods fails not only to make full use of the information, but also to remove the heavy correlation between spectral band-variables. Partial least squares regression (PLS) can solve the problem properly. In order to validate its feasibility in inland Case-Ⅱwaters, a PLS model was established to estimate Chlorophyll-a (Chl-a) concentrations and total suspended sediments(TSS) contents using hyperspectral data obtained from April to October in 2006 in the Reservoir Shitoukoumen. The results showed that: PLS model could make a relatively full use of hyperspectral data. Coefficients of different band-variables in the final model reflected absorption and scattering properties of Chl-a and TSS. Compared with bands ratio model and first-derivative reflectance model, PLS model overmatched the former two with high determination coefficients and satisfactory estimating results. It indicates that PLS is suitable for the inversion of water constituents in inland water Case-Ⅱ.