Abstract: The transparency,which can demonstrate the limpid or muddy degree of lake water also be used to evaluate water eutrophic state,is one of the water body visible degree indicators.The traditional transparency of surface water is observed with Secchi Disk.It is time-consuming and strenuous,and only represents some local information. The remote sensing technology, which can cover large area simultaneously and periodically,may deal with this kind of limitation effectively. In this paper,77 spectral reflectance data of different water sites in Chagan lake were collected during 6 field works,and the water transparency data were acquired simultaneously. the linear regression constructed with every single band of reflectance and derivative against the water transparency data;the band ratio model with the reflectance of 780 nm and 654 nm also established.Finally,ANN-BP model is established with diagnostic band reflectance and derivative as the input vector.The result shows that water trans-parency has an intimate relation with water reflectance,and correlation coefficient is about 0.5 in blue and green spectral region, it is about 0.6 in red spectral region;while it obtains the maximum value in short wave of near infrared region. As for derivative reflectance,it also obtains higher value in near infrared spectral region.The regression model established with single band gets the similar performance as that with band ratio as regression dependent variable. By comparison,the ANN-BP model performed best with determination coefficient(R²)of 0.98. It indicates that hyperspectral remote sensing models established for estimation Chagan lake water transparency can obtain the comparatively accurate result. The construction of hyperspectral models in Chagan lake will to help estimate the water body transparency with satellite images,and also provide a method for remote sensing monitoring of other inner water body with similar water status as that in Chagan lake.