The field spectral reflectance was measured with a spectroradiometer in Taihu Lake of China, and the water samples were collected in situ and analysed in the lab. The water quality results are correlated with the reflectivity to find out the spectral characteristics and evaluated for estimating chlorophyll-a (CHL) and total suspended matter (TSM), on the basis of which the most appropriate single and combined bands with different pixel window sizes are selected from Landsat sensor to develop estimation models. And then the two best precise models, one with ETM3 dividing ETM1 and the other with ETM4 dividing ETM1, are used respectively to approximate the CHL and TSM concentrations. Additionally, the pixel window size has an effect on the precision to estimate the water quality parameters, 7 x 7 or 5 x 5 is mostly suitable to estimate CHL concentration whereas 3 x 3 or less is more suitable for estimating TSM concentration.