According to the measured hyperspectral remote sensing data and synchronous sampling data about the water quality of Nansi lake from 21 to 23 of July,2014.The paper established and verified the single spectral reflectance model, the band ratio model, the first derivative model, three band model and four band model respectively. It made an attempt to use the continuum removal and reverse continuum removal method to set up retrieve chlorophyll-a concentration model. The results show that R*728 model,R′705 model,R*728/R*617 model,R734/R463model,［R-1rs (671)-R-1rs(730)］·Rrs(763) model,［R-1rs (682) -R-1rs(720)］·［R-1rs(761)- R-1rs(733)］ model have high accuracy to estimate chlorophyll-a concentration, the root mean square errors and average relative error of verified samples are less than 7 μg/L and ARE is in the scope of 15%. The linear models have the highest accuracy to estimate chlorophyll-a concentration compared with the other models,which is more suitable ways to estimate chlorophyll-a concentration of Nansi Lake;the model of continuum removal and reverse continuum removal method cannot be established.