By restoring plentiful spectral information from several bands, hyperspectral image reconstruction could provide more suitable data source to water environment remote sensing. This is significant for inland water color remote sensing. By using the HJ1 A-HSI and HJ1A-CCD image acquired on June 6th, 2009, the hyperspectral data was reconstructed from HJ1A-CCD data. The results show that: (1) The average relative error of HJ1A-HSI data and reconstructed data compared with measured Rrs at 660 nm-900 nm are 0.333 5 and 0.307 7, respectively; (2) The entropy and average gradient of reconstructed image are higher than HJ1A-HSI image. In additional, the three band model get higher accuracy when inversing chl-a concentration by the reconstructed data.