[Evaluation on the atmospheric correction methods for water color remote sensing by using HJ-1A/1B CCD image-taking...

HJ-1A/1B satellite CCD images have higher spatial and temporal resolution, making them of great potential in quantitatively monitoring the water quality of inland lakes. However, the atmospheric correction of the images restricts their application. Therefore, taking Poyang Lake, the biggest freshwater lake in China as study area , and using the in-situ data collected in 2009 and 2011, this paper compares the atmospheric correction results done by the four methods: FLAASH, 6S, COST and QUAC, and analyzes the influence of these atmospheric correction methods on the inversion accuracy of the total suspended sediments (TSS) concentration. The results indicate: (1) the band 1 (blue band) of HJ-1A/1B CCD satellite images should be recalibrated while being applied into water quality remote sensing. The accuracy of atmospheric correction done from band 2 (green band) and band 3 (red band) is higher than that of others , especially that of the correction done by FLAASH, 6S and COST is much higher while that of correction done by QUAC is lower. So the algorithms of QUAC should be pointedly improved. (2) The ratios done from band 2 and band 3 have a good match with in-situ data , with an average relative error of 8.2%, 9.5%, 7.6% and 11.6% respectively for FLAASH, 6S, COST and QUAC. Therefore, it would be better to use the ratio done from band 2 and band 3 as inversion factors in Poyang Lake. (3) It is found that the accuracy of directly building models by using the four atmospheric corrected results and the TSS concentration is higher than the models built by the in-situ remote sensing reflectance and the TSS concentration. The accuracy of the TSS concentration inverted by FLAASH, 6S and COST is much high with an average error of only 10.0%, 10.2% and 8.0% respectively, while the error inverted by QUAC is a little bit higher of being 18.6%. So it is suggested to build model with atmospheric correction results and the TSS concentration data, because it can avoid the cumulate error resulted from modeling by using the in-situ spectrum data. (4) Under a low requested situation, these four atmospheric correction algorithms can all be adopted; otherwise, the COST should be used in the case of lacking supplementary information.