Calculating NDVI for Landsat7-ETM data after atmospheric correction using 6S model: A case study in Zhangye city, China

Remote sensing of vegetation is developing into a highly quantitative science. 6S atmospheric correction model based on radiative transfer theory, with a definite physical meaning, but it provided a limited number of patterns which are difficult to reasonably describe atmospheric conditions in a specific study area, therefore, in the absence of measured data, it is necessary to select or estimate reasonable atmospheric parameters, the selection of parameters directly affect the result of atmospheric correction. In this study, firstly atmospheric correction parameters identified for landsat7-ETM remote sensing image in Zhangye city, in particular, were analyzed, china. And then the Normalized Difference Vegetation Index (NDVI) of desert, grassland and forest from May 2001 to November 2002 was calculated before and after making the atmospheric correction. Further more, the spatio-temporal changes in the NDVI were analyzed, thus the effect of atmospheric correction was evaluated for different land covers' spectral response characteristics of ETM satellite image. The result showed that firstly the intercept of the regression line, between the NDVI of before and after correction, was almost the same for each land cover type, while the slope was around 1.6 to 1.8 for grassland, it was smaller in the desert, where the range of the NDVI was limited. Secondly, the NDVI in dense vegetation was more sensitive to the correction, which was a result of the effect of path radiance. After correction, the range of NDVI increased, with the characteristic that the greater the NDVI, the larger was the atmospheric effect. Thirdly, in all of the desert, grassland and forest land, atmospheric correction made the NDVI larger by at least 0.05; in the desert, NDVI before correction less seasonal change than NDVI after correction; in the grassland, NDVI before correction and NDVI after correction showed the same seasonal trend; in the forest land, NDVI after correction increased more rapidly and decreased more gradually than the NDVI before correction before and after the peak. Fourthly, the aerosol optical thickness and elevation in each pixel were considered in this 6S model, in this way, the results of atmospheric correction can be more reasonable and more truly reflect the surface vegetation conditions.