Atmospheric Correction Assessment of Surface Reflectance for Multi-temporal Landsat Images in Cloudy and Rainy Urban Area

Two main atmospheric correction methods LEDAPS and FLAASH are applied to multi-temporal Landsat TM, ETM+ and OLI sensor images to complete an atmospheric correction assessment. It includes comparisons of 15 consecutive years atmospheric corrected images surface reflectance, Normalized Difference Vegetation Index (NDVI), Tasseled Cap components of Greenness and Brightness and class separability Jeffreis Matusita (J-M) distance in Guangzhou City before and after correction. The results show: two main atmospheric correction methods present the different spectral response characteristic in each band, and correction effect is the most obvious in the visible bands; both of the corrected time series images reflectance showed standard error variability, especially to FLAASH corrected images; compared with LEDAPS corrected images, FLAASH corrected images overestimate Near-Infrared(NIR) reflectance and related spectral indexes such as NDVI and Tasseled Cap components; according to paired t-test, two sets of images also show differences on the class separability of artificial vegetation cover and city areas. FLAASH corrected reflectance calculate higher tasseled cap components, which mainly come from insufficient correction of water vapor in shortwave infrared bands. For the cloudy and rainy urban areas like Guangzhou, water vapor has an important influence on atmospheric correction.