Considerations for Atmospheric Correction of Surface Reflectance for Soil Survey Applications

摘要

Remote sensing is increasingly used in soil survey efforts. Atmospheric interactions can have a large influence on the use and interpretation of remotely sensed data. Therefore, an appropriate atmospheric correction method is central to correct interpretation of remotely sensed data for soil survey efforts. Correction methods often require assumptions to account for varying atmospheric conditions, and incorrect assumptions can introduce error. The objectives of this study were to compare methods of atmospheric correction for a Landsat 7 ETM+ image from southeastern Arizona in coordination with a NRCS soil survey premapping effort. Two atmospheric correction methods were compared: (i) COST (cosine theta) dark object subtraction and (ii) Rayleigh scattering correction with 6S (Second Simulation of a Satellite Signal in the Solar System), which uses radiative transfer code. Comparison of corrected bands using COST and 6S methods indicated significant differences in band reflectance among corrected data and uncorrected apparent reflectance for all bands, except for band 3. Landsat band 1 was most sensitive to the 6S method due to effects of Rayleigh scattering. Pixel values corrected with the COST method were significantly different from apparent reflectance and Rayleigh correction in bands 4, 5, and 7. Lack of a true dark object in this scene likely contributed to differences in pixel values of the COST method; therefore, it was not selected for atmospheric correction. These data indicate that the 6S method provides the more robust atmospheric correction when reasonable constraint on atmospheric conditions is available and is suggested as the appropriate correction technique for incorporating Landsat data into soil survey.