Atmospheric Correction of Satellite Imagery Using Modtran 3.5 Code

When performing satellite remote sensing of the earth in the solar spectrum, atmospheric scattering and absorption effects provide the sensors corrupted information about the target's radiance characteristics. We are faced with the problem of reconstructing the signal that was reflected from the target, from the data sensed by the remote sensing instrument. This article presents a method for simulating radiance characteristic curves of satellite images using a MODTRAN 3.5 band model (BM) code to solve the radiative transfer equation (RTE), and proposes a method for the implementation of an adaptive system for automated atmospheric corrections. The simulation procedure is carried out as follows: (1) for each satellite digital image a radiance characteristic curve is obtained by performing a digital number (DN) to radiance conversion, (2) using MODTRAN 3.5 a simulation of the images characteristic curves is generated, (3) the output of the code is processed to generate radiance characteristic curves for the simulated cases. The simulation algorithm was used to simulate Landsat Thematic Mapper (TM) images for two types of locations: the ocean surface, and a forest surface. The simulation procedure was validated by computing the error between the empirical and simulated radiance curves. While results in the visible region of the spectrum where not very accurate, those for the infrared region of the spectrum were encouraging. This information can be used for correction of the atmospheric effects. For the simulation over ocean, the lowest error produced in this region was of the order of 105 and up to 14 times smaller than errors in the visible region. For the same spectral region on the forest case, the lowest error produced was of the order of 10-4, and up to 41 times smaller than errors in the visible region,