a Component Decomposition Model for Evaluating Atmospheric Effects in Remote Sensing
Abstract
A radiance value of a target pixel recorded by a remote sensor can be decomposed into three components: (1) attenuated target signature, (2) pure atmospheric radiation, and (3) the contribution made by the ground through the atmospheric scattering process. Given the meteorological and optical parameters of a layer-structured atmosphere, its transmittance and radiance distribution can be accurately calculated with a plane-parallel radiative transfer model. For a uniform surface, the third component can be obtained by comparing radiances for an atmosphere over a black but nonemitting surface and the same atmosphere with an underlying ground of given albedo or temperature. For an inhomogeneous surface, the first two components remain the same as long as the surface is a plane. The third may be estimated using the locally averaged top-of-atmosphere radiance. An atmospheric point spread function is calculated from a Monte Carlo approach and is used for retrieving the ground signature through a deconvolution procedure. By applying the component decomposition model in an atmosphere with an underlying homogeneous Lambertian surface, the band-averaged pure atmospheric radiance and the downward and upward direct, diffuse and total transmittances are calculated for the six reflective wavelength bands of the Landsat Thematic Mapper: 0.45-0.52 (mu)m, 0.53-0.61 (mu)m, 0.62-0.69 (mu)m, 0.78-0.90 (mu)m, 1.57-1.78 (mu)m, and 2.10-2.35 (mu)m. In the three shorter wavelength bands, the reciprocity between the upward and downward transmittances holds very well, indicating a small uncertainty in the band-averaged point spread function. For the other three wavelength bands, similar reciprocity does not hold exactly, but the insensitivity of the band-averaged upward transmittances to the solar incidence angle make it possible to construct a stable band-averaged point spread function for these wavelength bands. By applying the component decomposition model in an atmosphere-snow system, the band-averaged upward atmospheric transmittances from an anisotropic snow surface are calculated and compared with their counterparts for Lambertian surfaces of the same albedoes. The relative error in the atmospheric point spread function introduced by approximating snow by a Lambertian surface is 1.5% for shorter wavelength band and tends to be smaller for longer wavelength bands. Knowing the pattern of the surface anisotropic reflectance factor, an adjusted band-averaged point spread function can be calculated and be used to retrieve the upwelling radiance from an inhomogeneous surface with a stable anisotropic reflectance pattern.
- Publication:
-
Ph.D. Thesis
- Pub Date:
- 1985
- Bibcode:
- 1985PhDT........22L
- Keywords:
-
- RADIATIVE TRANSFER;
- SNOW;
- IMAGE PROCESSING;
- Physics: Atmospheric Science; Remote Sensing;
- Atmospheric Effects;
- Atmospheric Models;
- Pixels;
- Radiance;
- Radiative Transfer;
- Remote Sensing;
- Albedo;
- Lambert Surface;
- Landsat Satellites;
- Point Spread Functions;
- Snow;
- Thematic Mappers (Landsat);
- Thematic Mapping;
- Geophysics