Atmospheric effects are present in remotely sensed data. Atmospheric gases absorb the radiation in specific regions of the spectrum, molecules scatter the radiation, and aerosols scatter and absorb the radiation at shorter wavelengths. For observation of the Earth's surface, bands are located outside of the absorption windows, but scattering effects still occur in the visible to near-infrared bands and cannot be avoided. In the hypothesis of a uniform target (i.e., of infinite radius), the equation that describes the atmospheric scattering effect is nearly linear, and, provided that atmospheric constituent radiative properties are known, the remotely sensed signal could be easily inverted to provide the “surface reflectance.” In the case of a heterogeneous landscape, where the target radius is less than 500 m and significantly differs from its background, the inversion of the signal is more complicated and an atmospheric point spread function should be taken into account. We present here, first, the formalism of the atmospheric effect and, second, the implementation of the correction and the application to a Landsat 7/Enhanced Thematic Mapper (ETM) + scene.

Atmospheric constituents influence the signal in remote-sensing systems; thus, the data need to be corrected in order to retrieve surface properties. Gases absorb the radiation in specific bands and, whenever possible, the sensor channel is located in an atmospheric window to avoid those effects. Molecules and aerosols substantially scatter the radiation in the visible to near-infrared region of the spectrum. For a coarse spatial resolution sensor (1 km and above), the adjacency correction is small and may be ignored so the correction of atmospheric effects is done by treating each pixel independently. For a finer spatial resolution sensor and especially at shorter wavelengths, the environment of the pixel should be taken into account in the correction process as adjacency effects are large and clearly visible in the data. For an example, see the study by Kergormard and Tanre on SPOT 10-m resolution data.1 Using the atmospheric point spread function, we have developed a simple correction scheme for the adjacency effect that we have applied to Landsat 7/Enhanced Thematic Mapper (ETM) + data. We present, first, the theoretical background of the atmospheric effects and their correction and, second, the operational implementation and some results obtained using a Landsat 7/ETM + scene acquired over Washington, DC.