In-flight methods for satellite sensor absolute radiometric calibration.

Abstract

Three methods for the in-flight absolute radiometric calibration of satellite sensors are presented. The Thematic Mapper (TM) on the Landsat satellites and the HRV on the SPOT satellite have been calibrated using the three methods at the White Sands Missile Range in New Mexico. Ground and airborne measurements of ground reflectance, radiance, atmospheric, and weather parameters are made coincident with satellite image acquisition. The data are analyzed to determine inputs to radiative transfer codes. The codes compute the radiance at the sensor entrance pupil which is compared to the average digital count from the measured ground area. The three methods are the reflectance-based, radiance-based and irradiance-based methods. The relevant theory of radiative transfer through an atmosphere is reviewed. The partition of extinction optical depth into Rayleigh, aerosol and absorption optical depths is discussed. The reflectance-based method is described along with the assumptions made. The reflectance-based method accuracy is no better than the measurement of the ground reflectance which is made in reference to a standard of spectral reflectance. The radiance-based method is described. The standard for the radiance method is a standard of spectral irradiance used to calibrate a radiometer. The calibration of a radiometer is discussed along with the use of radiative transfer computations to correct for the residual atmosphere above the radiometer. The irradiance-based method is described. It uses the measurement of the downward direct and total irradiance at ground level to determine the apparent reflectance seen by a sensor. This method uses an analytic approximation to compute the reflectance without the use of an "exact" radiative transfer code. The direct-to-total irradiance ratio implicitly gives the description of the scattering normally calculated from the size distribution and assumption of Mie scattering by the aerosols. The three methods give independent results which should allow for the detection of possible systematic errors in any of the methods. All three methods give results within the estimated errors of each method on most calibration dates. We expect the results of our sensor calibrations are within five percent of the actual value.