An Error Budget for Cross-Calibration of AVHRR Shortwave Channels against ATSR-2

Cross-calibration of the AVHRR shortwave channels against the ATSR-2 is not straightforward because the spectral response functions of the sensors differ and because the sensors generally view targets from different directions, thereby requiring a priori knowledge of the bidirectional reflectance of the target. Evidence is presented to show that uncertainties caused by the directional mismatch may be reduced if the bidirectional reflectance distribution function (BRDF) estimated from AVHRR data is used to map AVHRR reflectances to the ATSR-2 geometrical configuration. A stochastic model is developed to study the impact on the correlation between ATSR-2 and AVHRR radiances caused by 1) the response of vegetation to rainfall in a semi-arid region, 2) errors in the BRDF model used to map AVHRR reflectances to ATSR-2 geometry, 3) spatial nonuniformity of the target, and 4) variations in the column water vapor. The model shows that spatial nonuniformity contributes most of the error, but that cross-calibration to 1% or better is possible provided that the coefficient of spatial variation is 2.5% or less and the sensor remains stable. However, determination of the rate of calibration drift is problematic, with errors of the order of 20% in the rate of drift even when the data span a full year.