Currently, radiance and reflectance methods are the most accurate methods for indirect calibration of the satellite-based visible wavelength radiometers. Using both methods to calibrate the same instrument and comparing the results provides an opportunity to detect and eliminate systematic errors (Slater et al. 1987; Slater, Private discussions with P. N. Slater, 1991, University of Arizona, Tucson, Arizona.). On days other than those for which absolute calibrations are made, relative calibration methods (Brest and Rossow 1991) can be used to interpolate between absolute calibrations.
Thus, we recommend that the ARM program make use of both radiance and reflectance methods as well as relative and absolute calibration methods. Programs are continuing through NOAA and NASA (Slater, Private discussions with P. N. Slater, 1991, University of Arizona, Tucson, Arizona.) to make the necessary absolute calibrations and through the ISCCP program to make relative calibrations (Brest and Rossow 1991; Whitlock et al. 1990).
Currently, typical absolute calibration accuracies for the reflectance method are 6.5% (AVHRR) and 5.0% (TM and SPOT), and for the radiance method, 5.0-7.0% (AVHRR) and 3% (TM and SPOT). Both methods are improving. Radiance methods are improving through better scene co-registration algorithms (navigation) and radi-ometer improvements (calibration, smaller spectrometer field of view, faster scanning). Reflectance methods are being improved by field measurements (particularly measurement of the aerosol complex index of refraction) that reduce the uncertainties related to aerosols. Expected improvements in the next three to five years should reduce uncertainties for reflectance methods to 5.0% (AVHRR) and to 3.0% (TM and SPOT). Radiance methods should give uncertainties close to 3.2-4.2% (AVHRR) and to 2.0% (TM and SPOT). These uncertainties are summarized in Figure 4.
Based on the foregoing discussion, absolute calibration of AVHRR to 3-4% uncertainty and TM and SPOT to 2.0% uncertainty seems possible in the same time frame as the ARM measurements. These levels of uncertainty are somewhat greater than those desired for the ARM program radiation budget analysis. Nevertheless, these reduced uncertainties, resulting from improved visible-wavelength calibrations, will lead to greatly improved TOA radiance measurements and much better radiation budget determinations than are now possible.