Table 3 and Table 4 below give detailed error budgets for the various reflectance and radiance methods recently developed by Slater and co-workers. The tables indicate the present achievable absolute accuracy and anticipated improvement over the next 3-5 years for high spatial resolution imagers, such as TM (Slater 1991, Private discussions with P. N. Slater, 1991, University of Arizona, Tucson, Arizona.).
As can be seen from Table 3, the uncertainty for the reflection method is dominated by uncertainties in the ground reflectance (due to radiometer calibration) and aerosol properties (complex index of refraction and size distribution). A small change in aerosol index of refraction from n = 1.54-0.01i to n = 1.54-0.005i, increases the computed radiance at the satellite radiometer by 3% (Kastner and Slater 1982). Slater (1988) reports that for TM calibrations, a change in the imaginary part of the complex index of refraction from 0 to 0.1 changes the radiance at the satellite by 8% for a surface reflectance of 0.25 and 23-km visibility. This percentage change is nearly constant over the range of reflectances of 0-0.5. In many rural areas the imaginary part is <0.005. This is significant in general, but specifically for calibrations at White Sands, New Mexico, because the reported value for the imaginary part is 0.01 (Jennings et al. 1978), which is high relative to other rural areas (Slater 1988; Herman and Browning 1975). It would appear that remeasuring the optical properties of aerosols at White Sands is in order (on calibration days, if possible) because of the importance of this region for satellite calibration. Changes in aerosol optical depths and visibilities have a much smaller effect.
As can be seen from Table 4, the dominant uncertainty in this particular radiance method (Slater 1991, Private discussions with P. N. Slater, 1991, University of Arizona, Tucson, Arizona.) which uses a high-flying helicopter, is the uncertainty of the radiometer calibration. This uncertainty is, in turn, limited by the uncertainty of the output of the calibration source (panel and lamp). Apparently, for the method that uses the ER-2 high-flying aircraft, pointing and scene co-alignment uncertainties are larger than for the helicopter. The reported scene co-alignment uncertainty currently contributes 3% uncertainty to the absolute calibration, but recent improvements in the scene co-alignment algorithm will reduce this to 1-2% (Slater 1991, Private discussions with P. N. Slater, 1991, University of Arizona, Tucson, Arizona.) Additional improvements in the calibration process are detailed in the following section.