= 0) and assume that they have the same reflectance from day to day throughout
most of the vegetation development period. The average values from forest and dark water pixels in AVHRR channels 1 and 2 are used to standardize the imagery.
Relative calibrations are useful for detecting drifts in calibration over time and for interpolating between absolute calibrations.
Relative calibrations are made by grouping scenes into several classes of targets of similar albedo and assuming that, as a class, their reflectance remains
essentially constant over specified portions of the year or on an annually sampled basis (Staylor 1990; Brest and Rossow 1991). The approach taken by Brown
and Manore (1986) is to select water bodies and forested areas near nadir ( = 0) and assume that they have the same reflectance from day to day throughout
most of the vegetation development period. The average values from forest and dark water pixels in AVHRR channels 1 and 2 are used to standardize the imagery.
A similar approach is taken in the ISCCP study by Brest and Rossow (1991), but many more types of targets are used (nine surface/vegetation classes and 28 specific geographical targets). Spatial and temporal averages are computed from the various scenes. Changes in these averages over time indicate changes in the satellite radiometer calibration.
The uncertainty in the methods described by Brest and Rossow (1991) are given in Table 5. The total uncertainty is ~5%, but drifts in satellite radiometer calibration on the order of 2% of the absolute calibration can be detected (Brest and Rossow 1991), and interpolation equations are developed that give the calibration for each satellite day anchored to absolute indirect calibrations.
Whitlock et. al (1990) briefly review interpolations for AVHRR and VISSR/VAS that are anchored to indirect absolute calibrations. Figure 5 shows their interpolation for NOAA-9 AVHRR Channel 1 gain calibration. Note, increasing values indicate the sensitivity of the radiometer is decreasing.
