A number of issues have been raised with respect to calibration of satellite sensors in general. The ERBE instrument calibrations may be affected by a number of these, however, one measure of the quality of the instruments is that several packages have flown for long periods, permitting an examination of accuracy and stability. (Note added in proof: Dr. Bruce Wielicki (NASA Langley) presented a discussion of errors in the ERBE data at the June 1990 JASON meeting in La Jolla, California. His discussion included instrument uncertainties as well as uncertainties due to viewing geometry and averaging times. Dr. Wielicki will prepare a future paper covering this subject in some detail.)
House (1989) has compared longwave radiation measurements from ERBE instruments on NOAA-9 and ERBS. The paper states: "Results indicate that the ERBE radiometers
are mildly sensitive to varying thermal loads from the spacecraft and/or the earth/space environment. Radiometric variations at the satellite and methods of
data interpretation contribute about equally to the uncertainty of radiant exitances from the earth. The rms uncertainty of observed differences between instruments
is about 
4 
for irradiances at satellite altitude, 7
for exitances at the top-of-the-atmosphere,
and 7 for monthly averages after time and
space averaging. Overall, the longwave observational consistency is about 3% for individual measurements and
3% for monthly estimates."
Lee et al. (1990) present data from the ERBS, NOAA-9 and NOAA-10 solar monitors, covering 10/1984 to 6/1989, 1/1985 to 6/1989, and 10/1986 to 4/1987, respectively.
The paper notes that "the ERBS, NOAA-9 and NOAA-10 measurements have not exhibited any signs of degradation or shifts in the instrument gains. Therefore, the
measurements which are presented have not been altered or modified." Furthermore, "all the derived values agree within the 0.2% (2.7 ) measurement accuracy,"
allowing " detection of the decreasing and increasing trends in the irradiance at levels of the order of 0.03% to 0.05% per year." Thus, the calibration and
stability of the solar measuring channel appears to be well in hand, though there is a possibility of a systematic data offset since the instruments were calibrated
in the same facilities.
Perhaps the documented excellent stability makes discussion of improvements to the total power calibrations and to the detectors themselves appear somewhat academic. However, extreme accuracies are required in the climate energy flux measurements: the climate is affected by small differences in the net energy flux, which in itself is determined from subtracting two large energy flux measurements. A few examples of possible concerns will, therefore, be listed here, mostly related to the wavelength-selected data and to the maintenance of the calibration on the satellite.
Guenther (1987) notes that NBS-traceable tungsten filament calibration lamps "are calibrated with an uncertainty of nearly 3% in typical use," so that the shortwave-channel calibration on the satellite is of limited accuracy. Indeed, an international comparison of spectral irradiance measurements in the 300 to 800-nm wavelength range showed a total spread between the seven participating laboratories of about 1.5% in the middle spectral range, increasing to about 6.5% at the extreme red and UV wavelengths (Gilham 1977). Guenther (1987) also notes that "There is little or no history yet of successful use of any of these [light] sources on satellite-based instruments." Furthermore, the optics used on the satellite are subject to degradation (see, Horan [1974]) for an example of significant in-flight degradation). It is noted here that the CERES design includes integral measures to protect mirrors and the solar attenuation plate, which are both sensitive to molecular and particulate contamination. Alignment may also be affected by vibration, thermal cycling, and warpage, to name but a few possibilities.
A fused silica filter is used to define the wavelength region for the shortwave channel of the wide-field-of-view ERBE. This hemispherical filter maintains neither a uniform nor constant temperature distribution as it responds to changes in radiation emitted from earth. Corrections for this effect have been provided (Cooper and Luther 1980). One remaining concern is that the optical transmission of the fused silica may change as a function of time due to radiation (Baur et al. 1981).
Paden et al. (1990) provide long-term documentation of this effect. They note that the NOAA-10 WFOV channel is the most rapidly degrading of the three instruments flown, though it appears to have reached its degradation limit. It is noted that the NOAA-10 orbit causes the filter to be exposed to more sunlight than those on NOAA-9 and ERBS. Paden et al. also present the calibration offset data for the other non-scanning radiometer channels, which show less severe, but significant adjustments in onboard calibration offsets over the years. A one-day unplanned power outage caused a discrete step in one calibration: no explanation is offered, and the post-outage data are satisfactory.
If for this, or other reasons, the spectral calibration shifts as a function of time, the on-board calibration devices have only limited capability in detecting such changes, since the calibration source, the transfer optics, and the detector may all be changing in time.
Several possibilities exist to help alleviate questions of long-term, in-flight calibrations, but they are not routinely applied. Slater et al. (1987) describe an in-flight radiometric calibration of the Thematic Mapper (TM) on Landsat 5 after significant calibration variations had been noted on the TM and Coastal Zone Color Scanner (CZCS) on Nimbus-7. Their method was to examine reflectance from White Sands, chosen because it is a flat, extended area, of highly uniform reflectance in the visible and near-infrared; it is close to a Lambertian reflector; and it has low atmospheric aerosol loading and high probability of clear weather. A second calibration was made by comparing the satellite data with those obtained on a helicopter flying over the same scene at 3000 m MSL. It is noted that "The value of a precise, independent second method, which can be used simultaneously with the first, is that when the two agree to within their error budgets, we can become more confident that their precision represents an absolute accuracy."
Recognition of the fact that one cannot ever be fully confident even of instruments with on-satellite calibration systems leads to the need for routine implementation of readily verifiable, independent cross-calibrations. A high-flying aircraft with instruments overlapping the satellite view probably provides the best cross-calibration. Such aircraft data are especially useful in that most of the atmospheric absorption, polarization changes, cloud and aerosol effects are measured directly, thus reducing the need for further extrapolation to the satellite. Such flights have already been reported (e.g., Hovis et al. 1985), and proposals to perform other such measurements have been made. The effective subtraction of atmospheric effects is important here since the radiance of light scattered by atmospheric gas varies approximately as the inverse fourth power of wavelength (i.e., Rayleigh scattering). While the corrections are clearly most important to instruments with high resolution, the effect of the atmosphere also impacts instruments that integrate over rather broad-band channels. An example of a computation of the atmospheric correction in the 8-14-mm band may be found in Byrnes and Schott (1986). Since the subject of radiation transport to the sensor is covered elsewhere in this report, no further details will be presented here.
