Improving Accuracy of Indirect Calibrations

Improving Radiometer Calibration

Future improvement in radiometric calibration requires the development of much more stable lamps or the use of an on-site calibration method based on detectors that have a well-characterized spectroradiometric response. Suitable detectors include very flat spectral-response detectors [(e.g., cryogenic electrical substitution radiometer (ESR)] and "self-calibrating" detectors [e.g., silicon-photodiode light trap (Hoyt and Zalewski 1991; Palmer 1988)]. The basic visible-wavelength calibration problem is that there are no well-characterized (theoretically and practically) radiance standards for the visible wavelengths in the same way blackbody sources are good standards for the infrared. But, there are good detectors for the visible (e.g., ESR). Thus, the practical problem of visible wavelength calibration becomes one of transferring the good characterization of the detector into a good characterization of a source [integrating sphere with lamp(s)].

The ESR method was used to calibrate the ERBE visible-wavelength radiometers to an absolute calibration of 1% (see Figure 3). (Private discussions with S. Carmen, 1991,TRW, Redondo Beach, California.) First, the radiance of a tightly thermostated blackbody source was measured with an ESR to establish the responsivity of the ESR to a known radiance (calculated from the measured temperature of the blackbody). Because the responsivity of the ESR does not depend on the wavelength of the light, it served as a "detector standard" for both infrared and visible radiation. The ESR was then used to measure the radiance from an illuminated integrating sphere that was also viewed by the device under test, the ERBE radiometer. This procedure provided an absolute calibration of the ERBE radiometer through a chain referring back to the temperature reference standard. A similar method might be used to calibrate the airborne reference radiometer for the indirect radiance calibration methods.

The indirect calibration methods for satellite radiometers must account for the spectral response of each of the satellite radiometer channels calibrated. This has been done by 1) selecting bandpass filters that have spectral transmissions as close to those of the satellite radiometer as possible, or by 2) using a scanning spectroradiometer, then convolving the spectroradiometer response with the measured preflight bandpass of the satellite filters. Because fabricating two matched filters is very difficult, the second method is generally more accurate.

Calibration uncertainties of 1% should be achievable for an aircraft spectroradiometer using the ESR transfer method (Private discussions with P. Jarecke, 1991, TRW, Redondo Beach, California.) employing an integrating sphere illuminated with lamp sources that are filtered with well-characterized bandpass filters. Using this procedure, the full spectral range of the radiometer is calibrated one spectral band at a time (each band ~100 nm wide). The accuracy of this technique depends on the integrating sphere's not altering the spectral content of the filtered light source. This is believed to be a good approximation and produces uncertainties comparable to or smaller than others in the calibration chain (i.e., < 0.5%). (Private discussions with P. Jarecke, 1991, TRW, Redondo Beach, California.)

An alternative to using a blackbody as a reference radiance source together with an ESR is to use a different kind of well-characterized detector, a "self-calibrating" detector such as the silicon light trap (Slater, (Private discussions with P. N. Slater, 1991, University of Arizona, Tucson, Arizona.); Hoyt and Zalewski 1991; Palmer 1988). The light trap is composed of three silicon photodetectors arranged in a light trap configuration so that reflected light from one detector is absorbed by another. This detector has good spectral response over the entire visible region (0.39-0.78 mm) and, except for a small region around 0.85 mm, has good responsivity over the entire range: 0.2-1.0 mm (Slater, Private discussions with P. N. Slater, 1991, University of Arizona, Tucson, Arizona.) The accuracy of this method relies on an accurate determination of the quantum efficiency of the light trap, a determination made from theory and measurements (Zalewski, Private discussions with E. Zalewski, 1991, Hughes Danbury Optical Systems, Danbury, Connecticut.) Through the use of the silicon light trap, a transfer radiance calibration source with well- characterized spectral properties can be referenced to the irradiance standard at the British National Physical Laboratory that is based on a cryogenic electrical substitution radiometer (Zalewski, Private discussions with E. Zalewski, 1991, Hughes Danbury Optical Systems, Danbury, Connecticut.) The absolute accuracy of this transfer radiance source is expected to be near 0.02%. This level of accuracy should make possible calibration of aircraft spectroradiometers to 1% (Slater, Private discussions with P. N. Slater, 1991, University of Arizona, Tucson, Arizona.)

Thus, either the ESR-transfer or silicon-light-trap techniques should make it possible to calibrate an aircraft spectroradiometer to an absolute accuracy of 1% in the near future. This is comparable to the other errors in the radiance method, so that the overall uncertainty for TM and SPOT/HRV, which is now ~2.8%, should be improved to ~2.0% (Slater 1991, Private discussions with P. N. Slater, 1991, University of Arizona, Tucson, Arizona.) as shown in Figure 4. TM and Spot/HRV can be used as reference radiometers to calibrate AVHRR (Nianzeng 1991; Teillet et al. 1990; Slater et al. 1987) with a transfer error of ~4% (Slater, Private discussions with P. N. Slater, 1991, University of Arizona, Tucson, Arizona.). Figure 4 shows the expected uncertainties for AVHRR using this method. When the reference radiometer used to calibrate AVHRR is on a high-flying aircraft (Smith et al. 1988; Kastner and Slater 1982) rather than a satellite, the absolute error previously reported is 7.0% (3.5% without radiometer uncertainty, Smith et al. 1988), with an anticipated improvement to 3.2% (Slater, Private discussions with P. N. Slater, 1991, University of Arizona, Tucson, Arizona.), also shown in Figure 4. The improvement comes from reducing the co-registration errors from 3.2% to 1.7% and the spectrometer calibration from 6% to 2.0% (Table 4).

Improving Scene Co-alignment

Scene co-alignment contributes to the uncertainty of the radiance method calibrations using the ER-2 aircraft. Although not reported in the literature, NASA/Goddard has already improved the co-registration algorithm; these improvements have been used to provide an indirect calibration for NOAA-11 (Abel 1991), which will reduce the present uncertainty from 3% to 1-2%.

Improving Aerosol Correction

Improvements in the aerosol correction needed for the reflectance methods are based on measuring the aerosol properties on the calibration day. This has been done by measuring the extinction and scattering of sunlight as a function of solar zenith angle and viewing zenith angle, for viewing directions close to the sun. The complex index of refraction and aerosol size distribution is then obtained by fitting the observations of the radiance from a high-flying helicopter to model predictions (Slater et al. 1987). The accuracy of this method relies heavily on accurate solar attenuation measurements, and work is under way to construct and evaluate an improved spectropolarimeter for making the necessary solar attenuation measurements (Slater, Private discussions with P. N. Slater, 1991, University of Arizona, Tucson, Arizona.). This should help reduce the uncertainty related to aerosols from 3.6% to 1.5% (Slater, Private discussions with P. N. Slater, 1991, University of Arizona, Tucson, Arizona.).