THE ABSOLUTE RADIOMETRIC CALIBRATION OF SPACE-BASED SENSORS.

Abstract

The need for absolute radiometric calibration of space-based sensors will continue to increase as new generations of space sensors are developed. A reflectance-based in-flight calibration procedure is used to determine the radiance reaching the entrance pupil of the sensor. This procedure uses ground-based measurements coupled with a radiative transfer code to characterize the effects the atmosphere has on the signal reaching the sensor. The computed radiance is compared to the digital count output of the sensor associated with the image of a test site. This provides an update to the preflight calibration of the system and a check on the on-board internal calibrator. This calibration procedure was used to perform a series of five calibrations of the Landsat-5 Thematic Mapper (TM). For the 12 measurements made in TM bands 1-3, the RMS variation from the mean as a percentage of the mean is (+OR-) 1.9%, and for measurements in the IR, TM bands 4,5, and 7, the value is (+OR-) 3.4%. The RMS variation for all 23 measurements is (+OR-) 2.8%. The absolute calibration techniques were put to another test with a series of three calibration of the SPOT-1 High Resolution Visible, (HRV), sensors. The ratio, HRV-2/HRV-1, of absolute calibration coefficients compared very well with ratios of histogrammed data obtained when the cameras simultaneously imaged the same ground site. Bands PA, B1 and B3 agreed to within 3%, while band B2 showed a 7% difference. The procedure for performing a satellite calibration was then used to demonstrate how a calibrated satellite sensor can be used to quantitatively evaluate surface reflectance over a wide range of surface features. Predicted reflectance factors were compared to values obtained from aircraft-based radiometer data. This procedure was applied on four dates with two different surface conditions per date. A strong correlation, R('2) = .996, was shown between reflectance values determined from satellite imagery and low-flying aircraft data. Of the 32 predicted reflectance values only six had a difference greater than 0.01. A mean difference of .0007 was obtained for the 32 cases. In addition, a procedure had to be developed to obtain uncorrected digital counts from processed satellite imagery.