Radiometric Calibration of Reflective Bands of Landsat 4 Thematic Mapper using Pseudo-Invariant Site Technique

Abstract

Historically, for purposes of calibration, the Landsat 4 Thematic Mapper (TM) sensor has relied upon prelaunch calibration data, internal calibration data and very limited post-launch data. Landsat 4 additionally has used Landsat 5 TM cross-calibration data. The availability of one more independent source, commonly called pseudo-invariant sites, can be used to improve the calibration of the sensor. Pseudo-invariant sites are mainly well characterized desert sites exhibiting the properties of high reflectance, high spatial and spectral uniformity, temporal stability and are nearly Lambertian to decrease BRDF and shadow effects. The successful update of the calibration of the Landsat 5 TM sensor using the pseudo-invariant site approach suggested using this approach to calibrate the Landsat 4 TM sensor. The pseudo-invariant sites selected to characterize the Landsat 4 TM sensor are Middle East sites at Path 166, Row 39 and Path 166, Row 40. Previously, calibration attempts have been primarily based only on the internal calibrator (IC) and prelaunch data. The Landsat 5 internal calibrator was found to be unreliable approximately three years after launch. The Landsat 4 IC data also become questionable especially knowing that the internal calibrators for both the instruments were essentially identical. Thus, it is necessary to characterize the system with another independent and reliable source-in this case pseudo-invariant sites. Unfortunately, due to the unavailability of an adequate number of scenes at the site used for Landsat 5, the Middle East site of Path 166, Row 39 and Path 166, Row 40 were selected. A large drawback of these sites is that they are the locations of the First Gulf War in 1991. This negates the assumption that the surface and atmosphere did not change during the study period. To account for the change, radiometric correction was done using the known Landsat 5 gain signature. A correction factor, which is the ratio of the averages of Landsat 5 pre-gulf war and post-gulf war data sets (at-sensor radiance), was used to scale the post-gulf war gain of Landsat 4. Both the IC data and pseudo-invariant site data suggested quadratic models for bands 1-4 and linear models for bands 5 and 7. The proposed models provide consistent calibration knowledge of Landsat 4 TM sensor with differences less than 6% in band 1, nearly 2% in bands 2, 3 and 4 and within 4% in bands 5 and 7, respectively.