Landsat-7 and Landsat-5 Thermal Band Calibration Updates

摘要

Landsat-7 ETM+, launched in April 1999, and Landsat-5 TM, launched in 1984, both have a single thermal band. Both instruments' thermal band calibrations have been updated: ETM+ in 2001 for a pre-launch calibration error and TM in 2007 for data acquired since the current era of vicarious calibration has been in place (1999). This year, the vicarious calibration teams have made regular collects of very hot targets, and have been able to make use of archived buoy data to extend the TM calibration back in time. The new data has made it clear that both instruments require slight adjustments in their thermal calibration coefficients. These new coefficients will be generated and put into the operational processing system to remove the calibration errors. The JPL vicarious calibration team has long operated automated buoys on Lake Tahoe for the purpose of vicarious calibration. This year, the Salton Sea station came on line. Salton Sea, located in southern California, gets far hotter than Lake Tahoe. Vicarious calibration results of the Salton Sea for both instruments added to the understanding of a small gain error that the Tahoe data had suggested. With the Salton Sea data, an ETM+ gain error became statistically significant. Though it causes errors as large as 1.2K at high temperatures (35C), at more usual earth temperatures (4- 20C) the calibration error is within the noise of the calibration methodology (+/-0.6K). With an ETM+ calibration update, the RMSE will be +/-0.6K for all temperatures. The RIT vicarious calibration team mined the archive of the NOAA National Data Buoy Center for sites on the Great Lakes and in the Atlantic Ocean where buoy data was regularly available between 1984 and 2007 and there were radiosonde data within close proximity to allow for atmospheric correction. Four Landsat scenes were chosen and the study made use of almost 200 separate acquisitions of these scenes. The technique was first tested with Landsat-7 data, and was shown to be as reliable as the standard RIT vicarious calibration methods. The TM calibration was largely unmonitored for most if it's lifetime. The buoy results suggest a lifetime error in gain and a change in the offset after 1997. The 2007 TM calibration update accounted for much of the offset error but was only implemented for data acquired after 1999. With the additional buoy data, the calibration will be corrected for the earlier time period and the result will be a consistent calibration to within +/-0.6K for the lifetime of the TM.