Historic thermal calibration of Landsat 5 TM through an improved physics based approach.

摘要

This investigation is motivated by the current need for a detailed post launch calibration of the Thematic Mapper (TM) thermal band (Band 6), aboard NASA's Landsat 5 spacecraft. The historical calibration spans the period from 1984 to 2007. Through fusion of environmental data sources (i.e. buoy observations, surface observations, and radiosonde observations) a vicarious calibration approach was implemented to construct the complete calibration record of the Landsat 5 TM thermal band. The vicarious calibration process took advantage of the long standing National Data Buoy Center (NDBC) moored buoy fleet to acquire historic ground truth measurements needed over the lifetime of Landsat 5. These measurements were propagated to the sensor through the use of physics based models to establish a predicted at sensor radiance. Through comparison of the predicted at sensor radiance and the actual sensor observed radiance, a calibration metric was established.;Results indicate the Landsat 5 TM thermal band, originally planned for a 3 year mission, has fluctuated only slightly (∼1 K) over the 24+ years in orbit. The calibration curve developed in this study is consistent with previous results from campaigns preformed in 1985 and post 1999. The data indicated that the sensor exhibited a clear gain issue (i.e. over estimates low radiance targets and under estimates high radiance targets) found to be approximately consistent over time. Additionally, an event occurring either prior to or during 1999, caused a discernible fluctuation in sensor performance (i.e. dominant cold bias) for all data post 1999. It is the recommendation of this vicarious calibration campaign that a linear (Dual: slope & intercept) correction be applied to the Landsat 5 data archive. As a result of the correction, the Landsat 5 TM Band 6 is radiometrically calibrated to within +/-0.488 K, in reference to a 300 K blackbody. This result was verified through an extensive error propagation analysis, which found the proposed methodology to have an expected error of 0.454 K. The proposed methodology was also verified by a comparison study to the traditional approach (i.e. non buoy derived ground truth) using the closely monitored and trusted Landsat 7 data calibrated using the traditional approach. The comparison found the two methods were not statistically different, which offered the confidence that this methodology could be applied successfully over the domain of this study. This comparison not only validates the calibration record of Landsat 5, but also demonstrates the utility of the method in future efforts.;This work has demonstrated that a successful historical vicarious calibration campaign can be conducted using exclusively free and easily accessible data. It has been established that the proposed methodology can be implemented to achieve a high level of radiometric integrity, which includes both historic and future efforts, in the calibration of remote thermal infrared systems.