Calibration of Landsat 5 thermal infrared channel: updated calibration history and assessment of the errors associated with the methodology

The Landsat 5 thermal band lifetime calibration is being updated based on an improved calibration method that uses water temperatures observed by buoys at deep water sites and thermal and radiative transfer models. An uncertainty propagation analysis was constructed to determine the expected uncertainty in temperature (one standard deviation) at the sensor for this historic vicarious calibration process. The historical calibration effort fused environmental data sources that feed a forward modeling vicarious calibration process. The process consists of three major modeling efforts: subsurface temperature to water skin temperature, atmospheric radiative transfer, and sensor noise modeling. Each modeling effort was investigated uniquely, and the results were combined to derive the total process error. The uncertainty propagation results indicate that the historic vicarious calibration process has an expected uncertainty of ±0.5 K. This conclusion is consistent with the observed root mean square error (RMSE) between observed and predicted values using this method. After the instrument calibration was updated, the difference between instrument-derived radiance (observed data spanning a 23 year period) and radiance estimated using the subsurface buoy temperatures was 0.6 K RMSE. This demonstrates that the residual error in the observed calibration results and the expected process uncertainty are essentially comparable. Using the error analysis results, the data from the historical buoy temperature study were combined with data from traditional surface temperature and radiance studies (1999–2000) to generate a lifetime calibration update for the Landsat 5 instrument. The final calibration uses data from the long established thermal calibration sites on the Great Lakes (Erie and Ontario), the Salton Sea, and Lake Tahoe, as well as a number of additional deep water sites where National Oceanic and Atmospheric Administration (NOAA) bouys and atmospheric sounding data provide adequate ground truth for the historical calibration approach. This updated calibration has been implemented in the U.S. Geological Survey (USGS) – National Aeronautics and Space Administration (NASA) processing system. These results indicate that the image data is calibrated to better than 0.67 K (one sigma) over its 25+ year record. While this work rigorously investigated the historic thermal vicarious calibration process for Landsat 5 Thematic Mapper (TM), the approach and the new study sites can be easily extended to the investigation of similar systems.