Landsat-5 Thematic Mapper (TM) has been imaging the Earth since March 1984 and Landsat-7 Enhanced Thematic
Mapper Plus (ETM+) was added to the series of Landsat instruments in April 1999. The stability and calibration of the
ETM+ has been monitored extensively since launch. Though not monitored for many years, TM now has a similar
system in place to monitor stability and calibration. University teams have been evaluating the on-board calibration of
the instruments through ground-based measurements since 1999. This paper considers the calibration efforts for the
thermal band, Band 6, of both the Landsat-5 and Landsat-7 instruments.
Initial calibration results for the Landsat-7 ETM+ thermal band found a bias error which was corrected through changes
in the processing systems in late 2000. Recent results are suggesting a calibration error in gain, apparent with high
temperature targets. For typical earth temperature targets, from about 5-20C, the gain error is small enough to be within
the noise of the vicarious calibration process. However, for very high temperature targets (greater then 35C), Landsat-7
appears to be predicting several degrees too low. Questions remain on whether the change happened suddenly or is
varying slowly, so the team will wait for another collection season before making any updates to the calibration.
The calibration efforts for Landsat-5 TM considers only data collected since 1999, though there are efforts underway to
extend the calibration history prior to the Landsat-7 launch. The latest data suggests that the Landsat-5 thermal band has
a bias error of about 0.65K too low since 1999. Studies early in the life of Landsat-5 show that the instrument was
calibrated within the error of the calibration process. It is impossible to tell, at this point, when or how the change in
bias may have occurred. A correction will be calculated and implemented in the US processing system in 2006 for data
acquired since April 1999.
Since shortly after launch the radiometric performance of band 6 of the ETM+ instrument on Landsat 7 has been evaluated using vicarious calibration techniques for both land and water targets. This evaluation indicates the radiometric performance of band 6 has been both highly stable and accurate. Over a range corresponding to a factor of two in radiance (5 to 55 C in kinetic temperature terms) the difference between the in-situ derived radiance and the image derived radiance is on average 0.5% or less. Water targets are the easiest to use but are limited to the temperature range from 0 to about 32 C. Land targets can reach 55 C or more but are far less spatially homogeneous than water targets with respect to both local surface temperature and spectral emissivity. The techniques used and the results are described.
An atmospheric correction tool has been developed on a public access web site for the thermal band of the Landsat-5 and Landsat-7 sensors. The Atmospheric Correction Parameter Calculator uses the National Centers for Environmental Prediction (NCEP) modeled atmospheric global profiles interpolated to a particular date, time and location as input. Using MODTRAN radiative transfer code and a suite of integration algorithms, the site-specific atmospheric transmission, and upwelling and downwelling radiances are derived. These calculated parameters can be applied to single band thermal imagery from Landsat-5 Thematic Mapper (TM) or Landsat-7 Enhanced Thematic Mapper Plus (ETM+) to infer an at-surface kinetic temperature for every pixel in the scene.
The derivation of the correction parameters is similar to the methods used by the independent Landsat calibration validation teams at NASA/Jet Propulsion Laboratory and at Rochester Institute of Technology. This paper presents a validation of the Atmospheric Correction Parameter Calculator by comparing the top-of-atmosphere temperatures predicted by the two teams to those predicted by the Calculator. Initial comparisons between the predicted temperatures showed a systematic error of greater then 1.5K in the Calculator results. Modifications to the software have reduced the bias to less then 0.5 ± 0.8K. Though not expected to perform quite as well globally, the tool provides a single integrated method of calculating atmospheric transmission and upwelling and downwelling radiances that have historically been difficult to derive. Even with the uncertainties in the NCEP model, it is expected that the Calculator should predict atmospheric parameters that allow apparent surface temperatures to be derived within ±2K globally, where the surface emissivity is known and the atmosphere is relatively clear. The Calculator is available at http://atmcorr.gsfc.nasa.gov.
Launched in April 1999, the Landsat-7 ETM+ instrument is in its fourth year of operation. The quality of the acquired calibrated imagery continues to be high, especially with respect to its three most important radiometric performance parameters: reflective band instrument stability to better than ±1%, reflective band absolute calibration to better than ±5%, and thermal band absolute calibration to better than ± 0.6 K. The ETM+ instrument has been the most stable of any of the Landsat instruments, in both the reflective and thermal channels. To date, the best on-board calibration source for the reflective bands has been the Full Aperture Solar Calibrator, which has indicated changes of at most -1.8% to -2.0% (95% C.I.) change per year in the ETM+ gain (band 4). However, this change is believed to be caused by changes in the solar diffuser panel, as opposed to a change in the instrument's gain. This belief is based partially on ground observations, which bound the changes in gain in band 4 at -0.7% to +1.5%. Also, ETM+ stability is indicated by the monitoring of desert targets. These image-based results for four Saharan and Arabian sites, for a collection of 35 scenes over the three years since launch, bound the gain change at -0.7% to +0.5% in band 4. Thermal calibration from ground observations revealed an offset error of +0.31 W/m<sup>2</sup> sr um soon after launch. This offset was corrected within the U. S. ground processing system at EROS Data Center on 21-Dec-00, and since then, the band 6 on-board calibration has indicated changes of at most +0.02% to +0.04% (95% C.I.) per year. The latest ground observations have detected no remaining offset error with an RMS error of ± 0.6 K. The stability and absolute calibration of the Landsat-7 ETM+ sensor make it an ideal candidate to be used as a reference source for radiometric cross-calibrating to other land remote sensing satellite systems.
The standard atmospheric correction algorithm for five thermal infrared (TIR) bands of the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) is currently based on radiative transfer computations with global assimilation data on a pixel-by-pixel basis. In the present paper, we verify this algorithm using 100 ASTER scenes globally acquired during the early mission period. In this verification, the max-min difference (MMD) of the water surface emissivity retrieved from each scene is used as an atmospheric correction error index, since the water surface emissivity is well known; if the MMD retrieved is large, an atmospheric correction error also will be possibly large. As the results, the error of the MMD retrieved by the standard atmospheric correction algorithm and a typical temperature/emissivity separation algorithm is shown to be remarkably related with precipitable water vapor, latitude, elevation, and surface temperature. It is also mentioned that the expected error on the MMD retrieved is 0.05 for the precipitable water vapor of 3 cm.
Calibration of the five EOS ASTER instrument emission bands (90 m pixels at surface) is being checked during the operational life of the mission using field measurements simultaneous with the image acquisition. For water targets, radiometers, temperature measuring buoys and local radiosonde atmospheric profiles are used to determine the average water surface kinetic temperature over areas roughly 3 X 3 pixels in size. The in-band surface leaving radiance is then projected through the atmosphere using the MODTRAN radiation transfer code allowing an at sensor radiance comparison. The instrument at sensor radiance is also projected to the water surface allowing a comparison in terms of water surface kinetic temperature. Over the first year of operation, the field measurement derived at sensor radiance agrees with the image derived radiance to better than plus/minus 1% for all five bands indicating both stable and accurate operation.
The advanced spaceborne thermal emission and reflection radiometer (ASTER) is a 14 channel high spatial resolution instrument selected for flight on the EOS AM-1 platform. This instrument has a 60 km pointable cross-track swath and five thermal infrared channels between 8 and 12 micrometers with 90 m spatial resolution. Correction for the effect of atmospheric attenuation and emission will be made using a radiative transfer model and atmospheric parameters either from the EOS AM-1 platform instruments MODIS (moderate- resolution imaging spectroradiometer) and MISR (multi-angle imaging spectroradiometer) or temperature and moisture profiles from global numerical assimilation models. The correction accuracy depends strongly on the accuracy of the atmospheric information used. To provide an objective assessment of the validity of the atmospheric correction in situ measurements of water surfaces under a variety of atmospheric conditions will be used to estimate the surface leaving radiance at the scale of an ASTER pixel. The procedure will use an array of continuously recording temperature buoys to establish the bulk water temperature, broadband radiometers to determine the near surface water temperature gradient and radiosonde and sunphotometer measurements and a radiative transfer model to deduce the sky irradiance. These measurements and the spectral emissivity of the water will be combined with the relative system spectral response to provide an estimate of thermal infrared surface leaving radiance for each ASTER thermal channel. An example of this approach using a multichannel thermal aircraft scanner as a stand in for ASTER is described. It is expected this approach will provide estimates of surface radiance accurate, in temperature terms, to better than 1 K.