Landsat 4-5 Thematic Mapper (TM) Calibration Notices

By Landsat Missions

April 25, 2017 - Changes to Landsat 4 and Landsat 5 Radiometric Calibration for Collection 1 Processing

Beginning in August 2016, several changes were made to both Landsat 4 Thematic Mapper (TM) and Landsat 5 TM calibration to improve the continuity of the Landsat archive. This includes updates to the radiance based absolute calibration of both instruments at specific cross-calibration points, lifetime gain trend adjustments for specific reflective bands and an adjustment of the reflectance calibration to tie it to the Landsat 8 Operational Land Imager (OLI) reflectance calibration (details in the following sections).

Landsat 5

The cross-calibration of Landsat 5 TM with Landsat 7 ETM+ was updated to account for the change to bias removal method (implemented in October 2012). The bias removal method affects the absolute radiometric calibration but the cross-calibration analysis linking TM to ETM+ was not updated in 2012 to account for the change in bias. This changed the calibration gain by as much as 2%.

Also updated were the lifetime gain models for bands 1 and 3. New work with PICS indicated that the previous lifetime gain trend was not accurately representing the degradation of those bands. There was no statistical difference between PICS and the lifetime gain trends in the other reflective bands. These changes result in slightly less radiance/reflectance in the 1990s, by up to about 1%, and a slight increase in radiance after the early 2000s (see Figure 3). The change to the gain in Bands 1 and 3 with this update varies throughout the TM life and is shown in the Figure below.

Landsat 5 TM old and new calibration gains for Bands 1 and 3. Where the blue line is above the red line, the products become darker. (Micijevic et al, 2016)

The reflectance calibration has been updated as well, to tie to the Landsat 8 OLI reflectance calibration. The new calibration transfers the reflectance calibration from ETM+ and replaces the solar irradiance spectrum with ESUN values based on OLI. Users converting Level-1 data to reflectance by applying the REFLECTANCE_MULT and REFLECTANCE_ADD parameters in the metadata file (MTL) will be making use of this updated calibration.

This update changes the reflectance by up to 4% (Table 3). The ESUN (ChKur) was used for the original reflectance calibration and ESUN (OLI) is used for the new reflectance calibration. The %Diff column shows how much the reflectance changes when using the updated calibration.

Percent difference in top-of-atmosphere reflectance between Landsat 5TM Level-1 products — Percent difference in top-of-atmosphere reflectance between Landsat 5 TM Level-1 products generated using the new OLI-based reflectance calibration (ESUN(OLI)) and the radiance calibration converted to reflectance based on ChKur solar spectrum (ESUN(ChKur)). (Micijevic et al, 2016)

Landsat 4

The absolute calibration of Landsat 4 TM is based on cross-calibration with the Landsat 5 TM pseudo-invariant calibration sites (PICS) data. Because the absolute calibration of the Landsat 5 TM was updated, the Landsat 4 TM calibration needed an update, too. The new Landsat 4 gains account for the changes in cross-calibration between Landsat 5 TM and Landsat 7 Enhanced Thematic Mapper Plus (ETM+). Since the Landsat 4 and Landsat 5 TM gains are tied to a single cross-calibration point, the Landsat 4 lifetime gain models were updated by a constant amount.

Band Percent Difference

1 0.35
2 0.90
3 1.11
4 0.70
5 0.20
7 -0.10

The reflectance calibration has been updated as well, to tie it to the Landsat 8 OLI reflectance calibration. The new calibration transfers calibration from Landsat 5 TM and replaces the solar irradiance spectrum with ESUN values based on OLI. Users converting Level-1 data to reflectance by applying the REFLECTANCE_MULT and REFLECTANCE_ADD parameters in the metadata file (MTL) will be making use of the updated calibration.

Percent difference in top-of-atmosphere reflectance between Landsat 4 TM Level-1 products generated using the new OLI-based reflectance calibration (ESUN(OLI)) and the radiance calibration converted to reflectance based on ChKur solar spectrum (ESUN(ChKur)). (Micijevic et al, 2016)

This update changes the reflectance up to 5% (Table 2). The ESUN (ChKur) was used for the original reflectance calibration and ESUN (OLI) is used for the new reflectance calibration. The %Diff column shows how much the reflectance changes when using the updated calibration.

Reference: Esad Micijevic; Md. Obaidul Haque; Nischal Mishra; Radiometric calibration updates to the Landsat collection. Proc. SPIE 9972, Earth Observing Systems XXI, 99720D (September 19, 2016); doi:10.1117/12.2239426. http://proceedings.spiedigitallibrary.org/proceeding.aspx?articleid=2554413

October 1, 2012 - Landsat 5 Thematic Mapper Quarterly Radiometric Calibration Update

In current Landsat 5 TM data processing, detector biases are estimated from shutter data acquired at the end of each scan. Prior to bias extraction, the shutter data (as well as the image data) are corrected for known instrument artifacts: scan-correlated shift and memory effect. Analysis of long term bias trends showed that variations in such estimated detector biases are merely scene content dependent and caused by an inadequate memory effect correction. While that correction is highly successful at typical image brightness levels, it is affected by detector nonlinearity at very low radiance levels causing an overcorrection of bias data. Analysis of bias trends extracted from images acquired during night, which do not exhibit significant transitions between image and shutter brightness levels, thus no memory effects, confirmed that overcorrection and revealed somewhat lower and much more stable Band 1-4 detector bias levels. For Bands 5 and 7, which are not affected by the memory effect artifact, there is no difference between night and daytime bias levels. Remaining variation in scene average detector bias values in all reflective bands is within 0.1 DN over the instrument’s lifetime. Consequently, it was decided to replace the current method of bias subtraction in L5 TM processing for all reflective bands from a scan-by-scan extraction to the use of constant per-detector biases estimated from the lifetime night data trends.

Distribution of bright targets (e.g. clouds) within an image may affect shutter data for forward and reverse scans differently. The application of constant detector biases will improve the visual appearance of L5 TM Bands 1-4 imagery through a reduction of residual banding caused by differences in bias estimated from calibration shutter data for forward and reverse scans.

The relative detector gains for the primary focal plane bands will also be updated to reduce striping caused by small differences in detector-to-detector responses. This update will be completed using lifetime trends of improved bias-subtracted detector responses.

This update will improve radiometric accuracy and significantly reduce or eliminate residual banding and striping in L1T products. Quantitatively, this change will lead to an increase in average brightness of primary focal plane bands; in Bands 1, 3 and 4 approximately 1% and in Band 2 about 0.7%. Depending on the scene content, the relative change in brightness of certain areas within the image could be much larger.

March 1, 2011 - Landsat 4 Thematic Mapper Thermal Band Calibration Update

Effective April 1, 2011, the calibration of the thermal band on the Landsat 4 Thematic Mapper (TM) was updated to correct for an offset error of -0.43 W/m2 sr µm, for all data acquired after April 1, 1986.

Using the network of National Oceanic and Atmospheric Administration (NOAA) buoys deployed in open water around the continental U.S., the vicarious calibration team at the Rochester Institute of Technology was able to determine that the Landsat 4 data acquired between launch and March 31, 1984 were calibrated to within ±0.99K. Data acquired after July 1, 1986 exhibited a calibration error of –0.43 W/m2 sr µm or about 3.1K at 300K. There was no data acquired between April 1984 and July 1986 due to several failures on board Landsat 4, one of which likely led to the calibration error. The thermal band calibration for the TM instrument relies on predetermined calibration coefficients. When the instrument was turned back on in 1986, it operated at much colder temperatures. The calibration error is likely caused by not adapting the calibration coefficients to the new operating temperatures.

The Calibration Parameter Files for the affected date range were updated on April 1, 2011. All Landsat 4 TM data processed after this date are considered calibrated to within ±0.80K. For data processed before April 1, 2011, users can add 0.43 W/m2 sr µm to the radiance image product and consider the data calibrated or reorder the data.

Reference: Schott, J.R., Hook, S.J., Barsi, J.A., Markham, B.L., Miller, J., Padula, F.P., Raqueno, N.G., “Thermal Infrared Radiance Calibration of the Entire Landsat 4, 5 and 7 Archive (1982-2010)”, Remote Sensing of Environment (submitted).

May 1, 2010 - Landsat 5 Thematic Mapper Thermal Band Calibration Update

Effective April 1, 2010, the calibration of the TM thermal band was updated to correct for a lifetime 5.2% gain error and to temporally extend the offset correction implemented in 2007. The gain error, which has existed since launch, causes a temperature dependent error where the instrument predicts too hot at low temperatures and too cold at high temperatures. The offset error had been discovered previously, but in 2007, it was only corrected for data acquired between April 1,1999 and the present, due to lack of supporting data prior to 1999. This update extends the offset correction back to the beginning of 1997.

The effect of this calibration error on the data is shown in the table and figure below. The table presents the estimated error in top-of-atmosphere brightness temperature for sample target temperatures. This error is present in the image data processed by USGS between April 1, 2007 and April 1, 2010. For errors in data processed prior to April 2007, please see the April 2007 - L5 TM Radiometric Calibration section lower on this page.

This update affects the calibration of all Landsat 5 thermal data acquired since launch. For data acquired between launch and Dec 31, 1996, this update changes the gain by 5.2%. For data acquired between April 1, 1999 and March 31, 2010, this update changes the gain by approximately 5.2% and makes a small modification to the offset of about 0.02 W/m² sr µm. For data acquired between January 1, 1997 and March 31, 1999, this update changes the gain by 5.2% and changes the offset by 0.11 W/m² sr µm.

All data processed by the USGS EROS processing system after April 1, 2010 are correctly calibrated.

More Information on the Calibration Error

The vicarious calibration teams, the Rochester Institute of Technology (RIT), and the NASA/Jet Propulsion Laboratory (JPL) have been collecting data from Landsat 5 since the satellite was returned to U.S. Geological Survey (USGS) management in 2001. The calibration of the thermal band had not been rigorously monitored since launch, and the available techniques between 2001 and 2007 only allowed monitoring of the current calibration, not back in time. In 2007, an offset error of 0.092 W/m² sr µm was detected and a correction was made in the processing system, but only for data acquired since 1999 because the teams had no data prior to that date [Barsi, 2007]. Since then, the teams have continued to collect field data but also expanded their methods: JPL added a calibration site on the Salton Sea, which can reach as high as 35C in the summer, and RIT developed a technique to use water temperatures from the National Data Buoy Center archive to extend the vicarious techniques back to launch in 1984. The additional data revealed that there is a lifetime calibration gain error of 5.2% and that the change in the instrument offset actually occurred in early 1997. This calibration update will correct the gain error and extend the correction of the offset error to 1997.

The errors were corrected by adjusting the calibration coefficients in the Calibration Parameter File (CPF). Users should note that with the current distribution system, the downloadable product may not have been processed with the latest CPF. The update was issued with the release of the April 1, 2010 CPFs. Please check the “Product Creation Time” in the metadata file (MTL) that comes with the downloaded product to ensure that the product was processed after April 1, 2010 with the updated calibration.

More details of the update can be found in [Barsi, 2010].

The table below displays error estimates for at-satellite radiances for data processed between April 2007 and April 2010. A positive error indicates the image data were too warm, negative errors indicate the image data were too cool. The calibration correction put in place on April 1, 2010 removes these errors.

Target Temperature	Appoximate Target Radiance [W/m² sr µm]	1984-1986 Calibration Error [C @ 300K]	1997-1999 Calbration Error [C @ 300K]	1999-2010 Calibration Error [C @ 300K]
273K / 0C	6.0	0.68	-0.19	0.87
279K / 6C	6.75	0.38	-0.47	0.56
290K / 17C	8.0	-0.18	-0.97	-0.01
294K / 21C	8.5	-0.39	-1.16	-0.22
306K /33C	10.0	-1.0	-1.71	-0.85

Reference: Barsi, J.A., 2010, Twenty-five years of Landsat Thermal Band Calibration. European Remote Sensing Opportunities: Systems, Sensors and Applications, Proceedings of the IGARSS Symposium held in Honolulu, HI, on 25-30 July 2010, (IEEE Geoscience and Remote Sensing Society).

Landsat-5 Thematic Mapper Thermal Band Calibration Update, Barsi, J. A. Hook, S. J., Schott, J. R., Raqueno, N. G., Markham, B. L., Geoscience and Remote Sensing Letters, IEEE Publication Date: Oct. 2007 Volume: 4, Issue: 4: 552 – 555

Error estimates for at-satellite radiances for Landsat TM data processed between April 2007 and April 2010 — Error estimates for at-satellite radiances for data processed between April 2007 and April 2010. A positive error indicates the image data were too warm, negative errors indicate the image data were too cool.

April 1, 2009 - Landsat 5 TM 'Heritage Format' Bumper Mode Calibration Files Discontinued

As of April 1, 2009, the USGS discontinued the preparation and delivery of any new L5 TM Bumper Mode ("Heritage Format") Calibration Files.The new full L5 TM CPF series (released in May 2007) will continue to be prepared and delivered via previous channels, and is expected to replace the previous bumper mode calibration file format.

The new L5 TM CPF provides a single, robust file containing calibration-related information needed for L5 TM Level-1 processing. Use of the new L5 TM CPF will provide many benefits, including:

Single source of calibration coefficients
Improved geometric accuracy
Improved radiometric accuracy
Enhanced Level-1 product quality
Mission-wide CPF readiness:
- Structurally very similar to L7 ETM+ CPF
- Likely basis for L4 TM CPF and L1-5 MSS CPFs
Additional information on the New L5 TM CPF can be found in the May 2007 Calibration Notice.

To access the new L5 TM CPFs visit the L5 TM CPF Download web page.

July 31, 2008 - Landsat 5 Thematic Mapper Internal Calibration HIstorical Trends

The files in this .zip file contain information that provides daily estimates for the historical radiometric internal calibrator gain (GIC), internal calibrator bias (BIC), processing gain (AlphaIC), and processing offset (BetaIC) coefficients for reflective bands 1-5 and 7 that were used by the National Landsat Archive Production System (NLAPS) in Landsat 5 (L5) Thematic Mapper (TM) product generation before the revised calibration procedure was implemented on May 5, 2003. Each of these files contain nine (3+6) columns. The first three columns are mission time related and represent Day Since Launch (DSL), Decimal year since launch, and the Day of the Year (DOY), respectively. The next six columns (one for each reflective band, 1-5 and 7) list the band average time-dependent calibration coefficients generated from the NLAPS trends.

Additionally, a file provides daily estimates for gain and offset used by NLAPS for the thermal band calibration (B6IC). This file has five columns. As for the reflective bands, the first three columns represent DSL, Decimal year since launch, and DOY, respectively. The remaining two columns represent the gain and offset, respectively.

For the reflective bands, these files can be used to recalibrate the L5 TM data processed before May 5, 2003, or to remove the lamp regression calibration and apply the gain and offset that reflect the current understanding of the system. The band 6 file could potentially be used as a look-up table for gain and offset. The thermal band does not require recalibration, but there could be an occasion that the raw data are available and these tables would give the user the information needed to generate a calibrated product without needing the calibration file.

May 1, 2007 - New Landsat 5 Thematic Mapper Calibration Parameter Files

As of May 2007, the U.S. Geological Survey (USGS) began generating new Landsat 5 (L5) Thematic Mapper (TM) Calibration Parameter Files (CPFs) that may be used for radiometric and geometric processing of L5 TM imagery. The new TM CPF files are structurally very similar to those that are currently generated for Landsat 7 (L7) Enhanced Thematic Mapper Plus (ETM+).

The new TM CPF files contain the same information and values that were previously contained in the L5 TM Lookup Table (LUT), as well as information from the previous bumper mode calibration files. However, the new TM CPF file contains much more detailed (daily) trended values for the bumper mode parameters, and also includes many additional groups and coefficients that may be useful for processing of TM imagery.

Calibration Parameter Files are available to download for the full archive, including historically acquired imagery.

April 2, 2007 - Landsat 5 Thematic Mapper Radiometric Calibration Update

Effective April 2, 2007, updates to the radiometric calibration of Landsat 5 (L5) Thematic Mapper (TM) data processed and distributed by the U.S. Geological Survey (USGS) Earth Resources Observation and Science (EROS) Center are available. The implementation of these processing changes enables radiometrically improved L5 TM data products that are more comparable to Landsat 7 (L7) Enhanced Thematic Mapper Plus (ETM+), and provide the basis for continued long-term studies of the Earth’s land surfaces.

Although this calibration update applies to all archived and future L5 TM data, the major improvements in calibration would be for data acquired during the first eight years of the mission (1984-1991), where the change in the instrument gain values is as much as 15 percent. Additionally, the radiometric scaling coefficients for Bands 1 and 2 have also been changed for approximately the first eight years of the mission. Users need to apply these new coefficients to convert the calibrated data product digital numbers to radiance. The scaling coefficients for the other bands have not changed.

The lifetime gain model that was implemented on May 5, 2003, for the reflective bands (1-5, 7) has been replaced by a new lifetime radiometric calibration curve derived from the instrument’s response to pseudo-invariant desert sites and from cross-calibration with L7 ETM+. Along with the revised reflective band radiometric calibration, an instrument offset correction of 0.092 W/ (m2 sr µm) or about 0.68 K (at 300 K) has also been added to all L5 TM thermal band (Band 6) data acquired since April 1999.

For detailed information and background on the reasons for this change, see the Revised Landsat 5 Thematic Mapper Radiometric Calibration (2007) and the Landsat 5 Thematic Mapper Thermal Band Calibration Update.