Dear James,

Here are some pointers (but no immediate answer) to deal with your question:

- Landsat 8 has had some problems with the thermal channels since launch. You may want to check the following web site to start with:

https://www.researchgate.net/post/What_is_the_temperature_saturation_range_for_Landsat_8_Thermal_Infrared_Sensor_TIRS

- All detectors have an effective range of usability. Given the importance of monitoring fires, volcanoes and other hot targets, I doubt the channel would saturate over temperatures as low as 40-80 C, unless there is a serious instrument malfunction.

- It is important to distinguish between the radiative or skin temperature and the bulk temperature of an object, and to remain alert about the scaling issue: A single pixel of that instrument actually collects radiation from some area on the ground (more correctly from a volume bound by the ground but including an atmospheric column), where there can be quite a high spatial variability in temperature, so the value retrieved from remote sensing is likely to be rather smaller than the hottest object in the pixel, if there is a lot of spatial variability.

- To gain confidence in the temperature retrievals, you may want to look at another scene, over a target for which surface measurements are available at the time of satellite overpass. In fact, if that can be done for multiple such sites, you might build your own calibration curve (relation between the actual temperature measured in situ and the value retrieved from the satellite measurements), and then apply that to your measurements in East Africa.

- Last but not least, you might want to contact directly colleagues who are specialist in this field, including:

* José A. Sobrino

* Darren Ghent

* Fred Prata

* Massimo Menenti

* Li Jia

I hope these leads may be helpful. Cheers, Michel.

Dear James,

Here are some pointers (but no immediate answer) to deal with your question:

- Landsat 8 has had some problems with the thermal channels since launch. You may want to check the following web site to start with:

https://www.researchgate.net/post/What_is_the_temperature_saturation_range_for_Landsat_8_Thermal_Infrared_Sensor_TIRS

- All detectors have an effective range of usability. Given the importance of monitoring fires, volcanoes and other hot targets, I doubt the channel would saturate over temperatures as low as 40-80 C, unless there is a serious instrument malfunction.

- It is important to distinguish between the radiative or skin temperature and the bulk temperature of an object, and to remain alert about the scaling issue: A single pixel of that instrument actually collects radiation from some area on the ground (more correctly from a volume bound by the ground but including an atmospheric column), where there can be quite a high spatial variability in temperature, so the value retrieved from remote sensing is likely to be rather smaller than the hottest object in the pixel, if there is a lot of spatial variability.

- To gain confidence in the temperature retrievals, you may want to look at another scene, over a target for which surface measurements are available at the time of satellite overpass. In fact, if that can be done for multiple such sites, you might build your own calibration curve (relation between the actual temperature measured in situ and the value retrieved from the satellite measurements), and then apply that to your measurements in East Africa.

- Last but not least, you might want to contact directly colleagues who are specialist in this field, including:

* José A. Sobrino

* Darren Ghent

* Fred Prata

* Massimo Menenti

* Li Jia

I hope these leads may be helpful. Cheers, Michel.

Dear Sir Michel,

Thank you for your comments. Your answers are always thought-provoking.

I've a few comments to add:

•  I'm aware of L8 TIRS issues. Did you mean this link http://landsat.usgs.gov/calibration_notices.php   Therefore, I only used Band 10 (10.60 µm - 11.19 µm) for LST.
•  The whole Landsat scene which I worked on gave an effective range of -11.15°C to 72.2°C. Therefore, the lower threshold seems to be still a reasonable figure because there are clouds in the image and they may have caused these anomalous low temperatures. The upper bound of LST from the image at 72.2°C is doubtful as Erta Ale lava lake may have temperatures over 1000°C. However, I agree with you that temperature range of 80°C is too low as TIRS may have been used for monitoring fire and hot targets. I will search for papers where TIRS is used for hot target monitoring. I am aware of MODIS which is used for a number of hot target monitoring with greater wavelength range from 3.6 to 14.3 µm (Band 20 to Band 36) at 1 Km resolution. But do you think there is a fundamental difference if I use, for example Band 21 (3.92 µm - 3.98 µm) from MODIS compared to Band 10 (10.60 µm - 11.19 µm) of Landsat 8 to calculate Land Surface Temperature? Is it possible to calculate the saturation range of temperatures which can be measured by these thermal sensors based on Wien's law, Planck's law & Stefan-Boltzmann's law or is it also dependent on the detector electronics range mounted on the thermal sensors or both?
• I completely understand that with LST I'm only measuring the first few centimeters or even millimeters of the crust. Therefore, the averaged out temperature over a 100 * 100 m pixel of TIRS which may have a varied sub-pixel temperature source is surely something which I'm aware of. Especially when these emissions have to travel through a column of atmosphere to reach the sensor. It seems atmospheric correction might be a crucial pre-processing step before LST is computed.
•  It is a good idea to look at another scene with available in-situ measurements at the time of satellite overpass. Thanks for the suggestion.  Are you aware of any possible sites which have continuous surface temperature monitoring systems installed anywhere in the world which can be accessed freely? Preferably, around Geothermal Sites. I would like to check if Yellowstone has any? I thought it might be helpful to work first on already installed measuring systems on the ground before going ahead with an own measuring setup.

In the past I've written to Dr. Sobrino but did not get any reply. May be I can try one last time. I will also contact others. Thanks for sharing the names of subject matter experts.

Howdy James,

Here some news from USGS on the Landsat 8 Thermal sensor.

- August 22, 2013 - Landsat 8 Thermal Infrared Sensor (TIRS) Calibration Notice Discrepancies have been noted between calibrated Landsat 8 TIRS Bands 10 and 11 data, and surface-water temperature measurements collected to validate thermal band calibration. Surface-water temperatures derived from TIRS data, after correction for atmospheric transmission and emissivity, are warmer than measured surface-water temperatures by 2K or more. These discrepancies also may not be consistent across the focal plane. This indicates a possible bias or other error in TIRS calibration that places the calibration uncertainty beyond the specified performance of 2 percent. Users are cautioned to be aware of potential impacts to their analyses and results. The calibration team continues to analyze TIRS data and compare results to surface-water temperature measurements to discover the source of the discrepancy. Updates to TIRS calibration coefficients will be incorporated into Landsat 8 data processing as soon as the discrepancy is sufficiently understood. Details will be provided on this website as they become available.

October 25, 2013 - Upcoming change in Landsat 8 Radiometric CalibrationIn the near future, calibration changes will be made that will affect the Thermal Infrared Sensor (TIRS) Band 10 and all of the Operational Land Imager (OLI) bands onboard Landsat 8. The calibration for TIRS Band 10 will be lowered by a constant 0.32 W/(m2 sr µm) for every TIRS Band 10 pixel. This adjustment is being made due to significant discrepancies as compared to surface water temperature measurements. Studies are ongoing to better characterize the source of the calibration errors, and if possible, provide a more accurate scene-dependent correction. No adjustment will be made to TIRS Band 11, as indications are that its calibration is further off and more variable. Until indicated otherwise, users should work with TIRS Band 10 data as a single spectral band (like Landsat 7 Enhanced Thematic Mapper Plus (ETM+)) and should not attempt a split-window correction using both TIRS Bands 10 and 11. Prior to this reprocessing effort, users can subtract 0.32 W/(m2 sr µm) from the TIRS Band 10 Top-of-Atmosphere (TOA) radiance data to improve the accuracy of their current image products or to avoid downloading a new image product after reprocessing occurs. Once a more accurate scene-dependent correction is determined, a second purge and reprocessing will take place.

The OLI radiance-to-reflectance conversion coefficients will be adjusted for the cirrus band (Band 9) to account for on-orbit performance. The prelaunch derived coefficients were calculated using heliostat measurements, which were expected to be in error because little sunlight reaches the ground at these wavelengths. This adjustment changes the reflectance by about 7 percent in the cirrus band. Additionally, the precision of the other spectral bands’ radiance-to-reflectance conversion coefficients will be increased, changing the reflectance by up to 0.3 percent.

The relative gains of single detectors on the edges of each OLI Sensor Chip Assembly (SCA) will be updated to correct slight striping that is typically not visible. This update will affect all OLI spectral bands.

January 6, 2014 - Landsat 8 Reprocessing Details Several calibration parameter updates will be implemented as part of the Landsat 8 data reprocessing on February 3, 2014 that will result in improved product quality. This reprocessing campaign includes all previously implemented calibration parameter updates that have been implemented since launch.  These changes are described in more detail below.

The OLI detector linearization correction coefficients have been refined for all spectral bands to decrease striping in imagery over dark uniform areas.  The absolute radiometric accuracy is not significantly affected. The OLI radiance conversion coefficients are also improved slightly by correcting for a slight error in pre-launch calibration that resulted in product radiance values with as much as a +2% error.  A slight improvement to the absolute accuracy of the reflectance values, mainly the cirrus band, will also be implemented.  The relative gains will also be updated to reduce striping in the cirrus band, but this will not affect the absolute radiometric accuracy. The previously mentioned TIRS temperature offset will also be implemented on February 3, 2014.  These offsets remove 0.29 W/m^2/sr/um from TIRS band 10 and 0.51 W/m^2/sr/um from TIRS band 11, relative to products processed prior to February 3, 2014.  These offsets will remove an average error introduced by stray light coming from outside the TIRS field of view and will improve the data accuracy for typical growing season data (10° to 30° C) where the surrounding areas are similar in temperature. The RMS variability of this correction is 0.12 W/m^2/sr/um for TIRS band 10 and but remains greater at 0.2 W/m^2/sr/um for TIRS band 11.  Due to the larger calibration uncertainty associated with TIRS band 11, it is recommended that users refrain from relying on band 11 data in quantitative analysis of the TIRS data, such as the use of split window techniques for atmospheric correction and retrieval of surface temperature values.

Based on the analysis of on-orbit data, the alignment between the TIRS instrument and the OLI instrument will be adjusted according to these time periods:

Effective Date Range; Change in TIRS alignment; Launch-date -> 09/19/2013; 12 microradians; 09/20/2013-09/30/2013; 25 microradians; 10/01/2013-11/27/2013;10 microradians

There is no change to the TIRS alignment in products after the calibration update on November 27, 2013, which already accounted for this adjustment.

January 23, 2014 - Immediate TIRS Data Adjustment. An incorrect bias adjustment for Thermal Infrared Sensor (TIRS) data was discovered in the latest release of the Landsat 8 processing system, which has caused a significant error in the TIRS products produced from Landsat 8 scenes acquired from January 14, 2014 to January 23, 2014. (The Operational Land Imager (OLI) bands are not affected.) All Landsat 8 scenes processed between January 14 and January 23 will be purged from the online cache and will become available for on demand order processing. An estimated 3 percent absolute radiometric error is observed and the errors do cause significant striping and banding in bands 10 and 11.

January 29, 2014 – Landsat 8 Reprocessing to Begin February 3, 2014 On February 3, 2014, the entire Landsat 8 archive will be cleared from the online cache and reprocessed to take advantage of calibration improvements identified during its first year of operation. All Landsat 8 scenes will be removed from the online cache at this time and these data will be reprocessed starting with the most recent acquisitions and proceeding back to the beginning of the mission. Data will then become available for download. Scenes waiting to be reprocessed will also be available for on-demand product orders. Reprocessing is expected to take approximately 50 days.

The Operational Land Imager (OLI) and Thermal Infrared Sensor (TIRS) instruments onboard the satellite have proven to be outstanding sensors; however, all sensors have required radiometric and geometric refinements to ensure good calibration and data continuity. This reprocessing campaign includes all previously implemented calibration parameter updates implemented since launch. These changes are described in more detail below.

Changes to Operational Land Imager (OLI) data

All spectral bands of the OLI sensor have shown minor striping, noticeable primarily in dark uniform areas; the striping is worse in coastal aerosol (CA) and short wave infrared (SWIR) bands. Striping has been reduced by making refinements to the detector linearization correction in the Radiometric Look-Up Table (RLUT). The relative radiometric accuracy is improved by this change; however, the absolute accuracy remains unchanged.

An additional change to the relative gains in the Calibration Parameter File (CPF) will be updated to further reduce striping in the cirrus band. This change will not affect the absolute radiometric accuracy.

The OLI radiometric calibration is being updated to reflect better knowledge of the prelaunch test data, on which the original calibration was based. The reflectance will change at most 0.8 percent, except for the cirrus Band 9, which will change by 7.3 percent (see table). This calibration update will make the reflectance appear brighter than when these data were processed with the previous radiometric calibration. These changes apply only to data converted to reflectance. An improvement to the absolute accuracy of the radiance values, around 2 percent for most bands, will also be implemented.

OLI Bands; Radiance Change (%); Reflectance Change (%)

1 (Coastal Aerosol) -2.1; 0.8

2 (Blue) -2.0; 0.4

3 (Green) -1.7; 0.0

4 (Red) -1.9; 0.3

5 (NIR) -1.4; 0.0

6 (SWIR1) -2.2; 0.4

7 (SWIR2) 1.5; 0.5

8 (Pan) -1.6; 0.2

9 (Cirrus) 0.7; 7.3

Changes to Thermal Infrared Sensor (TIRS) data

The TIRS temperature offsets noted on November 14, 2013 Calibration Notice on http://landsat.usgs.gov/calibration_notices.php is a primary driver for this reprocessing effort. These offsets remove 0.29 W/m2/sr/um (~2.1 K) from Band 10 and 0.51 W/m2/sr/um (~4.4 K) from band 11, relative to products processed prior to February 3, 2014. The offsets represent an average error introduced by stray light coming from outside the TIRS field of view and will improve the data accuracy for typical growing season data (10° to 30° C) where the surrounding areas are similar in temperature. The Root Mean Square (RMS) variability of this correction is 0.12 W/m2/sr/um (~0.8 K) for Band 10 and remains greater at 0.2 W/m2/sr/um (~1.75 K) for Band 11. Due to the larger calibration uncertainty associated with Band 11, it is recommended that users refrain from relying on Band 11 data in quantitative analysis of the TIRS data, such as the use of split window techniques for atmospheric correction and retrieval of surface temperature values.

TIRS offset changes

TIRS Bands Radiance Offset [W/m2/sr/um]; Temperature Offset [K @ 300K]

10 -0.29 +/- 0.12; -2.1 +/- 0.8

11 -0.51 +/- 0.2;   -4.4 +/- 1.75

Here ends the log of calibration notices of the Landsat OLI sensor. Evidently you can infer the temperature saturation range from these log data. And don't forget the info Michel Verstraete already provided.

Cheers and good luck with OLI.

Frank

https://www.researchgate.net/publication/285574324_Assessing_spatio-temporal_health_of_forest_cover_using_forest_canopy_density_model_and_forest_fragmentation_approach_in_Sundarban_reserve_forest_India

Article Assessing spatio-temporal health of forest cover using fores...

https://www.researchgate.net/publication/285574324_Assessing_spatio-temporal_health_of_forest_cover_using_forest_canopy_density_model_and_forest_fragmentation_approach_in_Sundarban_reserve_forest_India

Article Assessing spatio-temporal health of forest cover using fores...

Short answer: B10 upto ~400 K, B11 upto 370 K. These are according to pre-flight calibration (http://www.mdpi.com/2072-4292/7/1/1135/htm).

Longer answer: the sensitivity to sun angle and atmospheric water vapour seems quite high and requires quite complex corrections. High surface temperatures and high emissivity can lead to errors on the order -1.5 to -2 Celcius according to one study (here I extrapolate to LST > 50C) using a split-window approach. I assume this means the simple single-window estimate derived from the USGS is even more sensitive to errors.

Dear James,

Landsat TIR band only could detect maximum around 94-100 degrees celcius of the temperature of the active lava even in the field it reach 1000 degrees, This due to mixed pixel surrounding and also background temperature. I try the dual band method in order to extract hot component temperature from the hot target pixel, this the result of my research https://www.researchgate.net/publication/310952444_Analyzing_Active_Lava_Flow_Thermal_Using_Space_Based_Remote_Sensing_SWIR-TIR_Bands_During_2014-2015_Holuhraun_Eruption_Iceland

Combine SWIR and TIR band in Landsat to give insight of temperature.

It could help.

Kind regards,

Aufar

Conference Paper Analyzing Active Lava Flow Thermal Using Space Based Remote ...

 Dear Muhammad Aufaristama,

Thanks a lot for your precise answer to my question. As you can see, I asked the question in 2015 when I was just beginning my MSc thesis from the same Institute where you did your MSc. Strangely we never met, I wonder why? 

As far as the question is concerned, I realized last year that the answer was a simple math calculation. Theoretically Landsat 8 TIRS Band 10 (center wavelength 10.8 µm) & Band 11 (center wavelength 12 µm) can measure a maximum radiance of 22.00180 (W/m2/sr/µm) as indicated in the metadata text file (*_MTL.txt) of all Landsat 8 scenes that I analysed. Further, by using the Inverse Planck function maximum surface temperatures of ~ 94°C & ~ 110 °C can be retrieved for both Landsat 8 TIRS bands respectively. Ideally this may not be the best to study a feature which itself has a surface temperature of more than 1000 °C on the ground. Therefore, using Landsat 8 bands such as SWIR which can measure higher radiance is a better alternative in addition to using TIRS bands to get a figure closer to reality. Then you certainly have the issue of mixed pixel like you indicated in your comments. It is a complex problem but a lot of research has been already done on the subject as you may be well aware.

In any case, it is good to receive your comments. Very Interesting research results. I would like to follow the kind of research that you are doing. Are you continuing your PhD in Iceland after MSc in ITC? 

Dear James,

I dont know I either not find the answer there since in ITC mostly not dealing with the very high temperature area. Dealing with very high thermal anomaly in medium pixel resolution such in Landsat is common problem, due to mixed pixel problem of the area. I will give you example in lava even the temperature reach 800-1000 degrees in the field, in landsat will get around 94-100 and in SWIR band 1.6 µm is around 450 degress, thats why dual band method use to solve hot component temperature in that pixel by combining two bands until ratio between SWIR and TIR equal to 1, then this method also solve the thermal fraction. I will show you in case eruption in Iceland in Landsat 8, 6 September 2014 (attached). This method successfully derived the thermal component of lava with maximum around 940 degrees in the lava channel and vent (which is the hottest). We also need make an assumption of ambient temperature within the are, I will try using this method in eruption of erta ale.

Yes I'm continue PhD in Iceland after finish from ITC, I also done my MSc here so I got dual degree from ITC and also Iceland.

Kind regards,

Aufar

Muhammad, I am very glad to hear about your research progress. I wish you continued success in your PhD. I will get back to you soon to discuss on the topic when I have the time.

The LST equation is:

📷 (1)

Where BT represent Brightness Temperature, λBT Wavelength of the band, Ɛλ/LSE: Land Surface Emissivity and ρ is equal to 1438.

brightness temperature is:

📷 (2)

Where K1 and K2 represent thermal constants of the thermal bands in this study used only band 10. Therefore, Landsat-8 provides some constants to estimate LST such as thermal constant and rescaling factor; which is found it in metadata file of Landsat satellite images

the Top Of Atmospheric spectral radiance (TOA) is acquired. Equation (3) used that:

📷 (3)

Where ML refer to the band specific multiplicative rescaling factor (radiance_mult_band_10), AL represent Band specific additive rescaling factor (radiance_add_band_10) and Qcal as band 10 image

The formula of land surface emissivity is used equation (5).

📷 (4)

📷 (5)

Where ɛ = Land Surface Emissivity ,ɛSλ= soil emissivity, ɛVλ = vegetation emissivity

PV = proportion of vegetation and Cλ = surface roughness taken as a constant value of 0.009.