HI Martins,

All the steps you have mentioned are correct while the 3rd step is more end-user dependent (i.e. some researchers do while some don't). I think you don't need to calibrate the obtained Thermal image. As it is known that the Landsat TM and ETM+ Band-6 instrument calibration uncertainties in Top-of-Atmosphere temperature are

±1.0 and ±0.6 Kelvin, respectively.

It is necessary to perform the atmospheric correction of thermal band, I would suggest to read the attached articles. Further there is a web-based tool for the atmospheric correction parameters calculation (please see the attached link).

Hope this helps!

Dear Martins

It will be a Black Body Temperature image and not corrected for either atmospheric absorption or emissivity so after atmospheric correction the actual values would be increased by up to 5 or 6 degrees (or as suggested by Dr. Majid above)

Then, you need to do emissivity correction also. After emissivity correction the values will also be increased - except for water which has emissivity of almost 1, vegetation may increase a bit (around 2-3 deg) and man-made surfaces by more (around4-6 deg).

Also, have a look on this tutorial:

https://fromgistors.blogspot.com/2014/01/estimation-of-land-surface-temperature.html

Best Wishes,

This steps can be found in this slide deck on slide 19. Hope this helps.

https://www.researchgate.net/publication/301797360_Analyzing_land_surface_temperature_distribution_in_response_to_land_useland_cover_change_using_split_window_algorithm_and_spectral_radiance_model_in_Sundarban_Biosphere_Reserve_India

Article Analyzing land surface temperature distribution in response ...

Hi Patrick, 

I assume your aim is to derive LST. If you check the Barsi et al.(2003) paper attached with the answer by Majid they clearly state that atmospheric correction is crucial for deriving any landsat based surface temperature products. Currently USGS does not provide any atmospherically corrected thermal bands of landsat, therefore you can use the web based tool by Barsi et al. (2003). Once you derive the atmospheric transmissivity, upwelling and downwelling radiances from the web based tool you can use the radiative tranfer equation to derive the surface temperature. For the whole LST model to work, you would also need emissivity values which is necessary for LST, as mentioned by Sawaid too. One of the easiest method is to consider NDVI derived emissivity based on Sobrino et al. (2004), you can download atmospherically corrected (SR) products for NIR and Red bands in order to derive LST from USGS. But the method used for deriving emissivity depends very much on the accuracy required, objective of research, levels of acceptable uncertainties etc.

http://atmcorr.gsfc.nasa.gov/