Use Mono Window algorithm or temperature-emissivity algorithm.

If you need further info use Qin et al., 2001 paper very informative. You can also refer to few papers on my profile as well and a dataset for mapping hydrocarbon pollution using LST.

The radiative transfer is widely used and can give you a good result but you need to have the emissivity data to get at surface temperature. Landsat website has the parameters and equations for the different sensors to get at-sensor temperature. There are also a number of papers on this.

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 ...

Here comes some practical advice if somebody would like to use the radiative transfer method.

ToA or brightness temperature you get from here: https://espa.cr.usgs.gov

One possible source to get parameters for atmospheric correction would be:

http://atmcorr.gsfc.nasa.gov

- in ENVI: 1. use GeoTiff and mtl.txt and go to "ENVI Landsat calibration"

2. Apply Atm. Correction

3. Convert Radiance to Kelvin

Or if this data is not in the right format

____________

1. Convert DN to Radiance by

- Gain and Bias method

- Spectral Radiance Scaling Method

2. Atm. Correction

Radiance(atm. corr) =( (Radiance - upwelling Radiance) / (Transmissivity*Emissivity) )-  ((1-emissivity)/emissivity)*Downwelling Radiance)

--> please check equation for typo: From Coll et al 2010

3. Convert Radiance to Kelvin: for that you use the inverse of the Planck Function, and you need the constants K from the metadata,

see: http://landsat.usgs.gov/Landsat8_Using_Product.php