I'm attempting to calculate temperature from Landsat 5 and 8 thermal bands. I've already calculated NDVI, where the NDVI bands were corrected using dark object subtraction.

I see that many people use top of atmosphere reflectance to determine temperature. Would it be incorrect for me to use the DOS bands I calculated to determine temperature, or do I need to go back one step and use TOA reflectance?

Specifically, would it be inappropriate to use DOS radiance for the L lambda term instead of TOA radiance? The equation I'm referring to is the last one on the page I linked.


  • I calculated land surface temperature from Landsat 8 using both DOS and TOA corrections. The TOA temperature was 2.5 degrees higher than the DOS temperature. The histograms for both are about the same, although the TOA STD is 0.03 higher. – user5858 Mar 2 '17 at 8:44

It depends entirely on what the atmospheric conditions are in your area, and on what degree of bias you are prepared to accept. As you are using two dates of imagery this becomes more important.

An obvious solution would be to use the atmospherically corrected data from the USGS: select Landsat CDR on earthexplorer; https://landsat.usgs.gov/landsat-surface-reflectance-high-level-data-products

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.