I'm doing a temporal analysis of NDVI and temperature over 17 years, necessitating data from Landsat 5,7, & 8. I'm interested in using level-2 Landsat data available from https://espa.cr.usgs.gov/ordering/new/. I would be able to directly download NDVI as well as brightness temperature, which would save me a lot of time. However, the USGS is explicit that the algorithm used to calculate surface reflectance for Landsat 8 is not the same as the algorithm used for previous satellites.


My question is: are the level-2 products from Landsat-8 comparable on an apples-to-apples basis with Landsat 5 or 7 level-2 products, or does this kind of comparison require some sort of adjustment to the data? I haven't been able to find an answer to this question online.

  • 1
    I asked the USGS this question and got this response. I'm not really sure what to make of this: "They use different algorithms, yes. That doesn't mean the datasets are completely incompatible. It means you may need to provide additional information in your study." – user5858 Mar 14 '17 at 21:00
  • 1
    not much help here, but trying to do some classifications between L8, L7 and L5 images I found out some older images may (not neccessarily) be shifted by a range of kilometers. be sure to check that, it is caused by not using GCPs when calibrating older images (link: landsat.usgs.gov/landsat-processing-details). This info may be related - if you check whether your image was originally L1T or L1GT etc, it is possible your Surface Reflectance derived product has same characteristics - from what I know it can't be repaired, they just eg. didn't have some GCPs back then. – adamczi Apr 3 '17 at 21:13

Yes, you do need to correct Landsat 8 imagery to compare it against Landsat 5 and 7. The data between the three satellites are not directly comparable in their raw form.

This paper compares Landsat 7 to Landsat 8:

Roy, D.p., et al. “Characterization of Landsat-7 to Landsat-8 Reflective Wavelength and Normalized Difference Vegetation Index Continuity.” Remote Sensing of Environment, vol. 185, 2016, pp. 57–70., doi:10.1016/j.rse.2015.12.024.

For instance, from the paper, to compare Landsat 8 NDVI with Landsat 7 NDVI you need to apply the following correction to the Landsat 8 NDVI:

L7NDVI = 0.0029+0.9589*L8NDVI

Further, this paper shows that Landsat 7 NDVI is about 2.5% higher than Landsat 5 NDVI:

Teillet, P M, et al. “Landsat-7 ETM+ and Landsat-5 TM Cross-Calibration Based on Tandem Data Sets.” Remote Sensing of Environment, vol. 78, no. 1-2, 2000, pp. 39–54., doi:10.4095/219729.

  • Careful, this may not be the case for the "analysis ready" products. The USGS applies a dynamic range correction to the imagery to correct for wavelength variation. This has been a common correction since the deployment of ETM+7. I had not seen any published correction for cross calibration of dynamic ranges for OLI. You can use Roy et al (2016) band corrections for ETM but not TM5. I guess you could correct TM5 to ETM then to OLI using these coefficients. I just wish they would have developed their coefficient globally. – Jeffrey Evans Jul 27 '20 at 21:28

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.