I downloaded calibrated/corrected Landsat 8 data (Level 2) from ESPA (https://espa.cr.usgs.gov/) to do land cover classification. But there are many corrections and calibrations can be done as listed below and maybe more, such as:

 - Surface Reflectance
 - Top of Atmospheric 
 - sensor zenith and azimuth 
 - solar zenith and azimuth  
 - geometric 
 - radiometric

I am not sure all of them were processed on the scene. What are the must do corrections/calibrations for Land cover classification analysis?

  • As @Dan says, for a single scene is doesn't necessary. If you'll use predefined spectral signatures, you need a good surface reflectance product, otherwise, with TOA reflectance you'll be ok
    – aldo_tapia
    Dec 7, 2017 at 10:14

1 Answer 1


You can perform a land cover classification on a single Landsat scene without performing spectral and radiometric corrections. You will only need to do those corrections if you're trying to apply reference spectra to your classification, performing a classification that covers multiple scenes or performing a classification over a time series of the same scene.

You may need to apply a geometric correction if you find that the image has significant differences when compared to your reference data.

If you do want to apply a correction you will need to reference the image metadata and associated Landsat User Guide. In ArcGIS, the Spatial Analyst - Raster Calculator tool can be used to apply the appropriate correction formula to each band in the image.

  • Yes, I can do the classification based on single scene without cloud, but it is very scene dependent. Thus I am gonna use several scenes, from where all the clouds will be removed/masked.
    – Vandka
    Dec 8, 2017 at 1:39
  • Then yes, you will need to apply the corrections you mentioned. You might be able to get away with just doing geometric, TOA and radiometric...depends on the level of accuracy you need and how much time and effort you're prepared to put into applying all the corrections.
    – Dan
    Dec 8, 2017 at 1:44

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