This is for thesis work involving mapping of above-ground mangrove forest biomass using Landsat 7 ETM+ imagery.

I am assessing correlations between biomass with vegetative indices, e.g. NDVI, and then mapping biomass. I am using a 04/02/2013 Landsat 7 ETM+ image with Scan Line Corrector off, so I need to gap fill. I am using a 08/03/2013 image for gap fill.

Both images have been atmospherically calibrated before gap fill using DOS_improve. However, the reflectance of the two images still contrast to each other. I do not know why the reflectance is contrasting, or how to better match the reflectance of the 08/03/2013 image to the 04/02/2013 image.

  • Expanded the title a bit and did some formatting and linking. Not really my area, so left the technical alone (reflectance, contrast, DOS_improve). Commented Jun 5, 2013 at 19:38
  • For your dark-object subtraction, are you using the same location in each of your Landsat scenes, or the darkest pixel value in each scene?
    – Cyrus
    Commented Jun 5, 2013 at 21:50
  • I used darkest pixel value in each scene. Commented Jun 6, 2013 at 8:15

1 Answer 1


Changes in vegetation over the month between your scenes could be part of the issue. It is also possible that there is some haze over areas of your scene outside of your dark object location(s), and therefore this haze is not being removed during your atmospheric correction.

Another reason that you see contrast between the two scenes could be due to inconsistent dark-object subtractions (DOS) made during your atmospheric correction. This is very likely to be the case if you are using the darkest pixel in each scene rather than the same location. The reason is that the darkest pixel in one scene might be a shadow from a cloud over a forested area, while the darkest pixel in the second scene might be a cloud shadow over bare soil. This difference, even though the values are small, will be very apparent when you try perform your fill. In fact, you might find that there is less contrast when you simply fill with NDVI calculated from reflectance - without any atmospheric correction.

If possible, you might try a different method for your atmospheric correction such as FLAASH. If for some reason you can't use a different method of atmospheric correction, try to select the same dark-object location in both scenes (though this might be a problem because you won't have a lot of overlap between two Landsat 7 scenes).

Hope that helps.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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