3

I'm looking at NDVI changes for an area over 20 years. There seems to be a slight downward trend, however, upon switching from Landsat 5 in 2011 to Landsat 8 in 2013 I'm getting NDVI values that are suddenly a few values higher (0.03 - 0.06).

I was wondering if this has to do with the switch to Landsat 8? (higher spatial resolution, hence more accurate NDVI?)

  • Are you working with raw scenes or atmospherically corrected images? – aldo_tapia Nov 29 '16 at 1:25
  • Try identifying control sites for reflectance values (i.e. buildings, airport run ways - things that you know haven't changed over the years) and see if you get a similar jump in reflectance values. This will give you an indication if it is due to overall reflectance differences in the image or due to the sensor. Landsat 5 does capture a wider spectrum range in the NIR band than Landsat 8 which is something else to consider. – Liam G Nov 29 '16 at 2:13
  • I'm using LT05_L1TP - Tier 1, LE07_L1TP - Tier 1, and the LC80 from the pre-collection for Landsat 8. Should these be atmospherically corrected? – Vanludvig Nov 30 '16 at 4:18
  • You need only correct to TOA for NDVI as its a normalized index, although atmospherically corrected (surface reflectance) data is available from the USGS so you could use this. The same correction (TOA or surface reflectance) should be used for both sets of images before you calculate NDVI – Nathan Thomas Dec 6 '16 at 5:08
3

The dynamic range of each band has changed between the different Landsat missions, which means that the amount of energy captured in each band, even under identical conditions, is slightly different. The slight change in values you are seeing is likely a direct result of the shift in the band ranges.

There may be a published "correction" factor to account for this. Alternatively, if you have imagery from another sensor that covers both time periods, you could use it "calibrate" either or both sensor's NDVI time series, or use stable targets to calibrate the reflectance directly.

Atmoshperic correction is usually desirable, and any NDVI calculation should be on radiance or reflectance, not raw DN.

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.