I want to calculate EVI index from UAV multispectral survey. For satellite imagery it is pretty straightforward (after reading multiple posts on GIS.SE) but not sure how to approach this problem, when imagery is collected from approx. 120 meters and atmospheric correction is not required in such extent.

Can somebody give me a tip how to build the equation in QGIS raster calculator to get proper EVI values?

What vaules should I assign for the "G", "C1", "C2" and "L" coefficients?

EVI = G X ((NIR - RED)/ (NIR + (C1 x RED)-(C2 x BLUE) + L))
  • Could you please state the lat and long of your flight location?
    – AWGIS
    Commented May 22, 2018 at 13:36
  • Have you considered just using NDVI? If your AOI and altitude dont suffer from the conditions that EVI coefficients where designed to compensate for, then why bother using EVI at all?
    – Rex
    Commented May 22, 2018 at 14:21
  • Most of these equations use empirical calculations, so I guess try to do a regression with your image values or a PCA to test the weights of your area.
    – Ka_Papa
    Commented May 22, 2018 at 17:58
  • In this case I do not have specific location, but definetely working in Poland generally. With regard to NDVI, it does not provide results I expect and results vary depending on time of a day slightly even using calibration panels. This is why I am trying to explore other possibilities and EVI was advertised as lighting conditions independent, more reliable index.
    – proteus
    Commented May 23, 2018 at 11:02
  • George - I am not as good in remote sensing :) Can You provide me with a some guidance how to approach this? Any research paper? Would love to try this as well but do not know how to bite that stuff...
    – proteus
    Commented May 23, 2018 at 11:04

1 Answer 1


The common definitions for these coefficients when using MODIS are: L=1, C1 = 6, C2 = 7.5. The C1 and C2 coefficients are aerosol resistance terms that rely on the blue band to partial out atmospheric influence in the red band. The gain is commonly defined as 2.5

So, here is the thing. One would not expect the same atmospheric influences in UAV imagery as in satellite. These coefficients are fairly irrelevant in your case and I would be thinking of utilizing a different index or a variation of the EVI.

Alternately, you could use the two band EVI. This avoids signal-to-noise problems in the blue band and removes the necessity for aerosol resistance terms in the index.

two.band.evi = 2.5*((nir - red)/(nir + 2.4 * red + L)) where; L=1
  • Thanks for your voice in discussion! So the blue channel is the one which could be most problematic? I would like to survey berries, wine and some crops with Micasense Rededge. Can You suggest some more indices that would be reliable and yielding similarly accurat results as EVI (apart of EVI2)?
    – proteus
    Commented May 23, 2018 at 11:07
  • 1
    You are somewhat limited in the data from a UAV. Some of the more robust applications in this arena used data that at lease included SWIR and edge bands. You may want to consider some textural measures in the red, nir range. I would ask, why collapse the variation. Just use the [R,G,B,NIR] in a multivariate classification or feature extraction model? An index is not always necessary although, one that I commonly use in ag applications is the MSAVI because it corrects for background soil brightness. However, it relies on SWIR-1 (1.566 - 1.651mm). The SAVI is similar and only needs [RED, NIR]. Commented May 23, 2018 at 12:25
  • You are right, using uav data limits me definetely. But for my area using sattelite imagery is not possible. I require data that will indicate changes on the level of a single row, not whole field (this often happens due to parcels geometry and size). Can You recommend some research papers with comparison of indices for the ag area with emphasis on uav borne data?
    – proteus
    Commented May 25, 2018 at 5:50

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.