Correct, in theory NDVI ranges between -1 and +.

On mid-resolution satellite imagery, NDVI seems not go below -0.2 or above 0.8 (Gils et al. 2012).

Article Development of a two-band enhanced vegetation index without ...

go through the link and read the article for more detail

thank you very much for your kind help really appreciate

Good responses... However, please note that the values extracted may not often times get to the extreem (-1 or +1).

Cheers.

Hi,

The extreme limits should be between +1 and -1. However, theses are just the extremes. Your values could be in between.

Yes, spectral signatures is expected between -1 and +1

I have to disagree with all responses.

NDVI = (N-R)/(N+R)

N=0, R=1 => NDVI = -1

N=1, R=0 => NDVI = 1

=> valid range of NDVI is (-1, 1)

EVI2 = 2.5x(N-R)/(N+2.4R+1)

N=0, R=1 => EVI2 = -1

N=1, R=0 => EVI2 = 1.25

=> valid range of EVI2 is (-1, 1.25)

Lan Nguyen your reasoning is good but I think you have to replace the 1s with 255 (Maximum values of a digital number in a pixel) so that the maximum and minimum values of NDVI remain unaffected i.e. -1 to 1, while those of EVI2 will range between -1.04 and 2.488. To me this makes EVI2 less reliable compared to NDVI since it is not normalized.

Emmanuel: It is not my reasoning, it is how vegetation indexes are calculated. We use spectral reflectances in equations, not DN. Why? Because all equations were created on reflectances (0-1) for consistent physical meaning. Keep in mind that if you use DN then you will have different results depending on how many bits the image was created (see how it goes wrong here). Old image normally use 8 bits, newer image use 16 bits (and more as the quality keeps going up).

Now, talking about NDVI an EVI2 value range. Mathematically, NDVI: (-1,1) and EVI: (-1, 1,25). However, in real life, we normally have NDVI and EVI both are greatet than 0, excepts for some water pixels (can be slightly negative).

Another question is: can we get NDVI and EVI outside value ranges? Yes, we can. It is because during conversion of DN to reflectances we have reflectances are not within (0, 1) (over corrected). Just remove those bad pixels...

Thanks !!!