Since NDVI is calculated on a pixel-by-pixel basis and surrounding pixels do not affect the calculation, I don't think it is necessary to apply a mask to each image prior to calculating NDVI. I might use the following steps: Calculate NDVI for the Landsat scenes Layer stack the NDVI images Use Sum Data Bands to calculate the Mean NDVI from the 5 NDVI images (I think this is what you are looking to do - or perhaps you want the mean for an elevation region?) Then to apply DEM elevation levels to NDVI values, perhaps something like a Decision Tree or Raster color slices. I am not sure what end product you might want. If you are looking for Mean NDVI for an elevation region, then you might want to create an ROI of the elevation region from the DEM and then apply that ROI to the mean NDVI image. You can then get the mean for the ROI.

Good afternoon MariM, Yes I am looking for Mean NDVI for an elevation region. For example for the 1976 image, I would like to calculate the mean NDVI for each altitudinal range and add the value to a graph in Excel. Then I can compare Mean NDVI values between dates but also between dates and altitudinal ranges. I think that ROI should work. Do you mean that I should create an ROI for each elevation range? Like one for 0-3000 m, one for 3001-4000 m... Thank you so much for your time Sylvie

Yes, you can create the ROI on the elevation image and then move it to the NDVI image (it should convert on the fly if there are pixel/projection differences - this is for ENVI 5.2). Then you can get the mean for that ROI.

Perfect I will try this method and let you know. One last question though I calculated the NDVIs but I have values between 0 and 255 and not -1 to +1 like it is suppose to be with NDVIs. I probably missed something as I am a beginner with ENVI... Can you help me with this? Thanks so much. Sylvie

The NDVI tool defaults to an output of floating point but there is an option to output to byte. Did you select this option? What version of ENVI are you using? ENVI 5.2 has a Spectral Index tool that outputs to floating point as well.

I work with ENVI 5.2 and I used the option byte but my values are still ranging from 0 to 255. I just try to use the density slice tool but I am not sure if I have to enter the value manually for each of the 255... thank you

Are you using ENVI Classic or ENVI? Did you try the Spectral Index tool? If you select byte output, you will get 0-255. For the Raster Color Slice, you do not have to enter a range for every value in the image. It will only 'color' the values you give it so it can be like a quick classification and you can export to a classification image.

I am using ENVI but I can use ENVI Classic. I didnt try the spectral index tool. Should I do it before running the NDVIs? Thanks so much

The Spectral Index tool is not in ENVI Classic. It is in ENVI 5.2 SP1. It is an alternate method of getting NDVI that is based on the wavelengths assigned to the bands.

Here is what I just did: Band Algebra, Spectral Index and in the Index area I selected the Difference Vegetation Index. The quick stats show that DN values range now from -0.14 to 0.5. I was expecting values ranging from -1 to 1. My images are radiometrically corrected (reflectance and DOS). Maybe I did something wrong? Any idea? Thank you again Sylvie

Also for some reasons It doesn't work with my Landsat MSS images and I think I used the right NIR and R bands. 1975 Landsat 2 MSS: 6 and 5 1987 Landsat 5 MSS: 3 and 2 For the other dates (1994, 2002 and 2009), bands are 4 and 3 and the spectral tool is working... Thank you Sylvie

Those are reasonable values. You rarely get an NDVI value of exactly -1 or 1 (perfectly healthy, green).

oh good thank you! Do you have any ideas why the tool is not working with my two Landsat MSS?

I believe that people usually use the second band of MSS data (band 5 or 2: 0.6-0.7 microns) for red and the fourth band of MSS (band 7 or band 4: 0.8-1.1 microns) for near infrared in NDVI calculations. - Peg Exelis VIS

oh I see... Thank you so much!