Issue while exporting NDSI as .tif

I would like to essentially export the NDSI computed in SNAP as .tif file since when I try to compute NDSI from the band files (.img) in ArcGIS (using Raster Calculator) I get erroneous values (for example, snow in snow-free season) and I see the scale to be stretched between -1 and +1 whereas the histogram for the NDSI computed in SNAP shows the right range (not rescaled).

The issue with exporting as .tif is that when I try to just export the NDSI file, it prompts me to also export the referenced bands (B3 and B11). When this .tif is opened in ArcGIS the only layers present are the two bands and not the originally computed NDSI. I hope there is someone who has already been through the troubleshooting.

Before you can export a raster generated with the band maths, you have to make it permanent: Please have a look here: How can I export my band math result out as a tif or any image?

Thanks Andreas. One issue is solved but strange there is some bug with the naming convention since it seems to be not saving when I use a customized name and rather I need to go with what is suggested.

Now the other one I am looking at is the scaling that is happening while computing the NDSI. I am looking at an area where the temp is well above freezing. However, as can be noticed in the attached snapshot below, the river appears to be frozen (with high NDSI values, 0.98 here) but when one looks at the histogram, most of the pixels seem to lie between -0.81 to -0.16 (NDSI). Is it some kind of rescaling issue while computing the NDSI. Probably everything is fine with Sen2Cor 280 I suppose.

PFA a snapshot that explains the issue.

I don’t have much experience on the typical value range of NDSI, but if a value below zero indicaes snow-free land surface (according to this source) and your frozen lake is above 0.98 (also visible in the histogram), everything seems okay to me.

Or do I miss a point here?


Yea you missed a point here :smiley:
Like I said, the conditions on that day was well above freezing, especially when I looked at the acquisition time in the data file given in UTC, the scene was acquired around mid day which rules out the possibility of an early morning freeze. I am also quite happy with the histogram which shows prominently snow free pixels in the image. The issue is with the entire river that seems to be frozen which is not quite possible so I believe thats an artifact.

Hello @arnab

I would be interested in knowing the Tile and Orbit of your product. I presume the acquisition is the 17/09/2018.

Thanks in advance


Jan Jackson
S2 MPC/CC Operations Manager

Hello Jan

You are right, the product has been acquired on 17th September 2018
Here is the product I am looking at
Tile: 52VEP
Relative Orbit: 032

Dear @arnab: Sorry, I falsely interpreted your first message - now I understand your problem. Hope someone can help.

I guess the band combination suggested here are a bit confusing.
Here, the term VIR is denoted to B3 for NDSI and B4 for NDVI.

What I also don’t get here is the statistics as shown below

When the display range lies somewhere between -0.807 and -0.156, how is the pixel info window displaying the pixel values over the river around the range of 0.9?

The values in the color tab are not representing the full range. The view in SNAP cuts off 5% of the data’s extreme values for visualization purposes to enhance the contrasts: Differences between image without and with 95% or 100%
The data’s actual value range might be considerably larger, especially when there are extremely bright pixels in the image, but these values are neglected by SNAP so that the data has balanced contrasts.

If you want to get accurate values you have to select the statistics tool grafik