I have followed all the requisite steps to arrive at the resampled and reprojected format of S2 data after ATCORing it with SEN2COR. The final band files (Disk Image File) are then opened in ArcMap as layers and Raster Calculator is employed to compute the NDSI as per the band convention for Sentinel 2. The results seem to be erroneous since it indicates presence of snow in a snow-free season. Can somebody point out the mistake in the process flow?
I guess you should study the NDSI algorithm to see how it is calculated to understand better your results. You could apply it in SNAP to the original image using Band Maths and see if you obtain similar outputs.
The only place I can see the error happening is when I drag-drop the .img band files in ArcMap and compute the NDSI. The process flow in SNAP is pretty regular,
-ATCOR with Sen2Cor
-Resampling
-Subsetting
-Reprojection