I'm having some peculiar issues with Sentinel 2 data. I've downloaded two coterminous tiles (one L1C tile from EarthExplorer and one L1C/L2A tile from AWS) taken at the same time and date. I have done this to verify that my methods of atmospheric correction (ESA's sen2cor tool) are valid and give me precisely what I would get from simply downloading AWS' L2A product. The tile ID follows: S2B_MSIL1C_20190113T181719_N0207_R084_T12TWR_20190113T200650
The issue is, it doesn't produce identical results. When I go to calculate NDVI from my bottom-of-atmosphere product, I get different values when using AWS' L2A versus the L2A I produced using sen2cor.
I've done a little research on this and found reference to a "pilot" introduced to ESA's L2A_GIPP process in 2018. There seems to be a version 2.11 floating around out there, while the process in sen2cor (2.07) produces "L2Ap" instead of "L2A"
My question is: Does anyone know why an identical atmospheric processing chain would produce dissimilar results?