Noise in S2 results

[fsteinmetz]

Hi Quinten,

I have applied ACOLITE to S2 images (it works great, thanks !) but have noticed noise in the output - which is not a surprise.
Activating the resampling feature (res_ac_bands=1, res_size=32, res_dilate=16, res_fill=1 in the cfg file), as you explained here definitely helps a lot, but there are still randomly scattered NaN pixels at all bands with associated FLAGS=1. This is not something that seems to occur on your example (d) of Lake Balaton.
FYI my test image is S2A_MSIL1C_20161222T100422_N0204_R122_T33TUL_20161222T100606 over Venice. See a small screenshot of rho_w444 here:

Screenshot from 2017-09-19 16-33-34.png (44.38 KiB) Viewed 7986 times

Is this something you have identified and can be avoided somehow ? Just to check that I am using the optimal set of parameters.

Thanks,
François

[quinten]

Hi François

I think this may be caused by the threshold based masking - this is performed at native resolution of the 1.6 micron SWIR band (20 m), and not on the resampled dataset. It is maybe worth looking at the TOA for those masked pixels in this channel (you can export the rtoa product in ACOLITE) and maybe tweaking (increasing) the masking threshold (default 2.15%).

Hope this helps! Let me know what you find. I am at sea currently and unfortunately cannot look at your image itself in the coming weeks.

Quinten