Impact of varying SNR in L1C data on Sentinel-2 L2A products

In the frame of the Copernicus program, ESA launched the Copernicus Sentinel- 2 optical imaging satellites, which are fully operational since June 2017. Sentinel-2C- and -2D satellites will be launched following Sentinel-2A and -2B units with identical sensors. This paper reports on a sensitivity analysis of Sentinel-2 atmospheric correction / cloud masking vs Signal-to Noise Ratio (SNR) in specific spectral bands. Some Sentinel-2 L1C products are selected to study this effect. Noisy products are simulated adding noise to original L1C-data applying different Gaussian noise models. Finally, both original and noisy L1C-products are processed with Level-2A processor Sen2Cor and resulting L2A-products are compared. Results showed, that added noise to B10 is most critical due to performance reduction of cloud masking. Added noise to B01 is less critical because it does not lead to systematic changes of average surface reflectance. It results in increased scatter of surface reflectance. Added noise to B09 is found to be uncritical because the impact on water vapor retrieval is within uncertainty of validation method.