NDVI From S2 L1C

I am having some issue with atmospheric correction with some S2 L1C scenes. My final objective is to derive NDVI map. Will it be really inaccurate if I perform the band ratioing for the NDVI with converting the data to L2A?

The practical impact of atmospheric correction depends on the atmosphere and surface properties, so there is no universal answer to your question. If you have images where atmospheric correction is successful, and understand why it fails for some images (e.g., missing ancillary data), you can compare NDVI for the same image with and without atmospheric correction to determine the impact of the differences for your use case. If the failure is due to some differences in atmosphere properties (in N. America this summer, smoke due to wildfires spreading from the west coast to the eastern seabord) you may find some research into the optical properties of long distance smoke plumes (Intra-continental wildfire smoke transport and impact …, https://doi.org/10.1029/2018JD029878). If the failure is due to surface properties (in N. America this summer, bare soil due to extreme drought conditions) you may be able to say something about NDVI values for the problematic areas.

1 Like

In addition to what @gnwiii said: If you want to compare NDVI values of multiple dates or multiple areas, correction to L2A makes sense because some factors are reduced which could introduce unwanted bias.

1 Like

I am assuming that if it is a clear cloud free scene, computing the NDVI from L1C will not give me absolutely misleading results.


are subjective judgements. Good science needs objective measures to justify “clear cloud-free” and “absolutely misleading”. There are certainly scenes where the impact of atmospheric correction is small, but apart from extremely high plant biomass marine systems I’m not aware of systematic studies that provide guidance on when you can omit atmospheric correction.

Understanding the causes of

would be useful. Can you tell us more about this issue?

Seems like a lot of S2 scenes from 2016 have some issue like missing ancillary data. This puts the ATCOR processing on https://peps.cnes.fr/ to go on forever.

Hello @EarthObs

Can you be more specific in your statement, please? Anomalies for the S2A and S2B instruments are tabulated at Page anomaly list



S2 MPC Operations Manager

I am not aware of which issue is causing the delay in processing at PEPS but certainly there is an issue since the ATCOR model (MAJA) is running for the past couple of days, which generally happens when the L1C scenes have some issue.