Clarification on difference between L1C and L2A data

I want to make sure I am properly understanding the difference between the L1C and L2A data levels. From my understanding, the L2A is simply L1C data with some processing done to remove the effects that the atmosphere has on an image. That is, it removes discoloration caused by the air itself, or things in the air (such as smog).

Is this correct?

this is correct. From the official documentation:

The Level-2A processing includes a scene classification and an atmospheric correction applied to Top-Of-Atmosphere (TOA) Level-1C orthoimage products. Level-2A main output is an orthoimage Bottom-Of-Atmosphere (BOA) corrected reflectance product.

1 Like

@ABraun Thank you very much. I saw that in the documentation, but last week was the first time I ever used any remote sensing program other than Google Earth, so I am still learning the terminology.

So if I wanted to make the surface IR of an area (to look for temperature variations), I should use L2A?

What would the advantage of having L1C be?

absolutely fine - keep it up! :+1:

L1C allows you to perform atmospheric correction yourself. If you have good knowledge on cloud removal or the estimation of atmospheric conditions at the time of image acquisition (sun angle, water vapor) you can probably create a better product than the automatically created L2A, but most people (including me) are happy with L2A data.

1 Like

It looks like it has some strange effects in relation to correcting shadows on steep terrain. Although I don’t understand why.

For id: COPERNICUS/S2/20190726T110629_20190726T111240_T32VLP


Sentinel-2 2A


Sentinel-1C

Yes we have seen this too. It is due to an overcorrection according to the elevation.
It has bee reported in the S2 Quality Working Group just today.

1 Like


Also just notified a curious artefact here in the 1C image, it is not visible in the 2C.
Is that a plane with a vapour trail? and the very slight offset to do with the tiny time difference of the different light bands being captured?

This is a know “feature” of Sentinel-2.
It has been discussed here:

3 Likes

Ok thank you! How might I find out if this is corrected? Is there something I can follow?

For now I think I need to use both products to get good results - since I am interested in doing change detection on steep slopes and I get strange results from 2A, but I want the SCL band from 2A for masking.

Hello @erin.l Erin

This feature of the S2 acquisition is described in section 6.2.2 (Misregistration of High Altitude Objects) in the L1C Data Quality Report that is updated every month.

Cheers

Jan

Jan Jackson
S2 MPC Operations Manager

1 Like

Thanks Jan! Interesting to see.

How about the overcorrection effect on the slopes in the 2C product? - if you do a change to the way 1C images are processed to 2A - do you apply the changes to images collected in the past, or just to newly acquired images? I’m wondering if I should try my classification again with the ‘same’ 2C image in a month or two.

Hi @erin.l

Yes; we were impressed when we first saw it. Hence the addition to the Data Quality Report :slight_smile:

The general policy for correction of previously acquired images is to wait for a reprocessing of the whole archive, with any information on the issue - and its potential remediation by the User community - added to the relevant Level (L1C or L2A) Data Quality Report. It will also be raised as an Anomaly, and included in the Anomaly Database.

As noted by @marpet the feature you have identified is being discussed by the expertise at the S2 QWG, and I would expect an outcome of their analysis in the near future.

Cheers

Jan

Jan Jackson
S2 MPC Operations Manager

1 Like