Sen2Cor Topographic correction/cast shadows


#1

Dear Sen2Cor Users,

I have a question regarding the topographic correction in Sen2Cor.
I’m running Sen2Cor Version 2.4.0 on Linux. I am using a DEM with 10 m resolution to enhance the quality of the topographic correction compared to the default the default SRTM DEM. Since my region of interest is situated in the alps, topographic correction is mandatory. I’m quite new to topographic correction, therefore I am not a 100% sure how to interpret the results that I get.
I have run Sen2Cor without topograhpic correction at all, correction by the default SRTM DEM and correction with my 10 m DEM.
Visually the correction with the SRTM DEM looks the most appealing, however I suppose this is because of the relatively coarse resolution of the SRTM DEM smoothing out the terrain and thereby applying topographic correction to areas that lie within cast shadows but are excluded from the topographic correction when using the 10 m DEM.

The first part of my question concerns the topograhpic correction itself: Does the topograhpic correction in Sen2Cor account for cast shadows of mountains?
Through research I found out that Sen2Cor is based on ATCOR (http://elib.dlr.de/103120/1/Pflug-et-al_Living_Planet_SP-740_paper1248.pdf). The ATCOR website (https://www.rese-apps.com/software/atcor-3-satellites/atcor-2-3-method.html) states that the correction process accounts for cast shadows from surrounding topography. However the results I get with the 10 m DEM look like cast shadows are still visible and not corrected (e.g. in snapshot nr. 2 there appears to be an uncorrected part of a cast shadow at the bottom of the slope compared to snapshot nr. 3).
Is this
a) an error in the correction process or
b) a shortcoming of the algorithm which I have to live with?

Is there a way to tweak the topograhpic correction in Sen2Cor to adress this issue?
I would like to reduce the effects of cast shadows as much as possible. Since cast shadows are a common issue in mountainous regions, I suppose there should be a common way/best practice to handle them.
How do other Sen2Cor Users deal with this problem?


#2

Hi Andi,

I am not exactly a Sen2cor user, but it is not very wise to try to correct the cast shadows. The issue is that the radiance within cast shadows is very low, as no much irradiance is available. As a result, the corrected reflectance would have a very very high uncertainty, depending on the measurement accuracy of surface low radiance, the inaccurate estimate of the sky irradiance and the very rough estimate of irradiance coming from other slopes…

I am afraid these shadowed places will remain terra incognita for quite a long time for remote sensing, until we get observations at different hours in the day.

Best regards,
Olivier


#3

Hi Oliver,

thanks for clarifying!

Kind regards,
Andi