Influence of Sun Zenith Angle (SZA) in reflectance retrieved from Sentinel-2 images

Hi everyone! I’m using Sentinel-2 images. Well, I have used used iCor (as well as Sen2Cor and C2RCC) ir order to perform atmospheric correction of Sentinel-2 images of a Case 2 tropical estuary in Brazil. Before that, my research group obtained in situ radiometric data over the water (on various points) so we could make comparisons (correlation graphs):
Bottom-of-atmosphere Remote sensing reflectance (after correction) vs In situ Remote sensing reflectance.

Correlation graphs did not yield any good R² - either for iCor, Sen2Cor or iCor corrected images.

But… when I started filtering points by Sun Zenith Angles , R2 started to increase.

I would like to know authors who have already addressed the influence of Sun Zenith Angle in reflectance retrieval (what I know is that Sentinel-2 is more likely to be affected by sunglint but I don’t know why) . I really want to filter that data with SZA but I don’t know anyone who did the same as I did.

Hi @lorenajuca,
I just came across your thread.
Good idea to filter your data against SZA.
Actually S2-MSI is a near-nadir viewing sensor. For such configurations, you’ll increase sunglint probability when the sun gets closer to zenith.
Here is an example of the sunglint simulation given for a fixed sun elevation and resolved in azimuth and viewing angle:
The code is available here:


@harmel Thank you for your answer!

Hi Tristan, You have done a great work. But I have a questions. what does the the three figures(about I,Q,U) means? Can you give me some guide?

Hi @zhaohaiyanggit,
Thanks for your interest!
I, Q and U are the Stokes terms for radiance and linear polarization of the Stokes vector describing the light field. I is in reflectance unit, so if you want to get the top-of-atmosphere signal you can multiply it by the incoming Sun irradiance and the up/downward transmittances.

1 Like

Thanks for your explanation. I see it now.