For a project I want to use the view and sun angles of Sentinel-2.
While there were several discussions here in the forum before, it is still not clear to me how they relate to the other rasters.
I have seen a number of different implementations.
SNAP ignores the last row and coulmn, others simply interpolate to the raster sizes, and others are doing different interpolations.
But I haven’t found official information from ESA/Copernicus about it.
How do the 23x23 grid points relate to the 10m,20m and 60m image raster?
Does the centre point of the upper-left grid pixel correlate with the centre point of the upper-left pixel of the 10m image raster or do the upper-left corners match?
If you have some knowledge about it, I would be happy if you share it.
@Chenocq, @Jan Do you have more information about it?
Thanks
Marco from EOMasters - Mastering Earth Observation
Hey, About working with Sentinel-2 data. The 23x23 view and sun angle grid in Sentinel-2 data provides angular information at a coarser resolution than the image rasters (10m, 20m, 60m), so you’ll need to interpolate them to make it align with the pixel grids.
Different tools handle this in various ways. For instance, SNAP sometimes skips the last row and column and just interpolates from there, while other tools might use different methods. Generally, it’s assumed that the center of the upper-left grid pixel lines up with the center of the upper-left image pixel. But since ESA hasn’t given a definitive guideline, people end up using different approaches.
Here are a few methods you can use to interpolate the 23x23 view and sun angle grid to fit the higher-resolution rasters:
Please find the answer from Florian the S2 technical Manager from OPT-MPC project :
"Indeed, we have already identified a lack of information on this topic on Sentiwiki. A following text was already prepared and will be included on Sentiwiki soon:
Sun and viewing angles are provided in tile metadata. They are computed on a virtual grid projected on the tile ground surface. The top left node of the grid has the same coordinates as the top left corner of the tile footprint (top left corner of the top left pixel of all the rasters at different spatial resolutions). The grid contains 23 by 23 nodes, each node being separated by 5000 m. The grid fully covers the tile footprint. It should be mentioned that the nodes from the last line and last column of the grid are 200 m away from the bottom and right tile edges, respectively.
For each node of the grid, the zenith sun and viewing angles are computed based on the normal direction to the reference ellipsoid at the node coordinates. The azimuth sun and viewing angles are computed clockwise, based on the North direction at the node coordinates.
The viewing angles are provided for each spectral band and detector. An inverse location of each node of the grid to the detector/band viewing model is performed, so that only nodes lying inside the detector/band have their viewing angles computed. Therefore, if a node of the grid is located in an overlapping area between two consecutive detectors, two different values of viewing angles are assigned to this node, as being seen from two different viewing directions by two neighbouring detectors.
Both sun and viewing angles are provided in degrees."
Marco from EOMasters - Mastering Earth Observation