footprint / index issue WV product

[agrouaze]

issue spotted by https://github.com/owenod1 `Salut, when using xsar's .footprint function on sentinel-1 WV train metadata the returned imagette footprints appear to follow a different sorting than the file listing (e.g. footprints 3,4 and 5 belongs to imagettes 23, 21 and 22). When retrieving the imagette footprints from the subdataset directly the footprints are sorted correctly but this requires to open the subdatasets individually. Is there a xsar way to connect the correct footprints directly to files without having to open the individual subdatasets?

example of the issue:

train = '/home/datawork-cersat-public/cache/project/mpc-sentinel1/data/esa/sentinel-1a/L1/WV/S1A_WV_SLC__1S/2019/328/S1A_WV_SLC__1SSV_20191124T112814_20191124T113354_030052_036E86_6C90.SAFE'
right = [xsar.open_dataset(i).footprint.centroid.coords[0] for i in xsar.Sentinel1Meta(train).subdatasets]  # slow but correct
wrong = [i.centroid.coords[0] for i in xsar.Sentinel1Meta(train).footprint] # fast but unsorted

here a correct decreasing latitude is shown for descending train when loading the footprints individually, but when loading the footprints from the metadata directly the latitudes (and thus the footprints) follow an incorrect pattern `

[oarcher]

The problem is that for a multidataset, the returned geometry is a union from all geometries, that are unorderded:

https://github.com/umr-lops/xsar/blob/develop/src/xsar/sentinel1_meta.py#L364-L369

There is s1meta.manifest_attrs['footprints'] that should be an ordered list.

Can you try with

[i.centroid.coords[0] for i in xsar.Sentinel1Meta(train).manifest_attrs['footprints']`] 

[owenodriscoll]

This works, thanks

[oarcher]

Perhaps that there is a way to make it clearer for the user.

currently, s1meta.subdatasets and s1meta.manifest_attrs['footprints'] are zippable list.

We might consider to make s1meta.subdatasets be a geodataframe, with index from subdatasets, and geometry from manifest_attrs['footprints']

[oarcher]

The above solution won't works with TOPS SLC products, because there is only one footprint in s1meta.manifest_attrs['footprints'] ( the global one ).

What we can do for TOPS SLC is:

• repeat the same global footprint for all subdataset
• use an empty geometry
• remove the 'geometry' column.
• open all subdatasets to find the correct geometry

cc @agrouaze

[oarcher]

In #81, I've choose to recursively open all TOPS _IW_SLC_ subdatasets, to get their individual footprint.

This slow down a little the product opening, but I think it's an acceptable solution, as the subdataset count is low (3)

@owenodriscoll , can you try my fork (pip install git+https://github.com/oarcher/xsar)?

You should now use

[i.centroid.coords[0] for i in xsar.Sentinel1Meta(train).subdatasets.geometry]

cc @rmarquarlops, I think you will have to update you code after this PR is merged

[owenodriscoll]

@oarcher the code works using your fork, cheers