GRD: Consider using footprint from productInfo.json instead of safe:footPrint from SAFE manifest

[philvarner]

Is your feature request related to a problem? Please describe.

Currently, GRD uses the footprint element in the SAFE manifest (effectively a bbox) to produce a polygon.

e.g.,

             <safe:footPrint srsName="http://www.opengis.net/gml/srs/epsg.xml#4326">
                <gml:coordinates>66.097771,-174.028839 66.944458,176.813492 71.130539,178.793777 70.166634,-170.280487</gml:coordinates>
              </safe:footPrint>

This is not a very accurate footprint for the scene, and has at least one bug with it (#55),

Instead, consider using the productInfo.json "footprint" attribute. This has a MultiPolygon for antimeridian-spanning scenes and more accurate geometries of all scenes.

[philvarner]

@lossyrob do you have any insight as to why the bbox-like footprint from the manifest was using instead of the geojson footprint from productInfo.json?

[matthewhanson]

When I did https://github.com/sat-utils/sat-stac-sentinel/blob/master/stac_sentinel/sentinel.py it used the productInfo.json file and the footprint still had an issue of being not very accurate.

Then, sat-stac-sentinel spawned stac-sentinel where a lot of the stactools-sentinel1 logic was derived from. But that: https://github.com/stac-utils/stac-sentinel/blob/main/stac_sentinel/sentinel.py#L93 looks like it is only for the Sentinel-1 RTC dataset on AWS which is tiled.

So I'm definitely +1 to use productInfo, but also interested in more info about the differences between them. Do you find that the productInfo footprint is accurate @philvarner ?

[lossyrob]

@philvarner I don't have insights, but have seen instances of GRD footprints being off. If productInfo has a better footprint then I'd agree we should change to use them.

[philvarner]

Here's a comparison of a single item:

Current output of stactools-sentinel1, from manifest.safe coordinates:

footprint property in productInfo.json:

Current Earth Search item, calculated from the raster:

Current Planetary Computer:

[philvarner]

The existing package footprint is clearly the worse. Planetary Computer's footprint is the same as productInfo.json but with fewer decimal places, and Earth Search's recomputation from the raster is tighter, but not by much.