search

gbif / portal-feedback

User feedback for the GBIF API, website and published data. You can ask questions here. 🗨❓
30 stars 16 forks source link

Coordinates and Footprint WKT for sampling-event datasets #2166

Open ManonGros opened 5 years ago

ManonGros commented 5 years ago

Both Footprint WKT and coordinates can be included occurrence and sampling-event datasets.

However, I am confused by a few things when it comes to sampling-events:

Is there anything we could do to improve the visibility of Footprint WKT in the context of sampling-event datasets?

MortenHofft commented 5 years ago

The footprints aren't displayed at the event level, only at the occurrence level which also seems a bit counter-intuitive since they are entered for each event (not occurrence).

We don't model events, so we have no idea of knowing what the WKT is for the event. So everything on the event pages is taken from the occurrences. Which makes it quite fragile. We could take the WKT of the first occurrence with the event ID, but there is nothing hindering the occurrence from having another wkt than the event (eg a more precise location within the event area). So it is adding more fragile to the page. I assume that was the reasoning for not showing it.

MortenHofft commented 5 years ago

When only the footprints are given, the occurrences appear to have no coordinates (e.g. has_coordinates=False). This is not exactly true since the the footprints contain coordinates. However, since they are not included in the default download, it makes some sense. It is too bad that no-one benefits from this information.

This seems unfair I agree. A way to fix it during publishing could be to add a centroid with an uncertainty? Another approach could be that we interpreted the WKT and added the centroid and uncertainty during ingestion?

MortenHofft commented 5 years ago

When both coordinates and footprint are given at the same time, only the coordinates are displayed. And since the footprints aren't in the default download, the information might be overlooked.

That seems wrong, but I'm also unable to reproduce it. Can you give an example please. Below is one with a WKT and a lat/lon where both is shown. https://www.gbif.org/occurrence/1949673350

albenson-usgs commented 5 years ago

I agree this is problematic and for sampling event datasets I have been working with I have sometimes been including the lat/lon in the occurrence file in addition to the event file (example).

I could also see this being problematic in the work I'm doing with the biologging field. Specifically you might have individual points for a remotely sensed (or other automated method) animal but may want to represent all of the movements of the animal as one event with a WKT but have the individual points available as occurrences for modeling and other analyses.

ManonGros commented 5 years ago

That seems wrong, but I'm also unable to reproduce it. Can you give an example please. Below is one with a WKT and a lat/lon where both is shown. https://www.gbif.org/occurrence/1949673350

My bad, I made a mistake on that one, thanks for correcting me!

ManonGros commented 5 years ago

This seems unfair I agree. A way to fix it during publishing could be to add a centroid with an uncertainty? Another approach could be that we interpreted the WKT and added the centroid and uncertainty during ingestion?

The second option would be probably more intuitive for publishers (at they wouldn't have to calculate the uncertainty themselves). Would this be possible?