Issue 18240

Lat/long flipped showing polar bear in Indian Ocean

18240
Reporter: thirsch
Type: Feedback
Summary: Lat/long flipped showing polar bear in Indian Ocean
Priority: Critical
Resolution: Fixed
Status: Open
Created: 2016-02-17 17:12:49.367
Updated: 2016-02-18 12:56:03.814
        
        
Description: From the locality description and coordinates, it is clear that the lat and long values have been transposed in this record (the position would be plausible if the lat and long values were switched), and as there is no stated country no mismatch is found - in any case the position may well be outside national jurisdiction. Example of where our system is not catching obvious georeferencing errors.

*Reporter*: Tim Hirsch
*E-mail*: [mailto:thirsch]]]>
    


Author: mblissett
Created: 2016-02-17 17:34:43.211
Updated: 2016-02-17 17:34:43.211
        
Transposed coordinates wasn't being handled correctly before (only one or both being negated), but even after the deployments this week, this occurrence will still pass without issue — as you say, because there's no country.

Without the country, it's a lot less obvious...

    


Author: mblissett
Created: 2016-02-17 17:37:25.544
Updated: 2016-02-17 17:37:25.544
        
However, they do supply "Atlantic Ocean", and that's not really used.

    


Author: thirsch@gbif.org
Comment: I realise it's challenging and the main reason I noticed it is that the US ambassador's partner is a polar bear specialist and this will be a map we are likely to show them tomorrow (I'll try to start it on a zoomed projection to screen out the offending point!). There may not be any easy way of picking up this type of error other than flagging it as an outlier from the published range for the species  - but in this case I will send feedback directly to the publisher as it is coming through the main node institution for Norway, so hopefully we can get it corrected at source.
Created: 2016-02-17 17:50:10.483
Updated: 2016-02-17 17:50:10.483


Author: mdoering@gbif.org
Created: 2016-02-17 23:02:17.639
Updated: 2016-02-17 23:02:17.639
        
Intersecting expert species ranges with individual occurrences would allow us to flag suspicous outliers. The big question is how to get hold of such ranges. This was a prime goal from the early ChecklistBank development and we still store distribution records for species already. Most of them are listing countries or biogeographical regions. The piece that mostly was missing was to maintain shapes of these areas in postgis so we can easily do intersections. We called that Named Area Repository years ago but we never implemented it properly. But really this is not too much magic and we can easily deal with all those areas where we have accessible shapes already: http://rs.gbif.org/areas/

The other issue is knowing if a known species range is complete or just partial, e.g. for Europe only.

I would really hope we would get engaged in species ranges
    


Author: rdmpage
Created: 2016-02-18 12:56:03.814
Updated: 2016-02-18 12:56:03.814
        
[~mdoering@gbif.org] [~dhobern@gbif.org] Agreed, distributions are something I think GBIF should be involved in, if only to avoid cases like this. The Species Distribution Repository (SDR)  https://code.google.com/archive/p/gbif-sdr/ was, I guess, an attempt to do this. I'm assuming that if this was revived it could be done a lot more easily - species distributions as GeoJSON polygons and some spatial indexing, job done ;)

Given that there are sources of species distributions as polygons (e.g., IUCN RedList) surely it would be straightforward to import those and test for charismatic taxa whether we have outlying occurrences?

Alternatively, isn't this something that Map of Life http://mol.org could/should offer as a service? If they did this it would make a relationship with them actually useful.