Georeferencing broad localities and islands

[Archilegt]

Hi to all,
Thank you very much for the DwC Hour webinar on "Imagining a Global Gazetteer of Georeferences".
I am now dealing with old Caribbean localities and trying to curate old literature records and to link them to specimens. Very often I come across records which have only an island-level precision: Martinique, Hispaniola, Cuba, etc. I am using the Getty TGN as recommended by the DwC Quick Reference Guide for standardizing island names. The TGN also offers decimal point coordinates for islands but they do not give the shapes of the islands, and those are needed to get the "corrected center" and "geographic radial".
My questions for thoughts are: How should researchers deal with this type of broad locality? Should broad localities (e. g. islands) be "point-georeferenced" or shape-georeferenced or both? Could "smaller" islands be point-georeferenced (e. g. Saba) while "larger" islands (e. g. Cuba) shouldn't? When is an area/island "small" and when is it "large"? Is a point coordinate with a radius better for an island (or state, or province) than its exact shape on a map? Shall we aim to having a shape repository with PURLs to standardize and represent areas (countries, states, islands, etc.)? In case that "corrected center" and "geographic radial" for broad geographical/geopolitical areas are desired, shouldn't we run that process once in a GIS and create a standardized library (in GBIF or elsewhere) with those values?

Regarding my specific problem: If someone already has a standardized resource for "darwincoring" Caribbean islands (georeferences, uncertainties, etc.), I would be most grateful if you could share it to me.

Kind regards,
Carlos

Carlos A. Martínez Muñoz
Zoological Museum, Biodiversity Unit
FI-20014 University of Turku
Finland

[tucotuco]

Hi Carlos,

In VertNet we do prepublication data migration to try to standardize geography. The principles upon which the standardization are based can be found at

https://github.com/VertNet/DwCVocabs

while the actual geography lookup can be found at

https://github.com/VertNet/DwCVocabs/blob/master/vocabs/Geography.csv

Though the countries and territories of the Caribbean are not fully resolved yet, those that are should give you a good idea of how we treat geography as a whole, rather than as distinct fields to be standardized individually.

[Archilegt]

@tucotuco
Hi John,
Thank you for your reply. I read VertNet's standardization principles and I also downloaded and checked the Geography.csv file. I applied filters and explored the dataset for a few islands (e.g., Cuba, Martinique, Chatham I., Channel I.). I saw no standardized georeferences. Shall I assume that no coordinates ("corrected center" and "geographic radial") should be applied to localities that have only island-level resolution? That is my main question.

[tucotuco]

@Archilegt
Hi Carlos,

Your original posting is full of lots of wonderful questions. I would like to treat all of them fairly, but it will take a while to do so. Let me answer your latest posting first though, since it has your main question.

First, any lack of a standardizations in the Geography.csv file for islands in the Caribbean should not be taken as a sign of anything except the lack of time to get to them, yet. We have tried in VertNet to keep up with new geography combinations as they appear in our data migration (pre-publishing standardization in collaboration with the data publisher) using the VertNet migrator toolkit (https://github.com/VertNet/toolkit). We make strides as opportunities permit and have resolved raw geography to standard geography lookups for 152890 of 197994 distinct combinations encountered so far. Since the beginning of this project we have committed to resolving 100% of the geography combinations for the United States, Canada, and Mexico, and thanks to @pzermoglio try to keep up with all of South America periodically. Beyond that, updates occur when there is a specific demonstrated need or interest. We currently have 100% resolution on 33 distinct countries or territories. The current state of geography resolution in geography.csv is always given in the GeographyResolutionStatus.csv table (https://github.com/VertNet/DwCVocabs/blob/master/vocabs/GeogResolutionStatus.csv).

Shall I assume that no coordinates ("corrected center" and "geographic radial") should be applied to localities that have only island-level resolution? That is my main question.

No, you should not assume to not georeference localities that have island-level only resolution. Georeference EVERYTHING that doesn't fall into the do-not-georeference categories from the Georeferencing Best Practices (https://docs.google.com/document/d/1nTz6TMK7j3pdwcbneUrSzFMPlmgLnSCCaw7XgzhFM5E/edit?usp=sharing or eventually https://doi.org/10.15468/doc-gg7h-s853) or Georeferencing Quick Reference Guide (https://docs.google.com/document/d/1TySpxRrzks5nW-jnmFOu4mcnALMQoJXw8IHMHe34dWU/edit?usp=sharing or eventually https://doi.org/10.35035/e09p-h128), namely,

• Some Dubious Locations
• Cannot be Located
• Multiple Unrelated Features
• Demonstrably Inconsistent

Remember, the uncertainty is to help potential users know whether a record is usable for their purpose, therefore the uncertainty is ALWAYS useful to have, no matter how big it is.

[tucotuco]

Regarding my specific problem: If someone already has a standardized resource for "darwincoring" Caribbean islands (georeferences, uncertainties, etc.), I would be most grateful if you could share it to me.

I don't have a resource for systematically georeferenced Caribbbean Islands, but one could created from the spatial data available from https://pubs.usgs.gov/of/2004/1400/data/boundaries/caribis/caribis.htm.

[tucotuco]

Should broad localities (e. g. islands) be "point-georeferenced" or shape-georeferenced or both? Could "smaller" islands be point-georeferenced (e. g. Saba) while "larger" islands (e. g. Cuba) shouldn't? When is an area/island "small" and when is it "large"? Is a point coordinate with a radius better for an island (or state, or province) than its exact shape on a map? Shall we aim to having a shape repository with PURLs to standardize and represent areas (countries, states, islands, etc.)?

My opinion is that the ideal for any georeference is to have a geometry for it, and a point-radius derived from that following best practices. The size of the place shouldn't have any bearing on methods. Both are useful. If it is too much work, or if the appropriate tools are not available to make or store the geometry, then a point-radius is a good substitute, and possible no matter where you might want to store the data, and with free resources (Google Earth, Google Maps, Open Street Map, Natural Earth, etc.,) plus the Georeferencing Calculator (http://georeferencing.org/georefcalculator/gc.html) and GEOLocate (https://www.geo-locate.org/).

In case that "corrected center" and "geographic radial" for broad geographical/geopolitical areas are desired, shouldn't we run that process once in a GIS and create a standardized library (in GBIF or elsewhere) with those values?

Yes, that is exactly part of the plan for the broader services in the Biodiversity Enhanced Location Services (BELS) talked about in the webinar (https://github.com/tdwg/dwc-qa/wiki/Webinars#chapter-17-imagining-a-global-gazetteer-of-georeferences) and BBQs. A summary of the discussions in the BBQs is forthcoming from @pzermoglio, who so kindly hosted those activities.