mjy
commented
4 years ago Thanks @adrik29, all good questions that others have asked. We'll move these answers to taxonworks_doc at some point.
Why are there duplicate Geographic areas?
Managing geographic areas is hard, maybe as hard a taxon names. There is a great deal of homonymy, synonym etc. We build the GeographicArea according to a strictly algorithmic approach, i.e. no individual human curation is allowed. This is to prevent inconsistencies across broad scales to be introduced, it comes at the cost of things like "duplicate" records (but see below). The algorithm has the following principles:
- GeographicAreas are unique baset on the string of their name, and the string of their parents name. So
A Cis not the same asA B C. This ensure we we prevent accidental homonymy at the cost of replication. - When a GeographicArea is duplicate, i.e. two different gazetteers have the same name and parent names, then the shape for that area from that gazetteer is added to the prior entry, with an indication of what gazetteer it came from.
Does selecting a finer subdivision assign allow us to infer the parent subdivisions (i.e. does assignment "cascade" up).
Yes. We can determine based on the parent hierarchy, if available (note it might not be comprehensive) and spatial calculation where the entity might fit (if the entity has a shape). within the hierarchy.
Which GeogaphicArea should I use?
One with a shape. The TaxonWorks approach is get people to think spatially as a priority. We can re-calculate string names when countries change if we have shapes, we can not recalculate shapes if we have string names. As soon as we bind our data to spatial assertions we have made a huge leap forward in the long-term use of our data. Additionally:
- Picking prior-used geographic areas for your project will lead to slightly improvements in consistency.
- Picking entities, when possible, from the same gazetteer may improve visual layouts in the long run, though again given the TaxonWorks approach this is less important as we can ultimately match spatial records to a single gazetteer for display purposes.
Should I always use a GeographicArea in my CollectingEvent?
In fact, if you assign a point georeference you need not assign a GeographicArea as it will automatically fill in country, state, etc. string names in the DWC index. GeographicAreas do, however, prevent you from assigning point, or spatial georeferences outside the shape assigned to that GeographicArea, i.e. they act as a validation measure.
What is the scope of the GeographicArea table.
GeographicAreas are intended to cover the world until the second (third if countries are included) level of geopolitical subdivision. For example to counties (country, state, county) in the United States.
How do I suggest new gazetteers for missing entities?
Open an issue here. The issue should
- Identify a gazetteer with spatial entities that can be added. It is unlikely we will gazetteers without any spatial data.
- Documents whether that gazetteer is free for use and redistribution
- Identify the scope of the gazetteer and whether any entities already exist in TaxonWorks
Will I be able to edit my own GeographicAreas?
Ultimately yes, this will be the "Gazeteer" model.
adrik29
As far as I understand, TW uses a controlled set of toponyms.
(1) Some are duplicated, I don't know exactly why. It seems that it's because they came from different sources. Which one should I use? only "Brazil", or "South America: Brazil" for example?
(2) I don't know if they "cascade", that is, if they propagate upwards when used in "Asserted distribution". For example, if I mark species Aus bus as recorded from Clatsop County, Oregon, USA, should I expect that Aus bus name pops up when I list related data from Oregon or from USA as well?
(3) How can I notify missing areas? I'm not talking only of crazy countries that change names or are newly formed, but of traditional provinces that are simply missing from our list of controlled terms. For example: https://en.wikipedia.org/wiki/Department_of_Puno
(4) For huge and megadiverse countries such as Brazil, shouldn't we include also 2nd order administrative divisions such as done for the USA?