Wikidata:Events/Coordinate Me 2024/Tools/Suspicious coordinates

COORDINATE ME

MAY 2024

WIKIDATA COMPETITION FOR GEOLOCATABLE CONTENT

Start Participate Tools and resources Results and winners

Suggested for editing: "Suspicious coordinates" edit

Rationale and history edit

Emergence of rastered mountain ranges in Austria in 2020

In 2020 a special view on the Wikidata objects in Austria revealed a hidden property of certain subsets of Wikidata items: they had rastered coordinates (see also the German village pump article). It turned out that this was not due to quantization of coordinates, but due to the genesis of those subsets. Most of these Wikidata items and associated coordinates have been imported from the Cebuano Wikipedia and in smaller parts from the Swedish Wikipedia, from articles generated by a bot based on coordinates found in GeoNames. GeoNames however quite regularly does have coordinates rounded to full minutes only. Source and quality of coordinate information in GeoNames turned out to be unreliable and at least unprecise. Furthermore, some of the names from GeoNames looked as retrieved by OCR reading of old maps missing human correction. What made things worse, elevations (elevation above sea level (P2044)) have been calculated based on some digital elevation model using the rough coordinates. As half a minute on average makes a distance of some 100 to 1000 meters horizontally, calculated evaluations were far beyond what was acceptable.

Action edit

Sparql scripts and generated maps can visualize all the coordinates that look to be rounded to full minutes in latitude and longitude. This is an heuristic approach, as there are perfectly sound coordinates not needing seconds at all, either because they describe large objects (like administrative units, great lakes, mountain ranges, regions, protected areas etc.) or they describe an object that is located exactly at full minutes coordinates, even full degrees (e.g. a meridian monument or the poles or Null Island), be it by definition or by chance. Furthermore the maps generated from the Sparql scripts show how big the problem is.

Most of the listed Wikidata items with full minutes and no seconds in latitude and longitude are candidates for check, approval and in most cases, correction.

Find the object on a reliable map (some local GIS services are recommended for your country), take the coordinates from the GIS service (you may want to change the precision for educational purposes to seconds or 6 digits after the comma) and change the coordinate location (P625) property. Do not keep rubbish coordinates from GeoNames (In the end an object has a single location at any given point in time. Locations can be historized but this is different. In rare cases location might be disputed and then we should keep both coordinates with appropriate sources).
Add alias names, if you find some alternate names or spelling variation. Please search also for those alternative names in Wikidata to check for candidates for merging Wikidata items.
Always add the GIS service as a source of the new coordinates (no need to add a tedious url for the very location on the GIS service, as the service in combination with the coordinates is sufficient). Model the GIS service as a source for further use, e.g. Austrian Map (Q23196), if needed.
Do always check and correct the elevation above sea level (P2044) when items are sourced with GeoNames. This is especially important for pointlike objects like mountains, summits, gaps, etc. Elevations for extended objects like glaciers do not make sense in its simple form, the need a qualifier and for changing stuff like glaciers a reference date.
- If you cannot find a reliable source for the elevation of a pointlike object, and you change coordinates, you should assume that the old coordinates and the old elevation somehow depend on each other. It is better to remove such elevation values, especially for GeoNames coordinates. Elevation values can handle input syntax like 1200+-10 (meaning 1200±10).
For long objects like rivers and streets consider adding two coordinates for the source and the mouth of a river, or for both ends of a street. Use qualifier applies to part (P518) to qualify such coordinates (unless there are qualifiers the single value constraint for coordinates will fire).
Objects along the border often do not reflect the border location (on various levels: states, level 1 administrative subunits, municipalities, etc.), but are assigned via the originating Cebuano article to just one of the adjacent areas. So check country (P17) and located in the administrative territorial entity (P131) in case, both can have multiple values. It might be the case, that a real world object (like a mountain at the border - and mountains often define the border) are represented by two Wikidata items not related to each other, even with different names / labels in different languages. Eureka! Merge the two items to better represent the real world object.
Adopting coordinates might also effect located in the administrative territorial entity (P131) or even country (P17).
The types instance of (P31) derived from GeoNames are a bit heuristic and not always matching Wikidata modelling. E.g. in German the suffix -joch in mountainous areas can denote mountains as well as gaps and saddles. So check instance of (P31) accordingly.
If a Wikidata item has more than one coordinate location (P625) without appropriate qualifiers, consider to consolidate these values to the one you found on the GIS service.
If you are doubting the existence of the real world object represented by the Wikidata item (after profound research) you can mark it with instance of (P31) possibly invalid entry requiring further references (Q35779580).

Items will vanish from the maps as soon as they get more precision in coordinates. So, played as a game, you can create areas that are free of such suspicous coordinates.

Examples edit

Adoption edit

Sparql queries can be adopted to

restrict to a region

?item wdt:P131* wd:Q2189.

restrict to a special type

?item wdt:P31 wd:Q8502.