Incorrect county data for British Columbia, Canada

Question

I was using the "MapCompare" tool the other day, and noticed that when filtering to Canada / British Columbia / Greater Vancouver, the results included geocaches in the vicinity of Sechelt, Lund, and Powell River:

Indeed, inspecting the Maps generated by “ProfileStats” reveals that some regions are considered to be part of Greater Vancouver (note the colour):

Inspecting the Regional District polygons offered by OpenStreetMap (note: “Powell River Regional District” was renamed to “qathet Regional District” in 2018. Project-GC still appears to use the former name), it starts to become obvious what has happened:

I believe that this occurs because there are indigenous/First Nations land enclaves within these Regional Districts. Due to this, the system has chosen (either intentionally or possibly due to a bug) to geocode caches located within these enclaves to “Greater Vancouver”, despite the fact that these enclaves are independent from county administration.

Using OpenStreetMap’s “query features” functionality within one of these enclaves does correctly confirm that this is indigenous land, not part of any county:

Compare this to a query within the Regional District:

I see three possible scenarios here:

1. Project-GC has intentionally chosen to place these enclaves with Greater Vancouver, for reasons that are not published (or that I cannot find.)
2. Assuming that Project-GC is sourcing its county data from OpenStreetMap, it may be a logic error in the geocoding machinery.
3. However, this question from 2017 (which is still relevant) suggests that Canada's county data is sourced from Statistics Canada, in which case this may be an issue with the polygons provided by Statistics Canada.

For now, I have placed this question under "Support and help", assuming (1).

In the case of (2) or (3), there are two possible ideas that come to mind.

• Group these enclaves with the enclosing Regional District.
• Create a new "county" for British Columbia called "Aboriginal Lands" which includes all of these enclaves. (That said, Geocaching is generally not permitted in indigenous lands so the first option is likely better.)

Comments

I'd say the scenario is 4) "bad data gets bad results and this was likely the best that could be done at the time." Seems like the source material is better now so it would be a good time to take a look and update the definition. (I've forwarded this to admin with the information about the available source material).

---

Aha; that would make sense too. I hadn't realized the data hadn't been updated in a while.

Interestingly, it appears that *some* enclaves do *not* appear on the map: compare my profile map (https://project-gc.com/dimages/ps_map2.php?mapHash=c5679ceaa3fe5d431e6ccabc6a21cab956fe984d) to this screenshot (https://i.imgur.com/bYvVUMD.png) from Overpass Turbo which shows many many indigenous reservations.

Earlier today I spent some time trying to identify county boundaries (see https://overpass-turbo.eu/s/1w8H - may have to click "run"), with mild success. This is probably mostly useless for the administrators since it is simply a big list of ways as opposed to nice labelled polygons, but I am happy to go through and clean up and organize the data if the administrators would like.

---

Follow up to "some enclaves do not appear on the map": not sure how I missed this, but some of the indigenous reservations appear to be recognized and appear in grey on the map. See https://i.imgur.com/qW5Mm64.png. So there is partial correctness here...

---

Project-GC has a rule called "fill hole" that can be set for a region/county which usually solves the issue of these kinds of enclaves. The issue in this case was that there was no usable polygon for the Greater Vancouver-county at the time and it had to use a larger polygon and subtract lots of other polygons from it. I believe the issue is that "fill hole" doesn't really work in that case and that's what caused the issue. But since the smaller polygon is now available in OSM it should be solvable, but will likely take a little while to get done.

---

Ah that would make sense: though it is peculiar that Greater Vancouver county (the largest and most populous) didn't have a polygon but I am sure there are good reasons for that. Thank you for explaining!

---

I just wanted to say that we are working on this. It has been a lot more complex than expected. Canada is the country with the most complex configuration for polygon data. One of the reason was that a lot of data was missing in OSM. OSM has improved a lot since then though, so we are now working on making it simpler.

It's also a huge country with very detailed cost lines, which makes the number of vertices huge, which in turn makes every single test import very slow. This far we have spent 50-100 cpu hours of just importing test data.

But we are getting close to a final result.

Answer 1

So, I have now been working with this on and off for ~2 weeks. As mentioned in a comment, of all the ~250 countries Canada has the most complex polygon configuration in our system. In fact, it's the only country that isn't based solely on OpenStreetMap. The reason being lack of data, but also the fact that big parts of Canada doesn't seem to have "administrative boundary division" in the regular way. Therefore some "regions" are using statistical census zones instead. The whole setup has been a mix between OpenStreetMap (from 2021-03-15, YMD) and Census data (from 2020).

Since we created the current definitions another geocacher has added quite a lot of data into OSM. This will long-term be much better for us, but it required quite a lot of work in our upgrade process. The up-side is that the current configuration in the test definitions relies on OSM only, just as all other countries do. It helps a lot to have a more consistent work flow.

Regarding your assumption I see three possible scenarios here. It's probably as Pleu has mentioned, bad OSM data back then. Otherwise I agree with your conclusion that your solution number 1 would be best, and that is also what we in general try to do. Canada has several areas that do not belong to any "county", and if such area is surrounded by one, we just let it be included there with out fillHoles logic. If such area has more than one neighboring county we have to be more creative. Generally we just add it to one of them, the one where the result looks best geographically.

I am planning to push the new polygon data this week. The expected changes are:

Northwest Territories
* Removed Behchokǫ̀ Region (Now part of North Slave Region)
* Added Gwich'in Settlement Area (Splitted from Inuvik Region)

Ontario
* Sudbury and Greater Sudbury merged

British Columbia
* Powell River renamed to qathet Regional District

Once final the above change list will also be posted on the checker forum.

You (and others) are always welcome to Discord to discuss polygon issues. This invite link is valid for 30 days. I will have to do some reconfiguration on Discord to allow permanent invites again. If you have further question it might be easier to answer them in chat form.

Comments

Wonderful. Thank you so much for your hard work!