We need to talk about the Canvec elephant in the room

Communities Canada
41

I would agree to stop importing vegetation. Buildings and roads I haven’t been importing anyway.

42

Should Canvec forestry data be supplemented or replaced with higher quality local datasets? I know that the Government of Ontario maintains a forest inventory that includes species distribution, stand composition, and tree age.

43

I’m not able to add much to this conversation as I’ve been working a lot lately, just adding that I also have never imported canvec buildings or roads and stuff as it is simply not in the data on tiles that I work with.

1 Like
44

What license is this Ontario dataset under, and what is its quality compared to aerial imagery?

45

I’m assuming they mean the Forest Resources of Ontario 2021 dataset1 that is licensed under the OGL (Ontario) which is a compatible license2

1 Like
46

Is anyone currently importing buildings and roads from Canvec?

mboeringa started the thread about natural landcover/forests in unpopulated areas.

Then others commented that natural landcover data is out of date in populated areas, and some commented that building data and/or railways data is out of date. But that’s not what the first post was about.

Some back and forth about the natural landcover concluded, as far as I can tell, that the Canvec natural landcover data isn’t wrong, but our interpretation into OSM might or might not be, but it’s hard to tell partially because we don’t have any other detailed data for what’s there.

So does anyone want to discuss more about imports of natural landcover data in unpopulated areas? Or are people upset about lack of documentation in the wiki? Or about quality of individual contributors’ changesets? If you want something done or changed, please suggest actions.

47

@PcMouse1 Do you intend to look into this data and compile documentation for its import into OSM, or do you want others to look into it?

48

Thank you for moving this conversation forward.

I see two points of concern in this thread regarding the Canvec/OSM elephant in the room.

  • A massive import of Canvec data has taken place in recent months, which brings up the second concern…
  • The imported Canvec forest is very inconsistent/broken, as @mboeringa clearly documented.

Without this import, this thread might not have started. So what do we do with Canvec data? On the one hand, as @pnorman pointed out, a data source not updated since 2012 is no longer suitable for import. On the other hand, many participants pointed out that lakes and river beds are probably still valid.

So I suggest that from now on, the only Canvec/Osm features that can be imported are natural=water and natural=waterway.

If everyone agrees, I could edit all the wiki pages related to Canvec to add the appropriate warnings. Comments, suggestions?

1 Like
49

What about all the other data apart from forests, roads, railways, and buildings/POIs?

50

Per @Jarek suggestion, I proposed something for which most participants would agree.

Considering that most of the inhabited areas of Canada have been mapped, here are a few comments about the map features you refer to.

  • Forests: @mboeringa clearly documented its inconsistency, and I agree. Should not be imported.
  • Roads: updated just before 2012, most of it should still be valid - but newer segments will be missing. Needs to be updated before importing.
  • Railroads: digitized from paper maps, most are probably no longer valid. Should not be imported. If you see them, map them.
  • Buildings/POI: digitized from paper maps, most are probably no longer valid. Should not be imported.

So remains water and roads.

Would these map features be acceptable as the only data to import from the Canvec/Osm product?

51

Sorry, my question might be a bit confusing. I agree to stop importing forests, roads, railways, and buildings/POIs. I was asking about all the other data (power lines, pipelines, addresses, coastlines, cutlines, mountain peaks, residential areas, wetlands, sand, landforms, eskers, etc.)

52

I was asking about all the other data (power lines, pipelines, addresses, coastlines, cutlines, mountain peaks, residential areas, wetlands, sand, mountain peaks, landforms, eskers, etc.)

as you mention them 2 times, I agree peaks might be suitable for conflation, you could have a look into the data and see what is missing or where our positions could be worse than canvec.

53

I’ve had a quick look at all the Canvec/OSM map features. Looking for permanent or potentially still valid items, I propose to keep only the following for import:

Hydrography
natural=water
waterway=stream
waterway=riverbank (need to be converted as natural=water + water=river)

Wetland
natural=wetland

Place
place=*
natural=peak

Transportation
highway=*

Any comments? Any other suggestions?

1 Like
54

I would like for others to look into this potential import due to my inexperience with the import process.

1 Like
55

This needs to be decided on an import-by-import basis. In some cases the data is from the 70s and too much has been changed. This can be discussed when someone proposes an import.

56

I am OK with this suggestion, though I wonder about usefulness of some of the data:

  • if we consider forests to be too inaccurate, what are the chances of wetlands being accurate?
  • are there any actual place=city in Canada that are still unmapped and thus could be imported? or does this refer also to settlements much smaller than “cities”?
  • I have the impression that most roads in Canada are mapped. What kinds of unmapped highways are in canvec and thus could be imported? If this refers to things like logging roads, could these legitimately be out of date (blocked, rerouted, etc) after a decade?
57

Thank for getting back to my proposal.
Here are some answers and topics for discussion…

There is about 14K Canvec/Osm files available. Many of them have already been imported in southern Canada. So, it would all depend on “what an import is?”. I agree that it might be a good idea to ask those who want to import Canvec data to raise their hand. However, in order to keep the whole process simple, my proposal is to ask those who import Canvec/Osm data to limit their imports to “long-lasting” map features only. That’s why I’ve limited my proposal to seven feature types, from four of the eleven Canvec themes. We could also use the Metadata provided in each .zip file, but this would be much more complex to apply and verify.

If I remember correctly, roads and hydrography features were updated all over the country beginning in 2004. Furthermore, the update was completed in Yukon, Northwest Territories and Nunavut, just before the creation of the latest version of Canvec/Osm (2012).

That’s a good point. Wetland delineation is often fuzzy by definition, but those in Canvec/Osm may be annoying/irritating when updating them. I’m quite willing to remove wetland from the proposal if we all agree.

My mistake, I should have written place=*. Have a look Here. You’ll see that it also refers to settlements much smaller than “cities”.

Possibly, but we can’t assume that what remains to be imported doesn’t contain roads.

It’s possible for some of them. If you look at all available road types Here, those processed with the &RECREATION tag are often logging roads. These roads are tagged highway=track when surface=unpaved and lanes=1, but the rest are tagged highway=unclassified, like many other types. So highway=track could be removed from the proposal, and a warning about highway=unclassified in forested areas could be added to the wiki.

If we agree, and I would like comments from all who participated in this thread (particularly from @mboeringa @dmich9 and @jmarchon), I’ll document the whole thing in the Canvec wiki page, including a short procedure for preparing the data in JOSM, before an import.

58

I agree to stop importing everything except streams, rivers, peaks, and lakes, if that’s what the community would like.
I don’t think it makes sense to import roads or place=*.

59

This is not correct. At least in BC the CanVec water is extremely dated. I was told it was based on digitised topo maps from the 70s but don’t know for sure.

The government has changed the names of offensively named features. Why would we import decade-old CanVec instead of the latest data from NRCan?

60

I’m pretty sure the latest version of Canvec included the latest BC Hydro features (data from the province). However, I have contacted NRCan to verify and make sure this is accurate.