You will also need to contact the province to see the sources of it’s data.
Maybe not, metadata was transferred to NRCan when data was integrated into the Canvec product.
I see two reasons.
First, it is much easier to import (i.e., small chunks, same format) while still providing roughly the same data for all provinces/territories - if confirmed by NRCan. Even if nowadays opinions are not unanimous, the creation of the Canvec/Osm product, and its import into OSM, made consensus at this time.
Second, importing the latest NRCan data would not be the same as importing Canvec/Osm (e.g., different data models, formats, content). IMHO, it would therefore be necessary to go through the whole import acceptation process before moving forward - which could be acceptable if all agree.
We should be doing this for CanVec too. Things have changed in the over a decade since community consultation and even without things changing, it was never properly documented.
The Canvec/Osm import documentation may not meet current expectations, but this was discussed (Talk-ca) at the same time the import guidelines were written and import mailing list created (2008-2009). Some Canvec and Geobase Imports started in 2008 using different tools until the official Canvec/Osm product was made available in 2010.
But as you wrote, things have changed. I am not against examining Canadian import data sources using current requirements if agreed by the community.
Hi all,
I’m a bit late to the party but as a past CanVec importer I figured I should chime in. My focus was on the wilderness in western Canada, particularly in the coast range between Squamish and Bella Coola. I’d like to travel through these areas by foot, bike, kayak, and ski one day and couldn’t handle the blank spots on the map so got to work fixing that.
I feel that the CanVec data has a ton of value in remote areas like this. It isn’t perfect, but it exists, and these areas will never be mapped out manually. The data is absolutely good enough to use for wilderness navigation. No one is using it to figure out where the edge of a lake is; it’s used to tell you what mountain you’re looking at, where the confluence of two rivers is, and which glacier is feeding it. Broad strokes navigation and feature identification.
The data quality does vary by type though, and I don’t think all of it should be used:
- Streams & Water Bodies - Very useful and accurate enough. It would be nice if newer data were available, but it will be out of date again after the next freshet. Rivers change size, shape, and course seasonally. Some lakes change size as well throughout the year, so any data is going to be an approximation. However, they’re not properly connected. This is really the only reason I would like to see a better source used.
- Mountains & Saddles - These don’t change much. CanVec data is mostly still perfectly good. Any offensive names that get updates are usually updated quite promptly by the community in my experience.
- Mountain Ranges - These show up as peak nodes, so I had to filter them out and convert them to the little-used natural=mountain_range. Ideally they would be mapped as ways, but it would be trivial to find them and delete or update in the future so I included them as nodes.
- natural=land features - These are a mixed bag. They are often good to include, but you need to review each one and figure out what the tags should be changed to. I don’t think natural=land isn’t acceptable.
- Glaciers - Yes, they melt. But they don’t really move, and outdated glacier info is still perfectly fine to tell you which glacier is which. Like rivers, they are also a moving target. If we had a newer source of glacier polygons I would love to use it though. Otherwise I would like to see this data used as-is.
- Vegetation - It’s mostly garbage. I included it with my imports because it wasn’t conflicting with anything and I felt that it didn’t hurt, but I also wouldn’t be upset if someone deleted it all. If there was an inhabited place in the tile I was uploading I would often just omit it entirely though since those areas tend to be wildly out of date.
- Wetlands - Another moving target. The shapes are very approximate, which is impossible to avoid for something that changes seasonally. It’s not particularly accurate but I find it useful information to have to tell you that an area is kind of marshy. They’re very poorly mapped though and overlap with waterbodies in many cases.
- Roads - Main roads were often accurate but they were almost always already mapped out. I omitted these when uploading in almost all cases.
- Trails - Worse than useless. It seems NRCAN handed a crayon to a beaver and asked them to get mapping. None of these should be used.
- Addresses, Buildings, Docks, etc. - Where there’s houses, there’s people and where there’s people, there’s OSM contributors. CanVec data wasn’t needed in these areas and is probably inferior to other sources.
- Place Names - Inaccurate, usually already mapped, and shouldn’t be included en masse. But occasionally they do have ghost towns, old railway stops, and other unique places that should be included. See Murderer’s Bar in the Chilcotin for an interesting example, which I believe was where the Chilcotin War was started in the shadow of Mount Waddington.
If there was interest in putting together an improved dataset for import I would be all for it and might have some useful input
5 Likes
I’m getting the impression that almost everyone here thinks Canvec water areas and streams are still useful data. Would it be ok if I continued importing these? Or am I misunderstanding?
Yes, there seems to be general agreement that it’s useful, at least outside of inhabited areas where local contributors are more likely to be found.
However, I do think there are better sources. The CanVec files don’t connect watercourses inside of water areas (see Rivers - OpenStreetMap Wiki for more info). In BC you can download hydrography shapefiles and do a bit of processing on them to get waterways that are properly connected. I’m not sure what’s available in other provinces.
Okay, let’s try to wrap up this thread.
Importing entire Canvec datasets no longer makes sense for all the reasons listed above. However, there seems to be a consensus - not unanimity - that some features might still be useful for imports in sparsely populated areas.
I propose that from now on, only the following features (Josm querries) could be imported into OSM:
natural=water | wetland | glacier | reef | peak [1]
waterway=stream | waterfall | riverbank [2]
highway=* [3]
addr [4]
place=* [5]
[1] includes toponyms (nodes) which may need some edits;
[2] riverbank needs to be converted as natural=water + water=river;
[3] highway=unclassified might be out of date (logging roads);
[4] adresses interpolation - in case some roads exist;
[5] includes cities, villages, hamlets and other named locations.
Unless many people object, I will edit the Canvec wiki page to reflect this proposal in mid-September.
Discussions about other sources should be conducted in separate threads, as new import projects. If these discussions lead to Canvec being replaced (partially or completely) by another source, the Canvec wiki page will have to be modified accordingly.
1 Like
I asked NRCan help desk for the source of the water features in BC. Since I have not yet received a complete answer, I have looked at the Metadata.txt files included in each Canvec/Osm .zip file.
You were right, at least in BC the water features are pre-1985 (not post 2004), except where the 2010 Olympics were held. My mistake.
Brutal.
1 Like
My understanding is that they’re digitized from topos made in the late '70s but I don’t have a source for that. In any case, 40-50 year old features is why we can’t just say “water is okay” for all CanVec. When someone proposes doing an import somewhere, we can look at what is appropriate in that area.
3 posts were split to a new topic: Replacement data sources for Canvec/Osm
The Canvec wiki page has been updated to identify the Canvec features that are considered acceptable for importing in OSM.
A new topic has been create with the latest posts discussing potential replacement data sources for Canvec/Osm.
2 Likes