That’s what I thought as well.
An assumption could be that the steepness of a path and its surroundings (based on external elevation data) somewhat correlates with sac_scale and, more importantly, would give an indication how dangerous a path might be due to the risk of falling.
But we should probably discuss this in a new thread, separate from the tagging discussion?
I’d also liked so see such an analysis made. @mods-general be so nice and split post by @ikonor and referenced posts to a fresh topic? Maybe “Tooling to find paths missing sac_scale” (please invent a better title.)
On the new topic: Administrative GIS in public WMS has a slope layer, 1px = 5m colour graded in steepness quantiles. Something like that should be the most convenient reference data to perform the analysis. Its published ODG CC-BY. Not aware if e.g. Sonny offers that?
Still will be hard 
The analysis would have to use very advanced methods. There is no other source for T* gradings apart from openstreetmap almost anywhere on Earth. There are some routes in Switzerland graded by the SAC, but they do not grade paths, they grade routes, mostly from the PT stop to the mountain hut of theirs, and some easy summits close by. There is also hikr.org, but this is not a geocoded database.
Personally, I also not so happy to slap sac_scale on e.g. any informal path. This may change with a single thunderstorm and nobody to heal it. Highly volatile data.
PS: If I remember correctly, outdooractive used to copy sac_scale from openstreetmap into their free tier, later only in their paid programme, but as of now, they seem to have completely dropped that, I cant find it mentioned in their documentation. Tour guides of our local rambling club used and liked that. I am certain they are not aware that they can update openstreetmap data if they find something in error.
@Hungerburg Please let me know if I moved everything you wanted moved. Schöne Grüsse.
I think most tools assign raw elevation data to ways and do the slope calculation afterwards. My idea would be to use a high resolution terrain model in a test area like this one:
ALS DTM Höhenraster 1m Stichtag 15.09.2024 (11 GB GeoTIFF).
Routing engines like BRouter assign elevation and I found some other tools on a quick search:
But I don’t know if any of these would support higher resolution data or if there are better alternatives?
At least QGIS can open that 1m file and might do for some initial manual testing.
Thanks! So first of all, here is a link to a map of paths that don’t have the sac_scale tag, and also don’t have a paved surface, or the surface tag isn’t set. The query can probably be refined, and then something like this could be used to identify paths where it might be worth adding sac_scale or surface. But there are millions of them, it shows a lot of urban paths where it’s debatable how valuable it would be to add sac_scale to all of them. Identifying paths that don’t have the tag is the easy part, identifying the ones probably should have the tag and are priorities for surveying is the harder part.
Slope could be one way of finding the worst offenders, like @ikonor suggested. But that needs a high resolution terrain model, so I think that could be useful if you want to identify paths in your local area to prioritise. My original idea was more to get a rough measure of the coverage in different regions of the world, as an indication of mapping progress.
Something like x% of paths in the Alps that probably should have the tag have it but only x% of paths in the Himalayas have it and x% in the Rocky Mountains. If you want to do slope calculations for all paths on the entire planet that would need a lot of data and compute.
For a simpler measure, how about identifying paths at a high altitude? For that purpose we could probably get away with a much lower resolution terrain model. Of course, the altitude that a hiker would think of as “high” will be different in different places. As a crude approximation, paths above the treeline?
Another option could be to look at something like Strava data to identify the most frequently hiked paths that don’t have sac_scale tags. Or looking at paths where people get into trouble the most. You can do this manually, looking at news reports. A more automated approach might be possible if there’s data, for example of mountain rescue incidents, with coordinates. But there probably isn’t a global dataset like that.
In case it helps give anyone any good ideas, I wrote a while back about identifying steep streets that might really be steps/staircases but were not tagged as such - a different but somewhat similar use case.
For that objective, high resolution terrain data was essential, as a single set of steps normally covers a short horizontal distance, and an urban area with a lot of steps also usually has a lot of streets with high gradients that are genuinely not steps. Maybe the approach is overcomplicated for sac_scale - it might be better to look for “areas where we would expect every path to have sac_scale” (equivalent to the suggestions in the previous post) rather than “specific ways we would expect to have sac_scale”.
Anyway, here it is in case anyone wants to look into this approach (I don’t have time myself at the moment).
I guess you are on to something: The tooling already exists 
For completeness a screenshot with bikerouter (top) and editor (bottom) pasted.
The path runs mildly steep through rather steep terrain, red 40-45, light purple 45-50, dark purple 50-60° incline, terrain model 1m resolution. Even not knowing the very area, visual inspection from the arm-chair alone would tell me, this is a candidate. SAC scale (at least as the SAC understands it) is a lot about terrain, not only about ways.
The bikerouter graph overestimates steepness because the path is not mapped correctly, but in a short section off by some five to ten meters from where it is on the ground. I chose bikerouter because it is said to use the most accurate terrain models for routing.
Depending on the threshold, programmatic analysis might get to the same conclusion.
Thanks, I haven’t thought of it, but now that you mention it, I remember that as nice analysis.
That said, re-mapping of a remote path in mountainous area needs … a mapper going through the path.
Quite easy to find out that it was mapped 5 years ago, anyway it needs re-mapping.
One thing I have found useful is to look for paths in OSM that have a suspiciously low level of activity in the Strava global heatmap. It can be a sign that a trail has deteriorated to the point it is rarely used (although at the relatively low altitudes where I usually walk, it is more often related to access).
Of course that is not enough to update these paths without a survey - some perfectly good paths are just not that popular. But it’s useful to have an idea before setting out of paths to be investigated.
I don’t know if there are any tools that would make this easier. I seem to remember a tool that highlighted possible missing ways in OSM based on heatmap data, but that is the opposite situation.
Have you heard of https://mapterhorn.com/ ?
I hadn’t seen that before, thanks. Will have a proper look when I have time.
I guess paths that connect to other paths that do have a sac_scale tag are good candidates. Paths that are within national parks, nature protection areas, etc. also are.
As a quick proof of concept I have done this in QGIS. I downloaded elevation data from NASA’s SRTM for Austria and Nepal and “draped” all highway=paths (from Geofabrik extracts) over it so I could filter by mean path elevation.
Austria has 239,357 paths. 10,622 have a mean elevation of over 2,100m. 7,676 (72%) of them have sac_scale.
Nepal has 258,721 paths. 3,367 of them are over 3,800m, and only 905 of them have sac_scale (27%).
That’s just two countries, and there is more to consider, for example there are entire villages in Nepal above 3,800m, so it might be worth excluding paths near buildings from this count. But it’s still interesting how the coverage differs, this makes me think it would be worth doing for the entire planet (in Python probably). Then the next question is how to motivate more people to add the tag to those paths where it would really be worth it…
Hi, you query selects all the paths above timber line. I’d be curious to learn the summarised length of them. Would you share that?
Practically, this will miss a lot of paths where sac_scale useful. An example: Last April a path was deleted from data with the reasoning, too many unexperienced people hiking there and calling for mountain rescue. This was reverted two months later. 1500 m above sea level top there.
The paths were sac_scale is missing most sorely are those that allow planning round trips, that provide connections, shortcuts and such. So the reasoning of @rhhs not so bad, paths that connect to paths that have the tag.
BTW: The path mentioned had been tagged demanding_alpine_hiking three years ago by someone with apparent ties to the local tourism agencies. This made it disappear from outdooractive maps. The person to delete it seemingly ties to mountain rescue. sac_scale tagging enough?
My first idea was that we could do some live analysis in the browser with vector tiles and Mapterhorn terrain tiles, but neither the OpenMapTiles nor the Shortbread schema support sac_scale and it would require zooming in quite a bit. Probably less suited for larger batch processing.
Good question. I’ve checked: In Austria the combined length of paths over 2,100m altitude is 7,450km, but only 1,685km don’t have the tag. So the 28% of paths that are missing the tag represent 23% of total path length.
In Nepal, the combined length of paths over 3,800m is 4,016km, and 2,725km are missing the tag, that is 73% of all paths and 68% of total path length.
In other words, longer paths are slightly more likely to have the tag, but the difference isn’t huge.
Overall I was looking for an approach that is easy to automate for large regions and minimises false positives. I am sure it misses many paths that deserve the tag..
Is there anyone of this opinion this anymore ?
Obviously, it is a niche.
Off-topic alert: The mapper local to the path mentioned above did reply. There is also another “informal” path joining the deleted one, no sac_scale tagged, where mountain rescue gets called. Callers purportedly mostly long-distance-hikers, acc. to the landlord of the nearby hut. I once talked to one of such: She used maps.me. I then found, it does a great job of visualising trail_visibility.
On the back of my envelope, I calculate that 28% paths missing the tag represent less than 10% of total when considering distance.
