Temporary use tag to map potentially dangerous highway=path

alan_gr · 2024-10-22T07:04:14.034Z

dieterdreist:

If we want to make sure that people consciously show potentially dangerous ways on their maps, it would be safer to use a new highway class (or more) permanently, not temporarily

That makes sense. With a temporary tag, we are actually reducing the incentive for renderers to look at secondary tags, as a lot of dangerous paths would temporarily disappear from the map. Then as these paths are gradually surveyed and retagged, they will simply reappear in general purpose renderers.

Here are two paths as shown in the default renderer. One is a wide smooth path in an urban tourist area that provides a short cut between paved roads. The other is the route to the summit of Mount Everest.

The route to Everest is already tagged as sac_scale=difficult_alpine_hiking. If I understand your proposal correctly, nothing would change here. So I’m not sure the proposal really helps with the issue of dangerous paths in general purpose renderers (not just the fact they are rendered, but the fact they look pretty much the same as easy urban paths).

It seems you view the main problem as paths of unknown difficulty level that have not been surveyed on the ground. My impression is that many of the discussions around this problem have been about paths that have been surveyed and mapped with appropriate tags, but those tags are not reflected in general purpose rendering.

osmuser63783 · 2024-10-22T07:57:34.708Z

I agree with @rhhs that paths with unknown difficulty level that have been added without being surveyed are a problem (or maybe they have been added after survey, but that was 10+ years ago, when tags like sac_scale were not as popular as they are now.) There are lots and lots of bare highway=path without additional tags in mountain regions. It’s easy to say just don’t map them without adding a difficulty tag at the same time, but this is really about what to do when you notice that someone else has mapped one.

Relevant discussion on a related topic:

[RFC] Deprecating `highway=road` in Favor of `highway=unknown`

When hiking, I’ve noticed ways that had clearly been drawn from nothing but the Strava heatmap, or worse, nothing but a single hiker’s GPS trace. They were tagged highway=path. Sometimes there was an actual path there, sometimes not

[RFC] Deprecating `highway=road` in Favor of `highway=unknown`

Things for which the only source is Strava should not be in OSM.

If someone does it, please explain to them why it’s a bad idea (starting with what you’ve described above). If they persist, report them to the DWG so that we can persuade them.

But we’re talking about thousands and thousands of paths, form the Canadian Rockies to Tajikistan, so I am wondering if OSM needs a more scalable solution?

Some tag that somehow says “don’t trust this path” could be useful. (For an experienced mapper it’s clear that highway=path, in a mountain area, last edited 10 years ago, has this meaning, but this doesn’t seem to be clear to data consumers.)

dieterdreist · 2024-10-22T09:18:23.000Z

There are lots and lots of bare highway=path without additional tags in mountain regions.

it is quite clear that these are likely not comparable to paths in cities. Any path in a mountain region is implicitly different from one in built up areas, while there may be some exceptions you will have to expect obstacles, hazards and problems in remote areas that you are less likely to encounter in places with more human frequentation.

julcnx · 2024-10-22T09:59:23.230Z

osmuser63783:

Relevant discussion on a related topic:

[RFC] Deprecating `highway=road` in Favor of `highway=unknown`

When hiking, I’ve noticed ways that had clearly been drawn from nothing but the Strava heatmap, or worse, nothing but a single hiker’s GPS trace. They were tagged highway=path. Sometimes there was an actual path there, sometimes not

[RFC] Deprecating `highway=road` in Favor of `highway=unknown`

Things for which the only source is Strava should not be in OSM.

If someone does it, please explain to them why it’s a bad idea (starting with what you’ve described above). If they persist, report them to the DWG so that we can persuade them.

These issues aren’t just limited to paths; individuals and organizations often add bare highway=residential/unclassified/service ways using only imagery or GPS traces, and without any local knowledge. Some of these roads turn out to be private, some are blocked by walls, while others are too rough for standard cars (i.e., “dangerous”).

The core problem is that OSM lacks an effective system to mark roads as verified or unverified, and simply discouraging mappers from adding incomplete or unverified data isn’t a real solution.

SomeoneElse · 2024-10-22T11:45:38.434Z

osmuser63783:

[RFC] Deprecating `highway=road` in Favor of `highway=unknown`

Things for which the only source is Strava should not be in OSM.

If someone does it, please explain to them why it’s a bad idea (starting with what you’ve described above). If they persist, report them to the DWG so that we can persuade them.

But we’re talking about thousands and thousands of paths, form the Canadian Rockies to Tajikistan, so I am wondering if OSM needs a more scalable solution?

A more scalable way to get Strava-only “paths” into OSM? No, that would be actively harmful to the project.

Getting some survey details from people who actually use these paths regularly would help, but that’s not something that can be done solely from behind a keyboard.

SomeoneElse · 2024-10-22T11:47:42.915Z

julcnx:

Some of these roads turn out to be private, some are blocked by walls, while others are too rough for standard cars (i.e., “dangerous”).

As I said previously, if they persist in doing that after it has been explained to them that their work is of poor quality, report them to the DWG so that we can persuade them about it.

julcnx · 2024-10-22T12:27:29.087Z

SomeoneElse:

As I said previously, if they persist in doing that after it has been explained to them that their work is of poor quality, report them to the DWG so that we can persuade them about it.

Actually I’d suggest recommending the use of highway=road when mappers add new segments they haven’t confirmed on the ground. These geometries are still useful and encourage both mappers and end-users to verify and update them.

SomeoneElse · 2024-10-22T12:30:54.558Z

julcnx:

Actually I’d suggest recommending the use of highway=road when mappers add new segments they haven’t confirmed on the ground.

In cases where the only “source” is Strava, I wouldn’t.

julcnx:

These geometries are still useful

If there are other sources (perhaps imagery) to back Strava up, maybe. If Strava is the only source no. All it means is that “someone’s GPS thought that it went from A to B”. We have no idea if that corresponds to any sort of path.

We don’t have to fill OSM with crap just because we can.

julcnx · 2024-10-22T12:41:06.117Z

SomeoneElse:

In cases where the only “source” is Strava, I wouldn’t.

What if the only source is imagery ?

dieterdreist · 2024-10-22T13:03:41.894Z

SomeoneElse:

If Strava is the only source no. All it means is that “someone’s GPS thought that it went from A to B”.

on the heatmap you get the result of averaging lots of recent traces of “someone’s GPS”, although I suspect it may be true that some people are cheating strava by uploading faked traces (or is this not what you were hinting at?), possibly so often that they appear on the heatmap, from my experience this is not very common (at least not around here, I have never noticed indications for it)

SomeoneElse · 2024-10-22T13:11:53.861Z

dieterdreist:

on the heatmap you get the result of averaging lots of recent traces of “someone’s GPS”, although I suspect it may be true that some people are cheating

Where I have seen this it isn’t faked GPS traces, but people cutting corners on the ground, and inaccurate traces under e.g. tree cover

SomeoneElse · 2024-10-22T14:26:33.568Z

julcnx:

What if the only source is imagery ?

If it was a part of the world that I was completely unfamiliar with, I’d personally be very reluctant to map anything. However, I know that plenty of people do, and there are places like North Korea that simply wouldn’t get mapped at all if no-one mapped them remotely.

However, sometimes “the only source is imagery” but you can do a much better job of interpreting it because you’re somewhat familiar with the area, and can make a pretty good guess even about things like surface.

osmuser63783 · 2024-10-22T15:50:42.109Z

SomeoneElse:

A more scalable way to get Strava-only “paths” into OSM? No, that would be actively harmful to the project.

Getting some survey details from people who actually use these paths regularly would help, but that’s not something that can be done solely from behind a keyboard.

No, I mean precisely the opposite. A scalable way of making sure more of them are tagged with their difficulty, or removed if they don’t exist, or not added in the first place if we can’t be sure that they exist, or if they are added, don’t show up as paths in hiking apps.

Sometimes I become aware of such under-tagged paths in OSM, when I am out on a hike, or when I hear about a mountain rescue incident and look up the path on OSM and check how it’s tagged. I then leave a note or a changeset comment. I am sure others to do. But I was wondering if there’s a more scalable approach, or if we need to organise better.

@rhhs is suggesting that we mark paths as “I’ve noticed that this has been added as highway=path with no other tags, I have good reasons to believe that it is dangerous, I am not sure if it’s responsible to have this in OSM in this form, so it would be really helpful if someone surveyed it” and I was expressing support for that idea.

osmuser63783 · 2024-10-22T15:59:16.275Z

I think we all have immense respect for the work of the DWG, and it would be valuable to hear what you think we should do. I know you encourage people to go out and survey instead of sitting in front of keyboards, I do that too, but we have “paths” in OSM that are only used by a handful of mountaineers a year. There are limits to where we can go and what we can survey. What can OSM as a community do about those? Especially in areas where sac_scale coverage is poor, and it’s been years since a mapper last walked the paths, if ever.

arctic-rocinante · 2024-10-22T21:17:50.970Z

SomeoneElse:

If there are other sources (perhaps imagery) to back Strava up, maybe. If Strava is the only source no. All it means is that “someone’s GPS thought that it went from A to B”. We have no idea if that corresponds to any sort of path.

Isn’t there some nuance to this?

Consider the screenshot below. I definitely wouldn’t add the line circled in red to OSM, as it’s almost certainly a single person bush-bashing (or erroneous data).

For the line circled in yellow though, I’d be very surprised if there weren’t some kind of OSM-worthy path at that location.

IMG_66091242×1763 343 KB

(Apologies for the messy hand-drawn markings )

dieterdreist · 2024-10-22T21:34:11.413Z

yes sure, it really depends on the area as it seems, I have been using Strava for remote areas (mountains) for signposted hiking routes, so it was mostly clear because they were more like the broad line that is yellow in your example, those tiny faint lines are not reliable of course but they were also rare in the areas I looked at, there was either a bright broad line or nothing at all.

yvecai · 2024-10-22T21:54:38.560Z

Could also be a one time event with many runners. I already suppressed nordic ski pistes mapped through Strava in my area because of this. Strava is full of traps without local knowledge.

SomeoneElse · 2024-10-22T22:23:53.049Z

arctic-rocinante:

I’d be very surprised if there weren’t some kind of OSM-worthy path at that location.

So would I, but I’d still check it against some other source. Someone (or a group of people) doing the same run every day and jumping over a wall to save a few seconds will cause that effect.

on another note:

osmuser63783:

There are limits to where we can go and what we can survey.

I really wouldn’t underestimate what’s possible there… I can think of one local-ish mapper who set themselves the target of walking every public right of way on their “local” Ordnance Survey map and seems to have mostly completed that (99% with a mobile phone app).

Looking at it from the other side, putting tools into the hands of the people who actually do use these paths has to be the way forward. Outreach initiatives such as those by OSM US’ Trails Working Group are also hugely beneficial.

Vinzenz_Mai · 2024-10-22T23:00:44.604Z

SomeoneElse:

I really wouldn’t underestimate what’s possible there… I can think of one local-ish mapper who set themselves the target of walking every public right of way on their “local” Ordnance Survey map and seems to have mostly completed that (99% with a mobile phone app).

It’s a great way to resurvey an area and provides a lot of material for map changes. There’s a crazy mapper near where I live:

image897×835 171 KB

julcnx · 2024-10-23T03:53:25.914Z

arctic-rocinante:

For the line circled in yellow though, I’d be very surprised if there weren’t some kind of OSM-worthy path at that location.

Exactly! In my experience, thick, popular lines like this—often stretching kilometers—are always a well-used path.

arctic-rocinante:

Consider the screenshot below. I definitely wouldn’t add the line circled in red to OSM, as it’s almost certainly a single person bush-bashing (or erroneous data).

Who said mappers using the Strava Heatmap are the ones adding these edge cases? The same goes for a single GPS trace on OSM or a faint line visible on satellite imagery.

yvecai:

Strava is full of traps without local knowledge.

The same could be said about tracing from imagery alone! In my area, I’ve found that anything traced from Strava heatmap is way more accurate than something traced only from satellite imagery.

Of course, I’m familiar with the area, but I just don’t get the specific hate on Strava. It’s just one of many valuable sources, especially in places where high-quality satellite imagery and GPS traces are lacking.