I have seen a lot of notes and edits where people say it is based on Strava heatmap. It seems to me people are basing too much on one source that may not be completely accurate. I have looked in to Strava a bit and maybe I am missing something. Is there some reason we can use that as a reliable source?
I wonder, how do we know that the people (maybe sometimes a small number of people) are even sticking to trails or roads. Could a heatmap be based on one or two people who bike or jog in a certain area?
Can someone explain why Strava is being used as such a “reliable” source? For example, trails that are not visible through tree cover getting added because of Starva heatmaps.
Thanks for some insight on that. I learned about Strava because of editing on OSM. So it is pretty new to me. In your link, you say “I sometimes map based just on Strava… usually only in areas where I have previously been…”
Once you have been somewhere in person, it changes the context a lot. Even if you haven’t been down a road, you know the terrain and how that effects how road and paths are done and how people use them. I have seen notes of people armchair (nothing wrong with that ) editing. They usually don’t seem to know the area. I wish I could think of an example now… but I cannot.
EDIT: I am looking around Strava for the first time, in some places I am very familiar with. I can see why people put trust in it. But I am seeing some things that seem very off at the same time.
Strava can be highly accurate in showing the location of where people travel while exercising/training, with the exceptions listed below. However, it doesn’t tell you whether they are traveling on a road, trail, or something else. It doesn’t tell you if access is legal, it doesn’t tell you the surface, difficulty, etc.
Exceptions:
The heatmap is for the past year, so if a trail was recently closed, it will still show “heat” for some time.
Unless filtered, the heatmap is for all activities, including winter only activities done on winter only routes. These should not be mapped as highway=path/footway/cycleway/bridleway/track
For roads and tracks, if people preferentially spend more time on one side of the road than the other (e.g. they prefer doing a loop in one direction vs the other, or there is a hill that slows them down), the heatmap will be skewed to that side.
For wide features, such as roads with cycle lanes on either side, the heatmap will be a binomial distribution. If you are mapping the road you need to pick the middle of the “heat”, and not the “hottest” location.
The heatmap is an average, as such it is a low pass filter, and it will tend to smooth out sharp corners and switchbacks.
The heatmap is updated monthly, so new features, and reroutes of existing features will not immediately show up, and even after the first update or two there may not be much heat.
This isn’t directly related to your post, but I wanted to share this tool: Patchwork - a Strava-developed tool finding potentially-missing ways in OSM!
Looks like a neat site, but when I zoom/pan away from the default location I never see any orange lines indicating missing ways, even in areas where I know there are some. I tried it on both Firefox and Chrome.
when people talk about “strava”, I did never think someone was referring to these faint 1-pixel lines, rather it is about the wider and lighter, aggregated ways which show significant traffic (right bottom corner of your picture)
Some of that might be noise (all the more reason not to use them). When you start the GPS in your phone or watch, there will be a bit of time where the accuracy will be very poor.
I’m heavily relying on Strava heatmap for mapping trails. In my opinion, it is a great way for confirming that trails exist by seeing a substantial amount of activity that matches the same twists and curves.
Some thoughts:
Log in – You really need the most detailed heatmap you can get, so get yourself an account. If you’re using ID, I strongly recommend these add ons:
Filter for sports – I use a background map with ALL sports, and then an overlay with filtered sports. This isn’t perfect, but it’s really effective for filtering out winter and water sports. Sometimes winter sports and foot activity overlap, and you have to be pretty careful when mapping in those areas.
Default mode – Many users simply fire up Strava on their phone and start tracking, not changing the activity in the app to match what they’re actually doing. I see this a lot in cross country ski tracks and ski resorts. This can also happen with GPS watches of course. For example, sometimes people do a novelty ski touring trip and couldn’t care less if it’s logged as the appropriate correct in strava or not. They just want to record an activity, so sometimes you’ll get walk/hike heat in what is clearly just a cross country track
Sources - You should back up your mapping with other sources, both as confirmation that the trails exist and for getting as much detail as possible. Sometimes, tree cover can really make it hard to use satellite imagery, but every once in a while, the trail pops up in between the trees. I’d advise everyone to follow the trail using at least one satellite imagery, and sometimes you’ll have access to imagery captured at different times of the year. Trails are way easier to spot if there’s no leaves on the trees. LIDAR scans can also be really useful in these instances.
Offset - Strava heatmap can be more precise if satellite imagery has offset, but you should be careful.
Off trail heat – yes, two people can wander off together and create two single jagged lines. This is, however, not sufficient heat to confirm a trail. Often, off trail heat does not coincide as precisely as heat from a fixed trail. There are hiking routes that doesn’t stick to any trails, and there have been a lot of discussion on whether we should map out establishes routes that does not follow a path (crossing glaciers, scrambling, easy off trail ridges etc) or if we should only map what’s on the ground. I’m liking the latter.
Trails without heat – Strava heatmap is of course not a judge of whether a trail exists or not, and there’s plenty of trails with no heat. This can be found in areas with a low population density. All it takes is just a bunch of folks not really feeling the need for Strava when they’re outdoors. Using other maps of trails and sattelite imagery can be a good way of confirming these trails.
Game trails - Be careful not to map game trails if they’re not used by people. Game trails will sometimes go into thick bushes and come out of it on the other side, and if people had been using it as a trail the vegetation would be pushed back a little more.
Races – Coinciding lines in Strava heatmap doesn’t automatically equal a trail. A race with a lot of participants that goes off trail can look like a clearly defined trail in Strava heatmap. Consider the surrounding area – if these trails are the only ones with a lot of heat, you could maybe be looking at a race course.
Orienteering - Areas that are often used for orienteering will generate a lot of heatmap activity, often all over a pretty small area. Orienteering is usually an off-trail activity, and this heat does not necessarily equal trails. After a while, this orienteering heatmap pattern is easy to pick up on. Golf courses can also look a little similar in that sense.
Unsure? - If you’re not sure, you’re sure about being unsure. What’s worse – to be led down a mapped-out trail that disappears after a while, or to stumble upon and on purpose use an unmapped trail? Think of the user – the first can put someone in very unwanted, potentially dangerous, situations. The latter… well, at least it’s the user’s own fault. It’s way better to not to include a trail that exists than to include a trail that does not exist. Mapping trails that you’re not sure of can also reduce the trust in OSM as a map, which is a shame as a lot of people dedicate a lot of great work into it.
I live in an Island where trail running events occur multiple times a year. Logically some participating athletes use strava during the races.
The thing is that even though the majority of the race goes through public highway and trails, some sections go through private property (on farmlands and meadows) under an agreement with the land owners.
I’ve seen some paths being mapped under this circumstances, and that’s where I see that mapping using strava goes too far.
I’ve also seen valid instances where trails were mapped that I didn’t know about and it’s fun to discover.
I’ve been trying to correct the situations where a path doesn’t exist unless it’s a trail run day.
Also, pay attention to virtual cycling activities (Zwift or likes) on the Strava heatmap. These activities will generate a lot of heat at a few locations, usually around some large city, with absolutely no connection to actual terrain.
Around ZĂĽrich for instance, there is a major virtual track that crosses mountains at a straight angle, then dives into the lake.
By the way, is Strava data OK for mapping, in terms of licensing?
There is no general, global rule whether you can enter private land, it depends on the region, in some places you can go everywhere in the nature unless it is protected (freedom to roam) and in others, you may not enter private land without the landowner’s permission. Generally, in OSM we allow to include all highways that exist and we add appropriate access tags to them.
) editing. They usually don’t seem to know the area. I wish I could think of an example now… but I cannot.