Hello! Apologies if this has been answered elsewhere, I wasn’t able to find any answers in my searching.
A park near me is adding several new paths. I have ridden/walked these paths with a Garmin watch, and exported the GPX tracks. Based on these tracks, I’ve added the new paths to OSM to reflect the new trails (using construction= tags where necessary).
For a large part of the new trail, I did an out-and-back, so my GPX track covered the same path twice. In most cases, the “out” and “back” tracks were no more than 10-15 feet away from each other. So it’s not ultra-precise, but it’s not completely incorrect either. The tracks also lined up fairly well with existing data in OSM when they went over areas that OSM had a clear reference to (existing trails, water features, proximity to mapped roads, etc).
However… Today I came across some threads that basically said “don’t use GPX tracks, because GPS isn’t accurate enough, use satellite imagery instead”. But in my case, this path is brand new (built or started construction this year), so there is no aerial imagery available that I’m aware of.
So my question is: for a new path that I want to add to OSM, that’s new enough that it doesn’t show up on satellite/aerial imagery, what is the best way to get good, accurate data that I can use as a reference for drawing/updating the path on OSM?
A couple thoughts that came to mind:
Walk/ride the path back and forth multiple times, across multiple different days, and take the “average” of all the GPX tracks, using both my phone and my GPS watch to record tracks. Is that considered “good enough”?
Are there Android applications that are able to to get high precision data from a phone’s GPS receiver?
Imperfect GPX-based data is better than no data; as long as I believe it’s reasonably close to ground truth, so just update OSM with what I believe is correct and don’t fret over minor inaccuracies?
My experience with mapping from GPS tracks and then refining from aerial imagery or Strava heatmaps:
One track is usually sufficient to show the approximate route of the trail, but it’s vulnerable to local, sometimes-severe distortion. This is particularly true in areas with heavy tree cover, cliffs, or tall buildings.
Two tracks averages out much of the distortion, but a same-day out-and-back is less effective than two tracks from different days.
Four tracks from at least two different days is usually better than aerial imagery except for fine details such as sharp turns, and is almost as good as the Strava heatmap.
Apparently from some mapper(s) who weren’t contributing to OSM before we had generally good aerial imagery. What you did is fine.
If you can, it would be nice to follow the suggestions by @Carnildo and get some GPS tracks on different days when the satellite configurations are different to help average out the errors. If/when updated aerial imagery shows the trail the come back and refine it.
In the meantime any hiker following the trail with a hiking app will only have a GPS as accurate as yours and probably won’t notice any real errors in the trail location.
Or who have never mapped narrow trails in conifer forests - sometimes impossible to see clearly on aerial images in my experience, so that guessing from images is likely to be no better than relying on GPS.
I agree with what others have said and would just add that consistency with existing data is a key point, particularly joining or not joining existing trails or roads as the case may be. If you get that right it is not too hard to improve the alignment later - you could make a note to yourself to check again when imagery becomes available, if you’re concerned about it.
this advise is wrong and some misunderstanding happened at some step (or multiple of them)
Though locally it may be true where we have excellent, available well aligned LIDAR and aerial (not satellite!) imagery they can be superior in quality to GPX tracks. And easier to obtain.
But even there, GPS may be needed for up-to-date measurements.
Passing the trail two times, I’d say, due diligence taken care of. GPS accuracy should be reflected in the result. So not a node every meter, keep it coarse, will make life easier for late comers. Strava should be the first to get you more GPS readings. The most accurate will be the laser scan, that might take years. No need to wait for either.
@maples37 , you are certainly “on the right track” with your GPX uploads. I’ve been uploading GPX traces to OSM from my GPS (Garmin 60CSx, with 16-channel satellite reception giving accuracy down to 3 or even 2 meters) for decades. Sometimes I’ll be able to see these actual trails from satellite or aerial imagery, but because my hiking / biking is often in areas with tall tree cover, most of the time I cannot see such trails. In those cases, uploading to OSM my GPX tracks of “10 feet or less” of accuracy is fine by me, even as I know there is likely a small amount of “wander error.” But, it isn’t major, and might also be accompanied on the ways with a fixme noting that it is “Approximated” or “Might benefit from more data averaged to improve accuracy.”
Sorta like a B+ (or even A-) is still very much “a passing grade.” Of course, we all love an A+ (very little or even no “wander error”), but I’ll take a B, B+, A-, solid A or even an A+ all day and every day. For those “borderline” B-level data, a fixme or note tag that some improvement might be mixed-in by a future mapper is a welcome addition to flag to other contributors that a way could use some sharpening up to A+ level. If your data are say, 85% or more accurate, please add them to our map. 99% or 100%? Oh, yeah!
Edit: Let not perfection be the enemy of the (very) good.
I haven’t seen the Strava Heatmap Extension mentioned, this will allow you to map using the Strava Heatmap and is perfect for adding trails. I will record my own GPX track when adding a new trail, but for accuracy I also check and see if it’s on the heatmap. Since the heatmap aggregates activities, it naturally converges on the center of the trail.
The heatmap will act like a low-pass filter on the single inputs. When on the ground there is a sharp zig-zag line, a smooth s-shaped-line will emerge in the heatmap. Probably not as much as cutting switchbacks, but similar in vein.