As far as I understand, Strava uses a stacking algorithm to create those nice colored streaks.
It basically means that no matter which direction a road is traveling in, everything will be applied and piled up towards the center. Additionally, the maximum zoom that can be obtained is 16 from which the various mapping and visualization tools zoom digitally.
If you told me that a track layer, whose maximum data display level is 16, was accurate enough to be used as a mapping source, I would be a little suspicious of that statement.
There is also a small problem that perhaps many have not paid attention to and that is the size of the strokes versus the size of the ways, this is particularly noticeable in narrow roads, where Strava tracks overflow or exceed the width of the roads, giving the impression that they are suitable for mapping. No, no athlete was running, walking or cycling over the fences on either side of the roads, So Strava is showing a bias in its data caused by an algorithm whose only purpose is to stack tracks and give them a nice color.
An 11 meter wide road with Strava on top that is approximately 40 meters wide.
Basically if you have a steady hand and aim at the center of the track youâll think youâve made a good mapping, assuming the image is off by a couple of meters (something common in mountainous areas), that could lead the mapper to believe that the mapping he made is quite accurate? who made his mapping following the center of the Strava track. Again, a 40m wide Strava track on a road that is a maximum of 11-12m wide does not strike me as a reliable mapping source.
But very usable to see if thereâs a missing road in OpenStreetMap and relate that to the aerial imagery, no?
Due to all stated facts in this thread the road shows up wider on Stravas Heatmap than it is in reality, but weâre primarily mapping roads ways and not areasâŠ
to me this looks perfectly fine. Yes, you cannot tell whether there is one way or maybe parallel ones, and I agree you shouldnât base your mapping solely on Strava, but if you know the place these traces are very useful, especially where you canât see the ground on aerial imagery, or to check the imagery for alignment and distortions in steep terrain
Sometimes the absence of a trace in Strava can also be very useful. I quite often find tracks and paths mapped in OSM from aerial images that are in fact private but not tagged as such - that problem didnât start with Strava. These often show up in the Strava heatmap as suspiciously activity-free routes in an otherwise busy area. As with many of the other examples above, if you know the area, this kind of information can really help improve the map, or at least suggest areas worth surveying.
Strava can be also useful for improving impecise geometry of paths invisible on satelite. It usually does a good job of noticing zigzags and if the source of the trail is one GPS track, it can be imprecise, strava averages dozens or hundreds, is is more precise in my experience.
It is very useful when combined with other resources, they complement each other. I have used it when I am following a road and I find an area with clouds in all the satellite images (Bing, ESRI, MapBox).
As you say, we are not mapping areas, however the fact that Strava is showing such a large area in proportion to reality is an indication that something is not right. Regardless of whether or not transparency is used to see the satellite image below, Strava can cause mapping errors. Are you sure that the extra 400% width of a road has no effect? The result is 2 wrong sources, satellite images and Strava traces.
Returning to the example of the 11-meter-wide road, what would be the correct position of the road? The center, a little more up or down from the center of the Strava trace? this in itself defeats the purpose for which Strava was being used, I mean, seeing that whole area of colorful tracks and not knowing where to draw the road. I remember a little conversation I had with a guy from a paid mapping campaign where he brought up that he had used Strava for the alignment of a road he had modified in a strange way, he sent me screenshots showing me that my previous version was a few meters out of sync with Strava, I just replied that which part of that whole Strava area was my previous version out of sync with? You see, Strava is used but the mappers donât really know where to draw a road.
This is not correct. The GPS traces upon which Strava is based are individually not very accurate. Sometimes they are going to be in error in one direction, sometimes in another. However, the average of many traces - which is what the heatmap is showing us - is going to give the accurate location of the trail or road (with the limitations that I an others have pointed out in this thread). The correct location of the road or trail is in the center of the Strava âheat.â
Someone mentioned âbiasâ: Strava generally doesnât have a âbiasâ, the individual traces have errors, but those errors are unbiased, that is just as likely to be in error in one directions as another. Unbiased errors can be removed by averaging, which is essentially what the heatmap is doing.
I have compared Strava to the USGS 3DEP data, which the USGS says is accurate to within one meter, and they agree very closely.
Mostly agree, but this one not: Strava turns ZigZags into SigSags. I noticed when this three nodes ZigZag got turned into a 10 nodes SigSag citing Strava as a source:
On the ground there is an improved path strictly forming kind of a Z. The DTM shows it very good. Strava shows an S. Perhaps Strava users do not walk the path? Perhaps it comes from the algorithm? Change made me wonder if the mapper was there at allâŠ
Said that: Strava is very good when paths got relocated. Aerials and LIDAR scans only come every so often, Strava is much quicker in taking up. No need to rely on a single GPX of mine alone.
I just checked and today this is not the case anymore.
A while ago, selecting âall sports on wheelsâ included virtual activities. Today the problem seems to have been solved. Virtual activities will appear only if you select âvirtual cyclingâ.
Slightly OT and nitpicky: naturally you can have a bias in GPS locations, urban and natural canyons and other situations with limited view of the sky or/and situations that affect reception (for example by reflections).
Because of the way GPS traces are captured, this is to be expected. Think of how the data is being collected - a GPS watch will perhaps record once every second. If youâre running through a Z shaped turn in only 2-3 seconds, what are the odds of the timing of recording matching up with the turns? More often than not, the points a GPS device captures will be placed at random points in the Z turn, and when you draw a line through these points you will often get an S rather than a Z. When you travel through the Z, the switchbacks are the outmost extremes. Unless you happen to record a point at the exact same time as youâre in that spot, the GPS trace will not be a 100% precise rendering of the way you traveled.
This is one of the limitations in using GPS traces, and unless youâre capturing a lot of points with an incredible precision, all GPS tracks somehow leave out some of the pathâs details. This is why we use other sources (aerial imagery, LIDAR etc.) to increase the level of detail rather than just importing GPS tracks.
