From memory** what happened was that the I was on a trail that took a fairly gentle left turn, but noticed that the GPS trace had a kink in it. So far, so not very unusual - but then I noticed that the other traces uploaded to OSM had the same left-right kink, and when I went back a couple of weeks (or months**?) later the same think happened.
It wasn’t the sort of multipath confusion that you get with large buildings where a GPS may be all over the place, it was a specific and reproducible offset. Whether modern receivers would show the same issue I’ve no idea.
I was lucky enough to be at a RSGS talk last week with Ordnance Survey professionals describing how they map with GPS and how they map when GPS is not sufficient. They used the term “canyons” (which could be natural or caused by tall buildings). The GPS signals are more imprecise in “canyons” because of “reflections” and “shadows”. In such cases they fall back on laser line-of-sight trigonometrical techniques, triangulating back to a known datum.
From the discussion it seems there might be at least two problems in this question:
fundamentally: does OSM map what can be verified OTG or some geographical “truth” - let’s ignore continental drifts for the moment.
how to resolve the conflict on the ground for the user, who neither knows exactly where they are because they cannot know whether the map or the GPS are wrong, or even if anything is wrong (except for cliff situations) and hence would carelessly leave their triangulation equipment and lasers at home.
There are probably OSM uses where the true location vs verifiable coordinates is not relevant (can’t think of one though) but that would be just as much a specific use case as the other.
This might be academic in excellently mapped areas but is more of an issue where official maps are not trustworthy to begin with.
We are mapping reality (=truth in this context?) that can be verified OTG.
We are not mapping GPS signals, we are not slavishly mapping reality exactly as presented by GPS receivers.
We are using GPS signals to map reality, we are using GPS data and our brains and other sources if present to improve map data.
We are using GPS data (and other sources) to gather info about reality, but these are our tools.
If we map some rock at specific coordinates from GPS - we do it because GPS is our best source about reality, not because GPS display is sacred.
And yes, in many cases we use on the ground info collected not by ourselves - but also collected by other sensors ranging from GPS to aerial imagery (both visible light spectrum and LIDAR and in some cases more exotic ones), satellite imagery, reports from friends, photo recording of streets, government databases etc.
(obligatory note: some of nice resources cannot be used, for example we cannot use aerial in Google Maps)
OTG does not mean that we are not allowed to map while on a boat or that we are obligated to map GPS signal anomalies as reality or that we cannot use aerials or that we cannot import address databases on a compatible license.
But if imported address database claims that address does not exist while it in fact exists and is in use it can be mapped. Yes, also if such address use without reporting it to government database is illegal.
Similarly, if government database recorded single address 45 times it does not mean it should be mapped the same in OSM, if on the ground truth is that it exists once.
If GPS mistakenly claims that peak is 20 m toward north it does not override reality, if we can get more precise info about OTG situation from elsewhere.
“The source” is understanding how trilateration and mathematical error work.
When you have a quarter-sphere reception situation, everything from another quarter-sphere is blocked (strongly attenuated) by the wall, it’s fairly easy to demonstrate how the error distribution peak would be biased on the open side (centered roughly 1/2 of the full error range from the true location).
You are trying to pretend that uncertainty doesn’t exist.
There’s no “absolute” ground truth. There’s a spectrum of methods with various levels of errors. (I know, for some people it’s really hard and uncomfortable to realize.)
Attempts to prioritize one method unconditionally and declare it equal to ground truth are unproductive.
The solution is to be able to evaluate multiple suitable methods of mapping, make an educated logical choice, and indicate the source one way or another in your changeset.
Because “the same” method isn’t actually the same. You are going to use a two-decades-old Garmin, someone is going to use a multi-frequency, multi-constellaction receiver, someone else is going to use an older L1/L2 receiver with differential corrections, yet another person is going to do the same but in a wrong CRS without transforming their data.
Same with satellite/aerial data, but it’s even worse, because the exact same image might have different levels of correction, correction can be done using different DEM data (fresh or outdated LIDAR, fresh or outdated SAR data from different instruments, etc.)
This is yet another reason why it isn’t a “life-saving” question because collaborative maps with unknown uncertainty level should never be trusted unconditionally in life-critical situations.
And yes, there is a spectrum of method to measure it - in the same way as specific shop is existing at the specific location but various mappers may come to a bit different conclusion which shop= value is applicable and different GPS receivers will record different position.
Is someone setup GPS spoofing next to object it does not move its actual location, only distorts GPS receiver output.
If you can’t measure it with no error or a completely negligible error (since we are talking about spatial information in this topic, not about taxonomies of shops), its existence is a complete abstraction that can’t be recorded.
Therefore, for the purpose of OSM, it doesn’t exist. Only more or less accurate records of more or less accurate measurements of it do.
You might be able to claim it does in a philosophical, abstract sense, but you can’t even define it as a mathematical limit to record it as a number or an algebraic equation. But again, I acknowledge that some people, due to certain psychological “inclinations,” can’t tolerate uncertainty to the point of feeling deeply uncomfortable anywhere near it, so it’s understandable that it’s hard for them to deal with this concept.
However, my point isn’t about the existence per se, it’s about unproductive attempts to identify and declare a certain measurement method as identical to ground truth. I hope nobody’s desperate enough to argue about that.
Since any map is a projection of the rugged surface of a planet that is not even a sphere into a 2D plane, no distance on a mountain slope is true on the map etc. These are old geography problems that tickled some great mathematicians. Every map is an approximation only, hence, no map should be used blindly. OSM is anyway far from correct (whatever that means) in most areas. The cliff example was obviously only meant to illustrate the issue of a possible conflict of information, maybe for the “philosophically inclined”.
The underlying question is, does and should OSM aim to map objects to the geometry of earth’s surface towards some sort of theoretical correctness given the available information (plus non-geometrical information), or should it reflect what users experience while in some location with the (best) tools available at that place and time? Mappers integrate whatever information from a variety of sources in front of their screen but if reproducible and consistent conflicts should occur between this bird’s eye perspective of the map and OTG information (GPS signal in the toy example), which version is to be preferred for OSM? What is the value of a map being “correct” when that is not verifyable or even conflicting and misleading in the use case? Personally, I think a map should first of all be useful and robust, which is mostly the case when it is geometrically correct but not necessarily so.
It might help if you expanded on exactly what you are proposing in your toy example. Suppose the path is a straight line, but GPS traces have a systematic kink. Are you saying you would map the path as if it had a kink? And the parallel cliffs as well?
Generally as a few other people said, I think the ground truth of how observable features are positioned relative to each other is important, and I can’t see how you could preserve that if you try to align with local GPS anomalies.
That would be a problematic consequence but since one could see the straight line OTG it would be possible to identify GPS as the problem and resolve the conflict.
Right, but in hilly or rugged terrain OTG distances and angles anyway cannot possibly be the same as in a map’s projection. The 3D surface area is much larger than the 2D map.
Ignoring (or not) such GPS anomalies isn’t a solution but a decision for the OSM community to make. I don’t have a satisfactory general proposal, just a user + mapper perspective, which is why I started this thread to find out if the community maybe had reached a consensus years ago.
My summary is that there is no defined standard operating procedure for such conflicts. Since most discussion contributers tend to a literally higher perspective on the map, practical OTG verifyability, sometimes cited like an OSM commandment, can have 2nd priority over somewhat theoretical correctness in such cases. Fine with me, maps cannot be accurate in all respects anyway and there are bigger problems for the time being (not only in OSM).
Thanks everybody for your interesting contributions.!
