Mapped features eventually change. The average frequency varies from one kind of feature to another, maybe even for each tag of a feature differently. The frequencies can also differ by country.
The number of changes in specific keys can be an indicator, but more likely reflect mapping activity.
It is a bit unclear to me what you’re trying to ask here. Are you talking about just features changing in general? because that can be anywhere from decades to minutes. It really depends on very specific circumstances. Without further clarification, I don’t think that question can be answered.
If you’re asking about how frequently objects should be changed. As a general rule of thumb, if a change lasts shorter than a few months, you probably shouldn’t add that temporary change to OSM.
For things like constructions, it is fine to add a construction area to OSM, but it is generally not adviced to try to keep track of construction progress on a day to day basis. Of course this will vary a lot per case. A big highway going through a city is likely to get a lot more attention than some random suburb being constructed.
I think all you would be able to say is that the frequency of changes to local objects (so explcuding objects like municipal border), will be roughly proportional to the amount of users in the area. But even then, changes could be anywhere between a minutes apart, likely due to some streetcomplete editor adding a whole bunch of info in a short amount of time, to several years.
Do you mean how frequently the mapped OSM objects are updated in OSM? (This could be determined from historic OSM data - although you might need to think about how to define a significant change to an object.) Or do do want to know how often the actual entities undergo a significant change in the real world? (This might be much harder to determine.)
I’m interested in changes in the real world grouped by OSM features, keys or tags, not mapping activity (though the later could be an indicator).
In the past, I think I had seen a study by a GPS company about traffic related features.
PS: Interestingly, park benches tend to be placed (or replaced) at the same spot and allow comparing ortho photos over the years. So the frequency may be much lower.
There is no guaranteed relation between real world changes and OSM edits. There is no distinction between edits that reflect recent real world changes and edits that just improve or adjust existing features.
Taking a bench for example: It could be that a specific bench has been there for 10 years already before getting mapped. It could be that a bench gets replaced in the real world, without OSM being updated. It could be that in OSM someone at some point decides to add more details to the bench, without any real world changes having prompted this edit. It could even be that a bench was removed years ago, but no one noticed that it was still mapped in OSM, so no one removed it from the map
Maybe the opposite needs to be done: try to find studies about real world features and their rate of change and then try to determine how this translates in changes of features/keys/tags.
I’m not aware of anything previously published on this, but I’m sure that there will have been - maybe someone will chip in with previous examples.
What’s in OSM is only a proxy for what’s in the real world - OSM data changes as more things are mapped, it changes as people sometimes decide to use different tags for the same thing, or “remap it better”, or “move a shop across town as the shop has logically moved”, and it also changes as a result of the real world changing.
The “historical state of OSM” is easy to obtain - there are historical “planet” downloads and historical extracts (and here “historical” means both “a regular OSM compressed XML file at a point in time” and “a special historical format containing all data”).
For broad-brush key/value changes, taginfo has a few tag history links that can be useful and you can also do more complicated Overpass queries at a date in the past.
I’d suggest clearly defining a tiny initial query as a thought experiment of the things that you are interested in, then define what that would mean in terms of changes to OSM tags and values (for example, do you need to check that node XYZ still exists, or that something with name XYZ and certain tags exists in a certain place?), and then seeing what tools you need to do that.
No we didn’t, the changes that we missed (and some haven’t been fixed yet) were errors erroneous assumptions in the original border dataset and later policy changes that were not announced by the relevant government agency.
that would only help in cases where mapping is and was consistently to a known degree of reality. Municipality borders are a good sample in principle (coverage is and has been high and changes are fairly well published, at least most of them).
One could easily expand this to the name of municipalities, then the website, population (where one bothered adding it), etc.
The previous month’s project for OSM.ch mentioned 800,000 street lamps in CH. One could then check how many are added in 1 year, how many changed. (coverage by OSM is estimated at 6%, so changes within OSM are unlikely to help).
I tend to notice that stable features that we wouldn’t expect to change much tend to get mapped and tagged once and stay the same, I would put things like roads, buildings benches into that category.
I was more curious about shops so I did look at how long since the last edit for the shop= tag across Scotland:
I think some of the peaks probably correspond to organised mapping efforts to map/update POI’s. If I get a chance it would be interesting to see how this compares to other POI types and also across different regions.
For shops in France, I came across https://www.complete-tes-commerces.fr It uses a public registry of business locations (compulsory I guess). Stats on this “SIRET” by sector could help determine rates of changes to be expected.
For shops that are cross-referenced in OSM, in the tools, there is a tab “commerce clos” showing which ones no longer exist, but are still in OSM. It also suggests replacements at that location.
Not sure what the delay is between physical closing and appearing there, but it’s better than the occasional long out of business nodes I found here.
I don’t think it has a details on what percentage of business that would be good to have in OSM are actually mapped (initial mapping needs to be done otherwise).
