Maximum size of Openstreetmap

General talk
1

Today I looked at the OSMstats of Neis and started wondering. How big will Openstreetmap eventually get? How close are we to map all of the buildings in the world? How close to all of the roads?

And to go further in. How big will the openstreetmap database get at each milestone? We are at 100gb now for the planet.osm.pbf.

This is surely a philosophical or guessing question, there’s always more to micro map but I wonder whether anyone went down that rabbit hole already? Maybe even a presentation or talk?

2

According to a recent study from 2022, the estimated completeness (based on machine learning) of building outlines in OpenStreetMap is around 24% worldwide.
Source: Herfort, B., Lautenbach, S., Porto de Albuquerque, J. et al. A spatio-temporal analysis investigating completeness and inequalities of global urban building data in OpenStreetMap. Nat Commun 14, 3985 (2023). A spatio-temporal analysis investigating completeness and inequalities of global urban building data in OpenStreetMap | Nature Communications

The study does not give an exact global percentage for the completeness of the road network in OpenStreetMap. They used a so-called mapping saturation method to assess the completeness of the road network. In some regions (e.g. South Asia: 10.5%, Middle East & North Africa: 10.2%, Sub-Saharan Africa: 5.9%) many roads are still in the early stages of mapping.
However, another study from 2017 claims that 80 % of the world’s road network has already been completed.
Source: Barrington-Leigh C, Millard-Ball A (2019) Correction: The world’s user-generated road map is more than 80% complete. PLOS ONE 14(10): e0224742. Correction: The world’s user-generated road map is more than 80% complete | PLOS ONE

At least for me, the question often revolves around the core of completeness. What does complete mean? I would argue this is so subjective. In my area, there hasn’t been an OSM meet-up for over a decade. When I asked a few ‘old hands’ if there was any interest in a revival, the surprising response was that OSM had already been ‘played through’ here – or rather, there was nothing left to map.

Yes, of course, when I look at the district here, the road network is complete (except for a few private driveways with a maximum length of 10 meters). All the houses that can be seen on the available aerials are also mapped including addresses, and new ones are always being built. This is an ongoing process. Speed limits are tagged, surfaces of main roads and public transport routes are there…

But is it complete? Is every single shop included, with opening hours, website and wheelchair access? Public Transport platform heights, bicycle parking with number of spaces, every single narrow stream, landuse down to the last millimeter, dog waste stations, manhole covers?

And that’s the flagpole only for today. Every week I write or translate another call for RfCs summary for weeklyOSM. Every week there are new ways to map or tag things. It’s like an endless computer game where there will never be an end level :wink:

My conclusion: There will never be a ‘complete’ version. You can never have all the buildings mapped. Just by looking out of the window in my tiny village, I can see six construction cranes that are already building new ones, the local council has just approved the next development plan. The world will grow, new things will be constructed, old ones will be demolished, and we will try to map all this hustle and bustle on a map.

9 Likes
3

I forgot to post two additional links:

1 Like
4

I was a bit surprised by that completeness analysis showing some areas in Switzerland as <80% because I wouldn’t have thought that such areas still existed in the country. So I checked out one such place I know well and that area includes a quarry with a lot of rectangular gravel hoppers. So I’m guessing their ML algorithm counted those as buildings that weren’t in OSM and that’s why the statistics show that place as incomplete.

1 Like
5

If you are referring to the first study (Herfort, Lautenbach. et al.), then according to the script they obtained the building footprints from Microsoft. You could check the Rapid Editor to see if buildings are offered there as an integration.

6

Yep, Rapid Editor prompts me to add gravel hoppers, shipping containers, a gravel storage area and even a stack of logs as a building in that quarry.

8 Likes
7

Yeah, I remember some heavy discussions about false positives. For example here, here or here.

1 Like
8

Just as a reminder adding buildings from the MS building dataset is highly problematic from a licence pov. Regardless of the quality

2 Likes
9

Your post is the first I’m hearing of this - do you have a link to previous discussions on this topic?

10

All datasets licenced on ODbL terms are problematic, on the one hand any such licenced data would be incompatible with -any- future license change own the other hand, by the OSMFs own interpretation of the ODBL any licensor of duch data could require attribution just as OSM does.

11

This should definitely be added to the first sentence here.

3 Likes