By these numbers, Massachusetts and Maryland had relatively less mapping (for their size) than the surrounding states, so they appear very squished. Not necessarily a fair visualization because Suffolk county (small) is the 10th most edited county in the US in 2020 in edits/sq. km.

I’m playing with a county-level version of this that highlights a few different trends, such as all of that mapping in Boston.

Congratulations on all the mapping, Adamant1! I fixed your username, apologies!

Thanks Joost! This is a great analysis; it will be very interesting to see how this breakdown will continue to evolve in the coming years. Another dimension of interest (though less important than geographic diversity, imo) could be mapping subcommunity: Hobby, Humanitarian, Professional, etc. of course, there are many mappers that are a part of all of these, but quantifying any shifts in this breakdown among OMSF membership would be illuminating. Additionally, what % of these members have been active on the osmf-talk list? (this will be telling in its own right) Thanks for this! Gets me thinking…

Thanks - I recommend using Pascal Neis’ tool to lookup individual stats (as well as active contributor status), if you’re curious, such as: http://hdyc.neis-one.org/?IpswichMapper

As for all of the Active User stats, I hesitate to publish anything that’s individually-identifiable (as opposed to aggregated) due to GDPR. Perhaps I could run these for the past year and put them behind an OSM login…

Hi Benoitdd, in my brief investigations into the 2017 spike, it appears to be more correlated with Maps.me allowing editing than the Pokemon Go surge - but Pokemon Go is definitely included in these graphs.

Additionally, Golfing video games, Microsoft Flight Simulator, and Tesla have been major inspirations for folks to get involved in OSM in recent years. Not that these fully explain the peaks, just fascinating to see!

Apologies skquinn! I have fixed those links, let me know if you have any questions!

Cheers! Jennings

Yes, the “empty space” is the real question at hand… and nontrivial to compute, but a few ideas:

If you calculated the bounding box / convex-hull of each object in a changeset, and subtracted the total area from the changeset area itself, you might get a measure of “empty space” in a changeset. That said, I really have no idea what that measure would mean, and could be kind of meaningless. The “empty space” measure between someone adding a bunch of new buildings and someone updating the opening hours on various businesses would look extremely different and it’s unclear what they actually represent.

Anyways, as for bots, I usually start with this list: https://wiki.openstreetmap.org/wiki/Bot and I’ve found that searching for ‘bot’ in the username is usually pretty good, as well as _repair, _cleanup, or _import, depending on what you’re after. If focusing on US stuff, especially pre 2010, remember TIGER and TIGER cleanup as terms :) There’s also the mechanical=yes tag on the changeset itself for some of these ?

Good luck!

Really interesting work! Another source to look at could be the OSM Public Dataset on Amazon where you can use Athena to query all of the changesets in the cloud and then download only the results you’re after.

My next questions (since you have the data :) ) would be to check the density of some of these changesets (num_changes / area) as well as the number of users submitting these changesets?As the number of monthly contributors increases, so does the rate of large changesets? Also, how many of these larger changesets are from bots?

Very cool stuff!

Yes, @Taktaal: So this visualization is a colorized, recreation of the visualizations in this paper by Bégin et al. with newer data.

That paper attributes the 2008 increase events to “The German journal Der Spiegel comparing OSM to Wikipedia” in May. Also, the mass dropping of users in mid 2011 is the removal of users / redaction work for those that didn’t accept the new license terms (as identified by Bégin et al.).

As for editors leaving in May 2013, I’m really not sure? Any ideas?

Thanks! Jennings

Thanks Rovastar!

You bring up some great points here, and I think you’re right, it’s a combination of maps.me and various games. Regardless, the change in mapping contribution patterns is dramatic.

Sorry to mislead with the y-axis; here’s another version of the graph that’s not time-bounded

• Jennings

Thanks @tyr_asd, I updated the post to mention the oshdb as well! :)

Thanks GOwin and Tomas, these are both great ideas. This is also a general concern with some tile-reduce comparisons, that tiles are not all the same area. Normalizing (especially between tiles) if doing side-by-side comparisons by area is important.

At the Country level, I wonder how these numbers may change. My hunch is that the US will go down and Countries like Germany will grow.

Thanks Martin, that’s really interesting… staying tuned.

Martin, these are so cool!

I’m trying to re-run some of these with mbtiles and mapbox-gl for smooth zooming: http://www.townsendjennings.com/osm-node-density/#3.88/47.28/13.66

My resolution, however is nowhere as good as yours! What were the settings that you used when you generate your highest zoom-level png?

Thanks!

Very interesting work, Martin! I look forward to seeing the full write-up, best of luck with the review process.