Hacker News new | past | comments | ask | show | jobs | submit login
OpenStreetMap node density map 2014 (tyrasd.github.io)
139 points by lelf on June 28, 2014 | hide | past | favorite | 51 comments



Looks like The Netherlands is one of the most dense countries in the world. I wonder why that is, we (speaking as a Dutch person) hardly seem to use OSM while in Germany it's very popular.

One explanation is the recent addition of the government's address and building data, which gives us close to perfect outlines for every building in the country plus an extra node per address. Or we just have a high number of people per square kilometer regardless of the recent (now almost finished) data import.

Edit: Looks like it's simply our density in general and not this import; last year the Netherlands also stood out while we hadn't even imported 10% yet. From last year: http://www.openstreetmap.org/user/tyr_asd/diary/19549


The buildings have a huge impact. Look at New York, Chicago or San Francisco. They also had building imports and they shine quite brightly.



That and 3dshapes[1] helped a lot of course, but since we started importing BAG data in March the map size (looking at the OsmAnd map file) roughly doubled. Thought that might also influence this density map a lot but it doesn't seem so.

[1] https://wiki.openstreetmap.org/wiki/3dShapes


My guess is that this might be cycling related; that an area where there's likely to be public-spirited cyclists with GPS devices is well placed to get good OSM coverage.

In support of my theory, I point to Cambridge being a little bit of a hotspot in the southeast of England. (Not so much as Clacton though ... hmmm ... maybe the theory needs some work.)


Isn't Cambridge where OpenStreetMap started? That's a bit of a confounding factor!


I believe Cambridge was the first place to be declared "complete".


"Isn't Cambridge where OpenStreetMap started?"

No. London.


Well, kinda. A lot of the coding and early work was done in Cambridge but the first node I think was in Regents Park.


The site runs over https but loads leaflet over http, so that it get's blocked at least by recent versions of Firefox, Chrom[e/ium] and Internet Explorer.

https://developer.mozilla.org/en-US/docs/Security/MixedConte...


Actually the link was an http link. You likely have HTTPS Everywhere that changed that.


It’s still nice, if all the content is available over both http and http, to use protocol-relative URLs, i.e. without http or https before the //: part. It looks like the author has read your comment, and updated his source code to say <script src="//cdnjs.cloudflare.com/ajax/libs/leaflet/0.7.3/leaflet.js”> for leaflet.

http://www.paulirish.com/2010/the-protocol-relative-url/ describes how protocol-relative URLs work.


You are right, I missed that. But you should have protocol relative urls on servers which run https.


It's not my site. I'd force everyone to https if it was



I found the map of greenland fascinating, since there is pitch black internally. Everything is along the coast.

At the lowest zoom level what are geometric lines that criss cross Greenland and the curvilinear lines and rings off the coasts?


Do you mean like this one:

http://www.openstreetmap.org/way/28421901#map=9/59.1125/-5.9...

That's just territorial waters.

The lines across Greenland are administrative borders:

http://www.openstreetmap.org/way/238745790#map=4/71.15/-44.1...

I guess there is so little mapped there that they can stand out.


Another interesting note: I would have expected North Korea to be dark, but it's brighter than neighboring China.


And South Korea is not nearly as bright as its population and wealth suggests.


The whole of Massachusetts is brighter than the surrounding states, and grows brighter the closer you get to Boston. I wonder where all that data came from. Was it a corporate or state data donation?


For the state, it's "MassGIS" data. Many of the points are from road center lines that are a lot denser than is typical in OSM, but there is also land use/classification data (like wetlands or whatever).


Thanks for the information.


It would be interesting to see this map adjusted to population density.


Here you go!

https://imgur.com/a/b6BjV

edit: Um, it seems Imgur does lossy compression now? Sorry about that -- here's a sample of the original (2.5 arcminutes, limited by the census data)

https://i.imgur.com/hBNEGNT.png


Well count the building tags and you got a rough population estimate for any area.


It won't work, since the average number of people living in one building varies with the country. I guess it should be around 10 people per building in USA and 100 people per building in Russia, for example.


I tried to overlay this map with a map showing population density from Google Images and then normalize it with a division filter. I couldn’t find an map with a matching projection however. :(


Check back here in a bit, I'll try to make one. I already have the OSM database and some related glue code, it shouldn't be too far out the way.


I'm sorry, but I don't see any much use in this visualization. It's pretty, but gives almost no information at the global level, for two reasons:

1) It's not adjusted for population ( http://xkcd.com/1138/ ). They can normalize the data to fix this.

2) The map is ludicrously distorted because of the Mercator projection. Greenland looks bigger than the entire South America, when it should be an eighth of the size. This directly affects the perceived density in each area. Suggestion: if you are using an interactive geographical visualization, why not display a globe?

However, it does look stunning.


That what I thought at first but the visualization is actually useful. For instance, I have the offline OpenStreet map of Germany on my phone, which is about 1GB. In comparison, the map China is about 110MB, despite having a population 15 times bigger. The node density map does actually reflect this kind of differences. I find the difference between Brazil, Chile and Argentina also quite interesting.


There is no claim that the density means anything, so there isn't any need to adjust it for population. I imagine the title "OpenStreetMap node density map 2014" was somewhat carefully chosen, and it manages to say what is shown.

If you read the link that Vik1ng posted, the author seems to be fairly clear that they were just making a pretty picture (and quietly points out the problems with the projection).


Density here is number of nodes by area, and if you distort the area it stops being density. It's like comparing the weight of different objects on different planets. "My shoe on Earth weights twice as much as my sandal on Saturn" isn't very useful.

And thanks for mentioning the link, but I don't see where they say it was made for a pretty picture.


Yes, I said "seems to be" to suggest that it was my impression of the page, rather than a quote from it or whatever.

Part of that impression is that you can't really zoom in far enough to do anything super useful with it, so it's probably just to look at.


@1 I think the main goal is simply to display OpenStreetMap node density. Nothing more and nothing less. For mappers it's kinda interesting. He did't want to make a map where the most active mappers are.


But if it's the same as a population map, what good is it?

If I had asked you about nodes density before you saw this picture, your guess would probably use the population density. If your guess remains the same after seeing the picture, it didn't give you any useful information.

I think this is gorgeous, and the process of making a visualization is interesting, but I'm questioning the usefulness of the information portrayed.


But it isn't just a population map. You can see where users have imported data (e.g. the parts of Canada that are lit up in small squares is CANVEC data) and you can see the relative brightness of the map in Europe vs. United States -- even in areas of higher population density the US is dimmer than most of Europe because there are more mappers there.


I hadn't noticed the small squares before. I stand corrected in that front, that is useful.

But on the Europe vs USA case note how you are automatically adjusting for population. Because of this you can only interpret data for places you know the demographics, and even that is error prone.


> I'm questioning the usefulness of the information portrayed

It does give you information on which parts of the world are more active in the project: the 3 biggest economic regions are the more mapped regions.

Besides, does everything need to be useful ?


I'm getting frustrated here.

    It does give you information on which parts of the
    world are more active in the project: the 3 biggest
    economic regions are the more mapped regions.
That's exactly what the grandparent said, and I tried to rebut in the second paragraph. It might be a communication problem, but I feel like you didn't read my reply to his message.

    Besides, does everything need to be useful ?
That's also explained in my reply above, third paragraph. It doesn't have to be useful, I'm not advocating the removal of this webpage from the internet, or preventing people from creating more visualizations like this. What I'm saying is that visualizations usually portray useful information, but I'm arguing that this one does not. That's all.

<rant>

This thread is making me sad. I feel like I made a valuable contribution, a little negative maybe, but useful nonetheless. I pointed out subtle but important problems, gave suggestions on how to fix them. I applauded the visual graphics.

Then I'm flooded with downvotes and replies that, from my point of view, are not correct. The only one I felt that helped change my mind a little was the CANVEC data squares, but all the others felt like communication failures. If I feel like you didn't read the message you're replying to, one of us made a terrible mistake.

This whole affair bummed me out so much that I'm considering logging out and not contributing anymore.

</rant>


I've experienced the downvote piling on before, it's definitely irritating to make a good faith effort and then have a bunch of people express their disagreement with clicks.

I do think that most of the replies are simply politely disagreeing with you, not missing the point you made. You mention the CanVec imports, another issue with the adjustment for population density is that OSM activity around the globe is not uniform; it's no doubt changed since this posting, but it gives an idea:

http://neis-one.org/2012/11/active-users-osm-nov12/

The activity is substantially higher in some countries, per population, and also per area.

There are a lot of other imported data sets that aren't terribly connected to population. The U.S. would look dramatically different without the TIGER road import, and the import of node heavy NHD data doesn't particularly follow a population distribution, in addition to not being completed. For example, if you look here:

http://tyrasd.github.io/osm-node-density/#7/41.034/-93.142

and then zoom in on the OSM map, it's pretty evident where an import was done and where an import was not done:

http://www.openstreetmap.org/#map=13/40.9580/-93.4140

But there's no connection between that dividing line and population density, it's from however the source data was split up (probably separate watersheds).


I agree the replies are in good faith, and I'm happy people took the time to respond to me. But somehow the good faith makes things frustrating. Trolls can be ignored, downvoted and banned, but communication problems are complicated. If we have a discussion and nobody changes their mind, something's wrong.

About your first link: http://neis-one.org/2012/11/active-users-osm-nov12/

They did precisely what I was advocating: adjust for population and area (though a little coarse). And it makes the data much more insightful: look at that bright Greenland, or Brazil lagging behind the rest of the continent. That's not visible on OP's map.

And I stand corrected on the import information. I hadn't thought of that. So the information on the map is not useless, but it could be much better.


It's a visualization of different data though, counting the number of users making a change each day.

I see what you are saying about being more interested in a visualization that tried to account for some regional variances (and it's a fair point about Mercator distorting things), but I guess the thing people are taking issue with is the statement that it would be better, when it would probably be different enough to not even invite comparison.

(What I mean with that last bit is that "here's the nodes" maps simply and directly back to the database, whereas any normalization is going to involve picking some basis for the normalization and figuring out how to apply that basis and so on.)


   or Brazil lagging behind the rest of the continent.
Except that when it comes to map data it they aren't lacking behind and that's exactly what is shown on OP's map. Adjusting for mappers would remove the essence of OP's map: How much has been mapped in a certain area.

If you could track every car and make a map showing how dense traffic is at a certain time you would not adjust for population either, because that would just distort you data.

   but it could be much better.
I disagree. Every adjustment you make I would have to consider in addition when interpreting the map. For example then I would have to know if Netherlands or Germany is more densely populated or has more contributors to figure out which one has more map data. Right now I just look on the map, see which one is bright and know it has more nodes.


By the way, West Wing had an excellent piece where they explain the need for accurate map projections: https://www.youtube.com/watch?v=n8zBC2dvERM


Interesting how brightly Bakersfield and Fresno glow.

I wonder if it's because the streets are so gridded and therefore not as prone to aliasing.



Each pixel represents quite a lot of area so aliasing probably isn't coming into it.

Bakersfield has a bunch of buildings (which lines up with the theory in my other post in this thread). It looks like imported land parcels make up a lot of the data in the Fresno blob:

http://www.openstreetmap.org/#map=16/36.7805/-119.9009&layer...

(Probably a good idea to turn off the data layer there before zooming out)


You can also see the border of Massachusetts perfectly clearly, likely due to the import of MassGIS data. Surrounding states are all much lower density, until you hit Manhattan.


What are the rectangular features in northwestern Canada?


On one side "CanVec" data has been imported and on the other it has not. If you zoom in from here you see all the small streams that are showing up in the visualization:

http://www.openstreetmap.org/way/157650655

That doesn't explain the differences on the imported side, but I guess that could just be more imports vs less imports.


Perhaps satellite or aerial photography.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: