Population Mountains
December 11, 2018 10:01 AM   Subscribe

What stands out is each city’s form, a unique mountain that might be like the steep peaks of lower Manhattan or the sprawling hills of suburban Atlanta. When I first saw a city in 3D, I had a feel for its population size that I had never experienced before. 1000 words from Matt Daniels at The Pudding with lots of images.
posted by cgc373 (32 comments total) 52 users marked this as a favorite
 
This is brilliant!
posted by twilightlost at 10:28 AM on December 11, 2018


Fantastic post.
posted by Cool Papa Bell at 10:32 AM on December 11, 2018


Nice to see a map projects that doesn't suddenly stop at the US border and acknowledges the rest of the planet.

Too many mapping/GIS posts that show up on the blue act like nothing on the other side of that border exists. Most egregious was the river basin map that sawed four of the five great lakes in half.
posted by thecjm at 10:38 AM on December 11, 2018 [4 favorites]


That large population belt in northern India and Bangladesh is essentially the Ganges basin, pretty neat to see it so clearly.
posted by peeedro at 10:50 AM on December 11, 2018 [2 favorites]


Heat maps are just not a very good way of representing quantitative information. People have a really hard time mapping hues and shades onto a scale in their heads, and that's if you're lucky enough to be looking at a heat map with a sane color scheme. This seems better, at least for this application.
posted by Anticipation Of A New Lover's Arrival, The at 11:10 AM on December 11, 2018 [4 favorites]


Neat. And thanks peeedro, I was wondering what that swath of high-density was.
posted by Kadin2048 at 11:12 AM on December 11, 2018


wow, this is incredible. So many people in the world must have a life so radically different than me. I kept finding myself thinking "These huge pillars must be inaccurate, there is probably a mistake that makes it look so populated." I would notice myself thinking that and get a funny feeling.
posted by rebent at 11:47 AM on December 11, 2018


Wow, these are great and yeah, drive home how much I have to learn about the world.
posted by LobsterMitten at 12:03 PM on December 11, 2018 [1 favorite]


To me it looks badly in need of a logarithmic scale or the like. Mountains are cool and all, but a topographic map of the world that shows only the mountainous parts is missing many features of geographic interest. I've been to up-state New York, and much more of it isn't the uninhabited wilderness one might guess it to be from that map. An NYC that rises out of the elevated plains of the north-east would have a more realistic feeling to me than one that just shoots up from nowhere into MEGACITY as if the world outside cities was all unimportant sea-level nothingness.
posted by sfenders at 12:32 PM on December 11, 2018 [5 favorites]


Also, do dense cities reliably develop suburbanization with age, so that the new spikes will look like the older mountains given time?

... which reminds me of similar visualizations, applied to daily commutes, e.g. in Manhattan, the Bay Area, and (different geometry) the London Tube. (Couldn't find a visualization for Kinshasa; some numbers, some qualitative discussion.)
posted by clew at 12:42 PM on December 11, 2018 [2 favorites]


A logarithmic scale would lose the way this treats equal populations as equal volumes. True, people are sometimes bad at appreciating the volume of a shallow bowl vs a tall glass. But I don't see a way to compensate for that without introducing other distortions.
posted by RobotHero at 1:01 PM on December 11, 2018 [5 favorites]


Population / 1,000 sq. meters

Pretty sure they mean /km^2 otherwise in the max density regions they're cramming 40 people into every square meter.
posted by GCU Sweet and Full of Grace at 1:03 PM on December 11, 2018 [1 favorite]


A logarithmic scale would lose the way this treats equal populations as equal volumes.

How would that be a meaningful loss? If people want to simply know which big cities have the numerically largest populations, that information is not scarce or surprising. Estimating the total population size of LA versus Jakarta by looking at this type of map this seems like a remarkably inefficient and difficult way to do a not very interesting thing. Losing it in service of more clearly conveying the shape of the landscape would be preferable to me. Estimating the volume of mountains isn't usually the best use of topographic maps.
posted by sfenders at 1:25 PM on December 11, 2018 [2 favorites]


It's a meaningful loss if you think a person with a lot of neighbors should have the same representation as a person with fewer neighbors.
posted by clew at 1:33 PM on December 11, 2018 [3 favorites]


People don't usually estimate the volume of mountains because they can only access the surface. So the rock that's underneath 1000 metres of other rock seems less important than whatever's right on the surface. Though if I was going traveling I might still prefer a linear display so equal slopes would be equal.

It's not just comparing cities as a whole, but looking at the shape of them for how much of a city's population is a big spike downtown vs. a wider, flatter spread, etc. They do some of this in the article.

And like, looking at Paris vs. London. Paris has this massive spike, so that's some very dense neighbourhoods, more than twice, almost three times as dense as anything in London. I can make that comparison at a glance because it's linear. If it was logarithmic I'd be like, "It's bigger. How much bigger? Lemme think, that is this big, and this is that big, and ..." and I'd give up.
posted by RobotHero at 2:05 PM on December 11, 2018 [2 favorites]


No, a person with fewer neighbors should not have the same representation as a person with a lot of neighbors on a map of population density, that's literally what it is for
posted by phooky at 2:09 PM on December 11, 2018 [1 favorite]


It's not just comparing cities as a whole, but looking at the shape of them for how much of a city's population is a big spike downtown vs. a wider, flatter spread, etc.

Yes, that is what I'm suggesting is the interesting bit. And it looks more spiky than it needs to be to usefully convey that information. Paris and London you can see some good information, and it's great, but that's true here only of the largest megacities in the world. All those more numerous unnamed towering one-dimensional peaks in the more densely-populated parts of the world, and all the less-than-gigantic towns and cities that barely appear in less amazingly crowded parts are left shapeless and uncharacterized when perhaps they needn't be, if in fact they didn't already try a logarithmic scale and find it too cluttered or otherwise difficult.

Unlike in mountain climbing, we've no way of directly measuring or experiencing the population density of what are still fairly large areas on a human scale while we're hiking through them. It seems less than realistic, in that unlike on the map, being in some kind of semi-suburban area just outside a city is a lot closer to being in a city, for most practical purposes, than it is to being out in the middle of the ocean or something.
posted by sfenders at 2:24 PM on December 11, 2018 [1 favorite]


Love the maps, but the text and analysis was really rather disappointing. I already knew that Paris, Hong Kong (and environs), and Lagos are big.

What I found much more interesting was the distribution of each city's population and how it interacts with its environs. For example, the analysis points out that Kinshaha (unlike Paris) is a "stand-alone mountain, surrounded by few settlements" -- which is a great point, but only raises the question, why? And this analysis didn't answer those questions.
posted by crazy with stars at 2:48 PM on December 11, 2018 [3 favorites]


Perhaps if you look at a map of the population density of the world when you're interested in the subtle population shadings between semi-rural and exurban and small town upstate New York, it's not the map's fault you aren't seeing what you want.

For what it's worth, this other visualization of the same data has their highest density category at 80,000 people per sq km, a density only meaningfully found in India and Bangladesh. If the map used a log scale, the half-way point between 80,000 people per sq km (and there are denser places!) and 1 person per sq km would be at 283 people per sq km, which is basically exurban densities, in the range of 2 to 3 acre lots (think of somewhere like this). That would also be interesting, I suppose if you were primarily interested in low-density suburbia and exurbia, which is largely an American phenomenon, but it would be wildly misleading about how the global population is distributed, which I gather was the goal of the author.
posted by Homeboy Trouble at 3:03 PM on December 11, 2018 [5 favorites]


It's not only low-density suburban America which is made to appear completely flat, it's also most of Europe, most of Africa, and indeed almost all of the world as a whole. So yes, it had the effect of making me curious about how lower-density populations are distributed around various places, including in the most populous parts of the world in-between those one-dimensional spikes. To me it seems like a less than optimal use of the 3d effect. If anything it's more potentially misleading this way rather than less, although that is not the point; what it does show is clear enough either way, I just think it could show more.
posted by sfenders at 3:39 PM on December 11, 2018 [1 favorite]


I think a compromise between a pure logarithmic and a pure linear scale could be made, such that the megacities would look suitably mega, but the more subtle contours of their surroundings could be discernible. The current map does pretty much erase anything smaller than a large city, which is not very realistic. The world is more complex than just isolated arcologies surrounded by wilderness.
posted by Anticipation Of A New Lover's Arrival, The at 4:05 PM on December 11, 2018


If you actually browse the visualization, you can see more detail when you've zoomed in. It's clear that they remove some data for the sake of 3d rendering performance, and that there's a 'zoomed-in' view with more detail.

See upstate NY.
posted by suedehead at 4:14 PM on December 11, 2018 [4 favorites]


Yeah, I'm wondering if some of you are looking at a different map than me. There are other smaller bumps in there. I wish more of these were labeled but I can easily pick out Rouen and Le Havre in the Paris map, for example.
posted by RobotHero at 4:20 PM on December 11, 2018 [2 favorites]


Too many mapping/GIS posts that show up on the blue act like nothing on the other side of that border exists. Most egregious was the river basin map that sawed four of the five great lakes in half.

Some public mapping data sources (such as from the USGS or Canada's GeoGratis) stop at national borders.
Full information: I know it's that way for GeoGratis. I don't know if it's that way for USGS data but I suspect so.
posted by Quackles at 4:24 PM on December 11, 2018


Ah, I completely missed the "interactive" link. That's neat.

I'm now looking at Manitoba, and while the megacity of Winnipeg is pretty big, you've still got a bit of detail for other cities like Oakbank and Tyndall.

I don't think I like how they change the scaling when you zoom in and out, though. I feel like that should stay consistent.
posted by RobotHero at 4:29 PM on December 11, 2018


I missed that too. Now I want a log scale for an additional reason, getting detail in large cities. Try zooming in on the wrong area around New Delhi, Mexico City, or Toronto, and it's an incomprehensible forest of individual peaks. On the other hand, Rome and Tokyo look pretty good.
posted by sfenders at 5:03 PM on December 11, 2018


This is nicely formatted and probably much better data (and updated for salient global trends), but sort of an old framing:

R Florida, The World Is Spiky
posted by Reasonably Everything Happens at 5:35 PM on December 11, 2018 [1 favorite]


The software to make these (and to have the neat interactive maps) is way more accessible now, but I can remember being shown density maps of the relative "shapes" of cities in my first year of grad school, which is quite a few years ago. They are fun to see and tell you a lot about comparative patterns of urbanization.

I don't see why you would want to use a log scale, but maybe it would highlight things in an interesting way that I'm not visualizing. It shouldn't be a difficult exercise to convert the data if someone really wanted to do it.
posted by Dip Flash at 5:50 PM on December 11, 2018


Heat maps are just not a very good way of representing quantitative information. People have a really hard time mapping hues and shades onto a scale in their heads, and that's if you're lucky enough to be looking at a heat map with a sane color scheme.

I’m a little more positive about the value of heat maps as a visualization tool when the color scheme is sensibly chosen, but it does blow my mind in the worst way that the default color scheme for heat maps in a lot of visualization tools is the rainbow gradient. Like, why.
posted by invitapriore at 6:39 PM on December 11, 2018 [3 favorites]


Another reason to use a log scale is that populations are probably roughly distributed according to a power law. It's certainly true between cities (see here, for instance), and I would be surprised if it wasn't also true with respect to scales both larger and smaller than cities. The best way people have found for visually representing exponential data is on a log plot. They're extremely common, and if you're hung up on the idea that they distort the 'true' amount of something, here's an alternative way to think about them: log plots show order of magnitude, which I would argue is what affects the qualitative experience of population density anyway. I think the link provided by Homeboy Trouble supports this point of view, since it is a log plot, and for good reason.
posted by dbx at 11:03 AM on December 12, 2018


It really depends on your use case. Yes, a log scale is better if you want visibility of structure at all scales. However, it's much worse if you want to do certain things like compare one city to another.

On the linear scale, if (small, dense) city A appears as one very tall bar, and (larger but less-dense) city B is four shorter bars, you can still see which one has more total people. Just compare the total volume. (Or just take the four short bars and stack them on top of each other, then place them next to the tall bar.)

If the heights use a log scale, then simple things like addition become non-linear. This means there is no way to compare the population of one area to the population of another using just the visual information in the plot. In many cases a larger volume will represent a smaller population, just because it's spread out over a different area. And it will look larger even though it is not.
posted by mbrubeck at 9:24 PM on December 12, 2018 [3 favorites]


I think for a logarithmic scale I might advocate going back to a heat map because then I won't instinctively make those sorts of comparisons as much as I do with bar height.
posted by RobotHero at 9:30 AM on December 15, 2018


« Older O you, beloved of my twenty-seven senses, I love...   |   Orlando, FL is named after a man who may have died... Newer »


This thread has been archived and is closed to new comments