The Shapes of Cities

For a long time, I've been interested in the shape of cities and I suspect that if you're reading this you might be similarly afflicted. By 'shape' I mean their political boundaries as opposed to their general urban footprint. The latter can be seen by driving around or from a plane window, particularly when it's dark, but the political boundaries are much less obvious. This is particularly true of US cities. Take Houston, Texas - the first example below. 

The boundary of the City of Houston - Google Map

Look closely at this and - at least if you're not used to the political geography of American cities - you might be very confused by this fragmented, segmented mess of boundaries. Then go to Google maps and try different search terms, such as 'city of los angeles' or 'city of columbus' (Ohio) and you'll soon discover that Houston isn't that unusual at all. Try it for other cities and you'll see what I mean. Columbus, Ohio is a particular favourite of mine as I know it quite well having lived there for a couple of years in the early 2000s.

City of Los Angeles - Google Map

City of Columbus (Ohio) - Google Map

These unnatural-looking boundaries are the result of a complex mix of geography, history and politics that have real impacts on the ground. From education and transport to housing and waste management, the shapes of cities really do matter in this respect. Of course, this is a much-studied topic in urban studies, not least by Professor John Parr of the University of Glasgow. In Parr's economic definitions of the city, he outlines four types - but none of these explain the kinds of boundaries we see above. One major explanatory factor in all of this, of course, is tax revenue. But I'm not going to get into that now because it opens up a whole range of other topics, including white flight, suburbanisation, and schooling, amongst other things. The point is that the 'shapes' of cities are not accidental and who is included or excluded is inherently political. 

In the United Kingdom, we might not have such unusual city boundaries, but the political geography of our cities is far from perfect - perhaps one reason for the resurgence of the 'city-region' concept over the past decade or so. When we're talking about urban economies, it makes much more sense to think about the functional urban area than it does to use data associated with an arbitrary political shape. This is as true in the US as it is in the UK. The example below shows that the City of Atlanta has less people than the City of Liverpool and that it's only slightly bigger in scale. But anyone who knows anything about these places will understand that 'Atlanta' is much bigger than 'Liverpool and is vastly more sprawling, with a metropolitan population of around 5.5 million compared to less than a million in 'Liverpool' (by one definition). 

Atlanta vs Liverpool - which is bigger?

These kinds of issues are part of the reason organisations like the Centre for Cities use the Primary Urban Area definition of cities for the 64 largest urban areas of the UK. In a recent study, I used a definition developed by Geolytix which is based on the 'sprawl' of the urban area rather than political boundary and found this to be a much better fit than the administrative area. When conducting comparative analyses of cities, we need to ensure we are comparing like with like, and using a functional definition often helps avoid the kinds of underbounded/overbounded problems that arise when (e.g.) comparing places like Manchester and Leeds. The former is normally said to be 'underbounded' because the functional urban area is much bigger than the local authority area of the same name and the latter is said to be 'overbounded' because of its much wider local authority area, which extends beyond the core urban fabric. For a comparison of UK 'city' sizes, see this graphic I produced a few years ago:

All cities shown at the same scale

Surely there's a point to all of this? 

Yes, glad you asked...

For planners, politicians, residents and neighbours, the shapes of cities matter enormously. It might dictate which school your children can go to, whether your local facilities are well funded, whether you have a well-functioning local transport system, when your bins get emptied, how many pot-holes you have in your street and all sorts of other things. But let's not get into that now. Instead, I'll end with another city shape, this time for the City of Detroit (one of my favourite cities, but much-maligned).

Detroit - 8 mile boundary line to the north

From mega-regions to mega commutes: US commuting working paper

My previous post provided some images from a recent piece of work I did on mapping tract-to-tract commuting patterns in the contiguous United States. This post provides a bit more background and extracts from a working paper, plus some of the original map outputs from the project - which are different in style (kind of a night time view). The focus is also more on mega commutes and mega-regions (think Gottmann's 'megalopolis'). I also provide a bit more detail on the method and data.

A constellation of cities in the Midwest

Being a member of the Regional Studies Association for a good few years now, I've followed various debates about regions, city-regions and mega-regions - including the very interesting work on mega-regions by the America2050 project of the Regional Plan Association. I also have a longstanding interest in commuting flows (and mapping them) so I set myself the challenge of mapping micro-level commuting flows in the contiguous United States in the hope of identifying what I expected would be some interesting mega-region commutes. I also hoped, in the context of this data, that I would discover some of the mega commutes identified by Rapino and Fields of the US Census Bureau. On both counts I wasn't disappointed. The first map below shows the entirety of the lower 48 states and the commuting patterns come out quite clearly.

Journeys to work in the contiguous United States

Obviously, some areas are more interesting than others, so I zoomed in on various areas, including California and the Northeastern United States. The map below shows travel to work patterns in California, and you can clearly see the wider Los Angeles metro area as one large commuter region, the Bay Area as another (but more polycentric), and also the settlement and journey to work patterns in the Central Valley, from Redding in the north down to Bakersfield in the South. This shows the urban settlement patterns in the state of California, but also the spatial configuration of the commuting connections between places.

If you take a closer look at the working paper behind these maps you'll find out more about the data. What I found most interesting were the locations where 'mega commuting' was prevalent, so I looked at the top 20 Census tracts in the Northeastern US with the highest number of people commuting there - i.e. over 50 miles each way. As you can see from the table below, this is dominated by New York City, but Washington DC also features. The total volume may not seem much, but remember that these are quite small Census tracts, with only a few thousand people.

Mega commuting in the Northeastern United States

I then did something slightly different - I wanted to filter the data in a more scientific manner. Since the data provided by US Census Bureau includes a margin of error (MOE) value for each individual tract-to-tract flow, I calculated the coefficient of variation for each individual flow line (there were just over 4 million). These were based on a 90% confidence level, so the formula was simply:

((MOE/1.645)/Commuting Estimate) x 100

I used a rather generous cut-off and then displayed only those flows which had a coefficient of variation of less than 40. The results are shown in the map below. We can see the expected pattern of commuting but - hold on a minute - what are those really long distance lines? Surely people don't 'commute' vast distances like this. Well, it turns out that this might actually be true because many of these lines begin and end in military locations or other places associated with regular, long distance moves for work and since the American Community Survey asks respondents how they usually got to work ‘last week’, it's entirely plausible that a number of people will work away from home and that this will lead to the kinds of patterns we see below. Or, to put it another way, don't think of these long lines as journeys people travel every day! 

If you want to read more about it, you can click below to see the working paper, which also includes links to high resolution versions of the images shown here.

American Commute: working paper

Mapping the American Commute

Update, 20 September 2015: scroll to the bottom of the post if you want to download the data.

One of my summer projects this year has been attempting to map the American commute, following earlier work on a similar subject. Put simply, I've attempted to put together a map which shows commuting connections between locations in the contiguous United States, using the most fine-grained data I could find. Some of the results of this went into a recent piece in WIRED, and also CityMetric, and the larger piece of work it's based on is part of on-going research into the best ways of mapping commuting flows. The main images are below, followed by some more technical information. For now, all you need to know is that these images show commuting connections of 100 miles or less between Census tracts in the lower 48 states. You'll have to forgive me if your city isn't labelled! 

Higher resolution image available here

And now some zoomed in versions...

Zoom in of the west coast

Texas, and beyond!

Interesting patterns of connectivity in the Midwest

Look closely for some interesting inter-connections

The famous BosWash megalopolis

But this just shows where people live, doesn't it? Yes it does. But it also shows how the places where people live connect with other places from a functional economic point of view, at a fairly fine-grained level. It offers a slightly different view than just looking at the urban fabric alone which, I might add, is interesting in itself. Mapping flows like this is not exactly new, as this paper from Arthur Robinson (1955) on Henry Drury Harness (1837) demonstrates. Nonetheless, I haven't seen anyone map travel to work at this resolution for the United States, so I thought I'd have a go myself. 

If you spend some time looking at the big version of the map you can begin to see how places connect and where there are obvious disconnections, even between places that are not that far apart. One thing that you can pick up from the complete dataset (but not this batch of maps) is the growth of mega-commuting, as explained by Melanie Rapino and Alison Fields of the United States Census Bureau. 

Background information: the data I used is the most recent tract-to-tract journey to work dataset from the American Community Survey. This dataset covers journeys to work between the c74,000 census tracts in the United States and the complete dataset has around 4million interactions. I mapped this in QGIS, using methods I've described previously on this blog. The tricky bits were dealing with the messy FIPS codes, dealing with the size of the dataset, and trying to decide what to label. There is quite a bit of error in the dataset (as acknowledged by the ACS people) and each individual flow line has a margin of error value associated with it, from which I also calculated the coefficient of variation. This is explained in a more detailed working paper, which I expect to publish in the coming months.

Update, 20 September 2015: there has been quite a bit of interest in the underlying dataset I put together to create the maps, so I have decided to make the whole shapefile available here in the hope that others will find it useful and be able to produce some interesting analysis or visuals from it. I'm hoping someone will do a cool interactive web map of it, but it might be quite technically challenging. If you do use it, make sure you read the associated working paper, which explains the process and the underlying data. One word of warning: the uncompressed file is pretty big so you'll need a good computer.

Mapping the American Commute: download the data (213MB, zipped shapefile)

Urban footprints: some building outline data sources

This is an informational post about where to find building outline data, which I've used a lot in previous GIS projects. It might also be of interest to architects, engineers and anyone interested in urban studies and planning more generally. I like using this kind of data to explore cities as it gives us a good idea of the layout of the urban fabric, as in the example below (New Orleans). The links mainly refer to data from the US, Canada and Great Britain but other parts of the world are covered to various extents by OpenStreetMap.

New Orleans

Let's start big, with OSM... Steve Bernard has produced an excellent video which explains how you can get OpenStreetMap data directly into QGIS very simply - he uses Madrid in the example. The accuracy and coverage varies a great deal across the world, so you need to bear this in mind when downloading and using it - but on the whole it is a fantastic resource. The example below shows Mogadishu, where the coverage is incomplete for buildings but pretty good for the road network. 

© OpenStreetMap contributors

Another useful OSM-related resource with decent global city coverage is CAD Mapper, where you can download areas up to 1km square for free. However, I'm focusing on open data today so will not go into detail on this. The best OSM download source is I think GEOFABRIK (German for 'geo factory), a German GIS consultancy who extract and process OSM data and then make it available for free online. It's really nicely structured and easy to find what you're looking for. Here's the download page for New Zealand, for example - followed by the contents, where you can see the building data on top of a current OSM base map. At time of writing, the zipped shp folder for the whole of New Zealand was 146MB.

The New Zealand GEOFABRIK download page (20 July 2015)

Auckland, NZ - very good building coverage here

The OSM sources are great, since the licence is very generous and you can use the data for just about anything, so long as it's properly cited. However, many towns, cities and counties across the world also provide building footprint or outline data (the terminology varies from place to place) so I've put together a list below of ones I know about. Some of them (e.g. Detroit, NYC) cover land parcels or tax lots so are slightly different but in the main it's just building outlines. I've included visuals for some of the datasets, so you can get an idea of what they look like.

New York City - from the BYTES of the BIG APPLE website you can download the MapPLUTO dataset, for all 5 Boroughs in New York City. Tax lot level rather than building outlines, but it's an extremely rich dataset with loads of useful land use planning variables in it, including 'year built' and number of floors. A little sample of the data are shown below, for the area around Central Park.

A little sample of the data (using Qgis2threejs)

Chicago - the building footprints layer is avaiable in two versions online, one of which says it is deprecated but I've heard from the Chicago GIS team that this isn't the case. It's just that due to limited staff the dataset is only edited when necessary. Also contains a 'year built' and height variable.

San Francisco - another really good city buildings dataset, from SF OpenData. Also lots of useful variables in this dataset, including height. I really like this one.

Dallas - you'll probably get a disclaimer box in a pop-up when you go to download this. I've linked to the general GIS page and the file you want is called Structures (Building Footprints) in the Planimetric Data section - it's about 81MB to download and the unzipped file is well over 100MB.

Atlanta - again, I've linked to the GIS page, this time from the City of Atlanta and you need to download the 'Impervious Buildings' layer. If you're looking to map the sprawl of Atlanta, this won't work as it covers the City area only. Still, a very useful dataset.

Denver - excellent open data from Denver. This dataset covers all permanent structures and buildings for a 152 square mile area of the City and County of Denver. Available in a number of different formats.

Seattle - this dataset was created in 2009 by Pictometry International Corp but is now in the public domain. It is available via the City of Seattle's data website.

Los Angeles - this is a fantastic dataset for the County (not just the City) of Los Angeles, which is the most populous county in the United States (just over 10 million). Made available via the LA County GIS Data Portal. It is a little hefty (581MB) so be careful! In the example below I show all the buildings in LA County but the City of Los Angeles in dark shading, just to emphasise its crazy shape.

Boston - this was created in 2012 and is available via the City of Boston. Contains a number of different fields, including base elevation of the structures, the elevation of the highest point above sea level and fields on building type.

Detroit - like New York, not strictly a buildings outline file but instead a property lot level dataset. Very impressive dataset produced by Data Driven Detroit's Motor City Mapping project. I've used this data a lot in talks and teaching as it's a really good example of its type.

Now some links to further datasets which I know of but haven't used that much...

Washington DC - link is to the download page, but direct link to zip is here (559MB unzipped)

Baltimore - the top link on this page

Philadelphia - via OpenDataPhilly

Massachusetts - buildings for a wide range of towns and cities in the state

Boulder - this is from Boulder County, Colorado. Available in a number of different file formats.

Bloomington, Indiana - one of many smaller cities with excellent geodata

New Orleans - an excellent dataset, not just because of the unusual shape of the city!

Toronto - don't be confused by the '3D Massing' terminology here. Scroll down to the 'Data download' section

Vancouver - doesn't cover the whole city and they were digitised in 1999 but still a useful dataset.

Waterloo - this is from the Region of Waterloo and was up to date as of January 2014.

Hobart, Tasmania - an nice example of building data from Hobart in Australia. Contains a 'year constructed' variable.

Wellington, NZ - can't overlook New Zealand! I think you need to register to download this but it's Creative Commons 3 so still open. 

The list wouldn't be complete without mentioning OS OpenData for Great Britain, provided by Ordnance Survey. A new dataset with detailed buildings became available in March (the OS Open Map - Local) dataset. The building data is a very small part of this collection but one I find very interesting. I've patched together a few cities here to get the ball rolling but you can download your own. There's also a 'tile finder' to help you identify which OS tile you need to cover your area of interest. 

This could save you some time 

I think this just about covers it. Get in touch if you have any other great data sources for building outlines.

US Election 2012 County-Level Results

I've been experimenting with the 2012 US election results at the county level* published on the Guardian Datablog and comparing percentages for Obama and Romney. One of the most striking things is how Obama won in DC (over 91.4% of the vote) and in the Bronx, NY (91.2%). The highest percentage for Romney was in King County, Texas where he won 95.9% of the vote. The image below shows these patterns and also includes some information on race, with % Hispanic and % African-American mirroring, to a large extent, the percentages of voters choosing Obama. It's not really that simple of course, but there is a correlation. One interesting nugget here is the difference in total votes won in counties with the highest voting percentage for each candidate. In the 5 counties with the highest Obama percentage, almost 850,000 voted for Obama. By contrast, in Romney's top 5 counties the total was just under 20,000 voting for him.

Full size version here

What does any of this mean? It means that the Republican Party probably need to think about how to do better in cities, with Hispanics and with African-Americans, but they already know that. Romney was very successful in areas where not many people live but not successful enough in major cities. It's all pretty obvious but it stands out more when you look at it on a 3D map!

*(for the continental United States, so no Alaska or Hawaii for now)

Comparing Populations: China, US, Europe

Since the US and China have both recently (2010) conducted a census I thought it would be interesting to look at some of the population results. Given the vastness of the US and China, I thought I'd do a little experiment and compare Chinese province populations with US states and European countries. I've put this data into a spreadsheet as well and also produced a map graphic comparing various areas. The purpose here is to highlight the large populations of so many Chinese provinces.

Some interesting nuggets from the data...

• Guangdong province in China now has a population of 104 million, which is more than any European nation except Russia (141 million). The European part of Russia has 110 million. 
• Poland, Shanxi province and California all have around 37 million people. 
• Shanghai has 22 million compared to Romania at 23 million. 
• Beijing and New York state both have about 19.5 million people. 
• In total, ten Chinese provinces have a population of 50 million or more. The US and China have a very similar land area. 
• Be honest, did you know that Anhui province has about the same population as Italy (60 million)?

Finally, some non-Chinese comparisons. Greece and Ohio both have a population of about 11.5 million and Michigan and Belarus are similar at about 9.5 million. Wisconsin and Denmark both have about 5.7 million people, and Finland and Minnesota have about 5.4 million. At the lower end of the scale, North Dakota and Montenegro both have about 670,000 people.

United States Census 2010

The 2010 US Census was conducted in April 2010 and already the results are looking very interesting. By the end of 2010 there was a new total population figure for the US, indicating a growth of 9.7% between 2000 and 2010. The total population on the twenty third US Census day was 308,745,538. This is just over double the total population from 1950. For a more up to date population estimate, you can check the US population clock from the US Census Bureau. Because I'm interested in all this, I've produce a graphic which shows population density and some population data for the lower 48 states, at county level. A couple of nuggets here: Los Angeles county has nearly 10 million people and Loving County (Texas) has only 82 people. All other counties lie somewhere in between...

World Prison Populations

I was going to post something else on deprivation, and I will do again soon, but today it's time for something completely different. For one reason or another I've been thinking about prison populations in different countries in the world. Maybe it's because I've been watching Banged up Abroad on National Geographic! Anyway, I got interested in it and started looking at the data and publications on the International Centre for Prison Studies website. It's all very interesting and at times alarming but thanks to the people at ICPS we have a reasonably good idea about all this. The 8th Edition of the World Prison Population List puts the total prison population at about 9.8 million in 2008 - roughly the same as the total population of Sweden! A couple of quick 'top ten' charts now...

The United States, China and Russia account for almost half the total world prison population. The US total in 2008 was 2.3 million, which is more than the entire population of Latvia and more than half the total population of Ireland. The prison population is more than the city population of Houston and not far off the city population of Chicago. The United States also has the highest rate of imprisonment per 1,000 persons, as you can see below...



In 2008, the rate of imprisonment in the US was 7.6 per 1,000. Russia was next at 6.3 and then Rwanda at 6.0. European nations generally have a much lower rate, with the United Kingdom at 1.5, France at 1.0 and Sweden at 0.7.

You can play around with the numbers to discover lots of interesting facts. If the UK had the same imprisonment rate as the US then instead of having a prison population of 90,000 or so, it would have a prison population of 465,000. On the other hand, if the US had the same rate as the UK, they would have 458,000 prisoners and not 2.3 million. Food for thought...

Given the topic, I'll end this rather random post with this...