While I love the Emerging Megaregions map, one of the things that it has most inspired me to do is to take a stab at recreating the map on my own terms. Some elements of the map are a little… far-fetched. Literally. The “urban corridors” laid out in the report can be pretty heavy on the corridor and very light on the urban. This is the most common critique of the Emerging Megaregions map. For example, anyone who has ever taken a road trip in the Western United States can tell you just how ambitious some of the megaregions defined by America 2050 are. It is indeed quite difficult to imagine, for example, suburban sprawl someday spilling over the crest of the Rocky Mountains to join Salt Lake City with Denver, or the Gulf Coast industrial zone ever reaching all the way down the Texas coast to the Rio Grande.
I wasn’t there when Emerging Megaregions was being drawn up, but I think that it’s quite clear to anyone who has ever really analyzed the map that there was a rather large measure of artistic subjectivity (and maybe just a dash of an ideological penchant for ever-grander megaregions) applied to its creation. My ambition has thus been to create a better-in-some-ways megaregions map — one that is based on objective analysis as much as possible — and I think that I’ve finally done it! The only bit of subjectivity involved with my version is the naming of the megaregions. (I did my best… please go easy on me!) Other than that little bit of creative license, every other aspect of this map is purely a product of the precise methodology that I established when I began working on this project.
That doesn’t mean that there aren’t some results that some will find strange, annoying, or even downright offensive — just please know that I did not divide up the regions based on any preconceived notions, I did not choose their extents or the cities within them, and I did not at any point “cheat” and make an editorial decision when regions were either joined or separated by connections so small that you could probably jump over them, even when the results ran counter to my own notions of U.S. regions. If you’re unhappy with one or more of the results here, I apologize, but I am also preemptively passing the blame on to simple math and geography. (I will go into some of the oddities of this map below.)
So, thank you for enduring my disclaimer — now for the methodology! For this project, I used U.S. Census Tracts as my geographic units. If you’re not familiar with Census Tracts, think of a neighborhood in a large city, or perhaps a small suburb, or a cluster of neighboring farm towns. In the geographical hierarchy, they sit just below counties. If you’re familiar with the geography of U.S. counties, you know that they can be an awkward tool for delineating regions, with their arbitrary boundaries and wildly varying sizes. San Bernardino County in California, for example, is larger than nine entire U.S. states! Many of the enormous counties in the Western U.S., like San Bernardino, include both densely populated urban areas and vast uninhabited regions, meaning that if you try to use county boundaries to define metropolitan regions, you’re going to have to make some tough decisions — either incorporating large, empty areas into your metro, or leaving out the parts of said metro that happen to be just over the county line in a “rural” county.
Census Tracts avoid this problem by breaking counties down into their functional parts. There are about 74,000 tracts in the United States, so to get a sense of just how large a tract is, that averages out to about 4,500 residents per tract (330 million ÷ 74,000). Census Tracts are also great because the Census Bureau regularly updates their data at the tract-level, so recent and complete data is always available. For this map, I was able to use population data for 2019. Finally, the tracts are contiguous, meaning that you don’t have to mess with city boundaries that can be all over the place, or figure out how to deal with the many unincorporated suburban and exurban communities that make of so much of our inter-city corridors.
So, I collected all of the data, located all tracts with a population density of 20 persons per square kilometer (that’s 50 persons per square mile for you imperial hold-outs), and then clustered all of those tracts. If any part of two tracts touched, they were merged, regardless of any other circumstances. This method resulted in some enormous networks of clusters in the densely populated Eastern U.S. — namely three super-megaregions: the Northeast, the Great Lakes, and the Upper South. In addition to the three big boys, there were 18 other smaller clusters that each reached a population of 2 million. And that’s basically it! The result is a map of 21 urban megaregions, defined objectively by measurable criteria and free from the whims of the mortal geographer.
Okay, so here some of the weird, counterintuitive, and frustrating things that came up with this methodology:
- Almost-regions: The Puget Sound (Seattle) and the Willamette Valley (Portland) come within about 15 km of touching, almost-but-not-quite meeting at the Columbia River and forming a combined Cascadia megaregion. San Francisco Bay and Sacramento-San Joaquin come within about 10 km of touching in a couple of spots, almost forming a combined Northern California megaregion. The “Texas Triangle” megaregion is probably still at least a generation away from becoming a reality, but right now only about 50 km separate the San Antonio-Austin corridor from both Dallas-Fort Worth and Greater Houston. Much to my horror, New Orleans, the crown jewel of the Gulf Coast, is separated by a hair’s width from Baton Rouge and thus from the rest of the Gulf Coast megaregion. That one hurt. For the grand finale, the three super-megaregions of the Eastern U.S. — The Northeast, the Great Lakes, and the Upper South, all come within distances of about 10 km or less of all merging into an Eastern U.S. ultra-region (along with Western New York) that would be home to half of the U.S. population. In most of these cases, the distances and differences in density are so small that I would expect for most of these almost-regions to fully merge when I update this map with 2020 or 2021 data.
- Barely-regions: Some of the megaregions are nice and solid, like the Northeast Megalopolis, but some are only held together by a thread. The Chicago-Detroit region ever so slightly touches the Cleveland-Cincinnati region near Toledo to form the Great Lakes megaregion. Further south, Birmingham is barely hanging onto Atlanta and the Upper South megaregion, while Nashville is only there because it, in turn, is barely hanging onto Birmingham. And there are countless smaller cities and corridors that come very close to merging into their neighboring megaregions or forming new ones, such as in the Central Valley of California, which could eventually form the spine of a California super-megaregion.
- Regional names: Some were obvious, like San Francisco Bay. Some were not so obvious, and I did my best to give the regions names that would balance familiarity with geographical accuracy. A couple of the megaregions, like San Antonio-Austin and Sacramento-San Joaquin don’t even really have good names yet. Some lone-wolf cities, like St. Louis, don’t even seem to have names for their regions other than just the name of the principal city. And then there were some regional names that came close to properly identifying and describing one of the megaregions, but with some crucial flaw. (I’m sure that I’ll be called out for using “Upper South” to describe a region that definitely dips down into the Deep South in parts, but there is just no other name for it, and it was certainly a better choice than “Upland South”. I realize that Emerging Megaregions had a name for this region — “Piedmont Atlantic” — but that is a name that, as far as I can tell, is only used by people referencing Emerging Megaregions, and I wanted to avoid using “made-up” names as much as possible.)
- Finally, if you have a bone to pick about a city that was included or left out of the “Major Hubs” column, I had a methodology for that too. It is too boring to go into here, but basically the component cities were weighted based on their size relative to the principal city. So, smaller megaregion = smaller hubs; larger megaregion = only the very largest hubs. I have nothing against your hometown, I swear! Nor do I have anything against Hawaii or Alaska — they were included in the data set but just didn’t produce any megaregions.
This map definitely has its limitations. At the end of the day, even Census Tracts occasionally have arbitrary boundaries which will affect the outcome of the methodology. Going down to the next level, Census Block Groups, would be ideal, but would involve exponentially more computing power. The other major limitation also comes down to data availability: I would love to be able to include Mexican and Canadian border districts. I guess what I’m saying is that this is “version 1”, with updates to hopefully follow.
TLDR: Took all of the neighborhoods and towns in the U.S. that are moderately populated, clustered them together, identified the largest clusters, and put them on a map with funny names.
Data source: U.S. Census Bureau.
Tools: Excel for data management, ArcGIS Pro for mapping, and GIMP for final design.