Garrett Dash Nelson is the incoming curator of maps and director of geographic scholarship at the Norman B. Leventhal Map & Education Center at the Boston Public Library.
Density is one of the most important urban characteristics. A high-density city like New York looks, feels, and functions differently than a low-density city like Des Moines. Yet the textures of density within a single city can be just as varied as average densities between cities. Pockets of intense concentration or islands of relatively thin settlement sometimes bear little resemblance to the overall density of an urban area.
Since most people experience a city at the scale of a neighborhood or district, rather than the scale of metropolitan regions, paying attention to these fine-grained variations in urban form is more useful than broad-stroke statistics for understanding the everyday realities of population density. A new visualization I’ve put together helps picture these local patterns by examining density in American cities one square kilometer at a time. (Technically, these rectangles are 30 arc-second cells, which are roughly, but not exactly, one square kilometer.) This perspective lets us drill down to the micro-geography of density in American cities, with oftentimes surprising results, as in the densest grid cell of greater Salt Lake City (below), which isn’t high-rises but mobile home parks.
Salt Lake City
Most statistics that try to capture a single average measure, after all, tell us fairly little about spread or distribution. Consider economic indicators: If we know that the average income in a country is, say, $55,000, we don’t know anything about whether that country has a lot of middle-class earners or whether it’s polarized between extreme wealth and extreme poverty.
The same problem holds with density. Imagine one city where 50,000 people live in a 5-square-kilometer superblock of high-rises, surrounded by 20 square kilometers of parks, and another city where 50,000 people live in 25 square kilometers of low-density single-family homes. Both cities would have an average density of 2,000 people per square kilometer, but their actual morphologies could hardly be more different.
Density statistics can be even more misleading than averages like income or age, because the divisor in the equation is a geographic unit that’s inevitably drawn with debatable boundaries. For instance, the Los Angeles metropolitan statistical area (MSA) contains hundreds of square miles of mountains in the San Gabriel National Forest where nobody lives—so why shouldn’t the Boston MSA also include the hundreds of square miles of ocean in Boston Harbor where nobody lives? The Minneapolis–Saint Paul urbanized area includes several large lakes and even a National Wildlife Refuge, so why shouldn’t the Miami urbanized area include some chunk of the Everglades?
There’s no consistent principle about whether to include uninhabited water, mountains, and deserts when calculating density—not to mention airport tarmacs, railroad classification yards, or central business districts with few residences.
One alternative way of measuring density involves dividing space up into a regular pattern of equally sized cells, without regard to the underlying physical or administrative features, and then tallying up the population in each cell. That’s exactly what’s available in the Gridded Population of the World dataset produced by NASA. (The European Commission also has a similar dataset called the Global Human Settlement Layer.)
Although there are still some statistical flukes worth keeping in mind in regard to gridded population data (which I’ve noted in the visualization’s information page), this method of measuring population and density allows us, first of all, to set aside the problems of arbitrary boundary drawing and, second, to zoom in on small-scale patterns without having to fall back on broad averages.
In the interactive visualization, I’ve taken GPW data for a curated selection of American cities. Some have old, historic cores, and others are dominated by more recent development; some have constricting physical geographies and others lie on relatively flat, open plains; some were built for horse transportation and others for the automobile era. I’ve also included one rural region, the Upper Valley of New Hampshire and Vermont, where I currently live and work. For each area, I’ve selected out the 200 most highly populated grid cells from the GPW and matched these up with aerial photography.
One of the first things that stands out is how much the variation in the top 200 cells differs from city to city. The bottom of the visualization shows a “pyramid” of the square-kilometer cells in a given city, with the largest on the left.
San Francisco–Oakland has a strikingly top-heavy pyramid: Ranking #1 is an extremely dense cell of more than 25,000 people in the Tenderloin, but by the time we reach the #6 most dense cell, we’ve already more than halved the density, and by the time we reach rank 100, we’re down below 5,000 people.
Quite by contrast, Salt Lake City has a remarkably even spread of population. The densest cell is only a little more than twice as dense as the 200th most dense cell, and cells 40 through 143 are all within a range of 500 people from one another.
In this sense, San Francisco is an extremely unequal urban area in terms of density: It has a tight core of extreme density, but, outside of that limited area, its local-scale density isn’t very remarkable, and so life in most of the city isn’t significantly different from life in a city which has a lower average density overall. Other cities with historic cores constrained by geography, like Boston and Washington D.C., show similarly top-heavy density pyramids, as does Chicago, where the Loop and Near North Side stand out from the rest of the region.
The opposite is the case for cities like Salt Lake City, Minneapolis, and Detroit, which lack hyper-centralized downtown areas for both historical and geographical reasons, and consequently have a much more gradual contrast between core and periphery.
Even more remarkably top-heavy than San Francisco–Oakland is the rural Upper Valley region, where the densest cell, on Dartmouth College’s campus, has a density more similar to a major city than to its thinly populated immediate surroundings. (The #3 cell in this region is also on the Dartmouth campus, and the #2 cell is on the campus of Plymouth State University.) This raises the question of what we actually mean when we talk about the lived experience of density.
Upper Valley of New Hampshire and Vermont
College dorms, where each person might have just a few square feet of living area, are some of the most densely packed places in the country, and we can even see that phenomenon in larger cities. The 18th-densest cell in all of San Francisco–Oakland, which is also the densest in the East Bay, lies on the edge of the University of California, Berkeley, campus. Does that mean that the dorm clusters of college towns should therefore be the places that first come to mind when we think about big-city life?
There’s another, more sinister, phenomenon where we sometimes find highly localized spikes in density. The #2 most dense cell in Dallas (below) looked like an error to me at first, since it’s centered on the flood corridor of the Trinity River. But it turns out that the very eastern edge of this cell just overlaps the Lew Sterret Jail, a high-rise detention facility serving all of Dallas County. Famously sprawly Dallas, it turns out, does have a “neighborhood” that’s almost as dense as a coastal city—but the residents are all behind bars.
That should pose a question about whether the benefits of density are really just about a lot of people living in close quarters, or whether some other phenomenon is at play. Prisons and refugee camps, after all, are some of the world’s most astonishingly dense communities, yet it’s hard to imagine these as sites of rich intercultural dialogue, incubators of technological innovation, or seedbeds of economic opportunity, despite the role that density is supposed to play in promoting these virtues, according to many offhand theorists. Density, as a raw statistic shorn of context, can be deceptive.
Even among cells with similar population and density, the built environment can look strikingly different, the result of different cultural and economic factors in city-building across the country, and a tour of aerial photography gives us a window into the morphological diversity of American urbanity.
Denver, Minneapolis, and Los Angeles have endless neighborhoods of medium-scale residential neighborhoods following arterial roads and rectilinear streets. In Boston and Washington, D.C., leafy but still relatively dense inner suburbs like Arlington (Massachusetts) or Arlington (Virginia) are more typical. Atlanta’s top 200 cells are strikingly heterogeneous, and Miami’s include immigrant enclaves, beach developments, and even resort communities.
Another conclusion that we can pick up from this visualization is that New York City really is unlike all of its American peers—its density patterns belong in a category all their own.
The plot shows the population of the top 200 grid cells in six representative cities, and here you can see just how much of an outlier New York is. The city’s 200th densest cell is denser than the most dense cells of Boston, the Twin Cities, or Dallas. Even Los Angeles and Chicago’s densest areas barely crack into the bottom half of New York’s top 200.
From this perspective, it’s obvious both how extraordinary New York’s patterns are and also how varied the city is within itself. The #1 most dense cell, containing more than 42,000 people in Manhattan, is twice as dense as the #53 densest cell, with 21,000 people in Queens. Yet even that cell is still denser than the densest parts of Los Angeles or Chicago.
Exploring urban density on a cell-by-cell scale is also useful for exploring cities outside of the United States. European cities, for example, have density patterns that have much to do with historical trends in urban settlement, as my colleague Alasdair Rae has shown. The data designers at The Pudding have also visualized gridded density all across the world, allowing for a striking view of just how many people live in the arc stretching from Pakistan around to China and Japan.
The most provocative lesson to be drawn from thinking about population at the grid scale, however, is that it forces us to think about what we really mean when we talk about dense urban life. Is urban density about tall buildings, traffic congestion, and sharing the name of a city on your mailing address with millions of other people? Or is it more about the small-scale experience of everyday life, the number of neighbors you can reach within walking distance, the form and structure of a local community?
Just as average wealth and income statistics can’t take us very far in explaining how to fairly structure the distribution of economic gain in our society, the average population density of a vast metropolitan area does little to capture the many different ways that we might want to structure our cities and neighborhoods.