What Can Satellite Imagery Tell Us About Obesity in Cities?

A new AI can figure out which elements of the built environment might influence a city’s obesity rate

A satellite image of Los Angeles (Deco Images II/Alamy )
smithsonian.com

About 40 percent of American adults are obese, defined as having a body mass index (BMI) over 30. But obesity is not evenly distributed around the country. Some cities and states have far more obese residents than others. Why? Genetics, stress, income levels and access to healthy foods are play a role. But increasingly researchers are looking at the built environment—our cities—to understand why people are fatter in some places than in others.

New research from the University of Washington attempts to take this approach one step further by using satellite data to examine cityscapes. By using the satellite images in conjunction with obesity data, they hope to uncover which urban features might influence a city’s obesity rate.

The researchers used a deep learning network to analyze about 150,000 high-resolution satellite image of four cities: Los Angeles, Memphis, San Antonio and Seattle. The cities were selected for being from states with both high obesity rates (Texas and Tennessee) and low obesity rates (California and Washington). The network extracted features of the built environment: crosswalks, parks, gyms, bus stops, fast food restaurants—anything that might be relevant to health.

“If there’s no sidewalk you’re less likely to go out walking,” says Elaine Nsoesie, a professor of global health at the University of Washington who led the research.

The team’s algorithm could then see what features were more or less common in areas with greater and lesser rates of obesity. Some findings were predictable: more parks, gyms and green spaces were correlated with lower obesity rates. Others were surprising: more pet stores equaled thinner residents (“a high density of pet stores could indicate high pet ownership, which could influence how often people go to parks and take walks around the neighborhood,” the team hypothesized).

A paper on the results was recently published in the journal JAMA Network Open.

It’s hard to untangle certain urban features from the socioeconomic status of the people who live near them—wealthier people are more likely to live near parks, but is it the park that makes them less likely to be obese, or is it other features of their privilege, such as access to healthier food and more leisure time to cook? It takes more than just an algorithm to answer these questions.

“[T]his work points to how big data and machine learning can be integrated into clinical research,” writes Duke University’s Benjamin Goldstein, David Carlson and Nrupen Bhavsar, in a commentary on the work. However, they caution, “this does not mean analysis alone can provide all of the answers. At their core, these analytic techniques only point to features, and providing meaning to them requires subject matter insight.”

The University of Washington team has worked in the past on other projects using satellite data to predict health outcomes. One project involved looking at the number of cars in hospital parking lots during flu season to predict when outbreaks were beginning. They hope that this newest project will have applications beyond obesity.

“We’re hoping that it will be useful for people that study the built environment and its relationship to obesity but also other chronic conditions,” says Nsoesie.

A number of chronic conditions besides obesity are associated with lack of activity and poor diet, including certain cancers, heart disease and diabetes.

They also plan to look at data longitudinally—as cities change their features, do obesity rates change along with them?

“We hope this will be useful for city planners,” Nsoesie says. “We can think about the way we design neighborhoods to encourage people to go out and exercise.”

The project’s findings are supported by other research on the effects of the built environment on obesity. James Sallis, an expert on cities and public health at the University of California, San Diego, says a city’s walkability is well-known to be associated with lower obesity rates. Walkability is a product of many design elements, including streets that are connected (as opposed to dead-ends or cul-de-sacs), safe sidewalks and crosswalks, and destinations (as in, is there anywhere to walk to?).

But making changes is easier said than done, Sallis says, due to zoning laws that favor cars over pedestrians and sprawl over the kind of high density that promotes walkability.

“We know what to do,” he says. “But what we need to do is very different from what we’ve been doing for the past five or six decades.”

Tags

Comment on this Story

comments powered by Disqus