New research uses artificial intelligence to simulate how changes to zoning might influence how far people travel each day thereby reducing transportation emissions.
Many U.S. cities have zoning policies that separate where people live from where they shop, work, and access daily services. This often leads to limited access to amenities and long car trips for basic errands that ultimately result in higher transportation emissions, especially in low-density neighborhoods.
A new study co-authored by Arianna Salazar-Miranda, assistant professor of urban planning and data science at the Yale School of the Environment, examines how land-use reform could help reverse those patterns. Published in Cities, the research uses generative AI to simulate how changes to zoning, such as increasing the mix of residential and commercial land use, might influence how far people in more than 400 U.S. cities travel each day.
“Mobility is key when we think of climate impacts because longer daily trips typically mean more emissions,” Salazar-Miranda said. “And one strong driver of those long trips is land-use configuration — where housing, jobs, and commerce are located in relation to each other.”
This research builds on ongoing work at the Yale Livable City Lab, in which scholars examine how land-use design shapes mobility patterns, access to daily needs, and broader sustainability outcomes across U.S. cities.
CT municipality considers increasing year-round shelter beds. Why it’s a tricky political gambit.
Transportation is the largest source of carbon emissions in the country, according to the U.S. Environmental Protection Agency. Until this study, however, there was limited empirical evidence on how specific zoning reforms might affect travel behavior across different urban contexts. This new research helps bridge that gap by combining nationwide parcel-level land-use data with anonymized mobility data from millions of smartphone users.
“What makes land use especially interesting is that it’s also one of the key levers planners can change,” Salazar-Miranda said. “My previous work shows that when everyday destinations are closer to home, people tend to make more short-distance trips. In this project, I take that a step further by asking how different ways of organizing land use could shift those patterns and support more sustainable daily mobility. Or put differently, if we were to actually change land use, where would it matter most, and by how much?”
“What makes land use especially interesting is that it’s also one of the key levers planners can change.” Arianna Salazar-MirandaAssistant Professor of Urban Planning and Data Science
Using a type of machine learning known as a generative adversarial network, or GAN, the research team trained a model to learn the relationship between land-use patterns and the share of trips that can be completed within a 15-minute walk. Once trained, the model allowed researchers to simulate hypothetical zoning reforms, essentially testing “what if” scenarios at a national scale.
On average, increasing land-use mix by 20% was associated with a 7% relative increase in short-distance trips. The gains, however, varied widely across cities. Those differences often could be attributed to a city’s starting point, according to Salazar-Miranda.
“We looked at differences in baseline built environment conditions — land-use mix, density, street connectivity, and existing levels of short-distance travel. What stands out is that starting conditions matter a lot,” she said. “Cities with low baseline land-use mix, lower density, and weaker street connectivity tend to see the largest gains when mixing increases. In those cities, even modest reforms can lead to sizable shifts towards shorter trips.”
By contrast, cities that are already compact and well connected tend to see smaller gains — not because zoning reform is ineffective, but because it has less room to move the needle. “Compact and well-connected cities already support relatively high levels of short-distance travel, so additional mixing produces smaller marginal gains,” she added.
Citywide zoning overhauls are rarely easy — politically or financially — so researchers tested whether zoning reforms need to happen everywhere or if carefully targeted, localized reforms might still achieve meaningful results.
“What we found is encouraging. Targeting low-density or poorly mixed neighborhoods produces gains that are comparable to those of citywide reform,” Salazar-Miranda said, noting that their findings support the idea that zoning reform need not be sweeping to be effective and that well-targeted incremental changes can meaningfully reduce travel distances and are more feasible to implement.
Developer changes plans to excavate CT nature preserve. Advocates say new proposal still destructive
As cities across the U.S. revisit decades-old zoning codes amid climate, housing, and transportation challenges, the research offers a new way to understand where land-use reform may have the greatest impact and how targeted change could reshape daily mobility.
“I see these kinds of tools as decision-support tools,” Salazar-Miranda noted. “Planning is hard partly because it’s difficult to visualize what alternative land-use configurations might look like at the scale of an entire city. Generative AI can help make those alternatives more concrete.”
Bree Shirvell is an assistant editor at Yale School of Environment.