Key takeaways:
- River levels once expected only every 50 years are now occurring every few years as climate change intensifies extreme rainfall.
- An interdisciplinary team of Penn researchers has identified a tipping point at the 1-in-100-year threshold, when floodwaters escape banks and rapidly spread into the city.
- Philadelphia's impervious surfaces increase runoff, while buildings and infrastructure trap water, intensifying flooding across neighborhoods.
- Both the most affluent and most vulnerable communities face high flood exposure, though driven by different structural factors.
On most mornings, the Schuylkill River Trail is where many Philadelphians step out for a breath of fresh air. Runners check their splits, cyclists zip below bridges, and dog walkers clutch leashes and coffee.
But after the remnants of Hurricane Ida and seven tornadoes struck Philadelphia in early September of 2021, those routines were put on pause as parts of the trail disappeared beneath a sea of sludgelike brown water. The Vine Street Expressway—running between the Schuylkill and Delaware rivers—was closed as water flooded the roadway , nearly reaching the overpasses.
To better understand the impact of the storm on Philadelphia, an interdisciplinary research team from the University of Pennsylvania has developed a high-resolution model, paving the way for better forecasting to mitigate damage as extreme weather events become more common.
"We got to use Philadelphia as a natural laboratory to model how floods actually unfold at this scale, including rail lines and neighborhoods," says geophysicist Hugo Ulloa , who supervised the research. "It was a unique opportunity to use city data and local expertise, to serve the city better for the next events."
The researchers show that soaked soils, paved surfaces, and overburdened drainage systems are intensifying flooding, and river levels once expected in the 1950s to occur every 50 years now statistically happen every three years. They also identified the 1-in-100-year threshold as the "tipping point," when flooding escapes containment and spreads rapidly across the urban landscape. Their findings are published in npj Natural Hazards .
"Hurricane Ida exposed a growing challenge in how we predict urban flooding with a changing climate," says Ulloa. "Existing models are too coarse and broad; they simply cannot resolve the complex, compounding interactions between rainfall, river flows, tides, sediment, and the urban landscape, which limits effective mitigation in densely populated regions."
Mapping a disaster
Ulloa and climate scientist Michael E. Mann worked with geophysicist Douglas Jerolmack and roboticist M. Ani Hsieh to assemble one of the most detailed views of the Schuylkill River basin to date. They used data collected by autonomous surface vehicles that mapped the riverbed at high resolution, capturing features typically smoothed over in large-scale surveys.
Dingyu Xuan , a former undergraduate researcher in Ulloa's GEFLOW Lab , used that data—along with LiDAR terrain models, high-resolution 3D maps of the ground created using laser scanning, and land-use information—to create a street-level model that recreated Ida's wrath block by block.
The researchers found that Philadelphia's vast blanket of impervious surfaces severely restricts the earth from soaking up the rainfall, converting nearly all precipitation directly into runoff during a storm like Ida.
Simultaneously, the city's labyrinth of engineered infrastructure—buildings, roads, and levees—acts as a massive bottleneck, impeding drainage and trapping stagnant floodwaters in low-lying neighborhoods long after the storm passes.
The cost of riverside property
"We learned that Ida was not an outlier so much as it was a stress test," says co-author Leandro Pongeluppe, an economist. "The systems meant to hold the line are now being asked to perform under conditions they were never designed to handle."
What makes this an unusual story, Pongeluppe notes, is that it is the poorest and wealthiest Philadelphians most at risk of exposure to flooding. For high-income residents, exposure is driven by extensive impervious surfaces and the ground sinking under the weight of dense infrastructure. For low-income communities, the risk is a product of high housing burden—financial strain of spending a disproportionate share of their income on housing—insufficient flood preparedness, and systemic neglect of environmental protections.
Analyzing ZIP code-level loan data, Pongeluppe also found that approved federal disaster loans did not come close to matching the scale of the damage.
Preparing for the next flood
The danger doesn't vanish when the water recedes, Ulloa notes, as contamination from floodwaters that get trapped in low regions can cause long-term public health effects. Floods can carry sewage, human waste, and other river pollutants into low-lying city streets where the water gets trapped.
"Floodwaters are like chocolate, thick with suspended, fine sediment," he explains. "When the flood comes, it penetrates low regions where the water gets trapped. The sun comes, evaporates things, and then you get this dusty layer through the city that can be contaminated and can get airborne if the wind blows. These are serious considerations that are not usually discussed, and they have a more long-term effect in the city."
Since Philadelphia is too intimately built around the Schuylkill to retreat, Ulloa advocates for expanding green infrastructure—turning open spaces into highly permeable sponges to absorb floodwaters.
Looking ahead, the team is securing funding for building an AI-powered "digital twin" of their Philadelphia model. Trained on their physical model, this twin would be able to process shifting storm forecasts and calculate street-level flood predictions in a matter of minutes rather than days.
Hugo Ulloa is an assistant professor in the Department of Earth and Environmental Science at the School of Arts & Sciences at the University of Pennsylvania.
Leandro S. Pongeluppe is an assistant professor in the Management Department at the Wharton School at Penn.
M. Ani Hsieh is an associate professor in the General Robotics, Automation, Sensing, and Perception (GRASP) Lab and the Department of Mechanical Engineering and Applied Mechanics at the School of Engineering and Applied Science at the University of Pennsylvania.
Douglas Jerolmack is the Edmund J. and Louise W. Kahn Endowed Term Professor of Earth and Environmental Science in the School of Arts & Sciences and a professor in the Department of Mechanical Engineering and Applied Mechanics at the School of Engineering and Applied Science .
Michael E. Mann is the Presidential Distinguished Professor in the Department of Earth and Environmental Science in the School of Arts & Sciences at the University of Pennsylvania and the director of the Penn Center for Science, Sustainability, and the Media . He holds a secondary appointment in the Annenberg School for Communication .
Dingyu Xuan was an undergraduate researcher in Ulloa's GEFLOW Lab who is now a PhD student at Imperial College London.
Mackenzie M. Weaver, a PhD candidate and graduate researcher in the Mann Group, also coauthored the study.
This work received support from the University of Pennsylvania's University Research Foundation Award, Leonard J. Horwitz, the Wharton AI & Analytics Initiative, the U.S. Army Corps of Engineers Philadelphia District, and the National Weather Service Middle Atlantic River Forecast Center.
