A team of researchers from Penn State and the University of Pittsburgh are developing a new method for infrastructure design that accounts for the continual rise and uncertainty of severe weather events, based upon the cost to adapt and maintain infrastructure over its intended lifespan. Here, a woman walks on a median in the middle of a flooded highway in New Brunswick, New Jersey, after Hurricane Ida. IMAGE: ISTOCK/@SARAH J LEE
$2 million grant awarded for climate-adaptive infrastructure design
9/28/2021
By Tim Schley
UNIVERSITY PARK, Pa. — In 2020, the U.S. experienced a record number of billion-dollar disaster events, with damages totaling approximately $95 billion. It was the sixth consecutive year the National Oceanic and Atmospheric Administration recorded at least 10 such events.
This year, severe weather events such as Hurricane Ida are again taking their toll on the nation’s infrastructure, and they continue a decades-long rise in frequency and intensity. Yet, much of the new infrastructure designed today does not account for the intensification of future climatic hazards, according to Gordon Warn, associate professor of civil and environmental engineering at Penn State.
“When loads and demands are calculated in order to design the infrastructure — a bridge, a building, or a culvert — those loads come from historical data,” Warn said. “Part of the problem is that the historical data no longer reflect what we can expect in the future.”
Warn is the principal investigator of a four-year, $2 million project funded by the National Science Foundation’s Leading Engineering for America’s Prosperity, Health, and Infrastructure program. The project, titled, “Optimal design and life-long adaptation of civil infrastructure in a changing and uncertain environment for a sustainable future,” aims to develop a new approach for designing, adapting and maintaining the built environment using climate projection models and artificial intelligence (AI) to evaluate the likely long-term cost and environmental impacts of infrastructure design choices.
“Typically, the design, construction and maintenance of infrastructure are handled by different teams with little integration, and this can translate into high life-cycle costs,” Warn said. “The proposed framework addresses this by simultaneously integrating the design, maintenance and adaptation phases, while considering significant future uncertainties, to optimally satisfy life-cycle objectives.”
Co-principal investigators include Penn State researchers Chris Forest, professor of climate dynamics; Lauren McPhillips, assistant professor of civil and environmental engineering and agricultural and biological engineering; Kostas Papakonstantinou, associate professor of civil and environmental engineering; and University of Pittsburgh researcher Melissa Bilec, William Kepler Whiteford Professor of Civil and Environmental Engineering and co-director of the University of Pittsburgh's Mascaro Center for Sustainable Innovation.
The researchers proposed that climate model predictions could be incorporated into computational frameworks for life-cycle assessment, which account for future potential hazards while determining the optimal way to construct, adapt, repair and maintain an infrastructure through its intended lifespan. This is achieved through deep reinforcement learning, a form of AI where an autonomous “agent” intelligently learns to achieve certain objectives over time, such as minimizing life-cycle cost.
“It incorporates future uncertainty into the decisions and continuously updates the lifelong maintenance and adaptation actions based on new information that becomes available in time,” Papakonstantinou said.
In this approach, the design team evaluates multiple design alternatives using low-fidelity, or less detailed, models. When they identify the more viable alternatives based on their life-cycle priorities — keeping to a specific budget while minimizing environmental impacts, for example — these designs are evaluated again using higher-fidelity data, leading to more precise life-cycle estimations.
This “multi-fidelity” framework saves computational resources while allowing designers to identify certain decision-making trade-offs between safety, resilience, resource consumption and environmental impacts, according to Warn.
“For example, a designer could choose to ‘over-design’ bridge piers by making them exceptionally large in size to account for rising river flood levels,” he said. “But this adds cost and uses extra materials like concrete that contribute more carbon dioxide emissions.”
Instead of over-designing the infrastructure, the designers could consider an alternative to include provisions to adapt the structure in later years, or they might look to restore a nearby wetland to help mitigate flood height.
This type of “green infrastructure” solution is a critical tool for teams to consider as they design for future climate change, according to McPhillips.
“We are trying to find solutions that are both sustainable and resilient,” she said. “This means solutions built to last in the face of multiple stressors that will not exacerbate climate change based on their own carbon footprint.”
The researchers plan to demonstrate their approach through a variety of practical design scenarios, including a riverine bridge vulnerable to flooding and deterioration; an urban mid-rise building stressed by increasing temperatures; and a port system threatened by sea level rise.
“If you want to mitigate sea-level rise for a port, there are various options,” Warn said. “For example, should the existing wharfs be elevated? If so, how tall should they be? What is the storm mitigation strategy? Should there be coastal armoring or a managed retreat to higher ground because it is too costly to build? Overall, the framework is intended to thoroughly evaluate diverse design concepts and adaptation strategies while considering, based on the best science, where we think the climate demands will be in the future.”
Share this story:
