Coastal management model plays the long game against the rising tides

The problem? Future climate conditions might differ substantially from the used projections, according to Ashmita Bhattacharya, a civil engineering doctoral student at Penn State and first author of a study published in Nature Communications by an interdisciplinary team of researchers from Penn State and the University of Pittsburgh. Communities risk either overbuilding too much costly infrastructure -- the construction and maintenance of which also contribute excess carbon dioxide into the atmosphere, further exacerbating the rate of climate change -- or designing inadequate defenses. This can lead to excessive flood-related damages and even more costly repairs.

"The issue with the current state of practice in climate adaptation is the large uncertainty associated with how the climate demands us to evolve in the future," said Chris Forest, professor of climate dynamics in the Department of Meteorology and Atmospheric Science at Penn State and co-investigator of the study. Global temperatures shattered records quicker than expected in 2023 and again in 2024.

To help communities avoid making a misinformed, potentially expensive investment, with no ability to re-allocate spent resources if future conditions are not what was expected, the team created a model that provides decision support over time as more information becomes available, while keeping costs as low as possible, according to Gordon Warn, professor in the Department of Civil and Environmental Engineering at Penn State and co-corresponding author.

"Our approach suggests dynamic actions in time, responding to the actual evolving climate, while also considering possible future scenarios in an optimal sense," Warn said.

By suggesting adaptation actions over time rather than committing resources for a comprehensive protection system up front, the model's long-term strategic approach could potentially translate into significant savings for municipalities, according to Bhattacharya. The researchers tested the model on scenarios inspired by Manhattan and Staten Island in New York, finding that the model's long-term path of adaptation recommendations -- based upon observations of real-time conditions and long-term objectives -- resulted in lower overall costs compared to a conventional decision-making framework based on a static cost-benefit analysis.

The model is underpinned by advanced mathematical and computational techniques. One, known as "(Partially Observable) Markov Decision Processes," mathematically represents nature's unpredictability through "beliefs," which quantify the uncertainty about possible future states, assigning probabilities to each. The model constantly updates its understanding of these beliefs about future sea level rise scenarios as new data is collected, just as a chess player carefully studies the board after each move to understand the likelihood of an opponent's next maneuver.

The chess player may forego an immediate capture of a piece to keep another move in their back pocket for a later point in time. Similarly, the model may recommend a smaller seawall initially or not to take action unless certain conditions are met down the road. The model's sequence of suggested actions are generated through a technique called "dynamic programming," which optimally evaluates decision-tree like formations based on new data and the latest decision made.

"Key to this are either indirect or direct observations of the physical process -- for example, sea-level rise and storm surge, which are measurable via tidal gauges," Bhattacharya said. "This dynamic adaptation leads to lower costs in implementation, maintenance, damages and environmental impacts in comparison to the static cost-benefit actions."

The model also accounts for the environmental impacts of a potential build or maintenance action, added Bhattacharya. She explained that the researchers tied such actions as concrete seawall construction -- which generates a carbon footprint from cement manufacturing, mining, transportation, equipment use and future repairs -- to the U.S. Environmental Protection Agency's estimates of the social cost of carbon, the damage caused by each ton of carbon dioxide emissions released into the atmosphere.

In addition to traditional concrete-based infrastructure options, the research team also evaluated nature-based solutions that can be used on their own or in tandem with traditional "gray" infrastructure, like a concrete wall. For instance, a recommended action could include a smaller wall and the addition of an oyster reef, which could help lessen the impact of incoming waves at a much lower carbon footprint, or a salt marsh, whose carbon uptake properties also help offset the carbon dioxide released from the wall's construction.

In their simulation testing of New York City's coastlines, the researchers found that including the social cost of carbon encouraged earlier adaptation actions, and that adaptation actions were taken more frequently in comparison to cases that excluded it.

"This is because the goal of the adaptation problem is to minimize the overall costs of adaptation including damages and the emission of carbon that comes with it," Warn said. "What this result means is that by ignoring carbon emissions we are underestimating the overall cost of flood-related damages."

The team's forthcoming efforts center on scaling up the model and testing it against increasingly detailed scenarios and different coastal contexts. Eight of the world's 10 largest cities are located along coastlines, according to the U.N. Atlas of the Oceans.

"The framework remains the same, but the data to be used -- geography of the region, property values and so forth -- need to reflect the local regions," said Kostas Papakonstantinou, associate professor of civil and environmental engineering at Penn State and co-corresponding author.

He mentioned the model could eventually be used by the government or insurance companies to incentivize adaptation actions, likening it to when insurance rates were reduced for cars with the advent of anti-lock braking systems.

"Similarly, National Flood Insurance Program costs could reduce rates when protective measures are justifiably taken in time," Papakonstantinou said. "Insurance coverage after flood events can also be directly considered in our framework, particularly when paid by the entity that is also the decision maker, like the federal government."

Co-investigators from Penn State also included Lauren McPhillips, assistant professor of civil and environmental engineering and Digant Chavda, doctoral student in civil engineering. From the University of Pittsburgh, co-investigators included Melissa Bilec, the George M. and Eva M. Bevier Professor in Civil and Environmental Engineering and Co-director of the Mascaro Center for Sustainable Innovation; and Rahaf Hasan, doctoral student in civil engineering.

The U.S. National Science Foundation funded this project.