Computer models evaluate risks before floods

Before waters rise, governments, banks and insurance companies can help people who live in flood-prone areas, thanks to researchers in the UNC Gillings School of Global Public Health.

Professor Greg Characklis, director of the Center on Financial Risk in Environmental Systems, and his team used cutting-edge mathematical techniques to develop computational models that combine census records with environmental and financial data to evaluate risks. The first-of-its-kind modeling involves a series of advanced approaches, including machine learning, to estimate the probability of mortgage default and property abandonment down to the neighborhood scale. Characklis is the W.R. Kenan Jr. Distinguished Professor in the school’s environmental sciences and engineering department.

What the models show can aid policymakers and stakeholders in creating effective and equitable strategies for helping communities recover. After floods and other natural disasters, people sometimes abandon their damaged houses. They may default on mortgages because repair costs are either unaffordable or higher than the property’s value.

Greg Characklis

Improved information on who is at risk gives policymakers a better chance at reducing mortgage default rates through aid programs. It can also reduce property abandonment rates, which can occur when neither owners nor lenders take responsibility for damaged properties. In such cases, local governments often bear the burden of maintaining or demolishing them.

The models produce aggregated data reports based on census tracts, which are small, relatively permanent statistical parts of a county that average about 4,000 inhabitants. The reports pin-point high-risk neighborhoods that governments and entities can help by developing policies, providing resources, subsidizing flood insurance or creating buyout programs. The models use data on home proximity to streams and other bodies of water, home elevations, paved land, flood policies and claims. The models also evaluate census records, mortgage data and home values to estimate the amount of damage and estimate the risk of default or abandonment.

By estimating mortgage balance and home values, researchers determine which homeowners can borrow against their home’s equity to make post-flood repairs. “Those with lower mortgage balances and more equity in their homes are better positioned to borrow against the value of their home. Homeowners who have flood insurance are less vulnerable to default because they have the ability to draw on their policy to pay off some or all losses,” Characklis said.

For a 2023 study published in Earth’s Future, a model estimated $562 million in previously unquantified financial risks in eastern North Carolina from property value changes and uninsured flood damages associated with Hurricane Florence. Characklis and Antonia Sebastian, assistant professor of Earth, marine and environmental sciences, co-authored the paper with lead author Hope Thomson, a graduate student.

“All this financial risk hadn’t been quantified before,” Characklis said. “Our work has the potential to provide information that can be used to improve situations for people.”

Characklis has briefed some state government officials on the models’ capabilities, and he plans to inform N.C. General Assembly members. The North Carolina Collaboratory, established by the General Assembly, funds the lab’s work.

“Our government does not want people in FEMA trailers or in motels or hotels for extended periods,” he said. “By identifying financially at-risk groups, the state can better target aid.”