Experts from the University of California, Irvine, have found that national flood risk models used by government agencies, insurance companies, and disaster planners are not accurate at local levels, such as neighborhoods and individual homes.
Experts from the University of California, Irvine, have found that national flood risk models used by government agencies, insurance companies, and disaster planners are not accurate at local levels, such as neighborhoods and individual homes.
A recent study published in the American Geophysical Union journal Earth’s Future warns that newer nation-wide flood data fails to reflect local geography and infrastructure, which are crucial for understanding how floods spread in urban settings.
“Our analysis of Los Angeles County, which has a population larger than 40 U.S. states and encompasses over 80 municipalities, revealed that the overall flood exposure estimates using national data are surprisingly similar to those from our more intricate models. However, the predictions regarding which neighborhoods and properties are vulnerable differ significantly,” stated Brett Sanders, lead author of the paper and a Chancellor’s Professor at UC Irvine in civil and environmental engineering as well as urban planning and public policy.
“Moreover, the disparities between these models indicate a significant inequality in exposure among various social groups, including Black, white, and marginalized communities,” he added. “Identifying flood risk hotspots and understanding social disparities are essential for planning urban flood responses, and relying too heavily on current data may result in inadequate protective strategies.”
Sanders and his research team from UC Irvine and the University of Miami created a detailed modeling system called PRIMo-Drain, which enhances flood prediction accuracy by using high-resolution topographic data, details on levees and channel conditions, and specific information about stormwater systems like culverts, underground pipes, and street drainage.
“When we compared the risk assessments derived from national data models with those generated by PRIMo-Drain, we found that estimates varied significantly from city to city—by a factor of ten,” Sanders explained. “Additionally, there’s only a 25% chance that the national data aligns with UC Irvine’s data regarding which properties are at risk of experiencing over a foot of flooding from severe weather events.”
Sanders noted that federal initiatives aimed at mapping flood hazards across the U.S. have struggled to adapt to changes in land use and climate, which leaves governments and the insurance sector needing updated information to manage risks effectively.
“While new national data sources from the private sector have been developed to meet this pressing demand, these models unfortunately lack the detail required to accurately assess flood risks in urban areas,” Sanders remarked. “Future models should include more comprehensive details about drainage systems like levees, floodways, culverts, and storm drains, as well as detailed hydrological and bathymetric data.”
The team highlighted a new collaborative approach for enhancing data accuracy nationwide.
“By engaging in collaborative flood modeling, where scientists and engineers employ advanced regional models in partnership with stakeholders, we can achieve economies of scale. This would help extend coverage to less affluent and smaller communities while boosting flood awareness and preparedness among affected populations,” Sanders noted. “Awareness of flood risks is essential for participation in flood insurance programs; more accurate data can aid insurance companies in identifying insurable properties, and property owners will be more informed about effective flood-proofing strategies.”
Brett Sanders was joined in this project by Jochen Schubert, a civil and environmental engineering research specialist at UC Irvine, and Katharine Mach from the University of Miami. The research team received data assistance from the First Street Foundation and high-performance computing support from the NCAR-Wyoming Supercomputing Center, funded by the National Science Foundation.
