Saturday, May 25, 2024


Where your horizon expands every day.


Using Machine Learning to Enhance Flood Prevention Strategies

Two U.S. Coast Guard mariners push a red flat-bottomed boat through floodwaters in Baton Rouge, La., in 2016.

Source: Earth’s Future

The annual cost of flooding in the United States is over $32 billion. With the impact of climate change causing more extreme and unpredictable storms, this number is expected to grow in the future. According to predictions, flood risk is projected to rise by at least 26% by 2050. Unfortunately, due to disparities in flood prevention measures, flooding disproportionately affects urban areas with higher numbers of Black, Indigenous, and People of Color (BIPOC) residents.

In light of this situation, Veigel and colleagues utilize explainable artificial intelligence to gain a deeper understanding of the effectiveness of flood prevention methods. Their model was constructed using 400 factors related to behavior and socioeconomic status that impact disaster preparedness and mitigation. These factors encompassed individual actions, such as obtaining insurance or making property modifications, as well as larger-scale policies at the community level. The team utilized publicly available data from the National Flood Insurance Program and the American Community Survey conducted by the U.S. Census Bureau.

The findings revealed that many households only purchase flood insurance after experiencing severe floods. This means that individuals living in areas that are not frequently or severely affected by flooding are more likely to be uninsured. The authors also acknowledge that frequent changes in residents living in cities can lead to a lack of awareness about past flood events, making it difficult to mitigate and prepare for future disasters. Additionally, urban areas tend to have lower rates of insurance adoption.

On the other hand, policies at the community level such as the Community Rating System (CRS) of the National Flood Insurance Program take a proactive approach. The CRS incentivizes the adoption of insurance by reducing premiums for communities that implement measures for mitigation and floodplain management. The writers propose that the system could better tackle the issue of unequal flood risk by focusing on underprivileged and vulnerable communities.

The study confirms previous discoveries that certain groups are consistently more at risk of floods and could potentially improve their resistance. The evidence suggests that implementing top-down methods, like the CRS, can provide proactive solutions to address structural disparities in risk. While flood insurance is an important tool in managing risk, it is typically a reactive approach and only offers limited assistance without community-level initiatives. (Earth’s Future,, 2023)

“I am a science writer named Aaron Sidder.”

Reference: Sidder, A. (2023), New Insights from Machine Learning for Enhancing Flood Mitigation, Eos, 104, Published on 18 October 2023.

Text © 2023. AGU. CC BY-NC-ND 3.0

Unless specified, images are protected by copyright. Unauthorized use without obtaining permission from the copyright holder is not allowed.