Protecting Lives During Inland Floods: Innovative research behind the scenes

By Theresa Barosh | July 2025

Historic rainfall fell across central Texas, leading to catastrophic flooding that claimed many lives. Water vapor satellite imagery July 2025 | Credit: CIRA/NOAA

CIRA researcher Stan Rhodes is part of a team that has spent the last two years working to understand flood fatalities nationwide. “Roughly every two to three years we suffer the cumulative equivalent of this recent Texas flood in deaths.”  

Excluding the Texas Hill Country event, the United States experiences, on average, 88 deaths a year due to inland flooding.  

Rhodes says that understanding the Texas event will take time, effort, and a lot of care. He is quick to point out the distinction between flood events where everyday people are caught unaware by its sudden onset versus those where they are not. Rhodes’ recent work focuses on situations, unlike the Texas event, when people have time to respond to an upcoming flood and make decisions.  

“In short, if we build decent models of how human systems behave in the lead up to – and response to – inland flooding, we can provide tools that help people, like emergency managers, save lives,” said Rhodes.  

CIRA’s computational social scientist, Rhodes, specializes in behavior. He recently worked on a general framework to integrate work by cognitive scientists, risk communication scientists, psychologists, economists, sociologists and others interested in how weather information affects decisions individuals make to reduce risks. Rhodes said he hopes that his work will support NOAA’s National Weather Service in advancing research programs to protect life and property and enhance the economy. 

A New Framework 

Rhodes works backwards in time. He recommends a bottom-up approach to understanding what decisions lead to risks. He recommends starting with the result, such as an injury or fatality, and working backwards from there to determine what events and decisions brought the individual to that outcome. 

In the weather world, the process has been classically considered from the top-down: from the forecast, to the resulting warning, to protective actions taken by people at risk and ending with the result of lives or property saved. The basic assumption is that people both get and follow advice in messages and warnings from forecasters. Choices in the real world, however, are rarely so simple. What if someone’s children are still at school? What if they see someone else drive past a barrier and through a flooded road? What if the weather warning comes through while they are already out and driving around? 

Historic flooding permeates the Ohio and Mississippi valleys following monumental rainfall across the region. March 2025 | Credit: CIRA/NOAA

“What decisions, situational contexts, and other factors lead to these risks?” said Rhodes, “The reality is clear: if we want to improve outcomes, we also need to know where we can improve decision-making.”  

Building an understanding of the chain of events leading up to deaths provides an opportunity to identify intervention points. Interventions may include a specific type of communication, signs on roadways or blockades to reduce inland flood fatalities related to driving on flooded roads. 

Rhodes’ framework is unique even in the context of broader science around risk communication. Some areas, like epidemiology, employ similar strategies by starting with potential decisions and actions.  

“We’re using some tools from other disciplines, including econometrics, but it hasn’t really been an approach that people have taken very much in this area,” said Rhodes, “And so, that’s the reason why we’re innovating here. In some ways, it’s a simple concept but actually executing it well and doing it in a statistically rigorous way is very challenging.” 

Challenges  

One of the biggest hurdles to statistically rigorous research is getting a big enough sample size. Rhodes said that with currently available data, the models are unable to capture how people respond in varied ways to imminent floods. With more information from a larger variety of people who both did and did not make specific decisions during flood events, models could have stronger statistical power.  

“Particularly our piloting work, or initial samples, tend to be people that are very weather interested.” said Rhodes. People who respond to surveys tend to be like one another, a phenomenon known as self-selection bias. To get a better idea about the general public, researchers need large sample sizes with information from a variety of dissimilar people who made unique choices.  

What next? 

Rhodes hopes that his new decision-chain framework guides future research in the United States, helping to both improve data collection and analysis. The new approach also likely applies to other severe weather events. In particular, he and his team plan to apply the framework to wildfires and winter weather in the future.  

“There are some elements of winter weather that are similar to flooding in that you decide whether or not to go out. With winter weather, you don’t come to a flooded road: you hit ice when you didn’t realize you’re going to hit ice, so that’s different,” said Rhodes. “Fire is a little challenging because people often think about evacuation as the only important choice with fire. But there’s actually a lot of dynamics there. There’s people that decide to stay and defend [their home], and then there’s also the issue of smoke. There’s a lot of work there to be done to try to tune into those different hazards.” 

Read more about inland flood fatality research at CIRA. 

What is needed: Continued and enhanced data collection on weather-related decisions and outcomes, such as injuries and deaths, throughout the US to improve interventions to save lives and property. Funding and support for NOAA to enhance the U.S.’s state of readiness for floods.