A Dam Failure Warning Alarm sign in the event of overtopping. (Photo by melissamn on Shutterstock)
In A Nutshell
- A new study finds six U.S. dams, including sites near Waco, Texas and Junction City, Kansas, now have the highest overtopping risk in decades.
- Researchers tracked 50 years of water level data at 33 large dams and discovered a rising trend in critical overtopping probabilities.
- Two of the six most at-risk dams — Kanopolis and Milford — show statistically significant increases in overtopping probability over time.
- With aging infrastructure and intensifying extreme weather, experts warn urgent upgrades and risk reassessments are needed to protect downstream communities.
TALLAHASSEE, Fla. — Six major dams across the United States now exhibit the highest overtopping probabilities recorded over the past 50 years, with several located dangerously close to large population centers that could face catastrophic flooding if these aging structures were to fail.
An analysis of 33 dams nationwide reveals a troubling trend: the number of dams with critical overtopping probabilities has steadily increased over the past five decades, while those considered safe have declined. The study, published in Nature Communications, identifies six dams with the highest probability of water levels rising uncontrollably above their crests, a scenario that can significantly increase the risk of downstream flooding.
Among the most concerning are Whitney Dam in Texas, which sits roughly 50 kilometers from Waco and its 140,000 residents, and Canyon Dam, also in Texas, located just 30 kilometers from New Braunfels and nearly 100,000 people. All six high-risk dams are classified as both “large” and “high-hazard potential,” meaning their failure could result in significant loss of life and property damage.
Recent dam incidents show these risks. In 2024, Hurricane Helene caused Lake Lure Dam in North Carolina to partially overtop, leading to evacuations and more than 25 swift water rescues. The 2017 Oroville Dam spillway failure in California prompted over 180,000 evacuations, while the 2020 failure of Edenville and Sanford Dams in Michigan forced 11,000 residents to flee and damaged more than 3,000 homes.
Why U.S. Dam Infrastructure Gets Failing Safety Grades
The research comes as America’s dam infrastructure faces mounting pressure. The American Society of Civil Engineers assigned U.S. dams a D+ grade in 2021, estimating that $185 billion in investment is needed to address current dam safety issues. More than 16,000 dams nationwide are classified as “high-hazard,” and many were built decades ago using outdated design standards that may not account for today’s extreme weather patterns.
Researchers from Florida State University, Princeton University, and the University of California, Irvine analyzed water level data spanning 1973 to 2022 for 33 dams operated by the U.S. Army Corps of Engineers and the Bureau of Reclamation. These dams were selected because they had publicly available long-term water level records of 50 years or longer.
The study employed what researchers call “updated stationary frequency analysis,” which examined 30-year periods of data to track changes in overtopping probability over time. Rather than simply looking at water level trends, researchers calculated the actual probability that water levels would exceed dam heights, revealing risks traditional monitoring might miss.
Scientists classified overtopping probabilities into four categories: very low (less than 0.01%), low (0.01% to 0.1%), moderate (0.1% to 1.0%), and high (1.0% or greater). Based on Federal Emergency Management Agency guidelines, the latter three levels were considered “critical” because they exceed federal dam design criteria.
The analysis revealed significant shifts in dam safety over the study period. During the earliest 30-year window (1973–2002), 62% of dams fell into the very low probability category. By the most severe period (1991–2020), this dropped to just 45%. Meanwhile, dams in the low probability category nearly doubled from 10% to 22%, and those in the moderate category increased from 17% to 22%.
During the 1991–2020 and 1992–2021 periods, dams classified as critical actually outnumbered non-critical ones, a reversal that signals growing infrastructure vulnerability.
Where America’s Most Dangerous Dams Are Located
Geographic patterns emerged from the data showing dams in the Southern Great Plains, particularly Kansas and Texas, had higher proportions of critical probability compared to those in Northern Great Plains states like North Dakota, South Dakota, and Nebraska, as well as California.
Of the 33 dams studied, 12 showed statistically significant trends in overtopping probability over the 50-year period. Six exhibited increasing trends, while six showed decreasing trends. Four of the six dams with increasing trends were located in the Southern Great Plains.
Regional differences align with the Southern Great Plains’ exposure to extreme weather events, including hurricanes, severe storms, and intense precipitation from tropical systems originating near the Gulf of Mexico.
Six High-Risk Dams Threaten Nearby Communities
The study identified six dams with the greatest overtopping probability:
- Canyon Dam: New Braunfels, Texas
- Kanopolis Dam: Marquette, Kansas
- Milford Dam: Junction, Kansas
- Somerville Dam: Somerville, Texas
- Whiskeytown Dam: Anderson, California
- Whitney Dam: Waco, Texas
All were built more than 50 years ago, with Whitney Dam being the oldest at over 70 years.
Several pose immediate risks to substantial populations. Somerville Dam sits just 1.7 kilometers from the town of Somerville, Texas, home to about 1,300 people. Milford Dam is located 8 kilometers from Junction City, Kansas, with a population of 22,430. Two of these six dams — Kanopolis and Milford, both in Kansas — showed statistically significant increasing trends in overtopping probability.
Notably, only one dam showed a statistically significant trend in actual water levels over the 50-year period, suggesting that overtopping risk depends on factors beyond simple water level changes. This includes how water levels vary relative to dam heights, highlighting why traditional water-level monitoring alone may be inadequate for assessing overtopping risk.
The research aligns with broader studies showing increasing precipitation trends across the United States, particularly in Kansas and Texas. Previous research has documented significant increases in extreme precipitation events in these states.
However, researchers acknowledged they couldn’t establish direct causal relationships between precipitation changes and dam overtopping risks without more detailed analysis. Potential factors like precipitation patterns, temperature fluctuations, soil moisture levels, and dam sedimentation remain unexplored areas for future research.
Despite study limitations — including the use of 30-year windows over a 50-year period and focus on annual maximum water levels rather than multiple variables — the research gives us the most comprehensive analysis of dam overtopping trends across multiple U.S. dams to date.
With aging dams facing unprecedented hydrological challenges and millions of people living downstream from high-risk structures, infrastructure investments and updated safety protocols can no longer be delayed without potentially catastrophic consequences.
Disclaimer: This article summarizes findings from a peer-reviewed study. While the study identifies statistical increases in overtopping probability, it does not claim imminent failure of any specific dam. The authors caution that further research is needed to explore the root causes of these trends. Interpretations provided here reflect a lay summary of the data and should not be construed as emergency warnings or engineering assessments.
Paper Summary
Methodology
Researchers analyzed daily water level data from 1973 to 2022 for 33 dams operated by the U.S. Army Corps of Engineers and Bureau of Reclamation. These dams were selected based on having publicly available long-term records of at least 50 years and passing statistical tests for randomness. Scientists used an “updated stationary frequency analysis” approach, fitting statistical models to 21 different 30-year rolling windows to track changes in overtopping probability over time. They classified overtopping probabilities into four categories (very low, low, moderate, and high) based on federal dam design criteria and used statistical tests to identify significant trends.
Results
The study found an overall increasing trend in the number of dams with critical overtopping probabilities (low, moderate, and high categories) and a decline in non-critical (very low) probabilities over the 50-year period. The proportion of dams in the very low probability category dropped from 62% to 45% between the earliest and most severe periods. Six dams were identified as having the greatest overtopping probability: Canyon, Kanopolis, Milford, Somerville, Whiskeytown, and Whitney. All are classified as large, high-hazard potential dams, with several located near significant population centers. Dams in the Southern Great Plains showed higher risk levels compared to other regions.
Limitations
The study was limited by its 30-year rolling window approach over a 50-year period, which introduces uncertainty. The analysis focused only on annual maximum water levels rather than multiple variables like dam inflow and spillway capacity. Lack of publicly available data on detailed dam characteristics prevented more comprehensive risk assessments. The research detected changes in overtopping probability but didn’t investigate underlying causes like precipitation patterns, temperature changes, or dam conditions. Population risk assessments were basic and could be improved with detailed flood modeling.
Funding and Disclosures
The authors declared no competing interests. The paper was published under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, allowing non-commercial use and distribution with proper attribution.
Publication Information
“Historical changes in overtopping probability of dams in the United States” was published in Nature Communications, Volume 16, Article 6693, on July 21, 2025. The study was conducted by Eunsaem Cho, Ebrahim Ahmadisharaf, Gabriele Villarini, and Amir AghaKouchak from institutions including Florida State University, Princeton University, and the University of California, Irvine. The paper is available at https://doi.org/10.1038/s41467-025-59536-1.
