In the world of dam safety, lives can literally depend on the accuracy of a model. For decades, dam breach analyses have leaned on deterministic methods –choosing one “conservative yet realistic” set of inputs to predict breach outflow rates and the associated failure consequences. But as infrastructure ages, climate extremes grow more volatile, and our understanding of risk evolves, a more nuanced approach is needed.
It’s time we talk about probabilistic dam breach modelling – not just as a theoretical ideal, but as a practical, necessary evolution of dam safety practice.
The limits of determinism
Traditional dam breach analyses rely on fixed input parameters to describe the breach size, shape, and formation time. While these parameters are chosen using guidelines from agencies like FERC and USACE, they are ultimately estimates that carry a lot of uncertainty. To account for this uncertainty, regulatory conservatism has often skewed these estimates toward worst-case scenarios, but even then, they can be misleading.
The deterministic approach gives us a single output – a snapshot that fails to capture the uncertainty inherent in any real-world dam failure. This can lead to over- or underestimating flood impacts, which complicates everything from emergency preparedness to infrastructure investment.
And in an era when the stakes include climate-amplified flood events and aging dams built under outdated assumptions, relying on a single version of an uncertain reality is simply not enough.
Embracing probabilistic modeling
Probabilistic dam breach modeling reframes the analysis as a spectrum of possibilities rather than a singular prediction. Using statistical distributions of breach parameters and Monte Carlo simulations, engineers can generate thousands of dam breach outcomes, each with a probability of occurrence. The result? A robust, more informative picture of flood risk that assigns probabilities to a range of outcomes – from moderate breaches to catastrophic failures (Figure 1).
At East Bay Municipal Utility District (EBMUD), this method was recently put into action for Dike 2 at Camanche Reservoir in Northern California. Working alongside Kleinschmidt Associates, the team conducted a robust probabilistic dam breach analysis that revealed critical insights not captured by the deterministic study.
A case study with real impact
Camanche Reservoir is formed by several earthen embankment structures including the main dam on the Mokelumne River and Dike 2 situated along the southern perimeter of the reservoir (Figure 2). The traditional breach model for Dike 2, based on its modest height compared to the main dam, predicted relatively moderate consequences. But the probabilistic model told a different story.
Unlike the main dam, Dike 2 impounds a significant volume of water despite its shorter profile. The probabilistic approach captured a range of breach sizes that better reflected the site-specific risks. Notably, the analysis showed that the deterministic parameters initially selected likely underestimated the consequences of a breach. Informed by this data, the project team was able to update the deterministic inputs to reflect a more accurate and conservative breach scenario (Figure 3).
Why this matters
Probabilistic modeling isn’t about replacing deterministic methods – it’s about complementing them. When used together, these approaches give dam owners, emergency managers, and regulators a more comprehensive understanding of flood risk.
- Emergency planners can assign exceedance probabilities to inundation maps, improving clarity and confidence during response planning.
- Dam owners gain deeper insight into site-specific risks, leading to smarter investments in mitigation and infrastructure upgrades.
- Regulators are equipped with better data to guide oversight, balancing safety with practical implementation.
A shift in the industry
This case study underscores a broader shift in dam safety: moving from rigid, prescriptive modeling toward flexible, risk-informed decision-making. Probabilistic dam breach analysis is a tool fit for the future – a future where uncertainty is sought to be understood, not overlooked.
The field is evolving, and so must our methods. Incorporating probabilistic frameworks into standard dam safety practice isn’t just innovative – it’s responsible.
