Using Logic Trees to Characterize and Model Uncertainty in Risk Analyses
« All Publications

Dam owners are increasingly using risk-informed decision-making processes to manage the safety of their portfolio of dams.

These dam owners rely on risk assessments, which comprises analysis, evaluation, and decision-making regarding management of dam safety risk. The objective of the risk analysis is to characterize the risk of the dam system to individuals or populations, property, and/or the environment (e.g., ICOLD 2005). Of the various risk analysis methods, event tree analysis has a prominent place in many dam safety guidelines (e.g., USBR, 2019). In the current dam safety practice, event tree analysis is often carried out without the explicit distinction between different types of uncertainties.

One type of uncertainty is associated with the random variation of events. An example of random variation is the uncertainty in the outcome of a coin toss. The outcome of a coin toss is either heads or tails, and the probability of either outcome is 0.5 for a “perfect coin.” Random variation is also commonly referred to as aleatory uncertainty.

A different type of uncertainty is knowledge uncertainty, which includes uncertainty about the condition or behavior of a dam system or component (i.e., the “state of nature”). Knowledge uncertainty, also referred to as "epistemic uncertainty," can be attributed to a lack of knowledge about the system (e.g., details of embankment construction or soil properties) rather than randomness of events. Sources of knowledge uncertainty include limited empirical data, alternative engineering models that differ in their model predictions, and uncertainties with respect to estimates of model parameters.

For complex systems, distinguishing between random variability and knowledge uncertainty is often necessary to ensure sufficient focus on the latter and to correctly characterize the uncertainty of an undesirable outcome. A logic tree can be used to model knowledge uncertainties, and an event tree can be used to model events and their random variation (e.g., Hartford and Baecher 2004). This paper presents an example of a quantitative risk analysis for a backward erosion piping failure mechanism using a logic tree for knowledge uncertainties and an event tree for random variations.

Publication Summary

  • Geosyntec Authors: Glenn Rix, Lucas de Melo
  • All Authors: Glenn Rix, Lucas de Melo, Paul Slangen
  • Title: United States Society on Dams (USSD) 2022 Annual Conference and Exhibition
  • Event or Publication: Event
  • Practice Areas: Dams, Dam Safety
  • Citation: United States Society on Dams (USSD) 2022 Annual Conference and Exhibition at the Town and Country Resort in San Diego, California on April 11 through 14, 2022
  • Date: Wednesday, April 13 at 4:25 p.m. PDT
  • Location: San Diego, California
  • Publication Type: Platform Presentation