Probability of dam slope failure and decision making

Contemporary approaches to estimating slope failure probabilities of dams combine two main frameworks. One employs numerical models that represent input variability with statistical distributions. The other relies on empirical relationships and expert judgement to account for unquantifiable factors and limited knowledge. Whatever the analytical sophistication, any probability assigned to the one-off event of slope failure eventually incorporates a subjective degree of belief.

Despite its subjectivity, any framework used for safety-critical decisions should be internally consistent and logically coherent, so resulting decisions remain ethically and legally defensible. With this principle in mind, this paper examines the method introduced by Silva et al. (2008), a framework designed to inform dam-safety decisions and widely adopted in Australia and overseas. Using formal logic as the primary lens, the paper evaluates the coherence of the Annual Probability of Failure – Factor of Safety – Project Category approach without assessing data availability or judging practical convenience.

Our analysis reveals three key points. First, the anchor points that define the curves are not supported by a probabilistic model or empirical observations. Second, the fixed ordering of Project-Category curves does not fully reflect the key premise that more rigorous engineering reduces uncertainty. Third, the progressive flattening of the curves could discourage practitioners from adopting proven risk-control measures.

To strengthen its logical footing, this paper restates the framework as a formal theorem that makes every assumption explicit and outlines adjustments that resolve the inconsistencies. Well-documented cases where failures occurred despite apparently adequate Factors of Safety motivate an alternative probability estimate based on the likelihood of a critical engineering error. The paper concludes with a streamlined procedure for addressing foreseeable slope-instability risks within the Australian context.