A comprehensive review of the primary sources of uncertainty in stone armor stability

Coastal rubble mound armor stability prediction uncertainty is relatively high in the field of civil engineering. The present study aims to provide an in-depth review of the principal sources of stone armor stability uncertainty derived from laboratory experiments. The study delineates the contribution of each source and sub-class to the total uncertainty based on the body of knowledge from the literature and data analysis. Uncertainty is first classified into two main components: aleatory (intrinsic), which is irreducible and arises from the inherent randomness of natural processes, and epistemic uncertainty, which relates to limited knowledge of physical processes, observations, and predictive methods, and can be reduced with appropriate precautions. Epistemic uncertainty is further subdivided into three main categories: data uncertainty (waves and damage), predictive model uncertainty, and experimental errors. The focus is on empirical stability equations and the underlying data and experiments. For each category and sub-class, a semi-quantitative estimation of the coefficient of variation is provided to convey a sense of the magnitude of the component contribution to the total epistemic uncertainty in stability predictions. Results indicate that data uncertainty, particularly related to damage assessment, is the dominant contributor, followed by predictive model uncertainty, while error-related uncertainty have a smaller impact. The findings highlight the importance of improving data quality and standardization to reduce epistemic uncertainty, thereby enhancing the reliability of empirical design models, and supporting more consistent probabilistic design of rubble mound structures. • Comprehensive review of uncertainty sources in stone armor stability based on empirical equations and laboratory data. • Semi-quantitative estimates for uncertainty sources and sub-classes to evaluate their impact on armor stability prediction. • Damage measurement identified as the dominant uncertainty contributor, with CoV values up to 100%. • Enhancing data quality and standardization is more effective than refining design models to reduce epistemic uncertainty.