Improving an estimation model for dam failure-induced loss of life and customizing it for North America

The potential loss of life (LOL) resulting from dam failures represents a critical concern in dam safety and disaster management. Accurate estimation of LOL is paramount for informed decision-making, emergency preparedness, and the minimization of human casualties in such events. This paper proposes an improved model for LOL estimation associated with dam failures and shows how to customize it to specific regions, exemplifying with North America. The approach categorizes dam failure into subcases based on flood severity and the distance from the dam. Two empirical equations that serve as the calculation method for LOL formulated through multivariate regression analysis are derived using thirty-two dam failure subcases in North America. The datasets were split into train and test sets, yielding R² values of 0.9949 for low severity cases and 0.9955 for medium-high severity cases on the test sets. Graham's model was selected as a comparison benchmark due to its straightforward application, established use in LOL estimation, and minimal data requirements. The successful implementation of this model suggests its potential applicability for diverse regions, contributing to improved disaster preparedness and response strategies, as well as enhancing dam safety and community well-being downstream of dams.