Date of Award:


Document Type:


Degree Name:

Doctor of Philosophy (PhD)


Civil and Environmental Engineering


David S. Bowles


Catastrophic events such as dam failures or severe floods are considered to be of low probability, although their consequences can be extremely high and might include loss of life. Earlier studies have linked circumstances surrounding historical darn failure events to actual loss of life and produced formulations using statistical analysis of these events. Shortcomings of these methods include the inability to adjust life-loss estimates based on the type of darn failure, global averaging of population at risk, and ignoring the dynamics of the evacuation process.

The main objective of this research is to develop a practical and improved life-loss estimation approach for use in dam safety risk assessment and emergency planning. The methodology is specifically formulated to overcome the limitations of previous. purely empirical, approaches. The approach takes into account the spatial and temporal distribution of flood water depth and velocity, fate of buildings, simulation of warning diffusion, and tracking the movement of people from their original location towards safe shelters.

The model created, called LlFESim, is designed to serve multiple function s. First, it can be used in a Deterministic Mode using best estimate inputs to obtain point estimates, or to test different policies for evacuation as well as different times of the day and for different dam breach flooding scenarios. Second, the Uncertainty Mode represents input and parameter uncertainties to provide estimates of life loss, and other variables relating to warning and evacuation effectiveness, as probability distributions. These distributions of life loss can be combined with estimates of the uncertainties in other risk assessment inputs, to obtain estimates of uncertainties in risk assessment results, including evaluations against tolerable risk guidelines.

Two communities were used to demonstrate the model performance. Deterministic Mode results display the various possible model outputs. Sensitivity analysis for the Deterministic Mode shows that the effect of warning issuance time is the dominant factor in the estimated life loss. However, other factors play an important role such as the time of day, effectiveness of the warning system, and shelter location. Uncertainty Mode results demonstrate the effect of uncertainties in model parameters and inputs on the model results.