| The transport of dangerous goods (DG) results in risk. A release of the DG may cause harm to property, the environment or the public. Authorities need to understand the risk associated with the transport of DG, to make informed decisions regarding transport modes and routes. This thesis predicts rates of releases and fires for trucks in transit carrying DG loads. It is intended that others will use the DG incident rates in assessing the risks of using specific truck routes for transporting DG. The research produces a probabilistic model that predicts the release and fire incident rates using five databases from Canada, the USA and France.; This research provides a methodology for estimating DG release and fire incident rates that is better than previous methodologies. First, it examines and compares each of the potentially significant factors available in the data that may affect the input variables. Previous research has not adequately provided an overall analysis comparing potentially significant factors affecting release and fire incident rates. The model combines the input variables to produce release and fire incident rates. The research uses statistical analysis to identify significant factors affecting the input variables. This reduces the uncertainty in whether or not there is an effect that should be incorporated in the estimates of the input variables. The research uses further statistical analysis to define the uncertainty associated with the input variables.; Second, the research extends the treatment of uncertainty beyond that of previous research on the risk of transporting DG, which has included sensitivity analysis, low, best and high estimates, and confidence intervals. The analysis of uncertainty uses Monte Carlo simulations to propagate the uncertainty in the input variables through to the resulting release and fire incident rates. The analysis represents the uncertainty through probability distributions for the incident rates. Statistics on the distributions include mean, median, standard deviation, skewness, kurtosis, coefficient of variation, and percentiles. The distributions help to put the incident rates in context and allow for appropriate use of the rates in future quantitative risk assessment.; For accident-induced incidents, the thesis predicts the probabilities of release and fire, given a truck carrying DG and involved in an accident. For non-accident incidents, such as leaking valves, the thesis predicts the rates of non-accident releases and fires per billion vehicle kilometres (Bvkm). The thesis illustrates how we can combine accident and non-accident information to provide total expected releases and fires per Bvkm for trucks carrying DG loads. |
