Degree Name

Master of Arts (MA)

Semester of Degree Completion


Thesis Director

Suhrit K. Dey


This thesis is aimed at developing and applying advanced modeling tools in the prediction of risk to the general public from transportation of chemical waste on public highways. The modeling tools developed can then be used to compare alternative waste management scenarios. The application considered is related to the transport of hazardous waste generated by the United States Department of Energy (DOE) to current treatment, storage, and disposal facilities. DOE is currently considering four different scenarios.

The application considered can be more specifically defined as an analysis of the risk to the general public from transporting the 63 shipments of DOE generated hazardous waste designated as "poison by inhalation hazards" (PIH), potentially resulting in fatality, by the United States Department of Transportation (DOT) [Title 49, Code of Federal Regulations (Part 173.132)] in the 1992 fiscal year. The analysis is based on current transportation practices employed by DOE. Once the modeling tools have been developed to perform this analysis, similar analysis can be routinely carried out for other health end-points (carcinogenic effects and other adverse health effects other than cancer and lethality) and other waste management scenarios.

Two types of tools have been developed, deterministic and probabilistic. Using the probabilistic modeling tool, a cumulative probability distribution of number of people impacted can be developed (for this application impacted means number of individuals experiencing potentially life-threatening health effects due to inhalation of a DOE generated PIH released as a result of a truck transportation accident). The probabilistic modeling tool developed is based upon a Monte Carlo analysis accounting for the uncertainties in the variables involved in the modeling process.

The deterministic tool developed provides a simplified version of the probabilistic model such that the prediction will be one risk value which should approximate the mean of the cumulative probability distribution developed by the probabilistic model. Both the deterministic and probabilistic modeling tools require the modeling of the consequence of a release of hazardous waste. The consequence is the result of source term accident modeling (i.e., resulting from a truck accident spill) along with dispersion modeling.

The source term modeling employs the use of (1) distributions of meteorological data supplied by the National Weather Service at over 60 locations uniformly distributed around the continental United States and (2) a detailed study on the US DOT HMIRS (Hazardous Materials Information Reporting System) database which encompasses information on thousands of hazardous material transportation accidents since the 1970's. The study of the HMIRS database led to probability distributions on the release amounts (by transport container), breach fractions and accident time (by hour and month). A health criteria, presented at WM-94 by Hartman et. al. (1994), is used in the dispersion modeling to define human health impacts from the concentration history at each downwind location.

Reasonable single values for all modeling parameters were used in the deterministic model, whereas probability distributions for release estimates and accident meteorological conditions were used for release amounts and meteorology in the probabilistic model. Realistic scenarios for the transportation accident itself were developed accounting for mixtures of chemicals released as is likely to occur.

It was found that the cumulative probability distribution of the number of individuals with potentially life-threatening health effects, is highly skewed. The probability that no individuals will have potentially life-threatening health effects from these 63 shipments is greater than 99%. Therefore the median (the 50-th percentile) of the distribution is 0, and all of the non-zero potentially life-threatening health effects are contained in the upper tail of the cumulative probability distribution (less than 1% chance of occurrence). Table 1 below presents some summary statistics compiled from the distribution.

Only 3 (of the 63) shipments had the potential to affect more than 500 people in a single accident. Furthermore, only 14 shipments had the potential to affect more than 100 people. Eliminating, or at least altering the waste management of these shipments, could dramatically reduce risks by reducing the probabilities for a catastrophic accident in which more than 100 people are affected.

An additional observation is that the mean of the cumulative probability distribution, 3.48E-4, is located at the 99.947-th percentile and the result of the deterministic calculations of risk, 1.74E-4 (½ of the mean), is located at the 99.941-th percentile.

The Monte Carlo analysis helped to provide a great deal of perspective on the deterministic risk value. The fact that there is such a large probability of zero risk and an extremely small probability of a high risk scenario can be very useful in the decision making process.