This dissertation investigates the site assessment procedures used by the Environmental Protection Agency (EPA) for assessing hazardous waste sites. The most comprehensive environmental studies done by the EPA at the Love Canal hazardous waste site in the state of New York are discussed in detail. These studies provide information on the development of the site assessment methodology over the years, and point us towards statistical problems encountered while analyzing environmental data. The EPA methodology for determining habitability of the Love Canal is shown to be very dependent on the choice of control areas.

Drawbacks of the current site assessment procedure used by the EPA are also pointed out in this dissertation. A single threshold score of 28.5 is shown to be inappropriate for classifying sites on the National Priority List (NPL). The method developed by Lucas (1981) to estimate true exceedences over thresholds, by taking into account the analytical error, is used to estimate percentage of sites that could get misclassified from the NPL. Four other applications of this method are also developed in this dissertation. This method of estimating true exceedences over thresholds and false classifications should prove valuable to the EPA while performing site assessment studies. Remedial action at a waste site is expensive and thus a decision rule based method along the lines proposed by Wolfson et al. (1994), for determining remediation responsibility at a hazardous waste site is also provided.

In addition to evaluating the site assessment procedures used by the EPA, and providing the EPA with valuable methods for dealing with difficult environmental issues such as remediation responsibility, this dissertation also provides the EPA with a method for evaluating the performance of laboratories participating in the analysis of environmental data. This method is an improvement on the method proposed by the American Society for Testing Materials (1963). This dissertation also explains reasons for censoring in environmental data and discusses a method developed by Lambert et al. (1991) to deal with censoring in environmental data, along with its limitations.

Finally, some recommendations and concluding thoughts are provided for habitability assessment of future sites.