On Uncertainty and Data Worth in Decision Analysis for Contaminated Land

Licentiate thesis, 2003

Contaminated soil and groundwater is a problem that has received increased attention in the last decade. Decision-making about investigation strategies, protective actions, and remedial actions is based on sparse and uncertain information, primarily data of contaminant concentrations and geological information. Because of limited economical resources, cost-efficient decisions must be made. Risk-based decision analysis is a tool for evaluating the cost-efficiency of different decision alternatives. Identification of cost-efficient site investigations can be performed by applying data worth analysis. In such an analysis, the value of additional sample data is compared to sampling cost and if the data worth is larger than the cost it will be worthwhile to carry out the sampling. Because environmental samples are uncertain, this uncertainty should be considered in the analysis. An approach for estimation of uncertainty in soil sampling is presented. It is based on the sampling theory for particulate materials developed for the mining industry. The sample uncertainty is broken down into eight basic types of uncertainty and variability. An application of the methodology is presented for the problem of soil sampling with a drill auger. The result indicate that the uncertainty in sample data can easily be in the range of 30-40 %. The sampling uncertainty is believed to be much more important that the analytical uncertainty.

A methodology for including sample uncertainty in data worth analysis is presented. It is based on a Bayesian approach to data worth. The sampling objective is to estimate the mean concentration at a site. A MathCad computer application for the calculations is supplied. An application of the data worth estimation procedure is presented for a sampling problem at a former Ferro-alloy work in Gullspång, Sweden. A conclusion is that prior estimates of contaminant concentrations may have a significant impact on the result, as well as estimates of failure cost. It is recommended to use different estimates of failure cost to study its influence. Results also indicate that when sample uncertainty is increased, the expected net value of the sampling program will decrease moderately and relatively constant.

In situations where contamination has not yet occurred, cost-efficient protective actions need to be identified to combat environmental risks. A methodology for selecting costefficient protective actions for water supplies along railways has been developed. The risk object is railway transport of dangerous goods. Also for this problem, estimation of failure cost is believed to be important for the result.

The need for additional development of the methodology is identified. Estimation of uncertainty in soil sampling can be improved and the described theory extended. The methodology for data worth analysis for contaminated land should be extended to take additional sampling objectives into account.

contaminated land, data worth, decision analysis, risk, sampling, uncertainty