Monitoring Networks in Fractured Rocks: A Decision Analysis Approach

Abstract

AbstractHydrogeological decision analysis is applied to the design of a performance monitoring network at a waste management facility overlying fractured bedrock. The objective of the monitoring system is to detect contaminants before they reach a regulatory compliance boundary in order to enable early and less costly on‐site remediation, and avoid the potentially more costly consequences of failure. Features in the design of the monitoring network include the number of monitoring wells to be installed and their locations, where in each borehole to position discrete monitoring zones, and how often to take water samples. The decision model identifies the preferred monitoring strategy as the design alternative, among all those considered, that minimizes the sum of the monitoring costs and the expected costs of failure and on‐site remediation. At a given distance from the facility, the highest probabilities of plume detection are obtained when the fractures intersecting the borehole wall that carry the largest flows are monitored. Monitoring intervals centered on fractures with highest aperture, or regions of highest fracture density, yield intermediate values, while intervals located at predetermined depths yield the lowest probabilities of detection. The monitoring scheme with the highest probability of detection is not necessarily the preferred monitoring strategy. For monitoring options that have a higher probability of detection than the preferred monitoring strategy, the higher expected cost of on‐site remediation, when combined with the increased cost of monitoring required to provide the higher probability of detection, can outweigh the reduction in the expected cost of failure brought about by a higher probability of detection. When the probability of contaminants migrating to the compliance boundary is small during the compliance period, changes in the probability of detection brought about by a more intense monitoring effort do not affect the expected cost of failure much; in these instances, the decision model may point to a reduced effort in performance monitoring.