Applications of the concept of uncertainty from sampling show the benefits of this approach. The investigations of Contaminated Land sites was one of the first applications that has proved particularly effective. One such study is described in the following study.
Site Investigation 1
An area of land in the West of England (approximately 200m x400m) was thought to be contaminated primarily with arsenic (As) as a consequence of tin mining over the last century. The proposed end-use of the site was a development of residential housing.
Soil samples were taken from trial pits excavated across the site, and chemically analysed (primarily for As) within a commercial laboratory. The soil samples were removed by the commercial consultant from the side of each trial pit. This procedure is now forbidden, because of the safety risk from wall collapse, but did represent some current practice at that time (2004).
Measurements of arsenic on duplicated samples for the estimation of uncertainty
The measurement uncertainty was estimated as 64% relative to each measured concentration (at 95% confidence). For example, the true contaminant concentration is found within ± 64% of each individual measurement. The measurement uncertainty is generated by both the field sampling and the chemical analysis (e.g. measurement uncertainty = sampling uncertainty + analytical uncertainty). The sampling uncertainty accounted for most of the measurement uncertainty (94%) compared to the chemical analysis (6%).
For this investigation, a misclassification of false-positive misclassification was considered (i.e. when the site is incorrectly determined to be contaminated). The financial cost arising from this form of misclassification was calculated using the value of unnecessary remediation (i.e. by the mass of soil that would be removed and covered with clean material). The actual costs of the site investigation, such as those for the field sampling and chemical analysis, were also used within the OCLI calculation.
The OCLI curve allows a comparison between a range of uncertainty values and their associated expectation of losses (£). The range of uncertainty values includes the actual value estimated during the site investigation and the 'optimal' value (at the minimum) that produces the lowest expectation of loss.
OCLI curve for site 1, showing actual uncertainty (at ~130 mg/kg) is much higherthan the optimal value (at ~50 mg/kg)
The actual measurement uncertainty for total As concentration at this site gave an expectation of loss value of £1200 per sampling location (so misclassified). The optimal level of uncertainty calculated by the OCLI method gives an expectation of loss of less than £300.
The measurements taken as part of this site investigation are therefore not necessarily fit for purpose. There is a probability that sub-areas of the site will be uneccessarily remediated. The OCLI method indicates that future investigations at this, or similar sites, should reduce the sampling uncertainty. This can be acheived by taking within-location composite samples (i.e. a number of sample increments from the trial pit that are used to make each individual sample). In this case, taking two increments to make up each individual field sample is predicted to reduce the uncertainty to approximately 90 mg kg -1 As which gives a much lower expectation of loss value of approximately £600.
