Measurement Uncertainty from Sampling
Measurements of contaminant concentration in the environment inevitably have uncertainty. We never know the true values of the concentration, but only estimate values that contain both systematic and random errors. Once an estimate of uncertainty is known, it becomes possible to make more reliable decisions using the measurements. Examples of such decisions are the classification of levels of contamination in land or food, compared to threshold values.
Research Objectives for Uncertainty from Sampling
1) To devise practical methods to estimate uncertainty, whether this arises from the analytical technique or the field sampling.
Applications to contaminated land have shown large uncertainties on estimates of contaminant concentration (e.g. 50-90 %), which are limited by the methods of sampling. This is much higher than current estimates of around 10%, which are based soley on uncertainty from the chemical analysis.
2) To find objective techniques for deciding how much uncertainty is acceptable in measurements intended for particular purposes. This relies on an appreciation of the consequences of the uncertainty for the interpretation of the measurements, such as risk assessment. Application of this optimization of uncertainty approach include contamination of soils (research funded by EPSCRC, TSB, DTI & CLAIRE) and also of foods (funded by FSA).
3) To find the causes of uncertainty, such as heterogeneity of contaminant distribution in space and time. To consider other implications of this heterogeneity, such as the effect it has on the uptake of contaminants by plants, and the further implications this has for (a) plant resistance to herbivory, and (b) human health risk assessment.
4) To consider the implications of uncertainty and heterogeneity for the assessment of risk to human health Uncertainty in the assessment of hazard, exposure and risk [PDF 372.17KB]. This research, as part of making multiple links toward integrating teams for understanding disease and environment, was funded by NERC's Joint Environment & Human Health Programme (NERC, MRC, ESRC, BBSRC, EPSRC, EA, Defra, MOD, HPA and Wellcome).
Impact of this research
These ideas have been applied to improve the reliability of contaminated land investigations and to the assessment of contaminated food. The wider take up of these ideas has been helped by their recommendation in this European Guidance document,
Fig 1. Eurachem Guide that recommends the estimation of uncertainty from sampling using methods published by this group
This Eurachem Guide (Fig 1.) has been cited in several guidance documents to evaluate the quality in the sampling of water, food and soil:
1. Guidance for the EU Water Framework Directive,
2. Food & Agriculture Organization /World Health Organization - Methods of food sampling
- JOINT FAO/WHO FOOD STANDARDS PROGRAMME CODEX COMMITTEE ON METHODS OF ANALYSIS AND SAMPLING, Thirtieth Session, Balatonalmádi, Hungary, 9 - 13 March 2009, Agenda item 9, GUIDANCE ON UNCERTAINTY OF SAMPLING
3. British Standard BS10175
- BS 10175:2001 Investigation of potentially contaminated sites – Code of Practice. Annex D (informative) The assessment of measurement uncertainty from sampling. Revision for publication in 2011
Methods of Uncertainty Estimation
One of the approaches recommended in the Eurachem Guide is called the Duplicate Method. It is based on this balanced experimental design, where a small proportion (~10%, but at least 8) of sampling targets are sampled in duplicate, by reinterpretation of the sampling protocol. Each of the duplicate samples is then also analysed in duplicate in a balanced design ( Figure 2)
Fig 2. Balanced experimental design for the estimation of uncertainty from sampling
The values for the total measurement uncertainty, and the separate contributions from the sampling and the chemical analysis, are estimated from the resultant measurements using analysis of variance (ANOVA).
Applications of these ideas have been made to chemically contaminated soil and food, and radioactively contaminated soil and buildings.
