Sampling, Detection and Uncertainty in Environmental Analysis—Challenges and Solutions ^†

Analytical measurements are increasingly vital to inform our understanding of our changing global environment and to support the regulation of human activity that affects the environment. Environmental analysis, however, experiences particular challenges for measurement reliability. Most practical environmental measurements are carried out on relatively small samples and their results are taken as indicators of the much larger area sampled. Environmental monitoring can span decades, and the comparability and consistency of measurement results over time and across geographical regions is important for detecting real trends. Environmentally important contaminants—often the subject of regulations—are frequently present at very low levels, often stretching the detection capability of even today’s analytical methods and instrumentation. At low levels, with significant sampling variability, and perhaps especially when environmental measurement is contentious, it is important to understand and express uncertainties clearly and accurately so that reliable policy and regulatory decisions can be made. This, with any accompanying conformity assessment decisions, can be particularly challenging in the frequent cases where sampling and even measurement distributions are far from the familiar normal distribution. Finally, the regulatory framework controlling laboratory operations has evolved over time; for example, the validation of test methods has become increasingly important as new regulatory flexibility allows wider choice of measurement methods, including ‘in-house’ methods, subject to achieving specific performance criteria. Here, these issues will be discussed in the light of experience of some of the UK’s frameworks for environmental regulation and analysis, and with attention to some important Eurachem guidance for analytical practice.

Sampling is often the first step in many environmental measurements. It is also among the most variable, simply because environmental systems themselves are variable on scales from kilometres to centimetres. Understanding the variation due to sampling is key to developing sound sampling strategies and for planning an appropriate allocation between sampling effort and analytical work. A recently updated Eurachem guide [1] provides guidance on the determination of sampling uncertainty, including relatively simple and economical methods for assessing sampling variation using a simple method based on limited duplicate sampling and analysis [2]. The latest edition includes even more economical approaches, which use a ‘staggered nested’ experimental design to reduce analytical effort [3], and also includes recent methods for summarizing uncertainty when the measurement or sampling distribution is asymmetric [4].

Once sampled, attention turns to the application of an appropriate, validated, analytical method. Validation procedures for environmental analysis in the UK are specified, for example, in the MCERTS performance standard for soil analysis [5]. This sets out a specific set of validation procedures, together with criteria for acceptable performance for a wide range of analytes. A recent example of such a validation study, for an in-house modification of a standard method, [6] illustrated some of the problems of achieving reliable results across different soil matrices, even with very precise methods (Figure 1); clearly, some matrices can provide individual challenges (LGC6145 for nickel in Figure 1). These pose practical difficulties for achieving performance and for reporting results and uncertainty.

Detectability poses further problems, particularly for estimating averages and for summary reporting. For example, a preponderance of low levels and the wide use of ‘less than’ statements in summary emissions data for paper mills in the UK led to practical difficulties in assessing compliance with new environmental controls [7].

Finally, environmental analysis is not complete without a reported value, with associated measurement uncertainty and, where necessary, a statement of conformity. Measurement uncertainty evaluation is a requirement for accredited laboratories, even if it need not always be reported [8]. Originally a challenge for analytical laboratories unfamiliar with the concept, several guides exist to help with uncertainty evaluation, including a general Eurachem guide [9] and a NordTest guide specific to environmental laboratories [10]. There is also comprehensive guidance on the use of uncertainty in conformity statements, for those laboratories required to assess conformity with a requirement [11,12]. Recent studies show, however, that asymmetry of the kind found in environmental sampling distributions can adversely affect producer and consumer risks in conformity assessment [13], and some recent studies of the effect of asymmetry will be described.