Michael H. Ramsey FRSC CChem
Mike Ramsey has been Professor at University of Sussex for over 15 years, now Emeritus. After degrees in Chemistry & Geology (BSc Hull), Mineral Chemistry (MSc Birmingham) and Analytical Geochemistry (PhD Imperial College), he worked for 3 years in the Mining Industry in Zambia, and then 20 years in research and lecturing posts at Imperial College London. He has published over 160 scientific papers, many on aspects of uncertainty of measurement arising from field sampling, and the effects of this uncertainty on decision making.
Research projects have included investigations of uncertainty from sampling of radioactively contaminated land (EPSRC-NDA funded), on site measurements of chemical pollution (TSB) and contaminated food and feed (FSA), and Multiple Links Towards Integrating Teams for Understanding of Disease and Environment (NERC/ESRC). He’s a former Chair of the Society for Environmental Geochemistry & Health (http://www.segh.net/home/). Mike is currently Chair of both the Royal Society of Chemistry/ Analytical Methods Committee Sub-committee on Sampling Uncertainty and Quality, and the Eurachem/Eurolab/Citac/Nordtest/AMC Working Group on Uncertainty from Sampling.
LECTURE: Measurement Uncertainty from Sampling: Implication for Testing, Diagnostic and Inspection
The measurement process begins at the time of taking a primary sample, not just when an instrument gives a measurement value. The uncertainty of the measurement value (MU) therefore includes that arising from the sampling process (UfS). UfS is mainly caused by the small-scale heterogeneity of the analyte in the sampling target (e.g. a batch of material). ISO/IEC 17025:2017 now makes it clear that UfS should be included in an estimate of MU, unless specifically excluded. Specifically, Section 7.6.1 states ‘Laboratories shall identify the contributions to measurement uncertainty. When evaluating measurement uncertainty all contributions that are of significance including those arising from sampling, shall be taken into account using appropriate methods of analysis’.
Eurachem, in collaboration with EUROLAB, has recently published a revised Guide on the estimation of UfS*. The two main types of estimation methods described are either based upon modelling, or upon empirical measurements. In the later, duplicated samples are taken on a small proportion of the sampling targets (e.g. 10% of the batches of product), and duplicate measurements are made on both of these samples. Analysis of Variance (ANOVA) is then used to quantify the measurement uncertainty, and its component parts. The six worked examples include measurements made of a wide range of different analytes, in various media including food, feed, water and soil.
The revisions to the Guide include the estimation of uncertainty measurements made in situ, where an instrument (e.g. a Portable XRF) is placed on the surface of the test material in the sampling target, without removing a physical sample. The heterogeneity of the analyte concentration in the sampling target still causes UfS, partially because the material is not homogenized as it would be in the laboratory. This uncertainty is evident when the testing instrument is repositioned in nominally the same location, but gives quite a different measurement value. For example, the PXRF instrument can report an uncertainty of 5% but when UfS is included, the value can rise to 50%. Exluding UfS can cause misclassification of the sampling target and consequent financial losses. When MU is estimated rigorously, including UfS, it becomes possible to make much more reliable decisions on the compliance of the product. This approach will therefore have important implications for testing, diagnostic and inspection in general.