QUALITY ASSURANCE AND QUALITY CONTROL (QA/QC) FOR IN-SITU GEOCHEMICAL METHODS, ESTIMATION OF MEASUREMENT UNCERTAINTY AND CONSTRUCTION OF PROBABILITY RISK ASSESSMENT MAPS

Abstract

This document gives guidelines for taking duplicate measurements by in-situ analytical instruments, or collecting duplicate samples for analysis in the mobile field laboratory, to assess random errors originating from sampling and analytical procedures, and to estimate the uncertainty of measurements. Since, NORISC is dealing with the assessment of contamination of small-size areas within cities, and relies mainly on in-situ analytical methods, a cost-effective technique, using robust analysis of variance for the estimation of necessary quality control parameters, and measurement uncertainty, is explained with examples. Measurement uncertainty in the interpretation of contaminated land may have profound effects on the realistic assessment of the extent of contamination, because it reduces ‘misclassification’. This has legal, financial, and possibly health implications, from both the unnecessary remediation of “uncontaminated” land, and parts of the property regarded as uncontaminated that are in fact contaminated. The recommended technique separates the different components of variance (sampling, analytical, geochemical), and indicates which ones are unacceptably high, and may need improvement. Further, it estimates the combined measurement uncertainty, and its use in the probabilistic classification of contaminated land into: (a) uncontaminated, (b) possibly contaminated, (c) probably contaminated, and (d) contaminated. Thus, improving the reliability of classification of contaminated land, compared to the traditional deterministic geochemical approach. Regarding the use of geostatistics for the estimation of uncertainty in the NORISC approach, there is the question of collection of an adequate number of samples for the valid calculation of the necessary geostatistical parameters. However, the use of two times the kriging standard error of estimation as an estimate of uncertainty is explained, and its utilisation in the probabilistic classification of contaminated and uncontaminated land. CONTENTS Page Introduction Estimation of uncertainty due to sampling and analysis Practical detection limit and analytical precision Handling of values below detection limit Limitations of the geochemical data set: quality & reliability Effects of uncertainty and probabilistic risk assessment maps 2 3 6 7 7 8 Example 1: Classification of contaminated land with an uncertainty of 10% Example 2: Probabilistic risk assessment mapping 10 10 Geostatistics: two times the kriging standard error of estimation as an estimate of uncertainty Discussion and conclusions References Instructions for downloading program ROBCOOP4.EXE 11 13 14 16 NORISC: EVK4-CT-2000-00026 WP5 SPECIFICATIONS TO THE EVALUATION PLAN QUALITY ASSURANCE AND QUALITY CONTROL (QA/QC) FOR IN-SITU GEOCHEMICAL METHODS, ESTIMATION OF MEASUREMENT UNCERTAINTY AND CONSTRUCTION OF PROBABILITY RISK ASSESSMENT MAPS Alecos Demetriades and Helen Karamanos Institute of Geology and Mineral Exploration, Greece INTRODUCTION Estimation of both random and systematic errors in geochemical analysis and measurements for environmental studies has become an established part of good professional practice, and are, therefore, included in the routine field geochemical analyses and measurements of the NORISC approach. The separate estimation of sampling and analytical errors using analysis of variance (ANOVA) was pioneered for geochemical surveys in mineral exploration by Miesch (1964, 1967, 1973, 1976), Garrett (1969, 1973, 1983) and Garrett and Goss (1979), and has been described for pollution investigations by Holcombe (1988). Combination of robust ANOVA and sampling and analytical quality control has been applied to the evaluation of precision requirements for applied geochemistry (Ramsey, 1992), and the estimation of measurement uncertainty (Ramsey, 1997, 1998; Ramsey and Argyraki, 1997). The robust analysis of variance technique proposed by Ramsey (1998) requires a minimum number of eight (8) locations at which to take duplicate samples or measurements, in order to give a reasonably reliable estimate of component variances (sampling, analytical & geochemical). According to Ramsey (1992), and all researchers working on contaminated sites, the application of these techniques to environmental surveys is particularly appropriate due to the high degree of heterogeneity, often associated with anthropogenic contamination of the environment. The robust analysis of variance technique proposed by Ramsey (1998) is adapted for the purposes of the NORISC field geochemical methods, and requires two measurements to be taken for each determinand (Fig. 1): (a) at the routine location, and (b) at the duplicate location (to be at a distance of 1 to 2 m from the routine location). OR in case physical samples are collected, two analyses for each determinand should be performed on either two sub-samples of the routine and duplicate samples or, at worse, each of the routine and duplicate samples to be analysed two times in a random manner (Fig. 2). Measurement location Routine measurement location Duplicate measurement location Measurement 1Α Measurement 1B Measurement 2Α Measurement 2B Fig. 1. Balance hierarchical geochemical measurement scheme for the estimation of geochemical, sampling and analytical variance and random components of measurement uncertainty by the in-situ geochemical methods used in the NORISC approach. This is the case when measurements are performed in-situ at the actual sample location, and no physical samples are collected. Two measurements are made at the routine and two at the duplicate location.