Comparison between heavy metal concentrations in sediments analysed by two methods: Analyses on detection limits and data quality

Abstract

Chemical components in geological samples are often measured by more than one method to be able to measure high concentrations in resistant minerals as well as low concentrations in other mineral phases. For quality control, it is necessary to evaluate if the results from these methods are consistent or not. A total of 1884 sediment samples taken from the Leinster area of Ireland were analysed by both instrumental neutron activation analysis (INAA) and atomic absorption spectrometry (AAS) for 5 chemical elements (Co, Cr, Fe, Ni, and Zn). It is well known that INAA detects the total concentrations of chemical elements, but it has much higher detection limits than AAS. In this study, descriptive statistics, scatter plots, statistical tests and GIS mapping techniques were employed to compare the differences between the results from the two analytical methods, and to evaluate the impact of detection limits on data quality. Descriptive statistics and scatter plots were effective in showing the discrepancies between values of the two methods. Scatter plots showed outliers and the poor accuracy of INAA for concentrations close to or below the detection limits. It is suggested that errors may have happened during both INAA and AAS processes for some samples. Statistical results of both nonparametric sign and Wilcoxon signed rank tests showed that the differences between concentrations measured with the two methods were significant at the level of p < 0.001, and the average differences were 20–30% for the 5 elements. Meanwhile, more than 25% of all samples showed differences greater than 40–60% for these elements. After removing values lower than 3 times detection limits of both methods, only a slight improvement was achieved for the consistency between the two methods. Besides detection limits, the incomplete digestion prior to AAS may have played an important role. The higher AAS values of Ni than its INAA values may have been caused by multiple detection limits in different batches of laboratory analyses. The high detection limits of INAA also caused spatial patterns on the GIS maps showing differences between the two methods. The current geochemical database may be sufficient for some aspects of mineral exploration, but not for other geochemical studies, such as baseline database construction and environmental investigations. Better results could have been achieved by initial error checking and correction, improvement on the detection limits of INAA, and improvement on the acid digestion for AAS. It is possible that other geochemical databases contain similar data quality issues demonstrated in this one. This study demonstrates how straightforward statistical methods can be applied to check/verify the quality of geochemical databases.