Monte Carlo bottom-up evaluation of the uncertainty of complex sample preparation: Elemental determination in...

Many chemical analyses involve a complex sample preparation, and some, based on an instrumental method of analysis such as spectrometric or chromatographic methods, are affected by matrix effects. The objective interpretation of the results of these analyses performed in the framework of a research or a conformity assessment requires quantifying the measurement uncertainty. This work presents a novel methodology for the bottom-up modelling of the performance of complex analytical operations, such as sample digestion or extraction, by the Monte Carlo simulation of their performance independently of the performance of the other analytical steps. The simulation of between-days precision of complex sample preparation and mean measurement error observed from the analysis of various reference materials and their combination with models of instrumental quantification performance allow the detailed modelling of the measurement uncertainty. The developed methodology adapts to the complex distribution of observed measurement performance data avoiding the under evaluation of the measurement uncertainty by assuming the normal distribution of systematic and random effects. The developed methodology was successfully applied to the determination of total or acid-extractable As (following OSPAR or EPA 3051A methods, respectively) in sediments where measurement trueness was assessed from the analysis of one Certified Reference Material and two spiked samples. The evaluated uncertainty is fit for environmental monitoring considering performance criteria defined for Quasimeme proficiency tests. The developed measurement models were successfully cross-validated by randomly extracting data from the validation set subsequently used to check the compatibility between estimated and reference values for 95% or 99% confidence level. The observed success rate of these assessments is compatible with the confidence level of the tests.