An interlaboratory comparison of vitrinite reflectance measurement

A vitrinite reflectance (VR) round-robin analysis was undertaken by The Society for Organic Petrology (TSOP) Research Committee. Fourteen laboratories from five countries participated in the study which involved samples of one shale and three coals, ranging in rank (maturity) from lignite to high volatile A bituminous. The main objective was to evaluate interlaboratory comparability and accuracy of VR measurements in a context of practical data interpretation (e.g. maturity determination of petroleum source rocks). The between-laboratory standard deviations of the VR data are 0.033–0.048 for mean random and 0.037–0.049 for mean maximum measurements. The results suggest that, at a moderate level of confidence (68%), assuming normally distributed populations, a VR differential of 0.1% is significant in indicating different thermal maturities (within the rank range studied). A differential of 0.2% VR is required if a high level (95%) of confidence is desired. At the 5% significance level, between-laboratory consistency tests (h statistic) show that the random VR data of one out of fourteen laboratories for each sample are found to be inconsistent with the main population, whereas the other thirteen laboratories are consistent with each other. The maximum VR data are shown to be consistent (at the same significance level of 5%), except for one data point. The within-laboratory standard deviations (θ) are relatively high, from 0.023 to 0.110 with an average ranging from 0.037 to 0.070. The relatively high θ is likely related to natural variations in the vitrinite macreal which can result in the inconsistent selection of vitrinite particles for measurements by the operators. Maximum VR measurements are found to give rise consistently to slightly higher interlaboratory and average within-laboratory standard deviations compared to those of corresponding random VR results. This may be related to experimental errors introduced during maximum VR analyses (e.g. imperfect focusing of the sample and centering of the stage). However, this observation by no means suggests that random analysis is a more precise measurement.