QUANTIFYING MACERALS - SOME STATISTICAL AND PRACTICAL CONSIDERATIONS

Precision, accuracy and practical aspects of quantifying macerals have been investigated based on a series of round-robin analyses by the Canadian Coal Petrography Group (CCPG) using fabricated coal samples. The precision of point-count analysis progressively increases with points counted, however, the accuracy of most operators shows little or no improvement after relatively few points (200-300) and the difference in accuracy between analysts is very large. The stepping distance used during routine point-count analyses is important; the stepping distance must exceed the maximum particle size of the coal, otherwise a greater random variation and possible systematic error is introduced. Visual estimates of the relative abundance of macerals are highly variable and there is little correlation between the analysts' experience and their accuracy. Visual estimates of some operators are more accurate than the point-count results of others. Image analyses and autocounting (specimen plane scanning) with construction and dissection of reflectograms yields mixed results. One image analyzer/operator obtained good results comparable to point-counting by the best analysts whereas a second image analyzer/operator obtained results poorer than average. An autocounting system using 5-mu-m stepping distance obtained poor results whereas using a 50-mu-m stepping distance the results were comparable to point-counting by the average analyst. The poorer results with small stepping distances is a product of the loss of independency between successive points (i.e., a particle may be counted more than once). Both the image analysis systems and the autocounter are plagued with problems of mineral matter and reflections below the polished surface which may lead to a systematic error (Holmes effect) and problems in differentiating binder from liptinite or very low reflecting vitrinite. The published standards for maceral analyses currently followed by most organic petrographers are not satisfactory. Recommended point-counting stepping distances do not exceed the maximum particle size, recommended sample preparation procedures yield particles too coarse for an adequate sample on the average-sized polished section and the number of points recommended are greatly in excess of those needed for reasonable accuracy. The number of points required for a reasonable accuracy is study-dependent and is best determined by replicate analyses in conjunction with a plot of variation in relative proportion of the macerals with number of points counted. Even when analyzing a single pellet replicate analyses should be performed by dividing the sample into separate sets.