The composition of volcanic ash, which is a source of primary description data in volcanological study, is important information for estimating the eruption styles and sequences. However, its description under a microscope by human operation has difficulties in classification thresholds and time and effort-consumptions. This study demonstrates an accurate and rapid description of volcanic ash samples that consist of thousands of grains. We analyzed nine tephra samples (two magmatic (dry) and seven phreatomagmatic (wet)), which were produced in the 1983 A.D. fissure eruption event at Miyakejima volcano, Japan. Our dataset, which is consists of multivariate shape and transparency parameters, was rapidly obtained using an automated grain analyzer. In this study, we applied a two-step cluster analysis to objectively and quantitatively define grain type and classify samples. To define grain types, we referred to the statistically appropriate number of clusters of whole-ash grains in our samples. For our samples, the appropriate number of clusters for grain type was five. Each grain type is characterized by parameters and has different proportions among our samples. In wet tephra samples, grains that were categorized as transparent and highly irregularly shaped types were relatively abundant. Those grains can be considered as vesicular sideromelane grains, which are often found in products of phreatomagmatic eruptions. Such a standardized description of volcanic ash based on statistically determined grain type will contribute to initial descriptions before subsequent detailed analysis.