2021 Volume 38 Issue 1 Pages 67-80
Large-scale phytosociological survey data are gathered globally and used for environmental conservation. The Japanese government collects vegetation data accompanied by national vegetation mapping and publishes it in a database. However, classifying large-scale databases using traditional phytosociological table manipulation is time- and labor-intensive. Although the automated approach is expected to classify large-scale vegetation data rapidly, it is necessary to verify its consistency with the established classification system for practical use. Thus, I compared traditional classification and ISOPAM, a recently developed automated approach, for classifying Japanese coastal beach and dune vegetation for which community classification has already been established. The traditional table manipulation classified 42 herbaceous and 11 shrub communities, most of which corresponded to previously reported phytosociological vegetation units. ISOPAM automatically classified the same dataset into 16 vegetation units using the default settings. These vegetation units corresponded well to major communities with many data obtained through traditional tabular comparison. Plant communities with few data did not correspond to the ISOPAM classification, as they were integrated into major vegetation units. ISOPAM was therefore considered a suitable method for automatically extracting large-scale vegetation patterns using a large dataset. However, rare communities and outliers with few data might be difficult to detect with an unsupervised classification such as ISOPAM. Therefore, a high-quality labeled database needs to be developed as training data based on past phytosociological findings to classify large-scale data efficiently.