A new formal classification for Japanese forest vegetation based on traditional phytosociological concepts

Abstract Aims To propose a comprehensive classification framework for Japanese forest vegetation, an expert system based on a traditional system was developed, using a large‐scale data set covering the whole of Japan. Location The entire Japanese archipelago. Methods A data set of 12,720 vegetation plots from the National Vegetation Survey database was established. Then, an expert system for automatic hierarchical classification was developed, based on the traditional system. The classification effectiveness was verified using fidelity measures, the occurrence of traditional characteristic species, consistency with semi‐supervised K‐means and modified TWINSPAN, and the similarity between each unit. To investigate correspondence to the environmental variables, a detrended correspondence analysis with the “envfit” function was used. Additionally, the occurring taxa were collated and compared with the flora lists of areas surrounding Japan. Results The Japanese forest data set was classified into 34 alliances by the expert system, and 14 orders and six classes were distinguished. Most of these units had diagnostic taxa including existing characteristic species and the heatmap of the Bray–Curtis index showed the independence of each unit, and the similarity of the plot groups assigned to the same unit. Out of 34 alliances, clustering with semi‐supervised K‐means supported 21 alliances, and modified TWINSPAN supported 16 alliances. The four classes correspond to the thermal conditions and support the existing concept of vegetation zones. The lower‐level vegetation units correspond to a narrower range of environmental variables than higher‐level units, and species composition varied with limited environmental variables. A floristic composition comparison emphasized the complexity of the species composition of Japanese forests. Conclusions This study proposes the first formal Japanese forest classification at the class, order, and alliance levels, reflecting traditional phytosociology systems and based on large‐scale vegetation plot data. Our classification provides an important foundation for the future revision of formal phytosociological systems throughout East Asia.