Abstract: Many methods are used in vegetation classification; however, their applications in a single vegetation type are seldom empirically compared. Here, three commonlyused methods, including dominancetype classification, TWINSPAN and UPGMA, were used to classify the communities from 40 sampling sites in Hulunbuir grassland. The dominancetype classification divided these communities into 16 groups, while TWINSPAN and UPGMA divided into 10 and 11 groups, respectively. The coherence coefficient was 0.94 between dominancetype and UPGMA, 0.91 between TWINSPAN and UPGMA, and 0.87 between dominancetype and TWINSPAN. The highest coefficient indicated that the results of dominancetype classification and UPGMA were highly consistent. By calculating the variance of species richness, total aboveground biomass and average similarity coefficient of three methods, the cumulative variance of each method was obtained: 1131 for dominancetype classification, 976.8 for TWINSPAN, and 952.8 for UPGMA. The largest variance suggested that the results of dominancetype method had the highest variation within groups, and thus the results from dominancetype classification might need to be adjusted. The difference of within groups for UPGMA was the smallest. Results from the present study indicated that UPGMA should be more suitable for the classification of plant communities in Hulunbuir grassland, compared with the dominancetype classification that was widely used in the previous community classification literature. Our results had great implications for the future research focusing on the vegetation classification of grasslands in China.

Key words: landscape planning, low impact development (LID), low impact development technology, runoff control rate, coastal industrial park