Chin J Plan Ecolo ›› 2015, Vol. 39 ›› Issue (6): 586-592.DOI: 10.17521/cjpe.2015.0056

• Orginal Article • Previous Articles     Next Articles

A comparison of multivariate regression tree and two-way indicator species analysis in plant community classification

ZHANG Wen-Jing, ZHANG Qin-Di(), WANG Jing, FENG Fei, BI Run-Cheng()   

  1. College of Life Sciences, Shanxi Normal University, Linfen, Shanxi 041004, China
  • Received:2015-01-19 Accepted:2015-03-31 Online:2015-06-01 Published:2015-07-02
  • About author:

    # Co-first authors

Abstract: <i>Aims</i>

Multivariate regression tree (MRT) is a new statistical technique for plant community classification. Studies on advantages of MRT were still insufficient. Our objective was to compare the results of MRT with two-way indicator species analysis (TWINSPAN), which is widely used in recent years, and to provide suggestions for choosing an appropriate classification method.

<i>Methods</i>

We used both MRT and TWINSPAN to classify the forest communities in the southern part of Lüliang Mountain. The MRT analysis gave a tree with the splits based on additional environmental factors. The degree of consistency between the two classifications are compared through the coherence coefficient.

<i>Important findings</i>

(1) MRT divided 41 forest quadrats into four formations: the Platycladus orientalis formation, Acer davidii formation, Quercus wutaishanica formation and Diospyros kaki formation. The types of the formations are consistent with the analysis results from TWINSPAN. (2) Based on the quadrats that compromise the formations, these two methodologies provided a relatively high consistency, with the coherence coefficient up to 80.5%. (3) Compared with the TWINSPAN, the MRT can be seen as a form of constrained classification analysis that can relate different aspects of species composition to environmental data, so the clusters defined by the MRT define community types of the ecotone in an objective manner not available in other techniques. Finally, the TWINSPAN is a very effective classification technique in plant ecology research. However, the MRT has advantage over the TWINSPAN in terms of classifying continuous or transitional quadrats.

Key words: coherence coefficient, comparison, multivariate regression tree, southern Lüliang Mountain, two-way indicator species analysis