Advantages of the original clustering method of DRSA, or Distance-Ranked Sorting Assembling, for vegetation classification are discussed. Using ranks in determining distances between objects provides robust clustering in case of noisy and heterogeneous phytocoenotic data. Algorithm of objects agglomeration is based on ranking objects by the indices of freeness and connectedness as well as on assessing clusters within k-NN graph’s framework. Clusters are assembled iteratively for some time to be finalized at the maximum of cluster’s connectivity. We also consider in detail approaches to assess classification quality of phytocoenotic dataset including degree of cluster’s (phytocoenon) compactness-distinctness and amount of differential species. We propose using nominal correlation coefficients to evaluate concordance of phytocoenotic classifications and contingency tables to compare frequencies of common releves between different classifications. Phytocoenon’s compactness and distinctness are evaluated using well-known internal cluster validation indices, e.g. silhouette statistics. We introduced CDR-index (compactness / distinctness ratio) which is calculated from the score of average similarity of within-phytocoenon and between-phytocoenons releves. Total amount of faithful (differential) species and average amount of them per phytocoenon as floristic index of partitioning quality were used. We classified differential species on a statistical basis calculating specied-to-cluster fidelity index and selecting species with fidelity above defined fidelity’s threshold. Using the sample phytocoenotic datasets we proved that both internal and floristic indices of classification quality improve after the exclusion of transient releves with ecotonic species composition. In the DRSA method, noise detection is carried out during cluster agglomeration; this objectifies rejecting ecotonic releves according to Braun-Blanquet approach as well as increases amount of differential species and thus improves phytocoenons interpretability.

Keywords: DRSA, cluster analysis, Braun-Blanquet approach, phytocoenon, quality of classification

Full text: PDF (Ukr) 920K

1. Bruelheide H. J. Veget. Sci., 2000, 11: 167–178. https://doi.org/10.2307/3236796
2. Calinski R.B., Harabasz J. Communications in Statistics, 1974, 3: 1–27. https://doi.org/10.1080/03610927408827109
3. Chytry M., Horak J. Preslia, 1997, 68: 193–240.
4. Chytry M., Tichy L., Holt J., Botta-Dukat Z. J. Veget. Sci., 2002, 13: 79–90. https://doi.org/10.1111/j.1654-1103.2002.tb02025.x
5. Chytry M., Vicherek J. Lesni vegetace Narodniho parku Podyji/Thayatal. Die Waldvegetation des Nationalparks Podyji/Thayatal, Praha, 1995, 166 pp.
6. Chytry M., Vicherek J. Přírod. Sborn. Zapadomorav. Muz. Třebíč, 1996, 22: 1–125.
7. Cover T.M., Hart P.E. Inform. Theory, 1967, 13: 21–27. https://doi.org/10.1109/TIT.1967.1053964
8. De Caceres M., Font X., Oliva F. J. Veget. Sci., 2008, 19: 779–788. https://doi.org/10.3170/2008-8-18446
9. Goncharenko I.V. Analiz roslynnoho pokryvu pivnichno-skhidnogo Lisostepu Ukrainy. Monografiya. In: Ukr. Phytosoc. Col. (spec. issue), 2003, 1(19): 203 pp.
10. Goncharenko I.V. DRSA (distance-ranked sorting assembling) – metod sortuyuchogo klasternogo analizu. Svidotstvo pro reyestratsiyu avtorskogo prava, № 58837, publ. 26.02.2015, 2015a, Byull. no 36.
11. Goncharenko I.V. Reports of the National Academy of Sciences of Ukraine, 2015b, 9: 129–136. https://doi.org/10.15407/dopovidi2015.09.129
12. Goncharenko I.V. Vegetation of Russia, 2015c, 27: 125–138.
13. Goodall D.W. Numerical classification. In: Handbook of vegetation Science. Part V: Ordination and Classification of Vegetation. Ed. R.H. Whittaker, The Hague: Junk, 1973, pp. 105–156. https://doi.org/10.1007/978-94-010-2701-4_19
14. Halkidi M., Batistakis Y., Vazirgiannis M. J. Intell. Inform. Systems, 2001, 17: 107–145. https://doi.org/10.1023/A:1012801612483
15. Hennekens S.M. MEGATAB – a visual editor for phytosociological tables. Version 1.0. October 1996. Ulft., 1996, 11 pp.
16. Hill M.O. TWINSPAN – A FORTRAN program for arranging multivariate data in an ordered two-way table by classification of the individuals and attributes. Program manual, Ithaca; New York: Cornell Univ., 1979, 90 pp.
17. Hill M.O., Šmilauer P. TWINSPAN for Windows version 2.3, Huntingdon & Česke Budějovice: Centre for Ecology and Hydrology & Univ. of South Bohemia, 2005, 29 pp.
18. Kosman Ye.H., Sirenko I.P., Solomakha V.A., Shelyah-Sosonko Yu.R. Ukr. Bot. J., 1991, 48(2): 98–104.
19. Legendre P., Legendre L. Numerical ecology, 2nd English ed., Amsterdam: Elsevier, 1998, 853 pp. https://doi.org/10.1016/S0304-3800(00)00291-X
20. Ochiai A. Bull. Japan. Soc. Fish Sci., 1957, 22(9): 526–530. https://doi.org/10.2331/suisan.22.526
21. Oksanen J., Blanchet F.G., Kindt R., Legendre P., O'Hara R.G., Simpson G.L., Solymos P., Stevens M.H.H., Wagner H. Vegan: Community Ecology Package, 2010, available at: http://cran.r-project.org/web/packages/vegan/ (accessed 22 March 2016).
22. Rendon E. Abundez I., Arizmendi A., Quiroz E.M. Intern. J. Computers and Communications, 2011, 5(1): 27–34.
23. Semkin B.I. Ekvivalentnost mer blizosti i ierarkhicheskaya klassifikatsiya mnogomernykh dannykh. In: Ierarkhicheskie klassifikatsionnye postroeniya v geograficheskoy ekologii i sistematike. Ed. B.I. Semkin, Vladivostok, DVNTs AN USSR, 1979, pp. 97–112.
24. Sokal R., Sneath P. Principles of Numerical Taxonomy, San Francisco, CA: Wit. Freeman, 1963, 573 pp.
25. Tichy L. J. Veget. Sci., 2002, 13: 451–453. https://doi.org/10.1111/j.1654-1103.2002.tb02069.x
26. Tichy L., Chytry M., Hajek M., Talbot S.S., Botta-Dukat Z. J. Veget. Sci., 2010, 21: 287–299. https://doi.org/10.1111/j.1654-1103.2009.01143.x
27. Vasilevich V.I. Statisticheskie metody v geobotanike, Leningrad: Nauka, 1969, 232 pp.