Evaluation of statistical fidelity indices in determination diagnostic species in Hyrcanin Box tree (Buxus hyrcana Pojark.) communities

• Published in: Taḥqīqāt-i jangal va ṣanubar-i Īrān Vol. 24; no. 3; pp. 427 - 415
• Main Authors: Hamed Asadi, Mohsen Hosseini, Omid Esmailzadeh, Younes Asri, Habib Zare
• Format: Journal Article
• Language: Persian
• Published: Research Institute of Forests and Rangelands of Iran 01.09.2016
• Subjects: Association indices of species-site groups; diagnostic species; TFVI model; vegetation classification
• ISSN: 1735-0883; 2383-1146
• DOI: 10.22092/ijfpr.2016.107368

There are several available indices to assess the group associations between species and sites in order to identify diagnostic species. In this study, we introduced 11 indices of species-site groups and evaluated their quality by means of a TFVI model and four numerical criteria, including average positive fidelity, sharpness, uniqueness and crispness based on of 168 releve of Buxus hyrcana habitats. We used a modified TWINSPAN classification of 168 releve (Nine vegetation community/site groups) to determine the species fidelity value. The results showed that new classification obtained from assigning releve using TFVI model based on IndVal, UbinB and Ochiai indices returned the highest values of the fourfold criteria in evaluating classification quality. Moreover, they achieved higher qualities compared to other indices. Also results of cluster analysis and PCA diagram showed that 11 species-site groups’ association indices can be divided into four groups. The scatter plot of association indices along two first axes in PCA (eigenvalues 0.556 and 0.365, respectively) explained 92% of total variation, and species fidelity values based on IndVal and Ochiai indices were similar and noticeably different with other indices. Finally, the results of this study revealed that IndVal and Ochiai indices have the most priority in determining diagnostic species of vegetation communities compared to the other examined association indices.