Assessment of the stability of boundaries in flora-faunistic zoning

The publication has been prepared as a continuation of the previous article. The results of complex floro-faunistic zoning of Northern Eurasia have been described earlier. This region which covers the territory of the former USSR within the borders of 1991 was previously divided into 597 sections. For every section Jaccard coefficients of similarity were calculated separately for woody plants flora (1), for invertebrates’ fauna (2) and vertebrates’ fauna (3). The zoning was carried out on the basis of the analysis of the average matrix of similarity coefficients. To assess the representativeness of the used materials and stability of the boundaries, it was proposed to gradually increase the number of averaged matrices with additional attraction of the available data on other groups of organisms. This approach is realized by adding one more, the fourth matrix calculated from the flora of all vascular plants at a genus level to the three matrices of similarity coefficients mentioned above. The inclusion of additional matrix in the calculation of similarity increased the number of features from 2156 to 3799 and species and genii – two times. As a result, the previous classification representations changed ambiguously. Formally, the differences in the new classification covered 66 % of the sections. But in comparative assessing of the information content of the old and new classifications, the average deviation (in absolute value) did not exceed 0.8 % of the total variance of the similarity coefficients matrix. This indicates the reliability of the previously expressed judgments about the floro-faunistic heterogeneity of the region and enables to draw up several classifications that are different significantly in the composition of taxa, but equivalent in the proportion of the variance of the same similarity matrix. As in the previous work, in the course of the study we used non-traditional methods of applied statistics combining in one algorithm the approaches of cluster and factor analysis. The original method of linear qualitative approximation of the coupling matrices was also used, this allowed to take into account the nonlinearity of the coupling using the selected gradations of the factors revealed in the cluster-factor analysis.