Ábrahám Levente (szerk.): Válogatott tanulmányok VI. - Natura Somogyiensis 19. (Kaposvár, 2010)
HORVÁTH GY., HERCZEG R., TAMÁSI K. & SALI N.: Nestedness of small mammal assemblages and role of indicator species in isolated marshland habitats
288 NATURA SOMOGYIENSIS this esimator calculates nestedness from quantitative data matrices that include the number of events of each interaction (GALEANO et al. 2009). We used the "bipartite" package of R (DORMANN et al. 2008) to calculate the weighted-interaction nestedness estimator (WINE), where d ma x is the distance of the completely packed matrix. The WIN of the data matrix can then be normalized by comparing it to the average WIN of equivalent random matrices and to the WIN of the maximal nestedness matrix (WRIGHT & REEVES 1992) to obtain the weighted-interaction nestedness estimator (GALEANO et al. 2009): where the d w is the mean weighted distance of all its non-zero elements, d rn J is the average value of 1000 replication random matrix. The value of this estimator approaches (rjJ 0 when the WIN of the original data matrix is close to the average WIN of the equivalent random matrices. It approaches 1 as it gets closer to the nestedness of the maximal nestedness matrix. The estimator evaluates the relative position of the data matrix between the corresponding random matrices and the maximal nestedness matrix. Negatives values for this estimator can be found in some synthetic matrices that have been described as "anti-nestedness" matrices ( ALMEIDA-NETO et al. 2007). Negative values indicate that the synthetic matrices are less nested than the corresponding random matrices (GALEANO et al. 2009). The program WINE calculates the standardized static variable z-score to assess the significance of WIN. This score assesses how different WIN d w is from average random WIN d rn d. It is important to mention that the original data is just one combination of all possible permutations of registered events, z values below -1.65 or above 1.65 indicate approximate statistical significance at the 5% error level (one-tailed test). WINE calculates a weighted-interaction distance (d^), which estimates nestedness taking into account the number of events in the links, whose values are expressed in a colour scale of the weighted distance matrix. The colour plot is relevant in depicting the relative importance of each interaction, for instance, in the identification of idiosyncratic species or in the evaluation of extinction sequences in species distribution in fragmented habitats (GALEANO et al. 2009). Similarity of the species composition was calculated by the Bray-Curtis quantitative index, and similarity structure was analysed by hierarchical cluster analysis applying the UPGMA fusion method. The SYN-TAX package (PODANI 1993) was used for these computations. To find character species of the cluster hierarchy the IndVal method (DUFRÉNE & LEGENDRE 1997) was used: In the equation Aa stands for specificity where N t j is the mean number of individuals of the i' h species in the j' h habitat (habitat type), and Nj is the totalised mean number of individuals of the i t h species in all of the habitats (or habitat groups). B t j is the degree of fidelity where N (traps)i j stands for the number of traps in the j' h group in which the i' h species was captured. N (sites) i is the total number of traps in the j t h habitat group. This method combines the mean number of species individuals with its relative frequency of occurrence in the various groups of sites in the cluster hierarchy. The index has a maximum 1 rnd
