Page 3 - Voda_Dapporto_alii_2016
P. 3

                                 Island         Res. 1   Res. 2    Diff.
                                 Maltese           3.5      0.8    2.7
                                 Levanzo          − 0.3   − 4.3    4.0
                                 Lampedusa         4.5      1.6    2.9
                                Table 2.   The residuals scored in the Generalized Linear Models for each island having had contact with
                                mainland areas during the LGM, obtained either by removing (Res. 1) or including (Res. 2) the Pleistocene
                                connection as a categorical predictor. The difference between the two values can serve as an estimate for the
                                number of species that colonized the islands during the LGM and survived until present.

                                Figure 1.  The packed matrix minimizing the NODF metric for the butterfly species recorded on the islands
                                examined in this study. The rows represent the studied islands and the columns the species occurring in each
                                island. Widespread species showed a much more nested pattern than rare ones.

                                                   βnest/βsor  NODF  Null Model  Mean NODF  Z
                                                                     EE      51.673     9.664* * *
                                 All species        0.603   75.613   CE      60.560     4.739* * *
                                                                     FF      77.549    − 2.698* *
                                                                     EE      32.451     1.649*
                                 Rare species       0.271   39.491   CE      37.739     0.275
                                                                     FF      43.557    − 2.016*
                                                                     EE      71.095     2.378*
                                 Widespread species  0.814  82.013   CE      74.553     1.989*
                                                                     FF      82.206    − 1.196
                                Table 3.   Results for the ratio between the nestedness beta diversity index (βnest) and the unpartitioned
                                Sørensen index (βsor) and observed NODF for the nestedness analyses, using all species, only rare species,
                                and only widespread species. Mean NODF and Z values are provided for different series of 999 matrices
                                generated by different null models (EE, equiprobable rows and columns; CE, proportional rows and columns
                                total; and FF, fixed rows and columns). Asterisks represent associated P values.

                                two sub-matrices corresponding to widespread and rare species (occurring in more and less than half of the
                                islands, respectively), widespread species showed a nested pattern for the EE and CE null models, while rare
                                species showed a significantly anti-nested pattern with the FF model (Table 3). In particular, species occurring
                                on less than four islands showed a highly chequered distribution (Fig. 1). Accordingly, when all species were con-
                                sidered, nestedness and turnover components had similar contributions in determining faunistic dissimilarities
                                with a ratio between nestedness and Sørensen (nest/sor) indexes of 0.588. When divided into sub-matrices, the
                                widespread species showed a predominance of the nestedness component (nest/sor =  0.814), and the rare ones
                                a predominance of species replacement (nest/sor =  0.232). A multiple regression showed that species frequency
                                on islands was significantly associated with both frequency at source (n =  32, Est. =  0.037, t =  3.88, P <  0.001,
                                Fig. 2a, Supplementary Fig. S2a) and with dispersal tendency (n =  32, Est. =  4.607, t =  2.84, P =  0.008, Fig. 2b,
                                Supplementary Fig. S2b). The partitioning of variance revealed that dispersal tendency explained a lower fraction
                                of variation than frequency at source (16.4% for dispersal tendency and 29.5% for frequency at source).

         Scientific RepoRts | 6:28828 | DOI: 10.1038/srep28828                                                 3
   1   2   3   4   5   6   7   8