Page 8 - Voda_Dapporto_alii_2016

www.nature.com/scientificreports/

therefore these populations require immediate conservation actions to preserve the uniqueness of the Maltese
butterfly fauna.
Levanzo and Lampedusa also had connections with Sicily and Tunisia, respectively, during the LGM. Due to
the proximity to Sicily, most butterflies on Levanzo are probably part of a metapopulation and no positive rich-
ness residuals were found for this island in the GLM including contemporary determinants of richness (Fig. 1a).
An exception could be C. pamphilus, which was represented on Levanzo by three slightly differentiated (single
mutation) endemic haplotypes. On Lampedusa, all the rare species are highly dispersive and widely distributed in
north Africa but not in Europe, probably because of climatic restrictions or species interactions leading to mutual
exclusion 31,32 . However, the substantially divergent haplotype of P. machaon only found on this island should be
considered as a priority for local conservation actions.
Chequered distributions should also be taken into consideration for conservation decisions. The possibil-
ity that density-dependent and founder-takes-all mechanisms are at the basis of the maintenance of these pat-
terns 36,37 suggests that the populations existing in an area/island represent the main barrier to the colonization
by other lineages, which would change the original genetic structure, probably established after a series of unre-
23
peatable historical events . For example, the fauna of Pantelleria is mostly composed of widespread and undi-
versified species, but three of them (P. celina, L. phlaeas, L. megera) have different lineages in Sicily and Tunisia.
Interestingly, all three species are represented on Pantelleria by typical north African populations, although the
island was closer to Sicily during the glaciations. The similarity to Tunisia for these three taxa is the most impor-
tant characteristic of this island’s butterfly fauna, but its cause is unknown. The extinction of these populations
may result in recolonization from Sicily, thus erasing a key biogeographic signal on Pantelleria.
Our approach provides an example of how a series of comprehensive analyses on a wide area and large tax-
onomic group can test rarely assessed biogeographic principles like the links of genetic structure with dispersal
tendency and frequency on islands, or the relative effects of contemporary and historical determinants on island
populations. The challenge of integrating community ecology and phylogeographic approaches can also provide
the baseline information for developing conservation strategies that maximize biodiversity at both the species and
intraspecific genetic levels.

Methods
Study area and data collection. We analysed the butterfly faunas of 11 circum-Sicilian islands (Lampedusa,
Levanzo, Linosa, Lipari, Maltese islands, Marettimo, Pantelleria, Salina, Stromboli, Ustica, Vulcano), Sicily itself and
nearby mainland locations in southern Italy (Calabria), Tunisia and Algeria. Presence data were gathered from several
post 1980 literature sources and from field surveys carried out by the authors between 1999 and 2015. Specimens used
for genetic analyses were collected outside protected areas and were deposited at the Institute of Evolutionary Biology
(CSIC-UPF), Barcelona, Spain.

Determinants for island species richness. To evaluate the influence of different factors on species rich-
ness for the studied islands, we used Generalized Linear Models (GLM) with Akaike Information Criterion (AIC).
We also assessed the relative importance of potentially correlated variables using hierarchical partitioning of
variance, employing the lmg metric implemented in the “relaimpo” R Package. To place the richness of the study
islands in a broader framework, we used presence data on the butterfly fauna from all the western Mediterranean
islands (excluding Sicily, Sardinia and Corsica because their large size would provide an unbalanced contribu-
24
tion to the analysis). Islands were included in the study if at least four out of five migrant species (P. brassicae, P.
rapae, C. croceus, V. atalanta and V. cardui) were recorded there. These species are conspicuous and widespread
throughout Europe and north Africa, and their records establish a minimal surveying standard .
24
In the GLM, species richness was modelled against the following biotic, geographic and climatic predictors:
i) mean annual temperatures (MT), ii) annual precipitation (AP), iii) island area (IA), iv) maximum elevation of
island (EL), v) isolation (IS), vi) source richness (SR) and vii) the occurrence of a Pleistocene connection (PC,
factor variable). MT and AP were obtained by computing the mean values of the cells corresponding to islands
in Bio1 and Bio12 layers from Bioclim (http://www.worldclim.org). For each island, the faunistic source was
identified as the nearest 50× 50 km area, either mainland (southern Europe or north Africa), large island (Sicily,
Sardinia, Corsica), or an island at least ten times larger than the target one. We calculated SR as the number of
species reported from the source area and IS as the minimum distance between the target island and its faunis-
tic source. To linearize the relationships, we log transformed richness, area, isolation and source richness. We
computed a first GLM using only contemporary variables and a second GLM in which we included the PC. We
compared the AIC, explanatory power and residuals of the selected variables between the two models.
Community structure. We estimated the degree of nestedness of butterfly communities in circum-Sicilian
9
islands by applying the widely accepted NODF metric . A recent study has shown that the detected degree of nest-
edness depends heavily on the selected null models due to their different tendency to preserve the features of the
original matrix . To assess the significance of the observed NODF, we used the NeD program and computed the
38
12
z-values using 999 null matrices, built by applying a series of different models (equiprobable rows and columns,
EE; proportional rows and columns total, CE; and fixed rows and columns, FF) .
12
39
Prior studies have divided species into “core” and “satellite” species in which “core” species are those occur-
ring on more than half of the islands, and “satellite” those present on fewer islands. We used the same approach
and divided the 27 species occurring on the 11 studied islands into “widespread” (those occurring on six or more
of the islands) and “rare” (those found on no more than five islands). We also separately analysed the degree of
nestedness of the two subsets.
The overall faunistic dissimilarities among areas can be partitioned between a component generated by
40
nestedness and one determined by species replacement (turnover) . We calculated the relative contribution of
Scientific RepoRts | 6:28828 | DOI: 10.1038/srep28828 8