Page 4 - Tonelli_et_al_2016

164 Tonelli et al.
sa (2006), Lo Cascio et al. (2006), Arnone, (2010),
Fattorini (2010). The Tunisian data were taken from
Baraud (1985) and Errouissi et al. (2009) and reviewed
by Imen Labidi (pers. com., 2015) on the basis of
comparisons with the collections Henry Normand and
Muséum National d'Histoire Naturelle (Paris). Some
data from the Egadi archipelago (Euoniticellus fulvus
(Goeze, 1777); Cheironitis irroratus (Rossi, 1790);
Onthophagus taurus (Schreber, 1759); Calamosternus
granarius (Linnaeus, 1767); Calamosternus mayeri
(Pilleri, 1953)) were provided by Marco Dellacasa
(pers. com., 2014). The data for Malta were drawn
from Pivotti et al. (2011).
Analysis and interpretation
There are several methods for reconstructing dispersal
fluxes (Nathan et al., 2003). Some are complex and
financially expensive but very informative (i.e., phylo�
geography), while others are simple and inexpensive
but less accurate (i.e., pairwise comparisons derived
from a presence/absence matrix in homogeneous
areas; Legendre, 1986). In this study, we adopted
the latter as a preliminary and exploratory strategy
to evaluate the existence of any noteworthy patterns.
Data were analysed using the coefficient of disper�
sal direction DD2 (Legendre & Legendre, 1984). This
coefficient, rarely used in biogeographical analyses
(but see: Legendre & Legendre, 1984; Bachraty et
al., 2009; Borcard et al., 1995), measures the likeli�
hood of species dispersal between two areas using
species presence–absence data.
The formula of the DD2 is:
DD2 (x1à x2) = 2a 2a c (b – c)
+b+ a+b+c
where a is the number of species that two regions
have in common; b is the number of species found in
x1 but not in x2; c is the number of species found in
x2 but not in x1. The first portion of the DD2 coefficient
is the Sørensen index of similarity, while the second
portion measures the asymmetry in taxonomic com�
position. As Legendre & Legendre (1984) assert, 'the
first portion states that unless two adjacent regions
possess species in common, it would be difficult to
think of these two faunas as deriving one from the
other. The second portion creates the pictures of a
fauna waiting at the border to invade an adjacent re�
gion'; namely, the greater the number of species that
inhabit an area, the greater the likelihood that this area
acts as a source for neighbouring areas. Thus, DD2
measures the likelihood that species have dispersed
(b larger than c). A negative value (c
form x1 to xb2) indicates that, if dispersal occurred,
larger than
species might have migrated from x2 tqouxa1n(tiLtyeg(ben–drce)
& Legendre, 1984). In summary, the
would indicate the direction of dispersal flux, while
the DD2 value, which reacts to both the similarity and
the asymmetry between areas, would estimate flux
intensity (Bachraty et al., 2009). We chose to use the
DD2 index because, in analysing the possible disper�
sal fluxes, we think that presence is more important
than supposed absence, and then we decided to give
double weight to the common species. This choice
also allowed us to limit the biases and inconveniences
caused by possible local extinctions.
To evaluate the dispersal fluxes, we divided the
circum–Sicilian volcanic islands into two groups:
north of Sicily (Aeolian Islands + Ustica) and south
of Sicily (Pantelleria and Linosa). Then we hypoth�
esized two colonization scenarios: the first excluding
any continental islands as possible sources, and
the second including continental islands as possible
source areas (table 1). Of all the possible source
areas (significant values of DD2), we considered only
those with the highest value of DD2 as being likely
source areas, since this value indicates a greater
intensity of flux. When two or more possible source
areas had a relative difference in the DD2 value of
less than 5%, they were both discussed as possible
equivalent source areas.
The McNemar test was used to test the null hypoth�
esis that there is no asymmetry between two areas
(H0: b = c). We used a two–tailed test of significance
setting the probability of a type I error at α = 0.05. The
coefficients were calculated for each pair of areas.
The coefficients DD2 were calculated using the func�
tion bgdispersal of the Vegan Package (Oksanen et
al., 2012) for the software R (R Development Core
Team, 2011).
In order to identify the possible dispersal routes
by the use of DD2 coefficient, we assumed that: (a)
The volcanic islands were originally empty. The fauna
and flora now present in these areas are necessarily
dispersed from other source areas. This assumption is
consistent with the geological history of the concerned
islands. Indeed, they have originated from volcanic
events in the period between 1000 and 90 Kyr and
it is impossible that they had an ab origine fauna. (b)
The past dispersal events have necessarily left marks
on the present communities (Legendre & Legendre,
1984; Borcard et al., 1995; Bachraty et al., 2009). (c)
Dispersal comes from areas of high to low taxonomic
richness (Legendre & Legendre, 1984; Borcard et al.,
1995; Bachraty et al., 2009). And (d) Given the strong
homogeneity of the environmental parameters of the
islands, the similarity in the biodiversity pattern and
fauna between islands should be mainly related to
dispersal processes.
Results
In total, through the literature review, we identified
176Â dung beetle species as being present in the study
area: 18 Geotrupidae, 53 Scarabaeidae, and 105
Aphodiidae. On the volcanic islands alone, 48 species
are reported: 3 Geotrupidae, 24 Scarabaeidae, and
21 Aphodiidae. The species richness of the volcanic
islands ranges from 35 (Vulcano island) to one (Pana�
rea island). The species with the highest frequency
in the volcanic islands is Thorectes intermedius
(eight islands; see annex).
Tables 2 and 3 show the results of DD2 fluxes
with their McNemar and probability values. Figures 2
and 3 show these results graphically. The dispersal