Page 2 - capula_eal2009

218 MASSIMO CAPULA, SARA CHIANTINI, LUCA LUISELLI, ANNA LOY
methods, as (i) it provides independent description of
size and shape variation, (ii) it can be used with
multivariate statistics to test for statistical differences in order to avoid depletion of the few, very small and
shape (Bookstein 1991; Bookstein 1996; Marcus et al.,
1996), (iii) it makes it possible to relate shape variation
to other sets of independent variables, i.e. size,
latitudinal gradients, etc., and (iv) it can be used to
describe differences in shape among objects in terms of
deformation grids (Adams et al., 2004). This is possible
because the geometry of shape is preserved throughout
the analysis.
MATERIALS AND METHODS
All the samples analysed were from three central
Mediterranean insular areas, namely Sicily, Egadi Islands
and Aeolian Islands. For comparative analysis (biometric
analyses only), two samples of Podarcis sicula from the
southern Italian Peninsula (Calabria) were also used (Fig.
1). Details regarding geographic origin and number of
specimens analyzed are reported in tab. 1. All the lizards
studied were obtained from the collections of the Museo
Civico di Zoologia di Roma (Italy). The same lizards
analysed here were previously used for allozyme studies
(Capula, 1994a, 1994b, 2004, 2006a).
To avoid any problem in species recognition, the
identification of the individuals studied was based both
on (1) throat colouration pattern (P. raffonei: dark
markings on the throat; P. sicula: white and unspotted
throat; P. wagleriana: salmon red throat, in some cases at
least with small dark markings; cf. Arnold & Ovenden,
2002), and (2) recognition of alternative alleles at the Ada
electrophoretic locus (P. raffonei: Ada103; P. sicula: Ada100; warp scores). Ordination of specimens was obtained
P. wagleriana: Ada110), according to previous allozyme
analyses of the same individuals by Capula (1994a).
In this paper landmark based morphometrics has been
applied to analyse the cephalic scales of the three
Podarcis lizards described above. Lacertid lizards have
well-shaped cephalic scales which are easily
distinguishable and the geometric properties of which
make them suitable markers for taxonomic,
biogeographical and phylogenetic studies (cf. Arnold,
1989; Bruner et al., 2005; Bruner & Costantini, 2007;
Kaliontzopoulou et al., 2007). Twenty nine homologous
landmarks were recorded on the half configuration of the
cephalic scales of 109 adult males of the three species (P.
raffonei, N = 13; P. sicula, N = 47; P. wagleriana, N = 49;
see tab. 1; for nomenclature of the analysed scales see fig.
2). Females were not analysed in order to avoid problems
in the analyses due to sex-related morphological
differences. The low number of P. raffonei individuals
used for the analysis is because the species is at present
critically endangered (IUCN, 2007) and extremely rare
(Capula et al., 2002; Capula, 2006a); we decided to
analyse the specimens stored in the museum collections
only and to not collect other individuals in the field in
protected remaining populations of P. raffonei. Only
lizards with a snout-vent length (SVL) ≥ 5 cm, i.e
sexually mature lizards, were considered for the analysis.
Digital images of the dorsal part of the skulls were
recorded by placing the head of the lizard under a
binocular microscope connected to a digital camera and a
PC (Leica QuantiLite Image Acquisition System®), taking
care of distorsions and scale as described by Bruner et al.
(2005). A file with list and address of images was created
through tpsUtil (Rohlf, 2006) to allow the consecutive
recording of landmarks on predefined groups of
specimens. Cartesian x and y coordinates of landmarks
were recorded on digital images using the tpsDig
program (Rohlf, 2006). The centroid of each
configuration was translated to the origin, and
configurations scaled to a common, unit centroid size
(Bookstein, 1986). Centroid sizes were extracted through
tpsRelw (Rohlf, 2006) and stored in a separate file for
successive analyses. The configurations were optimally
rotated to minimize the squared differences between
corresponding landmarks (Gower, 1975; Rohlf & Slice,
1990) by using the Generalized Procrustes Analysis
(GPA, Rohlf & Slice, 1990) in the tpsRelw (Rohlf, 2006)
and Morpheus (Slice, 1994-1999) programs. Landmark
vectors of all configurations were first plotted and
visualised to detect wildly incorrect or mislabelled points
rapidly. Size variation among and within species was
examined through ANOVA and MANOVA of centroid
sizes. Shape differences were evaluated through the
analysis of the residuals from superimposition (partial
through Relative Warp Analysis run on the partial warp
scores (weight matrix), using the tpsRelw program, by
setting the parameter α to 0, as recommended for
exploratory analyses (Rohlf, 1996). The same program
was used to produce the thin plate spline transformation
grids associated with variation along the relative warp
axes. Relative warp scores with eigenvalues greater than
zero were used as data input for a canonical variate
analysis. The tpsRegr program (Rohlf, 2006) was used to
correlate shape coordinates with variation along the first
canonical axes, and to produce transformation grids of
shape changes associated with the patterns observed in
the canonical space.
Seven biometric measurements were also recorded
on 135 adult males (P. raffonei, N = 13; P. sicula, N = 76;
P. wagleriana, N = 46; see tab. 1) using a precision
caliper (± 0.1 mm), following the standards provided by
Guillaume (1987), in order to compare geometric and
biometric patterns of variation and to evaluate any static