Morphological Variation of the Green Toad, Bufo viridis, in Italy: A Test of Causation

Abstract

This study describes patterns of geographic variation in morphometric characters of the green toad, Bufo viridis, in the western part of its distribution, including the Italian Peninsula and the islands of Sardinia and Corsica. The causes of this variation were investigated by comparing observed patterns of interpopulational differences to hypothesized factors derived from climatic and geographic variables, through the use of canonical correlation analysis and partial Mantel tests. Among-population differences in morphometric characters include both size and shape. Island populations were found to have larger body sizes than mainland populations, and there was an association between climatic factors and morphological characters. Partial Mantel tests carried out on a number of possible causal hypotheses, both climatic and geographic, suggest that a large portion of this association should be interpreted as a spurious effect of genetic isolation between island and mainland populations and of the limiting effect of geographic distance in determining gene flow. urnal of Herpetology, Vol. 32, No. 4, pp. 540-550, 1998 pyright 1998 Society for the Study of Amphibians and Reptiles orphological Variat on f the Gr en Toad, Bufo viridis, in Italy: Geographic variation of genetically determined characters can be the result of both selection and restrictions to gene flow (Endler, 1982; Thorpe, 1987). Morphometric characters have been employed successfully in describing intraspecific geographic variation; it is assumed they are polygenically controlled and therefore can reflect the condition of a large number of loci (Atchley et al., 1982; Thorpe, 1983). Many studies on different vertebrate taxa have shown that the geographic pattern of morphometric variation can be explained either by natural selection for current ecological conditions, or by the influence of gene flow (Flischer and Johni i e etical y deterston, 1982; Malhotra and Thorpe, 1991; Malhotra, 1992; Castellano et al., 1994), i.e., either by ecogenetic factors (sensu Thorpe et al., 1991) or by historical factors (Thorpe, 1975, 1979). Historical caused geographic variation does not relate to current conditions but is explained by historical events, such as genetic drift occurring in small populations after isolation. Few studies on morphological geographic variation of vertebrate species have employed amphibians as a model (Calhoon and Jameson, 1970; Fishbeck and Underhill, 1971; Hedges, 1991; Mensi et al., 1992; Lee, 1993; Green et al., 1996). Nevo (1973) found clinal variation with 540 This content downloaded from 157.55.39.217 on Mon, 05 Sep 2016 04:00:32 UTC All use subject to http://about.jstor.org/terms GEOGRAPHIC VARIATION IN GREEN TOADS increasing body size from east to west in Acris crepitans and from north to south in Bufo viridis in Israel (Nevo, 1972). He interpreted these patterns as responses to increasing risks of desiccation, with selection favoring larger size in the more arid areas of both ranges. Lee (1993) found no significant association between body size and arid climatic conditions in Smilisca baudinii, but found a significant association between relative appendage length and total annual rainfall. Consistent with Allen's ecogeographic rule, Lee (1993) suggested that desiccating environments can select for relatively short appendages. Lotter and Scott (1977) found a significant correlation between climatic parameters and morphology in Plethodon cinereus, while a number of studies on anurans have analysed variation in size in relation to altitude (Berven, 1982a, b) and latitude (Riha and Berven, 1991). Most of these studies considered body size variation but did not test a gene-flow hypothesis as the causal explanation of the observed patterns. Bufo viridis is distributed throughout the Mediterranean area (with the exception of southern France and the Iberian Peninsula), Central and Eastern Europe, Central Asia, and Saudi Arabia. This broad area of distribution is characterized by highly varied spatial structure and varied ecological conditions, from coastal environments to mountain (over 2000 m in altitude), and from deserts to riverine forests. In the Italian Peninsula the green toad is restricted to lowland habitats (from sea level to 600-700 m), whereas in Sardinia its range extends to include the highest mountains (1500 m above sea level). In this paper we describe the geographic pattern of morphometric variation in a part of the range of Bufo viridis, i.e., the Italian peninsula, Corsica, and Sardinia. A number of causal hypotheses were simultaneously compared to determine their relative significance in explaining observed patterns of variation. MATERIALS AND METHODS Samples and Morphometric Characters.-In April and May 1993 a sample of at least 10 adult males was collected from each of six breeding populations from throughout the Italian Peninsula, and from one population in Pellestrina, a small island in the Venetian lagoon, one population in Corsica, and two populations in Sardinia (Fig. 1). Some of these localities were also visited during the 1994 and 1995 breeding seasons, for a total sample of 184 males. Specimens were captured at their breeding sites during reproductive activity. Body mass and 20 body proportion characters (Fig. 2) were measured on anaesthetized specimens. Measurements were made to the nearest 0.01 mm w th dig tal calipers. All morphometric data were collected by the same researcher to avoid inter-observer source of variation (Lee, 1990). Description of the Pattern of Geographic Variat on.-To minimize deviations from normality and distortion effects caused by allometric relationships, all body measurements and body masses were log-transformed. Univariate and bivariate parametric statistics were calculated. To analyze separately size and shape factors, we first defined a multivariate size factor and then adjusted all body proportion characters with respect to the body size component. Multiple group principal component analysis (MGPCA; MULT program, Thorpe, Bangor Univ., UK) was used to obtain a multivariate expression of the size factor. As described by Thorpe (1983), when applied to morphometric data, MGPCA makes it possible to find a first component which is highly correlated among characters and best expresses body size. Therefore, the largest principal component obtained by MGPCA on the correlation matrix of log-transformed morphometric characters was assumed to represent the size factor. Residuals of regressions of the size-factor and the log-transformed characters were considered new values of size-adjusted body proportion characters and univariate analysis of variance were performed on these variables. Generalized patterns of geographic variation were obtained by applying Canonical Variate Analysis (CVA) to the correlation matrix of logtransformed data. To avoid problems due to colinearity effects, the minimum number of informative variables was determined by using MGPCA as an exploratory device, following the methods described by Reyment et al. (1984). Causal Analysis of Geographic Variation.-Each locality was defined by geographic coordinates and by a binary variable distinguishing Sardinia-Corsica populations from mainland populations (including Pellestrina). Altitude was not considered in the analysis, because eight localities are at sea level and two (Pellice and Isolabella) are only about 300 m above sea level. Four climatic parameters also were used to characterize each locality: (1) mean annual temperature (MAT); (2) total annual rainfall (AR); (3) aridity index (AI); and (4) prolonged activity index (PAI). All climatic variables were measured directly from climatic diagrams reported in Walter and Lieth (1967) of meteorological stations within 40 km from the study sites and at comparable altitude. The aridity index was obtained by measuring the total area comprised below the monthly average rainfall curve and above the monthly average temperature curve, during the water deficit period, as proposed by Gaussen (in Walter and Lieth, 1967). The prolonged activity index was defined as the portion 541 This content downloaded from 157.55.39.217 on Mon, 05 Sep 2016 04:00:32 UTC All use subject to http://about.jstor.org/terms S. CASTELLANO AND C. GIACOMA FIG. 1. Map of the Italian Peninsula, Sardinia, and Corsica showing localities from which specimens were collected. of the year with a mean daily temperature higher than 15 C. This value corresponds approximately to the mean air temperature of April in the northernmost localities sampled (Pellice, Isolabella), when most of the animals become active after winter latency. Canonical correlation analysis (CCA) on all body proportion characters of all specimens was used to obtain a general overview of the relationships between independent climatic variables and dependent morphometric characters. However, one intrinsic limit to CCA is its inability to include simultaneously the effects of geographic proximity. To overcome this limit, the partial Mantel test of matrix association (Smouse et al., 1986; Manly, 1991), a randomization test, was employed (Manly's RT-MANT program, modified by Thorpe). This analysis calculates the partial regression coefficients among a dependent distance matrix and up to eight independent distance matrices. Distance matrices were obtained by calculating betweenspecimens taxonomic distance (Sneath and Sokal, 1973). All distance matrices were standardized to obtain partial regression coefficients. Dependent morphological distance matrices were calculated by employing: (1) the first six CVA scores; (2) the body size factor scores; (3) 542 This content downloaded from 157.55.39.217 on Mon, 05 Sep 2016 04:00:32 UTC All use subject to http://about.jstor.org/terms GEOGRAPHIC VARIATION IN GREEN TOADS