Morphological shifts in populations of generalist and specialist amphibians in response to fragmentation of the Brazilian Atlantic forest

Changes in morphological traits, such as body size, body condition, and leg length, are important indicators of changes to life history or habitat quality, which can affect the performance of individuals and therefore the persistence of populations under environmental change. Only very few studies assessed the effect of fragmentation on morphological traits. The few available studies on anurans found that in landscapes with less forest cover body size decreased. Therefore, we predict that body size should also be smaller in fragments compared to continuous forest. Body condition is a further trait closely related to individual performance and thus should decline with more adverse conditions, as is expected in fragments. We tested these hypotheses using snout-vent length, body mass, body condition, and tibia length as response variables. We collected data of a habitat generalist (Rhinella ornata) and a habitat specialist (Ischnocnema guentheri), both leaf-litter amphibian species, from three sites in a fragmented landscape (two isolated and one connected site) and one site in a contiguous part of the Atlantic Forest of Southeast Brazil. In the generalist species, snout-vent-length (SVL) and body mass were significantly lower in fragments compared to the contiguous forest control, whereas tibia length and body condition Nature Conservation 13: 47–59 (2015) doi: 10.3897/natureconservation.13.7428 http://natureconservation.pensoft.net Copyright Henning Steinicke et al. This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. RESEARCH ARTICLE Launched to accelerate biodiversity conservation A peer-reviewed open-access journal


Introduction
Habitat loss and fragmentation are a major cause of biodiversity loss (Sarre et al. 1996, Pimm and Raven 2000, Sala et al. 2000, Funk and Mills 2003, Henle et al. 2004).
Most studies examining their effects focused on community composition, patch occupancy patterns, population viability, and the genetic variability of selected species (e.g.Tocher et al. 1997, Wiegand et al. 2001, Bell and Donnelly 2006, Dixo et al. 2009, Tucker et al. 2014).In contrast to the large volume of literature for oceanic islands (e.g.Kramer 1951, Losos 2009), very few studies addressed changes in morphological parameters following habitat fragmentation.However, habitat loss leads to major changes in abiotic and biotic conditions and morphology is known to respond to abiotic and biotic conditions (Wade 2004).Exemplarily, Smith et al. (1997) found that differences in vegetation structure between African rainforest and the rainforest-savanna ecotone led to different selection regimes causing morphological changes in rainforest birds.
Changes in morphological factors are important indicators of changes in life history or habitat quality (Palkovacs 2003, Buckley et al. 2005, Lowe et al. 2006, Lomolino and Perault 2007).Snout-vent-length, body condition, and body proportions are important phenotypic factors that affect the performance of individuals in terms of their reproductive output and survival in numerous organisms (Stearns 1976, Roff 1992).In amphibians, smaller snout-vent length and lower body condition often translate into reduced survival probability (Altwegg and Reyer 2003) and cause maternal effects, such as poor clutch quality (Reading 2007), smaller clutch size (Funk and Mills 2003), and smaller egg size (Kaplan 1992, Laugen et al. 2002).These factors in turn lead to a lower number of hatchlings and metamorphosing individuals (Berven andGill 1983, Semlitsch et al. 1988) and affect the size at metamorphosis (Pakkasmaa et al. 2003, Räsänen et al. 2005), maturity, and fecundity (Smith 1987, Semlitsch et al. 1988).As a consequence of their important influence on demographic traits, body size and body condition may ultimately determine the ability of species to persist under changed environmental conditions.
Various environmental factors are known to influence morphology and body condition.In amphibians, resource availability, predation risk, and temperature during larval development have been frequently identified as major determinants of body size and body condition.An increase in resource availability can lead to better body condition and a larger body size (e.g., Du 2006, Jessop et al. 2006, Wu et al. 2006).Due to the costs of anti-predator behaviour, higher predation pressure can lead to early hatch-ing, lower growth and developmental rates, and thus to a smaller size at emergence or metamorphosis (e.g., Ball andBaker 1996, Laurila et al. 2002).In amphibians this further negatively affects limb dimensions and locomotor performance (Goater et al. 1993, Buckley et al. 2005).Warmer temperatures accelerate developmental processes and usually lead to an earlier metamorphosis, smaller size at metamorphosis, smaller size at maturity, and therefore overall a smaller adult size (Smith 1987, Semlitsch et al. 1988, Beachy 1995, Ståhlberg et al. 2001).
Habitat loss and fragmentation create changes in the abiotic and biotic environment, therefore leading to changes in the quality of the remaining habitat (Saunders et al. 1991) that can result in changing selective pressures on morphology (Thomas et al. 1998, Ewers andDidham 2006).Thus, changes in morphology may serve as an early and easy to measure indicator for negative effects of fragmentation on species (Sumner et al. 1999).Finding suitable early indicators of environmental stress (Helle et al. 2011) is important in order to apply effective conservation measures prior to severe declines of populations and of the species at the landscape level (Leary and Allendorf 1989).However, the contribution of habitat loss and fragmentation to phenotypic changes is poorly known, in particular in amphibians.Neckel- Oliveira and Gascon (2006) showed shifts in the body size of a habitat generalist (Phyllomedusa tarsius) in fragments, regrowth forest, and pasture compared to continuous forest, with larger individuals in the latter, and explained these shifts with changes in habitat quality.Leg-length in relation to snout-vent length was also affected by the degree of retained forest in one Neotropical frog species (Delgado-Acevedo and Restrepo 2008).Likewise, the prickly forest skink (Gnypetoscincus queenslandiae) of tropical Australia was smaller and had lower mass (corrected for differences in snout-vent-length) in fragments compared to continuous forests (Sumner et al. 1999).
Body condition declined and the level of stress hormones increased with the number of forest fragments in a study of the common toad (Bufo bufo) (Janin et al. 2011).Likewise, yellow-bellied toads (Bombina variegata) from forest ponds showed better body condition than those from ponds in the agricultural matrix (Scheele et al. 2014).In contrast, Püttker et al. (2008) found a negative relationship between body condition and patch size in two tropical forest small mammal species that are habitat specialists and no significant relationship in four species that are habitat generalists.The mechanism(s) driving these patterns remain unclear but likely involve multiple factors.
Since previous studies found that in more disturbed habitats and landscapes with less forest cover body size decreased in anurans (Pyastolova and Vershinin 1999, Neckel-Oliveira and Gascon 2006, Delgado-Acevedo and Restrepo 2008), we predict that body size should also be smaller in fragments compared to continuous forest.Due to the strong link among morphological characteristics, we further predict that leg length and body condition should be reduced in forest fragments compared to continuous forest.Here, we test these hypotheses.Moreover, habitat specialists are particularly sensitive to changes of microhabitat quality in fragmented habitats and are consequently more prone to extinction in the fragmented landscape than generalists (Margules 1996, Henle et al. 2004), we therefore assume a stronger response regarding the morphological shifts of habitat specialist species compared to generalist species.

Study area
The study was carried out in the Mata Atlântica, Brazil, in continuous forest of the Morro Grande Reserve and a fragmented landscape surrounding Caucaia do Alto (both approx.23°40´S, 47°01´W), located 40 km southwest of São Paulo in the municipalities of Cotia and Ibiúna, Brazil.The forest reserve and the adjacent study area are located on the Atlantic Plateau of São Paulo at an altitudinal range of 860-1075 m above sea level (Metzger et al. 2006).The original forest of the study region is classified as lower montane rainforest and forms a transition between the coastal Atlantic rain forest and the Atlantic semi-deciduous forest (Oliveira and Fontes 2000).All study sites, including the control site, were characterised by secondary forest and have remained undisturbed for at least 50 years.The fragmented landscape presented approx.31% of remaining, highly fragmented forest that was surrounded by urban areas and rural installations (15%) and open areas (i.e., agriculture, pasture, fallow lands, areas in early successional stages; 39.3%) (see Dixo et al. 2009).
We selected four sites: two isolated small forest fragments (Carmo Messias and Dito, referred to in the following as "iso1" and "iso2", respectively), one small forest fragment (Alcides, referred to in the following as "connect") connected by a forest corridor to a larger forest area, and a control site ("control") within the continuous forest.Forest fragments covered an area of 5 ha each and contained no permanent/ larger water bodies within the forest area or within a radius of 200 m outside the forest fragment.The continuous forest was about 9,400 ha in size.There was no permanent/ larger water body present at the control site within a radius of 200 m.

Selected species
We chose two leaf-litter dwelling forest species for comparison.We selected Rhinella ornata, a bufonid species, as a habitat generalist due to its high tolerance to matrix habitats.This species is distributed throughout the Atlantic Forest in the states of São Paulo and Rio de Janeiro (Haddad et al. 2008) and can be found in undisturbed continuous forest areas as well as in disturbed forest fragments (Heyer et al. 1990).It breeds in temporary and permanent ponds within forests and open areas (Izecksohn andde Carvalho-e-Silva 2001, Haddad andPrado 2005).Reproduction starts at the end of the dry season in late July and at the beginning of August (Dixo 2005).
We selected Ischnocnema guentheri for comparison.This species is distributed over large parts of the Mata Atlântica (Haddad et al. 2008) and occurs only within forest habitats although it can sometimes be encountered in vegetation close to the forest edge (Heyer et al. 1990).Thus, it has a more specialised habitat requirement.Females deposit egg clutches terrestrially, hidden below tree trunks and stones (Kwet andDi-Bernardo 1999, Izecksohn andde Carvalho-e-Silva 2001).Frogs develop directly within the egg capsule (Heyer et al. 1990, Izecksohn andde Carvalho-e-Silva 2001).

Sampling
We established three parallel transects of 100 m each on each site.We sampled frogs by hand during the rainy seasons October 2003 until March 2004 and November 2004 until March 2005.We surveyed each site for five consecutive nights (21:00 -01:00 local time) before sampling the next site.Once all sites had been sampled, the next round of sampling started again at the first site.The total search effort was 25 nights per site (2003/2004 15 nights per site, 2004/2005 10 nights per site).
We measured snout-vent-length (SVL) and tibia length (TL) of all captured individuals using a calliper with an accuracy of ±0.1 mm.We measured tibia length for the left body side.We weighed individuals using digital scales with an accuracy of ±0.1 g.We used photographical individual identification to eliminate data points of recaptured individuals.As juveniles were seldom captured, we only included the data of subadult and adult individuals in the analysis (R. ornata with a SVL ≥ 27 mm; I. guentheri with a SVL ≥ 10 mm).As knowledge on natural history of both species is scarce, we could not further separate subadult from adult individuals.Unless calling, sexes cannot be separated by external morphology in both species.

Data analysis
Opinions about which body condition index (BCI) to use are controversial in the literature (Jakob et al. 1996, Schulte-Hostedde et al. 2005, Peig and Green 2009).The most common approaches are to use the residuals from a regression of mass on the third power of length or of ln mass on ln size.Peig and Green (2009) developed a scaled mass index (SMI) which standardizes body mass at a fixed value of body size based on the scaling relationship between mass and body size using major axis regressions.This new index better reflects changing relationships between mass and body size considering different growth curves at different ages than alternative indices (Peig andGreen 2009, 2010).We used the SMI and additionally the more frequently used residuals of the regression of ln mass on ln size for consistency assessment.
We assessed differences in SVL, body mass, and BCI among sites by using an analysis of variance (ANOVA), after testing the assumption of normality of all variables used (Kolmogorov-Smirnov-Test: D = 0.5 -0.7, all α > 0.05).For tibia length, we used an analysis of covariance (ANCOVA) to remove a potential influence of SVL in the comparison of sites.For significant results, we used the post-hoc Tukey's Honestly Significant Difference test (Tukey's HSD) for multiple comparisons to assess differences among individual sites.

Results
We captured a total of 499 individuals of the selected species, comprising 116 R. ornata and 383 I. guentheri.After removing juvenile individuals from the data, 54 and 376 individuals respectively were kept for the analyses.An overview of captures and measurements per species and site is given in Table 1.
Snout-vent length and body mass of I. guentheri did not differ significantly among sites (ANOVA: F = 1.35, α = 0.26 and F = 0.84, α = 0.47, respectively).Likewise, TL did not differ significantly among sites when the effect of SVL was accounted for (AN-COVA: F = 0.84, α = 0.5).BCI based on SMI was highest in the control site and one isolated site and lowest in the other isolated site (Table 1), but the difference among sites was only marginally significant (ANOVA: F = 2.00, α = 0.11).When using the residuals of the regression of ln mass on ln SVL as BCI, qualitatively results were very similar though BCI was highest in the control site (data not shown).The difference among sites was marginally significant (F = 2.67, α = 0.09), but all comparisons between pairs of sites were not significant (Tukey's HSD; all α > 0.05).

Discussion
Our results partly support the predicted effects of fragmentation on phenotypic characteristics (smaller snout-vent-length, shorter legs, and lower body mass, and body condition in fragments compared to continuous habitat).Individuals of the generalist species R. ornata were smaller in fragmented habitats compared to the control site, while body condition did not differ among sites.For the specialist species I. guentheri body condition differed marginally significantly among sites and was highest in the control site and in one isolated fragment but lowest in the other isolated fragment.The other morphological traits did not differ among sites.Thus, the habitat generalist and the habitat specialist showed different morphological responses to habitat fragmentation.
The reduced body size in R. ornata and the lower body mass of I. guentheri in the most severely affected (small isolated) site(s) are in line with previous studies in lizards and amphibians that found a reduced body size in landscapes with low forest cover, in fragments, and in disturbed compared to undisturbed habitats (Pyastolova and Vershinin 1999, Sumner et al. 1999, Neckel-Oliveira and Gascon 2006, Delgado-Acevedo and Restrepo 2008).Neckel- Oliveira and Gascon (2006) assumed, but did not test, that the morphological difference was environmentally induced during development due to changes in the habitat quality of the larval habitat.A smaller size at metamorphosis frequently translates into a smaller adult body size in anurans (Smith 1987, Semlitsch et al. 1988) although data from tropical species are not yet available.This explanation may also apply to the observed morphological shifts in R. ornata, as it also undergoes development in aquatic habitats.
In fragments that do not retain suitable aquatic habitats eggs have to be deposited in water bodies in the matrix.Eggs and aquatic larvae have to deal with changes in biotic and abiotic parameters of reproduction ponds within the matrix.Ponds inside and outside of forests usually will differ in physical conditions (Becker et al. 2007).Ponds within the matrix will receive more insolation and should therefore generally be warmer than forest ponds.Warmer temperatures accelerate developmental processes and usually lead to an earlier metamorphosis, smaller size at metamorphosis, smaller size at maturity, and therefore overall a smaller adult size (Smith 1987, Semlitsch et al. 1988, Beachy 1995, Ståhlberg et al. 2001, Wells 2007).
For frogs with a direct development much less is known about the factors that determine development and the morphology of hatchling and subsequently adult anurans than for pond breeding anurans.Delgado-Acevedo and Restrepo (2008) found smaller individuals of two directly developing Eleutherodactylus species -a genus closely related to Ischnocnema -in disturbed and fragmented compared to undisturbed and unfragmented habitats.They further found an influence between the level of remaining forest cover in the landscape and the allometric change of leg length in one of the species.We did not find a change in snout-vent-length, leg length, or body mass in I. guentheri.Delgado-Acevedo and Restrepo (2008) did not provide an explanation for their observations but laboratory studies in Eleutherodactylus coqui showed that froglets from eggs that hatched early had relatively smaller legs than froglets in the control group (Buckley et al. 2005).More studies on directly developing frogs in fragmented landscapes are required before a clear pattern on morphological changes can emerge and mechanistic explanations can be derived.
The results for body condition suggest that the quality of the terrestrial habitat was not negatively affected for the habitat generalist R. ornata despite body mass being significantly lower in isolated fragments.For the habitat specialist I. guentheri habitat quality seems to have been negatively affected in one but not the other isolated fragment.This suggests that the study sites differed in food resource quality or physiological stress that translate into energetic costs for the latter species in some fragments but not in others.While fragmentation of tropical forests has a large effect on species composition and abundance in invertebrates (Didham et al. 1998), the majority of anurans are rather opportunistic in their food choice (Toft 1985).However, the study of Didham et al. (1998) suggested that in tropical forests common invertebrate species seem to be more sensitive to fragmentation than rarer species; and edges alter the trophic structure of communities (Laurence et al. 2011).Whether these changes translate into different overall abundances of invertebrates that are suitable as food for medium sized frog species, whether abundance changes differs among fragments and may depend on the matrix, and whether these changes differ for habitat generalists and specialists remain to be studied.
Edge effects on abiotic conditions, such as an increase in temperature and a decrease of humidity, affect a larger fraction of a small than a large patch or continuous forests (Saunders et al. 1991).Habitat generalists that are matrix tolerant may be better adapted to deal physiologically with such changes than forest habitat specialists that are intolerant to the matrix.This may explain why in our study body condition decreased in one isolated fragment in the habitat specialist I. guentheri whereas it was not affected in the matrix tolerant R. ornata.Similarly, Scheele et al. (2014) explained the higher body condition in the yellow-bellied toad (Bombina variegata) in forest ponds compared to ponds in pastures by a combination of higher resource availability, more humid micro-climatic conditions and greater water availability.
In conclusion, the effects of fragmentation on size and body condition differed between the habitat generalist and the habitat specialist.This suggests that the two species are affected by different processes driving morphological shifts in the wake of habitat fragmentation, and that for specialists these processes may differ among fragments.The observation that in the same study region as ours habitat specialists among small mammals showed an increase in body condition with fragmentation (Püttker et al. 2008) further corroborates that fragmentation may affect ecological processes that result in unexpected morphological shifts in species-specific ways.Further studies of morphological shifts of small vertebrates in fragmented landscapes and the processes that drive such shifts are required to improve our understanding of subtle effects of habitat fragmentation that in the long-run may lead to different levels of species sensitivity to extinction.

Table 1 .
Mean ± one standard deviation (SD) of morphological traits of the studied species in the four study sites.Abbreviations: control = control site, connect = connected site, iso1 and iso2 = isolated site one and two; BCI: body condition; N: sample size; SVL: snout-vent-length; TL: tibia length.[BCIcalculated as SMI]