Introduction

Spatial variation in environmental factors can lead to changes in the most optimal phenotype within the same species (Calsbeek et al. 2007; Kaliontzopoulou et al. 2010; Bolnick et al. 2011). Such intraspecific variation acts on many phenotypic traits, including behaviour (Herczeg and Välimäki 2011) and morphology (Calsbeek and Smith 2007). Intraspecific variation can be driven by both sexual and natural selection (Biro et al. 2005), and has been used to explore environmental factors driving the function of traits, as well as the evolutionary processes shaping them (Kaliontzopoulou et al. 2010; Ramos and Peters 2017a). Understanding intraspecific variation has recently gained attention as a potential tool to manage species against anthropogenic threatening processes by considering population level differences in tolerance to different threats (Kelly and Phillips 2016; Herrando-Pérez et al. 2019). This indicates the need for continued investigation in this field, particularly with respect to taxa that are poorly understood.

Behaviour can be highly plastic and may be shaped by a wide range of biotic and abiotic factors, such as predation pressure (Herczeg and Välimäki 2011), food availability (Biro et al. 2005), climate (Dell et al. 2011), resource competition (Herczeg et al. 2012), and habitat (Ramos and Peters 2017a). Understanding which environmental factors are driving behavioural change in a species is complicated by the fact that one environmental factor may result in numerous behavioural shifts (Magurran and Seghers 1991); conversely, many environmental factors may manifest themselves in only one behavioural adaptation (Herczeg and Välimäki 2011). Manipulating the environment (e.g. modifying the surrounding vegetation or environmental conditions) to determine the effect of different levels of variation is impractical given the number of relevant variables and the extent of their influence (although see Bian et al. 2018), and changes in phenotypic traits may take a long time to become detectible. Intraspecific variation represents a natural experiment in this regard, as one can determine how behaviour might change between populations due to selection and plasticity, and then place them in the context of the differing environment.

Intraspecific resource competition is an important behavioural driver in a range of taxa (reviewed by Svanbäck and Bolnick 2007). This often leads to an increase in intraspecific aggression, which manifests as male–male competition for resources and/or mates (Hews 1993; Peters et al. 2016). Such competition has also been observed in females, particularly in more resource limited environments (Wu et al. 2018). Increased body size is a common morphological adaptation to increased aggression and territoriality (Calsbeek and Smith 2007), with larger individuals out-competing smaller conspecifics (Persson 1985), which can also result in demographic changes (Howard 1981; Johnston 2011). Physiological and morphological adaptations represent a clear selective advantage to maintaining a territory. However, violent interactions are costly to all combatants, leading many taxa to develop non-violent behavioural displays to mediate physical aggression (Lailvaux and Irschick 2007). Such signals may seek to either advertise ownership of a territory or project a message designed to showcase fitness in order to avoid violent exchange (Lailvaux and Irschick 2007).

Territorial signalling is constrained by physiology, environmental factors and the sensory systems of receivers. Physiological and morphological factors, such as vocal amplitude (Podos 2001) or dewlap length (Ord et al. 2015), pose a physical limit to signal projection, while factors such as environmental background noise may mask or distort the signal (Peters et al. 2008; Steinberg et al. 2014; Számadó and Hemmi 2015; Ramos and Peters 2017a). Effective signalling requires detectable signals. However, increasing conspicuousness against the environmental background may incur an increased risk of predation, which may constrain signal contrast (Gibbons and Lillywhite 1981; Stuart-Fox et al. 2003). These factors can lead to rapid and high diversification amongst signal strategies (Tinghitella 2008).

Lizards are excellent models for understanding the evolution of both colour- and movement-based visual signalling strategies (Fleishman 1988; Nicholson et al. 2007; Fleishman et al. 2011, 2017; Peters et al. 2016; Ramos and Peters 2016). Movement from wind-blown plants has been documented to be an important source of environmental noise that may impact lizard movement-based communication (Peters et al. 2008), leading to alterations in speeds, durations and amplitudes of motion signals (Peters et al. 2007; Ord and Stamps 2008; Ramos and Peters 2017a). Different vegetation densities, as well as different plant structures, respond differently to wind (Peters et al. 2008), which in turn alters the environment through which signals must be projected (Peters et al. 2007; Ramos and Peters 2017a). This has the potential to drive significant intraspecific variation in signal structure between species widely distributed across a varying landscape.

Species that range over a wide area or that display allopatric distribution patterns often exhibit interpopulation morphological variation (Stock et al. 2000; Calsbeek et al. 2007). This can be a result of genetic drift or of local adaptation to novel environments (Calsbeek et al. 2007; Calderón-Espinosa et al. 2013). Such species are informative as morphological variation between populations suggests the possibility of further phenotypic or ecological differentiation. One such species is the peninsula dragon (Ctenophorus fionni) (Gibbons and Lillywhite 1981; Stuart‐Fox et al. 2004). This small agamid lizard is endemic to the Eyre Peninsula in South Australia (Fig. 1), and is well known for its territoriality and sexual dimorphism (Stuart‐Fox et al. 2004). Males are usually bigger and much more conspicuously coloured than females, which have a more cryptic colouration that tends to match their local substrate well (Fig. 1) (Stuart‐Fox et al. 2004). This species also displays a high degree of geographic variation in colour (Fig. 1a–c) (Stuart‐Fox et al. 2004), body size (Houston 1974) and territorial display behaviour (Gibbons 1979). Variation in predation rate has been suggested as the main driver of this polymorphism (Gibbons and Lillywhite 1981; Stuart‐Fox et al. 2004). Furthermore, the chemical compositions of the pheromones that are used in social communication by this species vary between populations (La Nafie et al. 1995). Thus, evidence suggests that this species is undergoing substantial divergence between its three populations. Interestingly, we know very little about interpopulation variation in ecology, physiology and behaviour in general. As with other members of the Ctenophorus decresii complex, this species is known to engage in intricate territorial interactions, including some of the most elaborate movement-based signalling performed by any Australian reptile (Gibbons 1979; Ord et al. 2001; Ramos and Peters 2016). The territorial displays performed by this species are composed of three main sequential motor patterns beginning with limb waving, followed by hind leg push ups coupled with tail coiling, and ending with head bobbing (Gibbons 1979). Signal diversification has been shown to generally occur rapidly under selection (Tinghitella 2008), so it is probable that the communicative behaviour of C. fionni has also diverged between the different populations.

Fig. 1.

Occurrence of Ctenophorus fionni (grey dots) on the Eyre Peninsula in South Australia, and study sites (arrows): (a) northern population, (b) central population, (c) southern population. The dramatic variation in colouration between populations is illustrated in the corresponding right hand side panels. Distribution occurrence data were sourced from the Atlas of Living Australia (2020).

The objective of this study was to gather further data on the extent to which the three populations of C. fionni differ in aspects of behaviour and morphology (other than colouration), and to place these differences in the context of the environments they inhabit. We addressed four specific aims:

We expected to find substantial variation in substrate and vegetation between the three sampled populations due to known differences in geology and climatic factors. It was unclear how population densities of C. fionni may differ between sites; however, studies have shown that variation in density and abundance of lizard species are often correlated with habitat variation (Monasterio et al. 2010). Intraspecific variation in morphometric traits has been observed to correlate with differences in habitat and microhabitat use in a range of lizard species (e.g. many species of Caribbean anoles: Calsbeek et al. 2007), so we also expected population level variation in the morphology of C. fionni. In view of anticipated climate and habitat differences between the three populations, we also expected variation in the proportion of time lizards engaged in different behaviours. Social interactions are also likely to be influenced by the population density of C. fionni (Stamps 1977). Communication strategies are constrained by the environment in which they are used, with wind-blown vegetation known to cause changes to the structure of signal displays in a range of social lizards (Peters et al. 2008; Ramos and Peters 2017a). We hypothesised that signal structure would vary between populations in a manner consistent with differences in vegetation composition. More specifically, we expected populations exposed to noisier environments to produce faster display movements. Although we did not measure motion noise, we expected the coastal habitat of Lincoln National Park (NP) would provide a noisier environment and lizards therein would produce faster displays (Ord et al. 2007).

Materials and methods

Data collection

Environmental variation was examined by quantifying habitat structure and substrate composition. Relative abundances of C. fionni at the northern, central and southern populations were estimated from encounter rates following standardised search protocols. To quantify behaviour, we undertook focal sampling of multiple individuals, behavioural sampling to document rare but functionally important events, and tethered introductions of ‘intruder’ males to resident males to film communicative displays. Morphometric data were also collected from all three populations.

Habitat structure

For each of the three study populations, 10 quadrats were selected using ArcGIS global imaging software. The coordinates of the quadrats were selected by overlaying a grid onto the study site and selecting 10 evenly spaced 10 m² quadrats to ensure even coverage across the site (selected coordinates indicate the north-east corner of the quadrat). Vegetation structure was assessed by subdividing each quadrat into 1 m² cells and using a touch pole placed at the north-west corner of every cell. At each point we recorded the substrate composition as solid rock, loose rock, plant matter or bare ground. Vegetation data were recorded at 0.5 m height increments to a height of 2 m, with vegetation higher than 2 m categorised as a tree. Plant species were grouped into five broad categories based on life form: herb and forb, grass, spinifex (Triodia sp.), shrubs and woody plants. Substrate and vegetation data were analysed separately but using the same analytical approach. An analysis of similarity (ANOSIM) was performed to determine whether these sites varied statistically from each other. A multidimension scaling (MDS) plot using Bray–Curtis similarities was then used to identify discriminating factors between sites. All statistical analyses were performed using PRIMER ver. 7. Solid rock cover and presence of small shrubs (<0.5 m) are considered important ecological factors for this species so an analysis of variance (ANOVA) was performed to specifically look at site variation in these factors.

Encounter rates

We adapted a standard method for sampling birds formulated by Loyn (1985) to sample lizards. The method entailed two people actively searching the study site for the target species for 1 h. As the study sites varied in climate, it was decided that instead of standardising by time of day or temperature, the survey would begin once two lizards had been found within 20 min of each other, as this would indicate that conditions were suitable for lizard activity. Once these two lizards were seen, the two surveyors walked haphazardly across the site as a pair and recorded the sex (male, female or juvenile) and GPS coordinates of every lizard observed. Two surveys were conducted at each study site over consecutive days under similar environmental conditions, with the highest count for males and the highest count for females from either day used in subsequent analyses. We calculated the distance between lizards across the landscape for each site and compared these data between sites using the permutational dispersion (PERMDISP) routine in the permutational analysis of variance (PERMANOVA) add-on, in PRIMER ver. 7 (Clarke and Gorely 2015; Anderson 2017). We examined overall significance as well as pairwise comparisons between sites.

Morphometric measurements

Morphometric measurements were taken for adult lizards of both sexes (sample sizes are provided in the results). Lizards were caught using a noose, made from fishing line and attached to a pole, and measured after being observed or filmed as part of the behavioural and display analyses (see below). An effort was made to catch even numbers of both sexes, but it was not always possible. Weight was measured using Pesola scales (to the nearest 0.5 g). Digital callipers were used to measure snout–vent length (SVL), head length, head height, head width, interlimb, femur, tibia and metacarpal of the rear limb, humerus and radius of the forelimb (to the nearest 0.01 mm). All measurements, except head width, were taken on the left side of the lizard. To determine how the morphology varied between populations, we regressed all morphometric measurements by SVL to remove the effect of body size and performed a multivariate analysis of variance (MANOVA) on the residuals, followed by pairwise comparisons to determine which populations showed divergence. In addition, SVL was analysed separately using an ANOVA. All morphological data analyses were performed in the R statistical environment (R Core Team 2018). Male and female morphometric data were considered separately.

Behavioural analysis

A partial ethogram of the behavioural repertoire of C. fionni was developed based on the literature (Gibbons 1979), pilot observations and discussions with herpetologists (Table S1,   Supplementary Material (ZO20091_AC.zip)). Focal sampling (focus on one individual per session: Martin and Bateson 2007) was undertaken for a period of 2 h, or until the lizard was lost from view. Continuous recording of activities for this period can be difficult so we used instantaneous time-sampling with a 2 min interval (Martin and Bateson 2007) to record the frequency of occurrence of each behaviour across the 2 h session. We observed three females at each site, while six males were observed at Gawler Ranges NP and four each at Pimba and Lincoln NP. At the start of each session, we recorded the date and time along with estimates for percentage cloud cover and distance to the nearest conspecific. Temperature was recorded every 10 min using a Kestrel 4000 anemometer (Kestrel Meters), which was also used to record the wind speed at the start of the session. Focal sampling data (frequencies) were converted to proportions and analysed separately for each behaviour using the BETAREG function from the BETAREG package (Cribari-Neto and Zeileis 2010) in the R statistical environment. Individuals that did not express a given behaviour were assigned a value of zero and the median temperature for the session. We fitted the models with predictor variables of population, sex and temperature, and transformed the dependent variable to account for absolute zero values, using the formula

where n is the sample size, p is the raw proportion and P is the transformed proportion (Smithson and Verkuilen 2006). We examined the coefficients and corresponding z-scores for sex and temperature, and for pairwise comparisons for population. Estimated marginal means were generated from the model and plotted for behaviours that yielded a significant effect. The occurrence of rare behaviours is often missed, or misrepresented when using time intervals, therefore behaviour sampling (frequency) was undertaken concurrently with focal sampling. A separate observer recorded any rare behaviours seen in the general population (including the focal lizard), including territorial push up displays, non-aggressive social interactions and predator avoidance. We noted the time of occurrence as well as ambient temperature. Behaviour sampling data were considered separately for males and females and analysed as count data in the R statistical environment (R Core Team 2018) using GLMER from the lme4 package (Bates et al. 2015) and fitting a Poisson error distribution. We used population as a fixed effect and session as a random effect to account for multiple observation sessions with the same population.

Display data collection

Films of male territorial displays were collected in the field at Gawler Ranges NP and Lincoln NP. While effort was made to record displays from Pimba, extreme temperatures coupled with a scarcity of male lizards thwarted this attempt. Displays were elicited by introducing a tethered male via a fishing pole to a free-living male. This lizard was caught from the corresponding populations before commencing filming. To record displays, we followed the protocols of Peters et al. (2016). Briefly, two Sony RX10 iii cameras were used to simultaneously record the displaying individual at 100 frames per second. After the lizard left the scene, a calibration object consisting of 25 non-coplanar points of known distance in relation to each other was placed at the location of the display and without changing any camera settings. This object was later used as reference to calibrate the cameras and combine their footage in order to reconstruct the motion from the displays in three dimensions, using specialised software (see section below).

Display structure

Territorial displays performed by C. fionni consist of three key sequential motor patterns: limb waving, hind-leg push ups coupled with a tail coil and head bobbing (see Ramos and Peters 2016; Movie S1,   Supplementary Material (ZO20091_AC.zip)). To investigate whether the territorial displays performed by C. fionni varied between populations, we first calculated the duration of each motor pattern within the display (limb wave, push up and head bob) and used an independent samples t-test to compare between populations. We then used 3D reconstruction of displays in Matlab (MathWorks Inc.) to precisely compare movement speeds between populations, using the method outlined by Hedrick (2008) and previously applied to quantifying lizard displays by Peters et al. (2016). Three points were selected on the lizard: the eye, wrist and base of the tail dorsal to the cloacal vent. These points were selected as they represent key components of the motor patterns in the display, and correspond to head bobs, limb waves and hind leg push ups, respectively. All points were manually tracked frame by frame for the course of the full display. The data were then combined along with the calibration coefficient derived from a calibration object to generate a 3D reconstruction of these motor patterns using direct linear transformation to generate x–y–z coordinates of each point in each frame (see Peters et al. 2016 for full details of this approach). Speed was computed separately for each point as the change in position over time. We divided the duration of display sequences into quartiles and obtained an average speed within each quartile for each point. The resultant data were then statistically analysed using a mixed effects model with quartile, population and point as factors and lizards as a random effect. We used the lme function from the nlme package (Pinheiro et al. 2018), where the significance of factors is provided by an F-test.

Results

Environmental variation

Habitat variation

The MDS (Fig. 2a) and ANOSIM (Table 1) for substrate composition showed that Pimba differed from the other two sites, while Gawler Ranges NP and Lincoln NP were not significantly different from each other. A vector overlay used to explore specific environmental factors driving dissimilarities showed solid rock cover as a key factor driving site divergence (Fig. 2a). We explored this further using an ANOVA and found a significant difference across populations (F_2,27 = 9.194, P < 0.001) (Fig. 2a inset), with pairwise comparisons suggesting that Gawler Ranges NP had a significantly higher proportion of rock cover than both other sites (P = 0.025 and P < 0.001 for Lincoln NP and Pimba, respectively) but Pimba and Lincoln NP did not differ from each other (P = 0.344).

Fig. 2.

Summary of the variation in habitat structure between the three study sites: Pimba (orange), Gawler Ranges NP (yellow), and Lincoln NP (blue). (a) MDS plot of the dissimilarity in substrate composition between the three localities with the bubbles sized according to the proportion of solid rock cover recorded at each quadrat. Proportion of solid rock cover is illustrated in the bar graph inset. (b) MDS ordination of the dissimilarities in vegetation structure with vector overlay showing the association of sites with vegetation variables. Variation in the proportion of low vegetation is illustrated in the bar graph inset

Table 1.

ANOSIM output showing the overall significance level and pairwise comparisons between substrate and vegetation compositions across the different sites

Asterisks (*) indicate significant values

The MDS (Fig. 2b) and ANOSIM (Table 1) for vegetation showed that sites differed in vegetation composition and structure, with Gawler Ranges NP most dissimilar to the other two sites. Much of this dissimilarity appears to be explained by Gawler Ranges NP being associated with branching plants and negatively associated with the lower vegetation categories (e.g. shrubs, herbs, dead wood). The clustering of points for Lincoln NP (Fig. 2b) suggests that it had a more homogeneous vegetation covering than Gawler Ranges NP or Pimba. Taken together, our results indicated a less diverse and closed understorey at Pimba and Lincoln NP, in contrast to more open area at Gawler Ranges NP. Further analysis on the presence of small shrubs confirmed differences across sites (F_2,27 = 9.208, P < 0.001) (Fig. 2b inset), with Gawler Ranges NP indeed having less coverage of small shrubs than either of the other sites (P < 0.001 and P = 0.010 for Lincoln NP and Pimba, respectively), but Lincoln NP and Pimba were not significantly different from each other (P = 0.626).

Encounter rates

The maximum number of lizards observed during the timed surveys was substantially higher at Gawler Ranges NP (n = 17: 11 males, 6 females), whereas the number at Lincoln NP (n = 10: 4 males, 4 females, 2 juveniles) and Pimba (n = 8: 0 males, 6 females, 2 juveniles) were more similar. An analysis using the PERMDISP routine in the PERMANOVA add-on for PRIMER revealed significant differences in dispersal (F_1,2 = 16.002, P < 0.001), with lizards at Pimba significantly more dispersed in relation to each other than those from Gawler Ranges NP (t = 4.828, d.f. = 23, P < 0.001) or Lincoln NP (t = 3.370, d.f. = 16, P = 0.004), which were not statistically different from each other (t = 1.039, d.f. = 25, P = 0.309).

Morphological and behavioural variation

Morphometric variation

Overall, significant variation in most morphological traits was detected between populations for both male (Table 2) and female (Table 3) C. fionni. Significant variation in SVL was observed between all populations for both male and female lizards, with individuals from Pimba being the largest and those from Lincoln NP the smallest. Males had significantly larger SVL measurements than females at Gawler Ranges NP and Pimba, but not at Lincoln NP (Lincoln NP: t = 1.372, d.f. = 4.654, P = 0.311; Gawler Ranges NP: t = –3.89, d.f. = 14.825, P = 0.002; Pimba: t = –6.419, d.f. = 2.851, P = 0.009). In regressed morphometric traits (e.g. tail length, head height), the Gawler Ranges NP population was the most divergent from the other two groups, differing in most traits from Lincoln NP lizards for both males and females. Differences between Gawler Ranges NP and Pimba lizards were less consistent (Tables 2 and 3 for males and females, respectively). Gawler Ranges NP males generally had larger morphometric features in relation to body size than the other two populations, whereas Lincoln NP males generally had the smallest morphometric traits in relation to body size. Small sample size made any trends difficult to observe in Pimba, but lizards from this population appear to have longer tails in relation to size. Although most trait variations were found to be consistent between males and females, size trends in females were less clear due to a greater degree of within-population variation.

Table 2.

Results of statistical analysis of variation in morphology of male C. fionni for regressed and non-regressed traits

The overall result (MANOVA) and pairwise comparisons for significant main effects are provided (Pimba (n = 2); Gawler Ranges NP (n = 12); Lincoln NP (n = 5)). Asterisks (*) indicate significant values

Table 3.

Results of statistical analysis of variation in morphology of female C. fionni for regressed and non-regressed traits

The overall result (MANOVA) and pairwise comparisons for significant main effects are provided (Pimba (n = 2); Gawler Ranges NP (n = 12); Lincoln NP (n = 5)). Asterisks (*) represent significant values

Focal and behaviour sampling

From the 16 behaviours defined in the ethogram (Table S1,   Supplementary Material (ZO20091_AC.zip)), only 10 were exhibited by lizards during focal sampling. We focussed our analysis on six of these, selecting those that reflect social interactions (i.e. push up display, head bobbing not part of a display, and limb waving not part of a display) as well as one passive (basking) and active (feeding) behaviour, and scanning (see Table S2 in the   Supplementary Material (ZO20091_AC.zip) for summary data on all behaviours). The β regression analysis (Table 4) revealed significant differences between sexes with respect to feeding, whereby females engaged in the activity more frequently than males across all populations (Fig. 3a). We also found a significant interaction between head bobbing and population, with individuals from Gawler Ranges NP performing head bobs more frequently than Lincoln NP or Pimba (Fig. 3a). Of the behaviours recorded during behavioural sampling, only push up displays occurred frequently enough to warrant further analysis (Fig. 3b). While females were also observed to perform push up displays, they did not display this behaviour frequently enough to analyse statistically. A generalised linear mixed-effects model for male data showed only that Gawler Ranges NP males performed push up displays significantly more frequently than males from Lincoln NP or Pimba (Z = –2.571, P = 0.010 and Z = –2.723, P = 0.006, respectively).

Table 4.

Results of β regression analysis to investigate variation in proportion of behaviours recorded during focal sampling. Z-scores and associated P-values are provided for comparisons between populations, males and females and temperature

Asterisks (*) indicate significant values

Fig. 3.

Behavioural comparison between males (black) and females (white) across study sites. (a) Mean (±95% CI) proportion of time spent feeding and head bobbing during focal sampling, as calculated from β regression analyses. (b) Mean (+s.e.) frequency for push up displays recorded during behaviour sampling and plotted from raw data in order to show values for female lizards

Signalling displays

The displays filmed at Gawler Ranges NP (n = 5) and Lincoln NP (n = 4) both matched the broad stereotyped signal movements outlined by Gibbons (1979). Mean durations for components of the territorial displays (i.e. limb waves, push ups and head bobs) at the two populations are shown in Fig. 4a. Lizards from Gawler Ranges NP spent significantly longer performing limb waves (t = 3.110, d.f. = 11.74, P = 0.009), whereas Lincoln NP males spent significantly longer performing the push up component of the displays (t = 2.687, d.f. = 10.18, P = 0.022). Head bobbing did not differ in duration across the two populations. Three-dimensional reconstruction of displays permitted computation of motion speeds (Fig. 4b) between Lincoln NP and Gawler Ranges NP lizards. Lincoln NP males performed their displays at greater speed than Gawler Ranges NP males for all digitised points across the total duration of the display except during limb waving (Fig. 4b). A GLM model was used to compare speeds between populations, quartiles (time) and points. Table 5 summarises the outcome of the model and indicates that all factors and two-way interactions were significant. Visual inspection of the data revealed that a significant population × point interaction is likely attributable to forelimb movements being equivalent in speed for all populations, whereas Lincoln NP speeds are faster for all other points. With respect to variation across time (population × quartiles), Lincoln NP lizards exhibit a steady decline in speed as the display progresses, while Gawler Ranges NP lizards are more consistent in speed throughout the middle part of the display.

Fig. 4.

Temporal data of discrete components from territorial push up displays: limb waves (LW), push ups (PU) and head bobs (HB). (a) Mean (±s.e.) duration of each discrete component of the display for Gawler Ranges NP (yellow; n = 8) and Lincoln NP (blue; n = 6) lizards. (b) Mean (±s.e.) speed predicted from the linear mixed effects model for Gawler Ranges NP (yellow) and Lincoln NP (blue) lizards, averaged across four evenly spaced time quartiles. Illustrations of each component are also included in the corresponding panel, indicating the point on the lizard that was tracked to calculate speed values (adapted from Ramos and Peters 2016).

Table 5.

Results of the generalised linear model of display speeds with population, time quartiles and tracked points on the animal as predictor variables

Asterisks (*) indicate significant values

Discussion

Intraspecific variation across populations of C. fionni was observed to be related to the environment, behaviour and morphology. Notwithstanding the preliminary nature of the study, we recorded environmental variation between habitats, both in terms of substrate and vegetation structure, and encounter rates of C. fionni. We also found significant differences in lizard size between populations, as well as several other morphometric traits in both sexes. Through focal sampling, we were able to identify significant differences in behavioural activities, most notably in the proportion of time spent head bobbing at Gawler Ranges NP, whereas behavioural sampling revealed a significantly higher frequency of push up displays by Gawler Ranges NP lizards compared with the Lincoln NP and Pimba populations. Furthermore, detailed aspects of signalling behaviour were observed to differ between populations in terms of signal component duration and speed. Overall, these data are consistent with previous studies of variation in the ecology, morphology and behaviour of this species. However, our results explore how this variation is influenced by characteristics of the habitat and population density.

We observed substantial variation in encounter rates of C. fionni between the different sites, with more lizards detected at Gawler Ranges NP. More precise measurements of population size and density are warranted than were possible in the present study, but our preliminary findings are consistent with environmental factors (Monasterio et al. 2010) and competition (Buckley et al. 2008) influencing lizard population size. Previous work on the species reported a population density estimate of 27 lizards per 10 000 m² (Johnston 2000), which appears consistent with our own observations at the Gawler Ranges NP population during the encounter rate surveys. A meta-analysis of population densities of lizards from several different environments found lizard body size to be negatively correlated with population density (Buckley et al. 2008). In our study, the lowest encounter rates were in Pimba, where the lizards are larger. Another explanation for differing population densities is the difference in availability of suitable habitat between sites. The site with the greatest proportion of rock substrate, with which these lizards are typically associated (Gibbons 1979), was also where encounter rates were highest (Gawler Ranges NP). We also recorded higher male abundance at Gawler Ranges NP, which is surprising given that sex ratios are generally predicted to be tightly constrained by sexual competition (Greenwood 1980). In at least one other lizard species exhibiting male-biased sex ratios (Lacerta vivipara), males showed increased aggression towards females leading to increased mortality, which in turn increases the male sex ratio bias (Le Galliard et al. 2005). However, males in the present study were not found to display overt intersexual aggression but it is probably because the observations were made after the mating season (i.e. August–October: Johnston 1999). Other factors that could be influencing the sex bias include population density, which in turn is influenced by crevice availability in this species (Johnston 1997), and also environmental temperature, given that C. fionni exhibits temperature-dependent sex determination (Harlow and Price 2000). Additionally, the more conspicuous colouration that males usually display compared with females may play a role in increasing their detectability at this site. Further consideration of the factors affecting sex ratios is important for wildlife conservation and management (Ginsberg and Milner-Gulland 1994; Wedekind 2002; Robertson et al. 2006), and our observations herein suggest that C. fionni may make a good case study for exploring these factors.

This study found variation in morphology between the three populations of C. fionni. Herpetologists and enthusiasts have long been aware of these morphometric differences; however, this information has previously remained mostly unpublished (Greg Johnston, pers. comm. 2018). Intraspecific morphological variation with respect to differences in habitat use has been documented in several lizard species (Aerts et al. 2000; Herrel et al. 2006; Calsbeek et al. 2007; Brecko et al. 2008; Dollion et al. 2017). Arboreal species are particularly well documented as following consistent morphological variation patterns with respect to preferred microhabitat use, such as limb length increasing with branch diameter (Irschick et al. 2005; Calsbeek et al. 2007). Morphological traits may also be linked to social signalling behaviour, as increased size may generate a greater visual contrast and project a more aggressive or clearer signal (Kotiaho et al. 1999; Bertram 2000). Increased body size has been postulated as a product of territoriality in several species (Persson 1985; Calsbeek et al. 2007). We observed significant differences in overall size between the three populations of C. fionni, but to what extent this is a product of aggression and the need to maintain a territory is unclear. Agonistic interactions were most common at Gawler Ranges NP, and while they were found to have proportionally larger morphometric features in proportion to body size, their overall size was smaller than that of Pimba lizards. Further investigation into the factors driving this observed morphometric divergence is warranted, as our preliminary results clearly show interpopulation morphological variation in C. fionni. Alternative explanations of the observed patterns in morphometric variability in this species are genetic drift, the founder effect or population bottlenecks (Nei et al. 1975; England et al. 2003).

The results of the focal and behaviour sampling found a significant increase in two territorial behaviours, head bobs and push up displays, among the Gawler Ranges NP lizards compared with Lincoln NP and Pimba. We postulate that this might reflect the higher abundance observed at this site, despite population density not being statistically different between all sites. This observation is corroborated by several published accounts of increased population density and abundance leading to an increase in the rate of agonistic interactions (Fletcher 2007; Knell 2009; Cooper et al. 2015). In addition, males were more common at Gawler Ranges NP than at the other two sites, which also may have played a role in this increase in agonistic interactions, as it suggests an increased competition for mates and resources (Le Galliard et al. 2005). We observed feeding rates to be significantly higher in females than males across all populations. This could reflect the need to sustain egg development (Rose 1982), although most females would have laid their eggs by this stage (Johnston 1999). Alternatively, it may be reflective of the males’ need to spend more time actively guarding a territory, and therefore less time feeding (Whitham 1986).

Analysis of territorial displays revealed that males from Lincoln NP and Gawler Ranges NP signalled for approximately the same total duration, but prioritised different components within their display and differed in speed characteristics. Males from Gawler Ranges NP performed the opening limb wave component for significantly longer than their Lincoln NP counterparts, and males from Lincoln NP performed the hindleg push up component for significantly longer than those from Gawler Ranges NP. Both species performed the head bob component for a comparable time. Limb waves generated the fastest movements of the sequence, but slowed down rapidly over time and were equivalent across populations. However, lizards from Lincoln NP produced faster hind leg push ups and head bobs, which was consistent with our hypothesis. Could this variation reflect differences in the signalling environment? If so, it adds to a growing body of scientific work suggesting a strong interaction between the communicative behaviour of an animal and the environmental context in which it is situated (Fleishman 1988; Brumm 2004; Wood and Yezerinac 2006; Peters et al. 2007; Ramos and Peters 2017a). Our analysis of the vegetation data has shown that Lincoln NP is strongly associated with low shrubs and strong winds, which generate considerable background noise at the lizard’s plane of view and so may make detection by a receiver more difficult. With this in mind, male lizards from Lincoln NP may be altering their display to improve signal efficacy. Faster movements have been reported previously in this context (Ord et al. 2007), while adjusting relative use of different motor patterns would be a novel finding.

Divergence in communication behaviour has the potential to lead to speciation if populations can no longer interact effectively with each other (Morton 1975; Leal and Fleishman 2004). Results from this study suggest that diversification in signal structure has occurred between two populations of C. fionni. Within-species variation in display structure can be a plastic response to the signalling conditions (West-Eberhard 1989; Peters et al. 2007; Ramos and Peters 2017b). However, some genetic and colour variation has been recorded between these populations, and our results have provided evidence supporting substantial morphological variation. The genetic basis for movement-based signalling is not known for any lizard species, but it is intriguing to consider the possibility of a genetic basis for display variation. Even if signal variation does have a genetic determinant, it may not preclude effective communication as the core stereotyped display (Gibbons 1979) is retained between populations. Barquero et al. (2015) observed that population variation in signal structure, genetics and morphology did not alter rival recognition or aggression in jacky dragons (Amphibolurus muricatus). Barquero et al. (2015) suggest further that the underlying core signalling behaviour of A. muricatus has a genetic basis, with the observed intraspecific variation being the product of behavioural plasticity to local environments. This may also be true for C. fionni, although A. muricatus does not show the same geographic variation in colour between populations that C. fionni does, which may affect recognition between peninsula dragon populations. Chemical cues are important recognition traits in several lizard species (López et al. 2003, 2006; Baeckens et al. 2015), and chemical analysis of femoral gland secretions of C. fionni demonstrated variability between populations (La Nafie et al. 1995). Additional studies to test whether species recognition has been maintained in the peninsula dragon may give further insight into the extent to which this species is undergoing speciation.

In conclusion, we report substantial variation in habitat, encounter rates, morphology and behaviour between the three populations of C. fionni, as well as in the structure of territorial signalling between Lincoln NP and Gawler Ranges NP. We add to an existing body of evidence suggesting that the three populations have undergone divergence, and habitat differences may be a key driving force. Previous studies have found geographic variation in colour, size and body shape (Houston 1974; Stuart‐Fox et al. 2004) in C. fionni populations, and the current study has recorded significant variation in morphology and behaviour and related it to aspects of their environment and relative abundance. Environmental factors are known drivers of intraspecific phenotypic changes, including behaviour (Magurran and Seghers 1991; Dell et al. 2011; Ramos and Peters 2017a). This is because particular phenotypes are better suited for specific habitats or habitat components. The differences we detected in both substrate type and vegetation between habitats could be driving the development of morphological, behavioural and social structure differences between populations, with the vegetation potentially promoting further behavioural and morphological differentiation as a consequence of population density. Clearly, this avenue of research has implications for conservation and population management, but it is also an important tool for understanding how species are responding to changes in their environment, including anthropogenic processes.

Ethics approval

All work was conducted in accordance with the ethical guidelines of La Trobe University under permit AEC18-022, as well as a scientific research permit issued by the South Australian Government Department of Environment and Water (E26771-2).

Conflicts of interest

The authors declare no conflicts of interest.

Declaration of funding

This project was funded by the Australian Research Council Discovery Project scheme (DP170102370).