Local adaptation and divergence in colour signal conspicuousness between monomorphic and polymorphic lineages in a lizard

Study system and reflectance measurements

The tawny dragon, Ctenophorus decresii, is a small (≤30 g), agamid lizard endemic to central South Australia and Kangaroo Island. We collected colour data from six sites within the polymorphic northern lineage of C. decresii (Aroona, Yourambulla Caves, Devil's Peak, Mt Remarkable, Telowie Gorge and Bimbowrie Station), and two sites in the monomorphic southern lineage (Morialta and Snake Lagoon; Fig. 1e) during late Spring and Summer in 2011 and 2012. Male lizards were captured when active, by hand or by noosing with a loop attached to a telescopic pole, and spectral reflectance measurements were taken within ten minutes of capture to ensure that lizards were at close to optimal body temperatures, as some agamid lizards are darker when cold (Gibbons & Lillywhite, 1981; Cooper & Greenberg, 1992). We measured reflectance of the throat, head, dorsum, neck and lateral stripe (Fig. 2) of 46 males (35 northern, 11 southern) using an Ocean Optics (Dunedin, FL, USA) Jaz portable spectrometer. The probe was held at a constant angle (45°) and distance from the surface of the lizard, and each measurement was expressed relative to a Spectralon 99% white reflectance standard (Labsphere, Inc., North Sutton, NH, USA). Three readings were collected and averaged for each colour patch, and all lizards were released at the site of capture following processing.

To quantify background colour, we measured the reflectance of rocks and lichens on which lizards were observed. We identified different substrate colours (e.g. grey rock, red rock) and again three readings were collected for each patch of colour, which we averaged where appropriate. This resulted in 1–2 samples of each background colour within each site. In the northern lineage rock colours are highly variable, ranging from dark grey to pale yellow; however, the principal rock colours are orange and red-brown scattered with pale green lichen (Fig. 3a,b). Furthermore, red-brown and orange rocks are not present in the southern lineage (Fig. S1b) where the principal rock colours are grey and pink, and the lichen is predominantly orange (Fig. 3c,d). Within each site, lizards were observed on a wide range of rock colours with varying proportions of lichen cover (ranging from 0–90%), and rock colour was variable within an individual's territory (C. A. McLean pers. obs.). A detailed study of microhabitat preference in polymorphic northern populations of C. decresii showed that male colour morphs do not differ in microhabitat selection, and choose territories with less vegetation cover and more large rocks relative to available habitat rather than based on substrate colour (Teasdale et al., 2013). Due to these findings, and because we were interested in how differences in the predominant background colours for each lineage may have influenced the evolution of male coloration, we determined the mean reflectance of orange rock (N = 10) and green lichen (N = 6) across all northern sites, and grey rocks (N = 7) and orange lichen (N = 2) across all southern sites (Fig. S1b). Additionally, we recorded irradiance in full sun under fine conditions, at approximately 11:00 hours within each site using a dedicated Ocean Optics Jaz-ULM-200 spectrometer with a cosine-corrected probe (Fig. S1c). For analysis, spectral reflectance data were smoothed by averaging over each 5 nm interval within the range 300–700 nm, the approximate visual spectrum of birds and most diurnal lizards (Vorobyev et al., 1998; Loew et al., 2002), and the mean irradiance across all sites was normalized to a maximum of one.

Visual models

We evaluated the contrast of throats and lateral coloration (neck and stripe) against backgrounds, and the internal contrast between throat colours, as perceived by a conspecific lizard. We also assessed the conspicuousness of dorsal coloration (head and dorsum) and lateral coloration (neck and stripe) against backgrounds as perceived by potential avian predators. We applied the model of Vorobyev & Osorio (1998), using calculations detailed in Siddiqi et al. (2004), to determine the ‘distance’ between two colours in units of just noticeable differences (JND), where one JND is the threshold of discrimination and the greater the value, the more discriminable the colours. Accordingly, a JND value of greater than one indicates that two colours can be discriminated, whereas colours are indistinguishable if they have a contrast value of less than one JND (based on the visual system of the receiver and the light environment). In general, JND values between one and three mean that two colours are unlikely to be discriminated, but values below six JNDs may also be difficult to discriminate in complex visual environments (Siddiqi et al., 2004). Lighting conditions are incorporated into calculations of JNDs and changes in ambient lighting will affect estimated values; however, the relative differences in JNDs of lizard colours against backgrounds are very unlikely to change with the relatively minor changes in the shape of irradiance spectra in terrestrial environments.

The visual model estimates chromatic (colour) contrast (ΔS) based on the four single cones (ultraviolet/violet sensitive (UVS/VS), short wavelength sensitive (SWS), medium wavelength sensitive (MWS), and long wavelength sensitive (LWS)) and achromatic (luminance) contrast (f_D) based on the double cone, which is thought to be responsible for detecting differences in signal luminance for both lizards and birds (Campenhausen & Kirschfeld, 1998; Hart, 2001a; Osorio & Vorobyev, 2005). Full details of model calculations for C. decresii based on spectral sensitivities of the closely related ornate dragon, Ctenophorus ornatus (Barbour et al., 2002), and for C. decresii's main bird predators, are detailed in Teasdale et al. (2013; see also Fig. S1). Preliminary data on spectral sensitivities from microspectrophotometry for both the northern and southern lineages of C. decresii show no differences between lineages (supported by opsin gene expression data), and confirm that spectral sensitivities are nearly identical to those of C. ornatus (M. Yewers, B. Knott, C. McLean, A. Moussalli, A. T. D. Bennet and D. Stuart-Fox, unpublished data). This is consistent with the phylogenetic conservatism of visual systems evident within lizard families (Olsson et al., 2013). Consequently, we assumed that the two lineages of C. decresii shared the same visual system. Common predators of C. decresii include the little crow, Corvus bennetti, and the grey butcherbird, Cracticus torquatus, which have ultraviolet sensitive (UVS) visual systems, and the Australian kestrel, Falco cenchroides, and the laughing kookaburra, Dacelo novaeguineae, which have violet sensitive (VS) visual systems (Gibbons & Lillywhite, 1981; Stuart-Fox et al., 2003; Hart & Hunt, 2007). Therefore, we performed analyses based on the spectral sensitivities of both avian visual systems, including the effects of coloured oil droplets associated with the different cones (Endler & Mielke, 2005). Results were qualitatively the same for both avian visual systems so we only report UVS results here.

Statistical analysis

We calculated the visual contrasts (ΔS and f_D) of each male colour patch (throat, head, dorsum, neck and stripe) against lineage-specific mean backgrounds (southern grey rock and orange lichen, northern orange rock and green lichen), and the visual contrasts between throat colours. For statistical analysis, we used the contrasts of blue, orange and yellow throat colours as these occur in discrete colour patches, and combined cream and grey contrasts (in 50 : 50 proportions) to account for the fine patterning consistently observed in individuals of the grey morph (Teasdale et al., 2013; Fig. 1d). A measure of dorsal conspicuousness was derived by combining contrasts of the head and dorsum for each individual, weighted by the proportion of the total dorsal surface that each patch occupied, which were 30% head and 70% dorsum (excluding limbs and tail; Fig. 2).

We used generalized linear models (GLMs, SAS 9.3; PROC GLM) to compare the conspicuousness of southern and northern males against native and non-native backgrounds, and the internal contrasts between constituent throat colour patches of each morph. Chromatic and achromatic contrast values were the dependent variables, and throat colour or lineage (for dorsal and lateral coloration) and background, and their interaction, were the independent variables in the model. We performed post hoc pairwise comparisons and applied false discovery rate (FDR) correction for multiple tests to assess differences between morphs/lineages and backgrounds.

Conspicuousness of throat coloration to conspecific lizards

From the perspective of a lizard, throat colours differed significantly in their chromatic and achromatic contrast against rock and lichen backgrounds, and there was a significant interaction between throat colour and background (Table 1). Not all possible pairwise comparisons (between all morphs on all backgrounds) were considered as only those assessing whether there are differences in conspicuousness for each colour morph on their native versus non-native background were of relevance to the initial hypotheses. With the exception of the grey morph, these comparisons revealed that throats were more chromatically conspicuous against native than non-native lichen backgrounds but less chromatically conspicuous against native than non-native rock backgrounds (Fig. 4a,b, respectively; Table S1). Specifically, northern orange and yellow throats were more chromatically conspicuous against green lichen (native) than orange lichen (non-native; P < 0.0001 and P = 0.008, respectively; Fig. 4a; Table S1), and southern blue throats were more chromatically conspicuous against orange lichen (native) than green lichen (non-native; P < 0.0001; Fig. 4a; Table S1). Conversely, northern orange and yellow throats were less conspicuous against orange rock (native) than grey rock (non-native; P < 0.0001; Fig. 4b; Table S1), and southern blue throats were less chromatically conspicuous against grey rock (native) than orange rock (non-native; P = 0.001; Fig. 4b; Table S1). All throat colours, irrespective of lineage, had a greater achromatic contrast against southern backgrounds than against northern backgrounds (P < 0.01 for all pairwise comparisons; Fig. 4c,d; Table S1); however, there was no significant difference in the achromatic contrast of southern blue throats against lichen backgrounds (Fig. 4c; Table S1).

Throat colour morphs differed significantly in the internal contrast between constituent throat colour patches (chromatic: F_{4, 77} = 15.33, P < 0.0001, achromatic: F_{4, 77} = 10.87, P < 0.0001). The chromatic contrasts between the primary and secondary colours of southern males (blue and yellow) and the orange morph (orange and cream) were significantly greater than the orange-yellow, yellow and grey morphs, whereas the achromatic contrast between the two colours of the grey morph (grey and cream) was significantly greater than all other morphs (P < 0.01 for all pairwise comparisons; Fig. 5).

Conspicuousness of lateral coloration to lizards and birds

Neck coloration of the two lineages differed significantly in chromatic and achromatic contrast against rock and lichen backgrounds (Table 2). There was a significant interaction between lineage and background, as perceived by both lizards and birds (Table 2), with the neck coloration of both lineages matching native lichen more closely than non-native lichen (Fig. 6a,c; Table S2). Specifically, the neck coloration of northern males was significantly less chromatically (bird: P < 0.006) and achromatically (bird: P < 0.0001, lizard: P < 0.0001) conspicuous against native green lichen than against non-native orange lichen (Table S2), whereas the neck coloration of southern males was less chromatically conspicuous against native orange lichen than non-native green lichen (bird: P = 0.0004, lizard: P < 0.0001; Table S2).

With the exception of the chromatic contrast against grey rock, the black lateral stripe contrasted highly against all other backgrounds (chromatic contrast JNDs > 3; achromatic contrast JNDs > 10). There was no difference between lineages in the chromatic contrast of the lateral stripe (Table 2; Fig. S2a). However, the darker black lateral stripe of southern males had higher achromatic contrast against all backgrounds than the lateral stripe of northern males (Table 2; Fig. S2b). Furthermore, the lateral stripe of both lineages was more chromatically contrasting against orange lichen (Fig. S2a) and more achromatically contrasting against northern backgrounds (green lichen and orange rock) than against southern backgrounds (orange lichen and grey rock; P < 0.0001 for all pairwise comparisons; Fig. S2b).

Dorsal conspicuousness to avian predators

To the visual system of a bird, there was no difference in the chromatic contrast of dorsal coloration between lineages (Table 3), although both lineages contrasted more against orange lichen than other backgrounds (Fig. S3a). Southern males had higher achromatic contrast (i.e. were more conspicuous) against all backgrounds than northern males (Table 3; P < 0.0001 for all pairwise comparisons; Fig. S3b) but both lineages were more achromatically contrasting against northern backgrounds (green lichen and orange rock) than against southern backgrounds (orange lichen and grey rock; P < 0.0001 for all pairwise comparisons; Table S3). Consequently, both southern and northern males had similar absolute achromatic contrast against their native backgrounds (Fig. 6d, S3b).

Conspicuousness of social signals

Lineage-specific throat coloration was generally very conspicuous (high JND values) to lizards, particularly against native lichens which grow patchily on rocks throughout the region. The highest chromatic contrast was between southern blue throats and their native orange lichen, which can be extensive on rocks in southern populations (Fig. 3c), and is almost absent in northern populations. By contrast, the orange and yellow throat coloration present in northern males may be difficult to distinguish from southern backgrounds because they more closely resemble the colour of orange lichen, even though they are conspicuous against southern grey rock. Additionally, southern throat coloration had a large internal chromatic contrast (between blue and yellow), which can increase conspicuousness at short viewing distances (e.g. Marshall, 2000; Bohlin et al., 2008), and conspecifics are expected to view each other at relatively close range. Yellow also had the greatest achromatic contrast with both grey rock and orange lichen. Therefore, blue and yellow together, which constitute southern throat coloration, may provide the most effective overall throat colour signal (compared to northern throat colours) against the combination of background colours present in southern sites.

Given that the northern lineage of C. decresii is polymorphic, selection for conspicuous throat signals may vary among morph types in association with correlated behavioural, morphological or life-history traits. In the northern lineage of C. decresii, orange males display higher aggression towards model intruders than the three other morphs (M. Yewers & D. Stuart-Fox, unpublished data). Therefore, this dominant behavioural strategy may necessitate a more conspicuous throat colour signal. Consistent with this hypothesis, of the northern throat colours, orange was the most chromatically conspicuous against the predominant native orange rock, but was also very conspicuous against southern grey rocks. Rock colour is particularly variable in the northern lineage and includes grey rocks as well as yellow, orange and reddish-brown. Consequently, orange may be the most conspicuous throat colour against the full spectrum of rock colours observed in northern populations, as well as against the native pale green lichen that finely speckles many of the northern rocks (Fig. 3a).

The relative conspicuousness of C. decresii signals is dependent on the background against which they are viewed (particularly rock versus lichen). Here, we focused on the most abundant rock colours in the two lineages: grey rock in the south and orange rock in the north (which also happened to be unique to the northern lineage); however, there are other rock colours present in these lineages. Therefore, males could potentially select substrates that maximize colour contrast during courtship and territory defence and maximize crypsis at other times (e.g. when basking). Sophisticated choice of display location, or modification of visual backgrounds to maximize conspicuousness has been demonstrated in multiple species (e.g. Endler & Théry, 1996; Heindl & Winkler, 2006). For example, golden-collared manikins, Manacus vitellinus, remove leaf litter and perform courtship displays on cleared earth to increase plumage conspicuousness (Uy & Endler, 2004). In C. decresii, however, throat colour morphs of the northern lineage do not differ in microhabitat preferences (Teasdale et al., 2013; M. Yewers & D. Stuart-Fox, unpublished data). Extensive field observations suggest that perching rocks are selected according to their size, presence of a suitable crevice and position (e.g. in relation to other territories and obscuring vegetation; C. McLean pers. obs.; Teasdale et al., 2013) rather than by colour; suggesting that choice of display location to maximize colour contrast is unlikely in this species. Additionally, C. decresii colour signals may be viewed from different angles, which will alter the background against which they are viewed (e.g. green vegetation or blue sky). Despite this, C. decresii is a rocky habitat specialist and will thus be predominately viewed adjacent to (if not directly against) rock and lichen backgrounds. Consequently, comparing the contrasts of lizards against the most abundant substrate colours seems appropriate in this system.

Lateral coloration (cream-orange neck and black stripe; Fig. 2) may also be a social signal as it is displayed during contests and courtship and is more pronounced in males than in females. Therefore, we expected lineage-specific lateral coloration to be shaped by both sexual selection and predation as it is a social signal which is located on a body region visible to predators. We found that neck coloration was conspicuous against rocks to both lizards and birds, but not against native lichen. Furthermore, lateral coloration is more extensive in southern males than in northern males (McLean et al., 2013). This may be because sexual selection on male traits (including coloration) can vary geographically. For example, females favour bright male dorsal coloration in some populations of the chuckwalla, Sauromalus obesus, but not in others (Kwiatkowski & Sullivan, 2002). In the polymorphic northern lineage of C. decresii, sexual selection may vary in relation to throat coloration and/or alternative behavioural strategies adopted by the different colour morphs (e.g. Thompson & Moore, 1991; Sinervo & Lively, 1996; Huyghe et al., 2009; Olsson et al., 2009, M. Yewers & D. Stuart-Fox, unpublished). Conversely, given that the southern lineage is not variable for throat coloration (i.e. is monomorphic), selection for more intense and extensive lateral coloration may be stronger than in the northern lineage.

Although cream-orange neck coloration more closely matched native than non-native lichen, it contrasted highly with the conspicuous black lateral stripe. Therefore, an interesting possibility is that lateral coloration is both a sexually selected signal and contributes to disruptive camouflage, by breaking up the lizard's outline. Disruptive camouflage is most effective when colours are highly contrasting, located at the body's margins and when one of the colours matches elements of the background (Stevens & Merilaita, 2009). In accordance with this, when viewed from above, the high-contrast black lateral stripe and cream-orange neck coloration of C. decresii occur at the body's margins, and the neck coloration is a relatively close match to the native lichen.

Dorsal conspicuousness to avian predators

Dorsal coloration (head and dorsum; Fig. 2) was chromatically similar between lineages and closely matched grey rock (low JNDs). Southern males were more achromatically conspicuous against all backgrounds than northern males; however, the dorsal coloration of both lineages contrasted more against northern background colours (orange rock and green lichen). Consequently, the absolute achromatic contrast of southern and northern males was comparable. In other words, divergence between the lineages has resulted in similar conspicuousness of dorsal coloration against the native background in each lineage (Fig. S3). Given that luminance (achromatic) signals can be particularly important for prey detection (Osorio et al., 1999; Hart, 2001b), our results suggest local background coloration has constrained the evolution of dorsal coloration in C. decresii. Nevertheless, the dorsal coloration of southern males did not maximize crypsis as they were more conspicuous than northern males against southern backgrounds (grey rock and orange lichen). Therefore, factors other than predation risk may affect dorsal coloration, in particular, dorsal coloration may also be influenced by sexual selection (e.g. due to correlated evolution with sexual signals) or by geographic variation in predator assemblages and/or abundance (e.g. Macedonia et al., 2002; Kwiatkowski, 2003). Furthermore, conspicuousness does not necessarily equate to higher predation risk, as individuals can compensate for increased detectability by altering their behaviour (e.g. Hedrick, 2000; Cabido et al., 2009; Fowler-Finn & Hebets, 2011), and it is currently not known whether the lineages differ in their response to predators.