Patterned, plain, and in-between: An assessment of ecogeographic divergence between colour pattern morphs of the common egg-eater Dasypeltis scabra

ABSTRACT

Intraspecific variation in colour patterns may reflect adaptive responses to local environmental regimes that favour selection of different trade-offs between visual communication, thermoregulation, and anti-predatory functions. Understanding the drivers of colour pattern variation within species can therefore provide valuable insights about adaptation. However, the ecogeographic correlates, and thus the ecological drivers, associated with the segregation of different colour morphs for most reptile species, including snakes, are not well known. Here, we examined differences in environmental niches associated with occurrence records of preserved museum specimens and georeferenced photographic images of common egg-eaters (Dasypeltis scabra) to characterise ecogeographic divergence between typically patterned, patternless, and faintly marked individuals. We gathered 1707 records that we assigned as either patterned, plain, or intermediate morphs. Most records were of patterned individuals, which were widely distributed in southern Africa. Climate niches differed between morphs as plain and intermediate snakes were restricted to high-elevation areas in the Highveld grasslands of South Africa, and woodland-dominated areas in the central and eastern parts of both Zambia and Zimbabwe, but there were no areas where only plain or intermediate snakes occurred in isolation from patterned snakes. Environmental niche modelling predicted minimal areas suitable for co-habitation of colour morphs. We speculate that plain or weakly-marked colour patterns likely developed as an adaptation to increase camouflage ability in open areas in grassland and certain woodland habitats.

KEYWORDS:

• Dasypeltis
• Colour polymorphism
• Geographic variation
• Serpentes

Introduction

Many studies on the adaptive significance of intraspecific variation in animal colour patterns have been conducted over the last few decades (Endler 1990; McKinnon and Pierotti 2010; Natusch and Lyons 2021). Such research has shown that animal colouration and patterning often facilitate important ecological functions such as intraspecific communication (Stuart-Fox and Moussalli 2009), homeostatic regulation (Lattanzio and Buontempo 2021), and anti-predatory mechanisms (Endler 1981). Thus, animal colour patterns are often essential for survival and reproductive fitness. In cases where species exhibit intraspecific variation in colour traits across different populations, this may potentially reflect localised adaptive responses to divergent selective pressures that favour different colour functions or trade-offs between them (Skulason and Smith 1995; Da Silva et al. 2016; Farquhar et al. 2022). Ecogeographic correlates like divergences in climate, (Perez et al 2018; Delhey et al. 2019), habitat use (Delhey et al. 2023) or ecological factors, like sexual dimorphism or ontogeny (Allen 2013), can each contribute to facilitating the rise and maintenance of colour pattern variation within species. Understanding which ecogeographical factors correlate with specific colour patterns can thus provide important insight into the adaptive processes of species evolution.

Several species of African reptiles exhibit colour pattern polymorphism (e.g. Šmíd et al. 2018; Barends and Scholtz 2023; Bates et al. 2023), and plain or patternless forms have been reported for some typically ornate snake species from various families. Some examples of these include several species of Dasypeltis (De Waal 1978; Bates and Broadley 2018), Naja annulifera (Broadley 1990), Psammophis crucifer, P. brevirostris and P. notostictus (Bates 1992; Branch 1998; Marais 2022), Psammophylax rhombeatus (Bates 1992), and Causus rhombeatus (Broadley and Blaylock 2013; Pietersen et al. 2021). Patternless morphs are also known for some species of typically patterned lizards such as Trachylepis capensis (Bates 1992; Branch 1998). However, little is known about how different colour morphs of these species vary in their geography or ecology. As a result, it is not always obvious if different colour patterns have emerged among these species as adaptive responses to differences in local environments or ecological regimes.

The African snake genus Dasypeltis includes 18 species distributed throughout much of Africa (especially the sub-Saharan region) and parts of southwestern Arabia (Bates et al. 2014; Bates and Broadley 2018; Trape et al. 2021; Bates 2023). All of these species appear to be obligate bird egg specialists that have evolved a host of adaptations to facilitate the consumption of hard-shelled eggs and a mainly liquid-based diet (Gans 1959; Bates and Little 2013; Barends and Maritz 2022). Some of these adaptations include decreased numbers of teeth in the mouth and a slender body form (Gans 1952; Broadley 1990). Consequently, these snakes lack the morphological apparatus required to bite or constrict (Gans 1961; Branch 1998), but avoid or deter potential predators through a combination of environmental crypsis and the mimicking of venomous viperids of the genera Bitis, Causus, and Echis (Gans 1961; Broadley and Bates 2009; Bates and Broadley 2018). These anti-predatory strategies are facilitated by adaptations of dorsal colour patterns that make the snakes resemble local substrates and mimic models.

Several species of Dasypeltis occur in two or more distinct colour pattern morphs, including patternless morphs, throughout their respective ranges (Gans 1959; Broadley and Bates 2009; Trape et al. 2012). In a study of Dasypeltis in north-eastern Africa and south-western Arabia, Bates and Broadley (2018) documented several patternless individuals of various species. They found that 38.9% of the high-elevation montane species D. abyssina and 26.8% of the savannah species D. confusa were plain brown, and 51.4% of D. medici from forests and woodlands were faintly patterned to uniformly coloured. Additionally, in north-eastern and central Africa, D. atra occurs as either patterned, plain brown, or black morphs (Gans 1959; Broadley and Bates 2009; Bates and Broadley 2018). By contrast, the Namibian species D. loveridgei occurs only as a patterned morph (purportedly mimicking the horned adder Bitis caudalis), as does the common or rhombic egg-eater D. scabra in that country (Bates 2023). However, D. scabra occurs as a patterned and ‘plain’ morph in other parts of Africa (Branch 1998; Broadley and Blaylock 2013; Pietersen et al. 2021).

Dasypeltis scabra is the most widespread member of the genus and occurs within various habitats throughout its range (Bates et al. 2014; Bates and Broadley 2018). Typically patterned individuals (hereafter ‘patterned’) are grey to brown dorsally and possess a series of darkly coloured rhombic, oval, squarish, rectangular, or even irregular-shaped vertebral saddles, similarly coloured lateral bars, and usually a characteristic forward-pointing dark chevron on the neck that sometimes extends onto the back of the head (Figure 1a; Branch 1998; Broadly et al. 2003; Bates and Broadley 2018; Pietersen et al. 2021). The markings along the backs of these snakes appear to represent an example of Batesian mimicry (Gans 1961), showing resemblance to the venomous rhombic night adder Causus rhombeatus or in some areas, the horned adder Bitis caudalis (Gans 1961; Branch 1998). Patternless or unmarked individuals (hereafter ‘plain’) are typically a uniform brown or greyish colour and either lack markings or have a darker vertebral stripe (Figure 1b; Jacobsen 1989; Bates 1992). In parts of Zambia and Zimbabwe, some specimens are mostly patternless, lacking chevrons on the head and neck, but have weakly indicated vertebral saddles and minimal or no lateral bars on the body (hereafter ‘intermediate’; Figure 1c). While patterned snakes occur throughout the range, plain and/or intermediate snakes are known from only a few areas in South Africa (Branch 1998), Zambia (Pietersen et al. 2021), and Zimbabwe (Broadley et al. 2003; Broadley and Blaylock 2013).

Figure 1. Illustrations showing examples of colour pattern variation in Dasypeltis scabra: (a) a typically patterned ‘5N’ individual, (b) a ‘patternless’ or ‘plain’, brown individual with dark vertebral stripe, and (c) an ‘intermediate’, brown individual with weakly marked dorsal saddles. Illustrations drawn by Faansie Peacock.

Intraspecific colour pattern variation in D. scabra offers an excellent system to study the adaptive significance of colour patterns in African snakes. Because their patterns are seemingly necessary to facilitate mimicry of venomous vipers, in areas where mimic models are rare or absent, resemblance to those models may be an ineffective predator deterrent. The existence of plain and weakly marked D. scabra thus likely represents a selective trade-off favouring other anti-predatory mechanisms like crypsis over mimicry. In addition, in open environments with minimal cover that lack a tree canopy and densely arranged vegetation, adaptation for increased camouflage ability through greater background matching may be advantageous. An important step towards quantifying the above is to first characterise the geographic extent and environmental conditions of the areas inhabited by each morph of these snakes.

In this study, we mapped and characterised the ecogeographic divergence of patterned, plain, and intermediate D. scabra distributed throughout southern Africa (defined here as Angola, Botswana, Eswatini, Lesotho, Malawi, Mozambique, Namibia, South Africa, Zambia, and Zimbabwe). We hypothesised that plain and intermediate morphs do not occur in isolation from patterned snakes but are geographically restricted to a limited portion of the species’ range with particular climatic and environmental conditions relative to those elsewhere in the range. We predict that the areas inhabited by plain and patterned morphs would have distinct climate and environmental regimes compared to the overall range of the species that could likely explain in part the emergence of these morphs. To investigate this, we used a combination of environmental variables and georeferenced occurrence data from community science platforms, social media, and preserved specimens in museum collections. Environmental niche models were used to predict areas with suitable habitats for each colour morph. We discuss the possibility that plain and intermediate snakes occur only in environments that favour selection of camouflage over mimicry.

Materials and Methods

Colour pattern assignment

For D. scabra, we gathered georeferenced occurrence records, with accompanying photographs or videos, from various community science platforms and social media (see Table 1). These platforms included Atlasing in Namibia (AiN; http://www.the-eis.com), iNaturalist (https://www.inaturalist.org), iSpot (https://www.ispotnature.org), and the ReptileMAP virtual museum (http://vmus.adu.org.za/?vm=ReptileMAP). Social media sources included three public Facebook groups: ‘Free State Reptiles and Amphibians (including adjacent areas and Lesotho)’ (https://www.facebook.com/groups/FreeStateHerps), ‘Snakes of southern Africa’ (https://www.facebook.com/groups/snakesofsouthafrica), and ‘Zambian snakes and other crawlies’ (https://www.facebook.com/groups/Zamsnakes). Images and videos from community science and social media sources were downloaded and saved electronically to ensure they remain accessible in case the original poster moved or deleted them or changed their privacy settings (see also Maritz and Maritz 2020; Bates and Stobie 2022). Data was also collected from preserved museum specimens examined by MFB or the late D.G. Broadley (datasheets accessed by MFB), comprising material housed at the National Museum, Bloemfontein (NMB), South Africa, and several specimens housed in other southern African, European, and United States museums (see Electronic dataset S1).

Table 1. Summary of occurrence records of Dasypeltis scabra colour morphs from various sources. See electronic dataset S1 for additional information.

Data source	Patterned	Plain	Intermediate	Sum	%
Atlasing in Namibia^1	11	0	0	11	0.6
Facebook groups^2, 3, 4	116	13	1	130	7.6
iNaturalist^5	358	3	1	362	21.2
iSpot^6	49	1	0	50	2.9
Literature	1	1	0	2	0.1
Museum collections	674	63	12	749	43.9
ReptileMAP^7	396	4	3	403	23.6
Total	1605	85	17	1707	100.0

¹ http://www.the-eis.com; ²https://www.facebook.com/groups/FreeStateHerps; ³https://www.facebook.com/groups/snakesofsouthafrica; ⁴https://www.facebook.com/groups/Zamsnakes; ⁵https://www.inaturalist.org; ⁶https://www.ispotnature.org; ⁷http://vmus.adu.org.za/?vm=ReptileMAP

Photographic records were inspected by each author and preserved specimens by MFB to score snakes as either patterned, plain, or intermediate. Patterned snakes were those identified as having the typical ‘5N’ dorsal pattern (see Gans 1959; e.g. fig. 5 in Bates and Broadley 2018). We recognise that patterned D. scabra may show geographical variation in their markings but as we were primarily interested in comparing marked snakes to plain and weakly marked snakes, we did not segregate patterned snakes into narrower groupings. Plain snakes were brown or sometimes greyish in colour, lacking dorsal patterns except for a narrow to moderate dark ventral stripe present in some specimens. Snakes that were not completely patternless but had faintly marked dorsal saddles and faint patterns on the body and/or head, were classified as ‘intermediate’. Only specimens, images, or videos with sufficient visual evidence to identify the colour patterns unambiguously were included in our dataset. Two specimens (photo-records) from eastern Zambia with intermediate patterns were excluded from the analysis because of possible confusion with D. medici which also occurs as a reddish, intermediate morph (Bates and Broadley 2018).

Figure 5. Maxent models of predicted suitable habitat for: (a) patterned, (b) plain, and (c) intermediate morphs of Dasypeltis scabra across southern Africa. Warmer colours represent higher probabilities of suitable habitat.

Geographic dispersion

Occurrence records from AiN, iNaturalist, iSpot, and ReptileMAP had associated GPS co-ordinates. Most Facebook records did not include GPS information but had locality descriptions which we used to estimate GPS positions. GPS positions of most museum records were similarly estimated from locality descriptions. Because reptile occurrence data are often subject to sampling bias and spatial autocorrelation (Barends et al. 2020), especially those from community science and social media platforms (Di Cecco et al. 2021), we spatially thinned our dataset to remove clustered records so that no record had a neighbour closer than 5 km. This was done separately for each colour morph using the spatial rarefication tool in the SDM toolbox (Brown et al. 2017) in ArcMap version 10.4 (ESRI 2016). All subsequent analyses were conducted on the spatially thinned dataset. Following Lattanzio and Buontempo (2021), we evaluated differences in the geographic dispersion of each colour pattern using a multivariate general linear model in SPSS software version 23 (IBM Corp. 2015) which compared latitude and longitude values of each record as dependent variables against colour pattern as the fixed variable. Additionally, we compared the proportions of observations of each colour pattern between countries where colour morphs occurred in sympatry using Two-Factor Chi-square tests.

Environmental niche divergence

We downloaded 19 bioclimatic layers (Bio 1 – Bio 19) representing different temperature and precipitation variables, and a global elevation layer from Worldclim 2.1 (Fick and Hijmans 2017; https://www.worldclim.org). We also downloaded a global biome, and a global ecoregion layer from the World Wildlife Organisation (WWF; Olson et al. 2001). All layers were downloaded at a resolution of 2.5 arc minutes. Worldclim bioclimatic variables are often highly correlated with one another (Maria and Udo 2017), and to account for this, correlated variables are typically removed from analyses. However, removing variables may reduce the overall climatic space of the geographic areas under investigation which can weaken analyses (Sau et al. 2023). To avoid this, we opted not to remove correlated layers but instead followed the approach of Šmíd et al. (2020) and performed two principal component analyses in R software version 4.1 (R Core Team 2022) to create new uncorrelated layers from the Worldclim dataset. The first of these analyses was restricted to variables related to temperature (Bio 1, Bio 2, Bio 4 – Bio 11) and the second was restricted to those pertaining to precipitation (Bio 3, Bio 12 – Bio 19). In each case, the first two principal components represented over 80% of the total variation (Table S1 and S2), so we retained only those respective layers to represent temperature (hereafter Temp PC1 and Temp PC2) and precipitation (hereafter Prec PC1 and Prec PC2) respectively.

A general linear model was used to compare climate niches between colour morphs. At each snake occurrence, log-elevation, Prec PC 1, Prec PC 2, Temp PC 1, and Temp PC 2 data were extracted using the point sampling tool in QGIS version 3.2.3 (QGIS Development Team 2020). These values were then used as dependent variables against snake colour morph as the fixed factor. In addition, because comparisons of differences between principal components do not allow for easy interpretation of the direction in which they may differ, we also included additional ANOVA tests comparing extracted values of log-annual mean temperature (Bio 1) and log-annual mean precipitation (Bio 12) in separate analyses. Lastly, we compared the number of different biomes and the number of ecoregions that each morph occupies using separate One-Factor Chi-square tests.

Environmental niche modelling

We used Maxent version 3.4 (Phillips et al. 2017) to predict areas of suitable habitat and thus the potential geographic range for each colour morph across southern Africa using snake occurrence data and our seven environmental predictors (i.e. biome, ecoregion, elevation, Prec 1, Prec 2, Temp PC 1, and Temp PC 2). Maxent model parameters were consistent for all colour morphs. We ran each model over 100 bootstrap replications using linear, quadratic, hinge, and product features, and set a regularisation value of 2 to reduce model overfitting (Radosavljevic and Anderson 2014). We allocated 80% of occurrences to train each model, with the remaining 20% used to test model accuracy. We used ‘area under receiver operating curve’ (AUC) scores to evaluate model accuracy (Elith et al. 2011), with AUC scores ≥ 0.75 indicating strong support (Phillips et al. 2017).

Results

Geographic dispersion

In total, we gathered 1707 occurrence records of D. scabra for which colour patterns could be assigned (Table 1). Most records were scored as patterned (n = 1605, 94%), with only a small percentage of the sample allocated as plain (n = 85, 5%) or intermediate (n = 17, 1%). Similar percentages remained after correction for spatial autocorrelation via spatial thinning of records (patterned = 92%, plain = 6%, intermediate = 2%, n = 964). The overall dispersion of colour morphs varied across geographic space (Wilks Lambda = 0.938, F_{4, 1924} = 15.329, p < 0.001), with differences being driven by both latitude (F_{2, 963} = 28.172, p < 0.001) and longitude (F_{2, 963} = 13.102, p < 0.001). Patterned snakes were observed throughout southern Africa, but plain snakes were documented in only a few areas in central South Africa, Zambia, and Zimbabwe (Figure 2), and intermediate snakes were found only in Zambia and Zimbabwe, where they occurred sympatrically or in close proximity with the other colour morphs (Figure 2).

Figure 2. Distribution records of patterned, plain, and intermediate Dasypeltis scabra across some sub-Saharan African countries from observations shared online via community science platforms (n = 826) and social media (n = 130), with data from museum specimens (n = 749), and one locality from the literature (Schmidt 1999: n = 2).

Plain snakes in South Africa were recorded in the Free State and Gauteng provinces in the Highveld grassland region (as defined by Olson et al. 2001) (Figure 2). In Zimbabwe, plain and intermediate snakes were present in southern Miombo woodlands and montane forest-grassland mosaics (Olson et al. 2001), especially in Manicaland in the east of the country, and in some central parts of the country around Harare and Mashonaland Central. In Zambia, plain and intermediate snakes were recorded primarily in woodlands in the western region of Lusaka Province (Figure 2), the Eastern Province, and a single record of an intermediate snake was from the Northern Province.

The percentage of all records of plain and intermediate snakes, relative to patterned snakes, was highest in Zimbabwe (18%) compared to 5% in South Africa (but 16% when only considering the Highveld region as defined by Olson et al. 2001) and 8% in Zambia (Figure 3a). Similar percentages were present for the thinned dataset (South Africa: patterned snakes = 95%; Zambia: patterned snakes = 85%; Zimbabwe: patterned snakes = 75%; Figure 3b). Proportions of observations of patterned versus plain and intermediate snakes were significantly greater in South Africa compared to Zambia (χ²_df = 2= 9.932, p < 0.05) and Zimbabwe (χ²_df = 2 = 11.101, p < 0.05) respectively, and were significantly greater in Zambia than Zimbabwe (χ²_df = 2 = 8.029, p < 0.05).

Figure 3. Percentages of occurrence records for the three Dasypeltis scabra colour morphs from South Africa, Zimbabwe, and Zambia based on (a) our full dataset data, and (b) our spatially-thinned dataset. Values above each bar represent sample sizes of snake observations.

Environmental niche divergence

Areas inhabited by snakes of each colour morph differed in overall environmental niche (Wilks Lambda = 0.912, F_{10, 1922} = 9.045, p < 0.001), with differences being driven by each of the tested environmental variables (F_{2, 959} = 4.822–23.878, p < 0.05 in all cases). Plain snakes occurred in areas that were on average higher in elevation, received greater annual precipitation, and e^xperienced colder annual temperatures compared to areas occupied by patterned snakes (Figure 4a–c; Table 2). The overall precipitation and temperature regimes also differed between areas occupied by these snakes (Table 2). Similar trends in environmental differences were present between areas occupied by patterned and intermediate snakes, with the exception being that patterned snakes occurred in colder areas (Figure 4a–c; Table 2). Plain and intermediate snakes occurred at similar elevations but slightly diverged in precipitation and temperature regimes (Figure 4a–c; Table 2).

Figure 4. Differences in (a) average annual precipitation, (b) average annual temperature, (c) average elevation, and (d) numbers of habitat types of areas occupied by different colour morphs of Dasypeltis scabra across southern Africa.

Table 2. Results of GLM post-hoc comparisons between log-transformed climatic variables and principal components for areas occupied by patterned, plain, and intermediate colour morphs of Dasypeltis scabra in southern Africa. Positive mean difference values indicate that areas occupied by the first colour morph in the ‘Comparison’ column had a greater average for that variable, and negative mean difference values indicate the opposite.

Variable	Comparison	Mean difference	p
Annual mean precipitation	Patterned vs Plain	0.054	< 0.05
(log)	Patterned vs Intermediate	–0.156	< 0.001
	Intermediate vs Plain	0.102	< 0.05
Annual mean temperature	Patterned vs Plain	0.022	< 0.05
(log)	Patterned vs Intermediate	–0.034	< 0.05
	Intermediate vs Plain	0.057	< 0.001
Elevation	Patterned vs Plain	–0.328	< 0.001
(log)	Patterned vs Intermediate	–0.306	0.116
	Intermediate vs Plain	–0.023	0.991
Prec PC 1	Patterned vs Plain	–0.159	0.231
	Patterned vs Intermediate	–0.585	< 0.05
	Intermediate vs Plain	0.427	0.088
Prec PC 2	Patterned vs Plain	–0.468	< 0.001
	Patterned vs Intermediate	–1.510	< 0.001
	Intermediate vs Plain	1.042	< 0.001
Temp PC 1	Patterned vs Plain	0.458	< 0.05
	Patterned vs Intermediate	–1.085	< 0.05
	Intermediate vs Plain	1.543	< 0.001
Temp PC 2	Patterned vs Plain	0.589	< 0.05
	Patterned vs Intermediate	–0.305	0.688
	Intermediate vs Plain	0.894	0.079

In addition, plain snakes were observed only in ecoregions dominated by Highveld grassland, southern Miombo woodland, or Zambezian Mopane woodland vegetation, whereas patterned snakes occupied a significantly greater number of ecoregions (χ²_df = 2 = 35.150, p < 0.001; Figure 4d). Intermediate snakes occurred in a similar number of ecoregions to plain snakes (χ²_df = 2 = 0.111, p < 0.739; Figure 4d). Patterned snakes also occurred in more biomes than plain snakes and intermediate snakes respectively (Figure 4d), but this difference was not statistically significant (χ²_df = 1 = 0.818, p = 0.366).

Environmental niche models

Maxent models produced strongly supported predictions of suitable habitats for patterned (AUC_training = 0.857, AUC_test= 0.850), plain (AUC_training = 0.969, AUC_test = 0.957), and intermediate (AUC_training = 0.975, AUC_test= 0.951) snakes. Habitat ecoregion contributed most to model-building (percent contribution: patterned model = 40.9%, plain model = 79.4%, intermediate model = 69.1%; Table S3) and the final modelled predictions (permutation importance: patterned model = 37.7%, plain model = 71.19%, intermediate model = 18.4%; Table S3). The model for patterned snakes predicted areas of high suitability throughout much of southern Africa. In particular, most of Eswatini, Mozambique, South Africa, Zimbabwe, Zambia, some parts of Angola and Malawi, and low elevation areas in Lesotho comprised areas predicted to be highly suitable for these snakes (Figure 5a). The model for the plain morph predicted that suitable habitat is concentrated in the Highveld of South Africa, most of Zimbabwe except the southwest, and in the central and south-eastern regions of Zambia (Figure 5b). Similarly, the model for intermediate snakes predicted high suitability in central and eastern Zimbabwe, southern and south-eastern Zambia, as well as some areas in north-western Mozambique and western Malawi (Figure 5c).

Discussion

Our results provide evidence suggesting substantial ecogeographic divergence between patterned, plain, and weakly marked Dasypeltis scabra. Patterned individuals are ubiquitous throughout southern Africa and occur within several divergent habitats across several ecoregions. Conversely, plain individuals are restricted to areas greater than 500 m a.s.l. dominated by Highveld grassland vegetation in South Africa and southern Miombo or Zambezian Mopane woodlands in Zambia and Zimbabwe respectively. Intermediate snakes are similarly restricted to a few areas in Zambia and Zimbabwe but were not recorded in South Africa. The areas occupied by plain and intermediate snakes differed in overall precipitation and temperature regimes and were on average higher in elevation compared to areas occupied by patterned snakes. Thus, our hypothesis that such snakes have a restricted range and environmental niche is confirmed. Importantly, plain and/or weakly marked snakes occur in sympatry with patterned snakes and do not represent isolated or distinct populations.

Our dataset showed that the frequencies of plain and intermediate snakes are dramatically lower than for patterned snakes. De Waal (1978) observed a low encounter rate of plain snakes (9.1%, 29 of 320 individuals) when sampling across the Free State. Subsequently, Bates (1992) also observed a similarly low percentage of 9.5% (33 of 343 individuals) for the same region based on the same collection but with additional material added. However, those encounter rates were based on datasets comprised of specimens collected primarily by active sampling and thus involved directly searching for snakes in refugia such as disused termitaria and under rocks. Thus, those datasets are not directly comparable to datasets such as ours that include community science and social media records which often comprise once-off observations of random encounters (Kosmala et al. 2016). Nevertheless, the low encounter rates of plain and intermediate snakes observed here and in those previous studies indicate that such snakes are almost certainly considerably less abundant than patterned snakes.

The scarcity of plain or weakly marked snakes in our dataset does not necessarily mean that these snakes are not present elsewhere in the distribution range of D. scabra in this region. Given the low numbers of photographs and specimens available for some countries it cannot be ruled out that patternless or weakly marked snakes do occur there but have yet to be reported. For example, only a single photographic record of a plain or intermediate coloured snake from Zimbabwe was available from community science or social media despite reports of these snakes occupying several areas of the country (e.g. Broadley and Blaylock 2013; Electronic dataset S1). Our predicted distribution models suggest that such snakes could be present in some areas in Mozambique and Malawi. Additional sampling from those countries could provide evidence to support these model predictions. In addition, it is possible that patternless snakes are present outside of southern Africa. However, in north-eastern and eastern Africa, an area encompassing most other parts of the range of this species outside of southern Africa, all 107 specimens examined by Bates and Broadley (2018) were patterned.

The causal factors that led to the evolution of plain D. scabra are not known, but our results suggest that differences in habitat are likely key. In particular, our findings suggest that ecoregion, and likely vegetation type, is an important factor. Plain and intermediate D. scabra are known only from open or exposed grassland and woodland areas. Such areas often include large patches of exposed substrate (Mucina and Rutherford 2006) and regularly experience fires (Strydom et al. 2016), which could lead to further reductions in vegetation cover. In such areas, plain snakes are likely to be better camouflaged to avoid predation as their appearance will better match the substrate than that of a patterned snake. A selective trade-off between enhanced crypsis and reduced mimicry may have occurred in some populations in those areas that led to the snakes losing their markings. Differences in vegetation cover could potentially also explain why intermediate snakes are present in the woodlands of Zambia and Zimbabwe but not in the Highveld grasslands of South Africa. These results are contrary to those of Allen et al. (2013) who suggest that colour pattern variation among snakes is associated with shifts in behaviour instead of adaptive responses to different habitats.

In addition, individuals of purported mimic models Causus rhombeatus and Bitis caudalis are absent or rare in many areas occupied by patternless D. scabra (see Figure S1). Causus rhombeatus is absent from most of the Free State (Bates et al. 2014), but there is broad overlap in distribution between it and plain D. scabra in Zambia and Zimbabwe. However, C. rhombeatus also occasionally occurs as a patternless or faintly marked plain morph in those regions (Broadley and Blaylock 2013; Pietersen et al. 2021). Bitis caudalis is largely absent from most of the Highveld, most of Zimbabwe apart from southern regions, and all of Zambia (Bates et al. 2014; Broadley 1990; Pietersen et al. 2021). Similar selective forces may have facilitated the emergence of plain morphs in Zambian and Zimbabwean populations of both species. Although we cannot rule out that plain D. scabra evolved to mimic plain C. rhombeatus in those areas, this would not explain the presence of plain D. scabra in the Highveld. It is more likely that these snakes converged upon similar dorsal appearances as a result of shared selective forces, most likely derived from similarities in their habitats. We therefore purport that it is the combination of open habitats and reduced overlap with mimic models, either because models are absent or also vary in dorsal colour pattern, that likely facilitated the evolution of plain and intermediate D. scabra. Additional analyses such as comparing the overlap in distributions of corresponding morphs of D. scabra and C. rhombeatus could provide important evidence to support or deny these suggestions.

In conclusion, our study provides insight on the ecogeographic divergence between patterned, plain, and weakly marked D. scabra. Our findings suggest that the existence of patterned and plain could be due to a selective trade-off of anti-predatory mechanisms (mimicry vs crypsis) although further work is necessary to make such a definite claim. Further sampling and sharing of observations of D. scabra may provide evidence that plain and/or intermediate snakes are more widespread than currently known, which could expand on our findings here. However, considering that these snakes are among the most common and widespread snake species in Africa, and observations of them are very frequently reported (see links to online platforms such as iNaturalist), especially in southern Africa (see Figure 2), our findings are likely a reasonable approximation of reality. Our study highlights the usefulness of community science and social media for investigating ecological questions on snakes or other reptiles. Successes using similar approaches for studies of African reptiles have been published elsewhere (e.g. Maritz and Maritz 2020; Bates and Stobie 2022; Barends and Scholtz 2023). We advocate for the continued use and integration of data available on community science platforms and social media to fill in knowledge shortfalls concerning African herpetofauna.

Acknowledgements

We thank Darren Pietersen and Rene Navarro for providing access to ReptileMAP data, the late Donald G. Broadley for his records, and the various contributors who shared observations of Dasypeltis scabra online. We thank Faansie Peacock for providing digital illustrations of snakes. Lastly, we thank our anonymous reviewers and associate editor for their helpful feedback. Data used for this study (Electronic dataset S1) can be found at Figshare (https://figshare.com/articles/dataset/Electronic_dataset_S1/24552133).

Disclosure statement

No potential conflict of interest was reported by the author(s).