A taxonomic revision of the south-eastern dragon lizards of the Smaug warreni (Boulenger) species complex in southern Africa, with the description of a new species (Squamata: Cordylidae)

A recent multilocus molecular phylogeny of the large dragon lizards of the genus Smaug Stanley et al. (2011) recovered a south-eastern clade of two relatively lightly-armoured, geographically-proximate species (Smaug warreni (Boulenger, 1908) and S. barbertonensis (Van Dam, 1921)). Unexpectedly, S. barbertonensis was found to be paraphyletic, with individuals sampled from northern Eswatini (formerly Swaziland) being more closely related to S. warreni than to S. barbertonensis from the type locality of Barberton in Mpumalanga Province, South Africa. Examination of voucher specimens used for the molecular analysis, as well as most other available museum material of the three lineages, indicated that the ‘Eswatini’ lineage—including populations in a small area on the northern Eswatini–Mpumalanga border, and northern KwaZulu–Natal Province in South Africa—was readily distinguishable from S. barbertonensis sensu stricto (and S. warreni) by its unique dorsal, lateral and ventral colour patterns. In order to further assess the taxonomic status of the three populations, a detailed morphological analysis was conducted. Multivariate analyses of scale counts and body dimensions indicated that the ‘Eswatini’ lineage and S. warreni were most similar. In particular, S. barbertonensis differed from the other two lineages by its generally lower numbers of transverse rows of dorsal scales, and a relatively wider head. High resolution Computed Tomography also revealed differences in cranial osteology between specimens from the three lineages. The ‘Eswatini’ lineage is described here as a new species, Smaug swazicus sp. nov., representing the ninth known species of dragon lizard. The new species appears to be near-endemic to Eswatini, with about 90% of its range located there. Our study indicates that S. barbertonensis sensu stricto is therefore a South African endemic restricted to an altitudinal band of about 300 m in the Barberton–Nelspruit–Khandizwe area of eastern Mpumalanga Province, while S. warreni is endemic to the narrow Lebombo Mountain range of South Africa, Eswatini and Mozambique. We present a detailed distribution map for the three species, and a revised diagnostic key to the genus Smaug.


INTRODUCTION
S. barbertonensis. Some non-types of the new species referred to below are housed at the American Museum of Natural History, New York (AMNH), Durban Natural Science Museum, Durban (DNSM), and Natural History Museum of Zimbabwe, Bulawayo (NMZB); and a few specimens of S. warreni are in the collections of AMNH and NMZB.
When collection co-ordinates (presented as degrees, minutes and in many cases seconds) and/or altitudes (m above sea level) were not available in museum documentation, these were estimated using Google Earth Pro.
In addition to the data presented in this article, comparative data consulted for the diagnoses of species and for the purposes of preparing a diagnostic key (see below) were obtained from NMB and NMZB specimens listed in the appendix in Mouton et al. (2018), and data in , Van Dam (1921), FitzSimons (1930, 1933, 1943, 1958, , Broadley (1962, 1966), De Waal (1978, Jacobsen (1989), Stanley et al. (2011) and Mouton et al. (2018).

Ethics approval
This project was approved by the National Museum Bloemfontein Ethics Clearance Committee (NMB ECC 2019/13).

External morphology
Measurements Snout to vent length (SVL) was measured from the tip of the snout to the vent after flattening the specimen on its back; tail length, from vent to tip of tail; using vernier calipers or a millimeter ruler. Head measurements (determined using vernier callipers and unless damaged, taken on the right side): Length, measured from tip of snout to ear opening; width, at widest point at about the level of the posterior borders of the parietals; depth, from middle of posterior sublabial to highest point of posterior parietal.
Scalation (examined by MFB using a binocular dissecting microscope, mostly a Nikon SMZ 745T; both sides of head examined) For the most part the morphological characters employed by FitzSimons (1943) were used, and in the same way, unless otherwise indicated. To avoid uncertainty, the following scale counts are described in detail: occipitals: large scales behind the posterior parietals, the outermost ones situated directly behind the elongated upper temporals; gular scales (often elongated and in longitudinal rows): counted transversely between posterior sublabials, the first row extending to the anterior end of the posterior sublabial; dorsal scale rows longitudinally: counted across the widest part of the body more-or-less midway between fore-and hindlimbs (scales of the most lateral rows are at least half the width of adjacent enlarged dorsals); dorsal scale rows transversely: counted from the first complete row behind the occipitals to the row that ends immediately anterior to the vent (when followed around to the ventral side; excluding rows less than three-quarters as long as adjacent ones); ventral scale rows longitudinally: counted across the widest part of the body, more-or-less midway between fore-and hindlimbs (lateral ventrals are rectangular or quadrangular, smooth or weakly keeled, flattened and at least half the size and width of adjacent ventrals); ventral scale rows transversely: counted from the first row (which curves anteriorly) behind the posterior part of the forelimb insertion to the row (which curves posteriorly) immediately in front of the anterior part of the hindlimb insertion (i.e. scale rows between axilla and groin); lamellae under 4th toe of right foot were counted from the first scale entirely or largely (>60%) anterior to the junction between 3rd and 4th toes to the scale behind the claw, and incomplete lamellae (i.e. those that do not extend to either side) were excluded.

Sexing
Males (>70 mm SVL) were identified by the presence of large femoral pores (usually with waxy plugs of secreted fluid) as well as differentiated femoral scales (generation glands). Females (>70 mm SVL) had minute pin-prick-like femoral pores without waxy plugs, and lacked differentiated femoral scales.

Osteological data
Osteological data was obtained from representative specimens of the S. warreni species complex via High Resolution X-ray Computed Tomography (HRCT). Specimens used were: S. warreni NMB R9292, AMNH-R-173381; S. barbertonensis NMB R9196 (topotype); S. cf. barbertonensis NMB R9201 (holotype of new species, see below), AMNH-R-173382. These specimens were scanned using a Phoenix v|tome|x S CT scanner at the American Museum of Natural History's Microscopy and Imaging Facility in New York, and GE Inspection Technologies, LP Technical Solutions Center in San Carlos, CA, or on a Phoenix v|tome|x M at the University of Florida's Nanoscale Research Facility in Gainesville, FL. Each specimen was scanned twice: once to recover the full body, and a second higher resolution scan to focus on the cranial morphology. Current, voltage and detector-time were modified to optimise the greyscale range, and specimens were scanned in sections to maximise resolution (Table S1). Raw data were processed using GE's proprietary datos|x software V.2.3 to produce a series of tomogram images which were then viewed, sectioned, measured and analysed using VG Studio Max 2.2 (Volume Graphics, Heidelberg, Germany). Individual skeletal elements and osteoderms were reconstructed separately for each scan, so as to facilitate osteological analysis. Tomograms and 3D mesh files for all datasets are available online at www.morphosource.org (see Supplemental Data for DOIs).

Statistical analyses
Univariate analyses of scale counts were conducted using Statistica v. 6. Principal component and linear discriminant analyses were run for three mensural characters (head length, width and height) and 13 meristic characters (supraciliaries, suboculars, supralabials, infralabials, sublabials, occipitals, gulars, dorsal scale rows transversely and longitudinally, ventral scale rows transversely and longitudinally, femoral pores, subdigital lamellae on fourth toe), taken from 72 museum specimens (>70 mm SVL; i.e. juveniles excluded to avoid the effects of ontogenetic growth) (Table S2), using the prcomp and lda commands in R {stats} and {MASS}. When scale counts were made on both sides of the head or on both hindlimbs (see Table S2), a mean value was used for the analyses.

Species concept and species delimitation
We apply a lineage-based species concept whereby a species is represented by an independently evolving metapopulation lineage (see Frost & Hillis, 1990;De Quieroz, 1998. The genetic distinctness described by  and morphological characters were the operational criteria for species delimitation. Although  did not employ coalescent species delimitation analyses, the topological consistency of the mtDNA, nuDNA and combined analyses strongly support the existence of three distinct lineages within S. warreni and S. barbertonensis.

Nomenclatural note
The electronic version of this article in Portable Document Format (PDF) will represent a published work according to the International Commission on Zoological Nomenclature (ICZN), and hence the new names contained in the electronic version are effectively published under that Code from the electronic edition alone. This published work and the nomenclatural acts it contains have been registered in ZooBank, the online registration system for the ICZN. The ZooBank LSIDs (Life Science Identifiers) can be resolved and the associated information viewed through any standard web browser by appending the LSID to the prefix http://zoobank.org/. The LSID for this publication is: (urn:lsid:zoobank. org:pub:490BDD66-155F-423F-A4E9-DEAEEB024CC5). The online version of this work is archived and available from the following digital repositories: PeerJ, PubMed Central and CLOCKSS.

Character analysis
Dorsal colour pattern Figure 2 Specimens from all three clades recovered by  are distinguishable on the basis of dorsal, lateral and ventral colour patterns. S. warreni has a medium to sandy brown (sometimes reddish-brown) dorsum with a series of 5-6 interrupted transverse bands between fore-and hindlimbs, each consisting of white or cream ocelli (spots or blotches) with dark (often black) borders. The dark-edges exaggerate the ocelli, but some specimens have only small pale markings which also lack heavy dark borders. The dorsum of S. barbertonensis is medium to dark brown (or even black), usually with 4-5 interrupted bands on the back formed mostly by transversely enlarged pale markings (rather than spots or blotches) with moderately dark edges. Smaug cf. barbertonensis is similar to the latter form, but there are usually 5-6 bands. However, in S. barbertonensis there is almost always a pale spot on the nape immediately posterior to the median occipitals, followed in close proximity by a distinct transverse band. In S. cf. barbertonensis, the spot on the nape is replaced by a pale band, followed after a distinct gap by another pale band (often divided medially) on the neck with a slightly posteriorly-directed curvature. S. warreni also has a pale band behind the occipitals, but the band that follows is seldom curved as in the case of the previous form. Bates  Ventral colour pattern Figure 2: In S. warreni the belly is generally white with the centre of each scale pale brown; the throat is usually mostly white with scattered small to medium-sized dark brown spots. In S. barbertonensis the belly is almost completely black or dark brown, with only a few pale markings on the sides; the throat is also almost entirely dark, with only occasional pale specks or blotches. In S. cf. barbertonensis the belly is white with 5-6 broad, dark brown 'cross-bands', interrupted mid-ventrally by six longitudinal rows of brown scales; the throat is white with bold, dark mottling or reticulations (sometimes forming transverse bands; often most of the throat is dark).

Lateral colour pattern
The flanks of S. warreni are often mostly cream with a few dark markings, but may consist of alternating light and dark vertical bands. In S. barbertonensis the flanks are primarily dark brown or even black, with a few narrow or moderate cream bands and/or spots/ blotches. In contrast, the sides of the body in S. cf. barbertonensis consist of large cream spots or blotches on a dark background; and in some cases the light patch behind the armpit is elongated (antero-posteriorly).

Scales at the edges of the ear openings
Smaug barbertonensis usually has generally elongated and spinose scales at the anterior edges of the ear openings (especially the central ones), whereas in most cases these scales are short and non-spinose in S. cf. barbertonensis and S. warreni.

Relative length of occipital scales
In all three forms in the complex there are usually six occipital scales, and the scales of the median pair are shorter and usually smaller than the others (although usually less distinctly so in S. warreni). In S. warreni the outer occipital is usually of similar size and shape to the adjacent inner occipital, but in S. barbertonensis and S. cf. barbertonensis the outer one is usually shorter and smaller. In S. warreni a small median occipital is common.

Quadrate variation
In S. cf. barbertonensis the quadrates have a pronounced ridge and concave region at the lateral edge of the adductor musculus mandibulae posterior origin, whereas in S. barbertonensis and S. warreni the quadrates have a less pronounced ridge and a non-concave region (Fig. 3).

Scale counts
The three forms are similar in terms of scale counts (Table 1), but S. barbertonensis usually has lower numbers of transverse dorsal scale rows than S. warreni and S. cf. barbertonensis (28-34 vs. 31-41; Fig. 4A).

Head width
Smaug barbertonensis, when compared to both S. warreni and S. cf. barbertonensis, usually has a wider head relative to snout-vent length (SVL) (Fig. 4B).

Spinosity
A recent study by Mouton et al. (2018) that investigated the relationship between generation gland morphology and armour in the genus Smaug found that those species with multi-layer generation glands (S. giganteus, S. breyeri, S. vandami) had relatively long (basal) tail and occipital spines, while all other species (including S. warreni, S. barbertonensis and S. cf. barbertonensis) had two-layer glands and relatively short spines. The latter two forms were found to be more spinose than S. warreni (i.e. longer occipital scales and proximal caudal spines).

Statistical analyses
Both principal components and linear discriminant analyses reveal clear separation in scale characters between S. warreni and S. barbertonensis, and S. cf. barbertonensis and S. barbertonensis (4% LDA mis-classification rate in both cases) (Figs. 4C and 4D; Tables S3 and S4). Smaug cf. barbertonensis and S. warreni display similar pholidosis and head proportions and cannot be consistently sorted by these characters alone (25% LDA mis-classification rate). The first two principal components explain 32% of the variation in the dataset.     Diagnosis. (includes 'additional material') Distinguished from all other cordylids (Cordylidae) by its unique combination of dorsal, lateral and ventral colour patterns (see descriptions and figures). Referable to the genus Smaug on the basis of its large size and robust body, enlarged and spinose dorsal and caudal scales, enlarged occipital scales, and frontonasal in contact with the rostral, separating the nasal scales. A medium to large species of Smaug distinguishable by the following combination of characters: (1) back dark brown usually with 5-6 pale bands (usually interrupted) between fore-and hindlimbs, each band consisting of pale, sometimes dark-edged, markings;
Its status as a new species is also supported by monophyly with high levels of support from three mitochondrial and eight nuclear markers (see ; using samples from NMB R9201-2).
It differs from the terrestrial S. giganteus by its smaller adult size (maximum SVL 145 mm vs. 198 mm), and possession of six moderate sized and weakly spinose occipitals, vs. four (occasionally five) large and distinctly spinose occipitals. Differs from other species of Smaug as follows: from S. vandami by having six (versus usually four) occipitals; from S. depressus by having only 10-13 (vs. 16-24) femoral pores per thigh in males; and from S. breyeri by having much less rugose head shields. It differs from S. giganteus, S. breyeri and S. vandami by having less spinose occipitals and tail spines, and two-layer (rather than multi-layer) generation glands. Differs from S. mossambicus and S. regius by having the first supralabial with moderate or no (vs. distinct) upward prolongation and lacking obvious sexual dichromatism (only males of the latter two species have bright yellow to orange flanks).
Most similar to S. barbertonensis and S. warreni, but easily distinguishable by its colour pattern (as described above) compared to S. barbertonensis (back dark brown with 4-5 pale bands between the limbs, pale spot or blotch on nape behind occipitals; flanks dark with narrow pale vertical markings; venter mostly dark brown or black) and S. warreni (back usually pale brown with 5-6 pale dark-edged bands between the limbs, pale band on nape behind occipitals; flanks pale with brown markings; venter with brown markings on most scales) (Figs. 2, 6 and 8, and others below); by usually having short, blunt, non-spinose scales at the edges of the ear openings (usually elongate and spinose in S. barbertonensis); and quadrates with a pronounced ridge and concave region at the lateral edge of the adductor musculus mandibulae posterior origin (no pronounced ridge or concave region in the other two species). Also differs as follows: outer occipitals usually shorter than the adjacent inner ones (of about equal length in S. warreni); head narrower than S. barbertonensis (head width/head length = 76-84% vs. 80-92% in adults); generally higher numbers of transverse dorsal scale rows (32-37 in 86% of specimens) than S. barbertonensis (29-32 in 81% of S. barbertonensis).
Description of holotype. NMB R9201. External morphology: Snout-vent length 138.8 mm, tail length (original) 187 mm, total length 325.8 mm, head length 40.2 mm, head width 31.2 mm, head depth 16.2 mm. Tail length/SVL = 135%; head width/SVL = 22.4%; head width/head length = 77.5%; head depth/head length = 40.3%. Head strongly depressed, head shields rugose and moderately striated over parietal region. Frontonasal 1.05 times as wide as long, in contact with the rostral and loreals, separating the nasals, latter slightly swollen. Nostril-with a large inner flap attached posteriorly-situated in the posterior part of the nasal and in contact with the loreal and 1st supralabial. Prefrontals in contact at their inner angles, separating the frontal from the frontonasal. Frontal hexagonal, slightly widened anteriorly, anterior sides curved slightly inwards. Frontoparietals slightly broader than long. Posterior parietals larger than anterior ones; interparietal between the four parietals, more sharply pointed anteriorly than posteriorly. Occipital scales 6, well-developed, bluntly spinose, the outer ones shorter and smaller than the adjacent inner ones, middle pair shortest and narrowest. Anteriormost of three upper temporals large, keeled at its lower edge. Gulars 23. Lateral temporals large, often bluntly keeled. Scales at anterior edge of ear opening (four on left side of head, five right) projecting outwards as flattened and somewhat spatulate spines, the lowermost one narrow and somewhat slender, the one above it distinctly spatulate and the largest; the other scales shorter and somewhat blunt. Supraoculars 4, the anterior one longest, the next (2nd) one broadest, posterior one the smallest. Supraciliaries 4, anterior one the longest, posterior shortest. Lower eyelid opaque, consisting of about 10 small, vertically-elongated, scales. Preocular at least twice the size of the loreal. Four large scales below the eye. Rostral 2.04 times wider than deep. Supralabials 6, 4th (longest) and 5th separated by a large suborbital shield (much narrower below than above). Mental 1.36 times as wide as long. Infralabials 6 (5th and 6th somewhat keeled), bordered below by five large sublabial shields. First pair of sublabials separated by an elongated scale in contact with the large mental and followed behind by two pairs of similar-sized scales and numerous small elongated scales that increase in size until about the middle of the throat, but then reduce in size posteriorly. Sides of neck with irregular erect spines, the largest about twice as high as wide. Dorsal scales large, rugose, often striated, forming regular (but not always aligned) transverse series; four vertebral rows with smooth scales (probably due to rubbing against rocks), other dorsals keeled, but dorsolateral and lateral scales keeled and moderately spinose. Dorsals in 34 transverse series (from first row posterior to occipitals to row above vent) and 20 longitudinal rows. Ventrals smooth, mostly quadrangular, occasionally pentagonal near middle of venter posteriorly, mostly broader than long, two outer rows moderately keeled and weakly spinose, some scales of the 3rd row also weakly keeled, forming 14 longitudinal and 26 transverse series from axil to groin (with an additional seven rows to base of throat). A pair of enlarged hexagonal preanal plates (slightly longer than wide), with smaller plates anteriorly and on the sides. Limbs above with large, keeled, strongly spinose scales. Femoral pores 21 (10 left leg, 11 right). Differentiated femoral scales 46 (21 left, 25 right). Fourth toe on each foot with 16 subdigital lamellae. Tail with whorls of large, strongly keeled, spinose scales; each whorl separated by a smaller whorl of small, moderately keeled and weakly spinose scales; two upper lateral caudal scale rows consist of especially large and very strongly spinose scales (spines project backwards at angles of about 45 ); subcaudal scales of the larger whorl elongate and narrow, those of the smaller whorl much shorter, mostly pentagonal (occasionally rectangular), moderately keeled, those in the centre only weakly spinose.
Colour: (similar in life and in preservative; Fig. 6) Back dark brown to black with cream to yellow markings forming five interrupted transverse bands (with only slightly dark borders in preservative) between fore-and hindlimbs, which continue along the tail, together with an incomplete pale band immediately behind the occipitals and another on the nape that is divided medially and curved slightly backwards. Belly cream-white with a brown longitudinal band medially (six ventral plates wide) and short, broad, widely separated brown bands on either side between the limbs (at least five on the left, four on the right) which are often confluent with the darker parts of the back. The joining of these dark ventral and dorsal markings decorates the flanks with large cream-yellow spots/ blotches. Top of limbs dark brown with numerous irregular cream to yellowish spots and blotches; underparts of limbs mostly cream-white with irregular brown markings, occasionally bands. Top of head brown with scattered irregular cream markings; throat mottled in dark brown and cream-white, most dark markings forming four irregular, wavy transverse bands.
Cranial skeleton: (Fig. 7) Scales of the dorsal and temporal regions of the skull and the ventrolateral aspects of the jaws are underlain with rugose osteoderms. These osteoderms fuse to the proximal parietal, frontal and postorbital bones, although the mesokinetic and metakinetic joints appear unobstructed and flexible. Lateral maxilla and anterior aspect of the premaxilla lack osteoderms. The parietal is pentagonal, with five osteoderms that underlie the parietal shields fused to its dorsal surface and a bifid medioposterior process that extends either side of the sagittal crest of the supraocciptial. Three large osteoderms are fused to the frontal, which is unpaired and clasped by the parietal at its posterolateral edge. The upper temporal fenestra is obscured anteriorly by a large osteoderm fused to the dorsal surface of the postorbital bone, and posteriorly by an unfused rectagonal osteoderm that overlies the squamosal. Premaxilla is unpaired and contains seven pleurodont teeth and five foramina, with a dorsal process that extends posteriorly to be clasped by the nasals, which themselves insert into the frontal. The maxilla is scinciform, with a deeply grooved crista dentalis, nine left or eight right lateral foramina, and 19 teeth. Teeth display pleurodont attachment and are unicuspid, with a slight concave surface where they connect with the mandibular teeth. No palpebral is present and the prefrontal connects directly to the anteriormost superorbital osteoderm. The jugal is triangular in cross-section and asymmetrically T-shaped, with a tapering anterior process and a broad, truncated posterior process that extends along and past the posterior edge of the maxilla. Lacrimal bone is small, flattened and oval. Pterygoids are edentate and extend back to connect with the quadrates, becoming C-shaped in cross-section posterior to the epipterygoid condyle. The squamosal is curved and blade-like, circular in cross-section anteriorly, becoming flattened posteriorly, where it articulates with the cephalic condyle of the quadrate and the braincase. Supratemporals are flattened, ovoid and not fused with the elongate paroccipital processes. The posterior aspect of the prootic is not fully fused with the oto-occipital, resulting in a deep groove along the dorsal aspect of the para-occipital processes. Quadrates are very broad with a pronounced ridge and concave region at the lateral edge of the adductor musculus mandibulae posterior origin. The supraoccipital has a strong sagittal crest that extends posteriorly to contact the ventral surface of the medioposterior process of the parietal.
The prootic bears an extended alar process and a well-developed, rhomboid christa prootica, and a very weak supratrigeminal process. Basipterygoid processes are well developed and flattened. The lower jaw possesses a large adductor fossa, a highly flattened and medially extended retroarticular process, a medially open Meckelian canal that is closed posteriorly by a large splenial and a dentary with a strong subdental shelf; 21 mandibular teeth, and nine dentary foramina.
Postcranial skeleton: (Fig. 7) Tail complete, 26 presacral vertebrae, 32 caudal vertebrae. The haemapophyses of the first caudal osteoderms extend laterally to fuse to the posteroventral edge of the parapothysis, forming a biphid rib. Four cervical, three sternal, two xiphisternal, six left and seven right long asternal ribs with ossified costal cartilage, then six left and five right short asternal ribs and one very short pair of ribs immediately anterior to the sacral vertebrae. Cervical ribs 2-4 are distally flattened and biphid, with the ventral processes more elongated. Pubis flattened and curved with a large, ventrally angled pectineal tubercle. Pubic symphysis flattened and triangular, separating the pubes entirely. Hyperischiam and hypoischium well developed. Illium triangular in cross-section, with a feeble iliac tubercle. Sternal plate broad with no fontanelle. Interclavicle cruciform, clavicles flattened dorsally. Epicoracoid connects the scapular ray to the primary and secondary coracoid rays, but not to the anterior process of the scapular. Phalanges display a typical pattern of 2-3-4-5-3 for the manus and 2-3-3-5-4 for the pes. Metatarsal 5 with elongated medial process at midbody.
Dermal osteology: (Fig. 7) Dorsal and lateral sides of the trunk are covered in circular, well-separated osteoderms, dorsomedially unkeeled grading to well keeled and mucronate towards the sides. The nuchal osteoderms are small, becoming highly spined posterior to the tympanic opening. Ventral osteoderms are delicate and plate-like and restricted to the gular and anterior pectoral regions. The forelimbs are covered in keeled, non-imbricate, circular/rhomboid osteoderms, while the hindlimbs are well armoured, except for the ventral surface of the thigh which lacks osteoderms. Osteoderms on the posterior part of the hindlimbs are heavily spinose. The caudal osteoderms are large, robust and arranged in imbricated whorls. Caudal osteoderms are feebly keeled and mucronate along the dorsal and ventral aspects, becoming more heavily spined laterally. Ventrals occasionally pentagonal (TM 73290,78918,83532), longer than broad on anterior part of belly (TM 73290, 78918) or mostly square (NMB R9195, TM 51376 and 83000). All ventrals smooth in TM 51376; in NMB R9194 and TM 83532 only the outermost row of ventrals is moderately keeled and weakly spinose, with rows 2-3 very weakly keeled only; some scales of the 3rd row also very weakly spinose in NMB R9194 (including first inner row) and TM 73290, 78918, 83000; all three outer rows weakly keeled in NMB R9195. Ventrals in 23-29 (28 in allotype) transverse rows (6-9 additional rows on throat), and occasionally only 12 (NMB R9194, TM 51376) longitudinal rows. Enlarged hexagonal preanal plates 3 (TM 78918) or 4 (TM 83532); median preanal plates (pair) pentagonal in TM 42531, 51376, 78931 and 83000, heptagonal in TM 73290 (left side) which also has two extranumerary plates posterior to the large pair, and irregular in TM 78921; no enlarged plates anterior to median and lateral plates in TM 78931; enlarged median pair of plates in TM 42531, 51376, 78931 and 83000 much elongated, each plate about twice as long as wide. Femoral pores 20-24 (10-12 on each thigh, 10 in allotype), appearing as small, shallow pits in females. Differentiated glandular femoral scales in males 19-61 (9-35 per thigh). Fourth toe with 16-19 (18 in allotype) subdigital lamellae.
Colour: Dorsum dark brown to black in preservative. Back with 4-6 (usually 5) interrupted transverse bands between fore-and hindlimbs, which are without dark borders or with only feeble indications thereof after preservation in alcohol. In TM 73290 there is a squarish cream marking on the nape between the pale band behind the occipitals and the band on the neck. Belly with 5-6 brown crossbands on either side of the median band (comprised of at least six longitudinal rows of ventrals, sometimes eight at places (e.g. TM 51376)) which is prominent in the centre of the belly. Throat with bold, dark mottling or reticulations; occasionally some markings form transverse bands, and sometimes most of the throat is black, especially anteriorly (e.g. NMB R9195 and TM 83532). Colour: Similar to holotype. Back with 5-6 (4 in TM 73285) interrupted transverse bands (sometimes with slightly dark borders in preservative). Belly with 5-6 brown crossbands on either side of the median band (prominent in the centre of the belly). Throat with bold, dark mottling or reticulations; occasionally some markings form transverse bands.
Etymology. Named for the Kingdom of Eswatini, the country where most of the species' range is located. Both 'eSwatini' and 'Swaziland' derive from the word iSwazi, after the name of an early chief, Mswati II (c. 1820-1868).

Distribution. Highveld and Middleveld of Eswatini in Hhohho, Manzini and Shiselweni
Regions, and adjacent areas in the South African provinces of (eastern) Mpumalanga (in Nkomazi municipality) and (northern) KwaZulu-Natal (in uPhongolo and Abaqulusi municipalies) (Fig. 9) at elevations of 462 to 1,139 m a.s.l.
Natural history. Diurnal and rupicolous, living in deep, horizontal (or gently sloping) crevices in granitic rock along hillsides, usually in the partial shade of trees ( Fig. 8A; see also Jacobsen, 1989). According to R.C. Boycott (in litt., 2019), rocky terrain in closed canopy bushveld is the preferred habitat in Eswatini. A specimen in Ithala Game Reserve in KwaZulu-Natal was photographed on a tree trunk (ReptileMAP, VM no. 152451). When grasped by the hind limb, an individual from the type series performed an unusual anti-predator behaviour by repeatedly flexing and extending the inhibited limb caudally, so as to pull the captors' digits directly onto the very sharp whorl of spines at the base of the tail (E.L. Stanley, 2008, personal observation).
It differs from other species of Smaug as described above in the diagnosis of S. swazicus sp. nov. (but maximum SVL in S. barbertonensis is 140 mm, and femoral pores in males are 8-11 per thigh).
Most similar to S. swazicus sp. nov. and S. warreni, but easily distinguishable by its colour pattern (see comparisons in diagnosis of S. swazicus sp. nov. above); by usually having more elongate and spinose scales at the edges of the ear openings (shorter and non-spinose in S. swazicus sp. nov. and S. warreni); and quadrates lacking a pronounced ridge and concave region at the lateral edge of the adductor musculus mandibulae posterior origin (with a pronounced ridge and concave region in S. swazicus sp. nov.). Also differs as follows: outer occipitals shorter than the adjacent inner ones (of about equal length in S. warreni); head relatively wider than the other two species (head width/head length = 80-92% vs. 73-84% in adults); lower numbers of transverse dorsal scale rows (29-32 in 81%) compared to S. swazicus sp. nov. (32-37 in 86%) and S. warreni (32-38 in 92%); and lower numbers of longitudinal dorsal scale rows (20-24, mean 21.7) compared to S. warreni (22-28, mean 23.6).
Colour: (Figs. 2, 12B and 12C) Back dark brown to black with cream to yellow markings (mostly transversely enlarged) forming 4-5 (usually 4, as in holotype TM 4273, Fig. S1) interrupted transverse bands (usually with slightly dark borders in preserved material) that continue onto the tail, together with a band on the nape and a cream spot, blotch or elongate marking (absent in NMB R9193 and TM 73286) immediately behind the median occipitals. Belly mostly brown (older preserved material; including the holotype, Fig. S2) to black (as in life) with a few cream patches or short 'bands' on either side, which are occasionally joined to the pale bands on the side of the back. The flanks are dark brown to black, usually with narrow, cream to yellowish, vertically elongated bars, occasionally spots (e.g. TM 51066). Top of limbs with numerous irregular cream to yellowish spots and blotches; underparts of limbs mostly cream with irregular brown markings, occasionally bands. Top of head brown or black with scattered irregular cream markings. Throat (including sublabials) mostly black or brown with occasional irregular scattered cream markings, but about half dark and half pale in TM 73283 and 73293, and mostly plain cream in TM 4275 (juvenile).
Cranial skeleton, postcranial skeleton and dermal osteology (Fig. 11) The cranial skeleton, postcranial skeleton and osteoderms are all similar to those described for S. swazicus sp. nov. However, in S. barbertonensis ventral osteoderms were absent, and quadrates lacked a pronounced ridge and concave region at the lateral edge of the posterior origin of the adductor musculus mandibulae (quadrates have a pronounced ridge and concave region in S. swazicus sp. nov. Natural history. Diurnal and rupicolous, living in deep, horizontal (or gently sloping) crevices in and between large granitic boulders, often in the partial shade of trees ( Fig. 12A; see also Jacobsen, 1989). For S. 'barbertonensis ', FitzSimons (1943) noted that the diet is similar to that of S. warreni (see below), but includes cetonid beetles and small land snails; usually five young are produced, and based on his examination of a series of females, fertilisation occurs in early spring, with young born at the end of summer. However, one of FitzSimons' (1943) localities ('Hluti-Goedgegun', Eswatini) is within the range of S. swazicus sp. nov., so it is not possible to know which species his data applies to. Computed Tomography scanning of NMB R9192 revealed four large embryos, and a large beetle in the stomach.
Distribution. Restricted to the Barberton, Nelspruit and Khandizwe areas of eastern Mpumalanga Province, South Africa ( Fig. 9)  Notes. Van Dam's (1921) type locality of 'Barberton' does not indicate the exact location at which the specimens were collected, so it is considered appropriate to treat both 'Barberton Townlands' (Jacobsen, 1989) and 'Barberton army base' as topotypic.
Most similar to S. swazicus sp. nov. and S. barbertonensis but easily distinguished by its colour pattern (see comparisons in diagnosis of S. swazicus sp. nov. above); by usually having shorter and blunter scales at the edges of the ear openings compared to S. barbertonensis; and quadrates without a pronounced ridge and concave region at the lateral edge of the adductor musculus mandibulae posterior origin (with a pronounced ridge and concave region in S. swazicus sp. nov.). Also differs as follows: outer occipitals and scales adjacent to them of about equal length (outer occipitals usually shorter than the adjacent inner ones in the other two species); head narrower than S. barbertonensis (head width/head length = 73-83% vs. 80-92% in adults); generally higher numbers of transverse dorsal scale rows (32-38 in 92% of specimens) than S. barbertonensis (29-32 in 81%); and greater numbers of longitudinal dorsal scale rows (22-28, mean 23.6) than S. barbertonensis (20-24, mean 21.7). . Frontonasal in contact with the rostral and loreals (often in narrow contact, especially on right side in TM 50130), separating the nasals, latter slightly swollen; nostril-with a large inner flap attached posteriorly-situated in the posterior part of the nasal and in contact with the loreal and 1st supralabial (TM 50130: loreal in very narrow contact with nostril on left side of head, and separated from nostril by upward prolongation of first supralabial on right; posterior part of both supranasals separated by a suture to form a small rectangular scale) (N = 16). Frontal usually separated from the frontonasal by a pair of prefrontals, but in broad contact in TM 47449, 50130 (Fig. 13), 50660-1 (N = 16), and NMZB-UM 30514 (D.G. Broadley, 2014, personal communication). Scales at anterior edge of ear opening usually 4-5 (3 on right side of head of TM 63567) projecting outwards as flattened, somewhat spatulate, spines, the lowermost one narrow and slender (especially so in TM 47449), the one above it distinctly spatulate and the largest; middle scales usually somewhat blunt and rounded, but generally long and somewhat spinose in TM 47449,50130,53869,70961 (N = 16). Preoculars 1; supraoculars 4 (2nd and 3rd on left side largely fused in NMB R10913); supraciliaries usually 4-5 (3 on left side of TM 78967, 4 left and 6 right in TM 13639, 6 on left side of 78969); loreals 1; suboculars 4-5, but 3 on left side of NMB R10911; supralabials (anterior to median subocular) usually 4 (3 on one side in four specimens, 5 on one side of TM 78967); infralabials usually 6 (5 on both sides of TM 582, 5 on one side of TM 47449, 7 on both sides of TM 15320, 7 on one side in six specimens); sublabials 5; occipitals usually 6, but an additional-often narrow and much elongated-scale medially in 41% of specimens (median scale large in TM 50660, granular in TM 78966); outermost occipitals and those adjacent to them of similar size and length, but scales of the inner ( Colour: (Figs. 2 and 15B) Back usually sandy brown with irregular, black-bordered, cream or white blotches (ocelli) forming 5-6 (seven in NMB R10911) slightly to greatly interrupted transverse bands between the legs that continue onto the tail, together with a band immediately behind the occipitals and another on the nape (N = 16). This colour pattern is often evident in live (Fig. 15B) and preserved specimens (Fig. 2, specimens preserved for over 30 years). Ocelli may be in close proximity and even set within a continuous black band. Occasional specimens have grey-brown backs with small, widely separated white spots lacking obvious black-borders (e.g. fig. 212 in Reissig, 2014). All six NMB specimens from Manyiseni region in KwaZulu-Natal (preserved for over 12 years) have medium brown backs and the pale markings have only moderately distinct borders. Belly white to cream, usually with numerous small, square, rectangular or irregular pale to dark brown markings; occasionally mostly without markings except for the sides (e.g. NMB R10898 and 10912) or with a large dark blotch on each ventral (NMB R9292). The flanks are mostly pale whitish, occasionally with some darker colouring and pale vertical bars. Top of limbs with numerous irregular cream to yellowish spots and blotches; underparts of limbs mostly cream with occasional scattered, irregular brown markings. Top of head tan/khaki brown with dark brown patches and scattered, irregular, cream markings (or small yellow speckles or blotches, observed in photographs of live specimens; also Ping, 2019); throat mostly white to cream with varying amounts of darker markings in the form of small spots and blotches (often extensive and bold (e.g. TM 53869), but less so than in S. swazicus sp. nov.).
A revised diagnostic key to the genus Smaug  (2020)

DISCUSSION
Examination of voucher specimens (NMB) used for the molecular analysis of , as well as most other available museum material of the three lineages, indicated that the 'Eswatini' lineage-including populations in a small area on the northern Eswatini-Mpumalanga border, and northern KwaZulu-Natal Province in South Africa-was readily distinguishable from S. barbertonensis sensu stricto (and S. warreni) by its unique dorsal, lateral and ventral colour patterns. FitzSimons (1943) had in fact noted differences in colour pattern between specimens of S. barbertonensis from the type locality of Barberton and specimens from 'Hluti-Goedgegun' in Eswatini (now referred to the new species), but this had been regarded as merely representing regional variation. Multivariate analyses of scale counts and body dimensions indicates that the 'Eswatini' lineage and S. warreni are most similar. In particular, S. barbertonensis differs from the other two lineages by its generally lower numbers of transverse rows of dorsal scales, more spinose scales at the anterior edges of the ear openings, and a relatively wider head. Also, the outer and adjacent inner occipital scales in S. warreni are of similar length, distinguishing it from the other two species which have the outer occipital usually slightly shorter that the adjacent inner occipital. High resolution Computed Tomography reveals differences in cranial osteology between specimens from all three lineages, with the 'Eswatini' lineage being remarkable in having a pronounced ridge and concave region at the lateral edge of the posterior origin of the adductor musculus mandibulae.