Rock island melody: A revision of the Afroedura bogerti Loveridge, 1944 group, with descriptions of four new endemic species from Angola

Four new species of flat geckos in the Afroedura bogerti Loveridge, 1944 group are described from south-western and west-central Angola. The description of these new species significantly restricts the distribution range of typical A. bogerti, a morphologically very similar species, from which they differ genetically by 5.9–12% divergence for the mitochondrial 16S ribosomal RNA gene. Morphologically and genetically, Angolan Afroedura are divided into two main groups: a mostly south-western coastal group and a west-central inland/highland group. These two groups are further divisible into three and two subgroups respectively, all geographically isolated, differing by a combination of the following features: colouration, average adult size, number of mid-body scale rows, number of scale rows on dorsal and ventral surface of each tail verticil and if nostril scales are in contact or not. All five Angolan species are morphologically distinguishable and in agreement with the molecular results. An updated dichotomous key to the Afroedura transvaalica group is provided. The new discovery adds to a growing number of endemic Pro-Namib reptiles described from Angola in recent years.


Introduction
The current reptile diversity in Angola comprises about 300 species Branch et al. 2019a; and subsequent descriptions listed below). New species are constantly being added (Conradie et al. 2012Stanley et al. 2016;Marques et al. 2019aMarques et al. , b, 2020Branch et al. 2019b;Ceríaco et al. 2020a, b, c;Hallerman et al. 2020) and more new species are expected in the country (Branch 2016;Branch and Tolley 2017;Branch et al. 2017a, b;Butler et al. 2019). As Angola is rebuilding after decades of civil war, more remote regions are becoming accessible to researchers, with the promise of additional cryptic species.
The Gekkonidae in Angola comprises eight genera, Afroedura (one species), Afrogecko (one species), Chondrodactylus (three species), Hemidactylus (eight species), Kolekanos (one species), Lygodactylus (five species), Pachydactylus (11 species) and Rhoptropus (seven species) (see Marques et al. 2018Marques et al. , 2020Branch et al. 2019a;Ceríaco et al. 2020a, c). A total of 15 species of geckos are considered to be endemic to the country. This includes the monotypic Kolekanos plumicaudus (Haacke, 2008), the recently re-discovered Afrogecko ansorgii (Vaz Pinto et al. 2019), six species of Hemidactylus (Ceríaco et al. 2020a, c), three species of Lygodactylus , three species of Rhoptropus and Pachydactylus angolensis Branch et al. 2019a). Afroedura bogerti is currently regarded as a near-endemic species as a specimen was also found in northern Namibia (Branch 1998;Griffin 2003). Loveridge (1944) transferred African Oedura to the new genus Afroedura and described A. karroica bogerti from Angola after examining specimens collected during the 1925 Vernay Angolan Expedition. He designated this material as the type for the new genus and regarded it as part of the 'halli/karroica' group. Later, Onderstall (1984) investigated relationships between Afroedura species after examining material from south-western Angola collected by Wulf Haacke in the 1970s. He placed A. k. bogerti in the 'transvaalica' group, alongside A. transvaalica (Hewitt, 1925) and A. loveridgei (Broadley, 1963) and elevated A. bogerti to full species status, based on differences in numbers of paired scansors per digit and possession of a verticillate tail. Such findings were supported by recent phylogenetic studies, confirming that A. bogerti is part of the 'transvaalica' group (Jacobsen et al. 2014;Makhubo et al. 2015;Branch et al. 2017a), with other species occurring to the east in Mozambique, Zimbabwe and South Africa. This relationship of the A. bogerti group to the 'transvaalica' group is of zoogeographic interest as it is separated from other members of the group by nearly 2000 km and occupies more arid and/or mesic environments.
Up to now, Afroedura bogerti was the only species of flat gecko known from Angola Branch et al. 2017b). In a recent phylogenetic study, in which a new species from the 'transvaalica' group was described from central Mozambique (Afroedura gorongosa), cryptic diversity within Angolan A. bogerti populations was documented (Branch et al. 2017a) and this was elaborated on in more detail in a subsequent study ). Here we take the opportunity to investigate the Angolan A. bogerti group in detail and describe four new endemic species.

Molecular analyses
We used molecular methods to support the identified morphological groupings. Recent studies on the genus Afroedura (e.g. Branch et al. 2017a) have shown that the fast evolving 16S rRNA gene, with its hypervariable regions, perfectly resolves the different species within  Angola (data for A. wulfhaackei sp. nov. also contain the morphologically-indistinguishable Afroedura sp. 5-7 clades). Confirmed locality records are indicated by solid circles, with type localities indicated by stars. Coloured polygons indicate potential distribution areas with more than 90% of climatic habitat suitability. this genus and even provides insights into the intraspecific structure of the different monophyletic lineages. Therefore, for our molecular comparisons, we combined newly-sequenced mitochondrial 16S data for specimens from different Afroedura populations in Angola and complemented these with previously published (Branch et al. 2017a) sequences from GenBank. Total genomic DNA was extracted from tissue samples using the E.Z.N.A. Tissue DNA Kit (VWR/Omega bio-tek) and the Qiagen DNeasy Tissue Kit, following the manufacturer's protocols. A portion of the mitochondrial genome (16S rRNA gene) was PCR amplified and sequenced, following the methods described in Schmitz et al. (2005). All sequences have been deposited in GenBank (Table 1).
DNA sequences were aligned using the original chromatograph data in the programme BioEdit (Hall 1999) using ClustalX (Thompson et al. 1997) and the resulting alignment corrected manually. We sequenced 589 bp of the 16S rRNA gene, including some fast-evolving, indel-rich loop regions. We deleted 6 bp that could not be unambiguously aligned, resulting in a total sequence length of 583 bp for each voucher. We used Bayesian Inference (MrBayes v.3.26; Ronquist et al. 2012) and Maximum Likelihood (RAxML v.7.0.4;Stamatakis et al. 2006), using the rapid hill climbing algorithm and the GTR+G model of nucleotide substitution, following Stamatakis et al. (2006), to assess phylogenetic relationships. The best-fit model of sequence evolution for the Bayesian analysis was the TIM2+I+G model with values of I = 0.5140 and G = 0.7180, selected using jModeltest v.2.1.7 (Darriba et al. 2012) using the Bayesian Information Criterion (BIC).
Bootstrap analyses (BS) with 1000 pseudoreplicates were used to evaluate relative branch support in the ML analysis. We regarded tree topologies with bootstrap values of 70% or greater as supported (Huelsenbeck and Hillis 1993). Bayesian analyses were run for 10 million generations using four chains, sampling every 1000 generations, with the first 25% of trees discarded as burn-in. Stationarity, convergence and mixing of the parameters (ESS values) for the Bayes runs were checked in Tracer v.1.7.1 (Rambaut et al. 2018). Clades with posterior probabilities (PP) ≥ 0.95 were considered strongly supported. In total, sequences from 47 Afroedura specimens from Angola were included in the phylogenetic analysis (Table 1) to corroborate our morphological analyses. Afroedura loveridgei (MF565446), shown by Branch et al. (2017a) to be a sister taxon, but still well outside the Angolan clade, was chosen as outgroup. Pairwise comparisons of uncorrected sequence divergences (p-distance) were computed in MEGA X (Kumar et al. 2018; Table 2).

Morphology
We examined material in the collections of the Port Elizabeth Museum (PEM), Ditsong National Museum of Natural History (formerly Transvaal Museum ~ TM), Museum für Naturkunde, Berlin, Germany (ZMB) and Instituto Superior de Ciências da Educação da Huíla (ISCED), Huíla, Angola (see Material examined under systematic accounts). High quality photographs of the type specimen of Afroedura bogerti were also examined.
As with many other gecko genera, Afroedura is morphologically conservative, with few discrete differences. The following characters (detailed in Jacobsen 1992) were assessed: 1) presence or absence of internasal granules between the nasal scales; 2) number of postmental scales; 3) number of scales in contact in a straight line between the anterior corners of eyes across the crown of the head; 4) number of scales between upper edge of earhole and rear margin of eye, counted along the shortest distance between them; 5) number of scales between nostril and front edge of orbit, excluding postnasal; 6) number of enlarged supralabials to the angle of the jaw at mid-orbital position; 7) number of enlarged infralabials to the angle of the jaw at mid-orbital position; 8) number of mid-body scale rows (MSR), counted at the widest part of the trunk; 9) number of scale rows on dorsal surface per tail whorl (counted 3-6 verticils, posterior to the cloaca); 10) number of scales rows on ventral surface per tail whorl (counted 3-6 verticils, posterior to the cloaca); and 11) number of precloacal pores in males.
The following measurements were taken in millimetres (mm) using a digital calliper (accuracy of 0.1 mm) with the aid of a Nikon SMZ1270 dissecting microscope: 1) snout-vent length (SVL -from the tip of the snout to the cloaca with the gecko flattened on its back), 2) tail length (TL, only original tails were measured); 3) head length (HL -tip of snout to retroarticular process of jaw); 4) head width (HW -widest point of head, approximately at the level of eyes); 5) snout length (SL -tip of snout to front of orbit); 6) eye diameter (ED -measured in horizontal orientation); 7) ear to eye length (EE -top edge of earhole to back of eye); 8) ear opening (EO -greatest length); and 9) internostril distance (IN -shortest distance between nostrils). All head measurements were taken on the right side of the head.
Measurements and scale counts were carried out mostly by WRB and WC incorporated newly-collected material to supplement the final dataset. Only adult material was used for additional comparative morphological analysis. Specimens were considered adult (sexually mature) at SVL > 38 mm (following Makhubo et al. 2015). Boxplots were created for all measurements and scale counts. Only the following measurements SVL, HL, HW, ED, EE and IN, number of scales rows on the dorsal surface per tail whorl, precloacal pores and MSR were included in morphological comparisons. The other measurements and scale counts were found not to be informative for further comparative analysis. These morphological data were tested for significant differences amongst species pairs using a Tukey multiple comparison of means test. Differences were considered significant at P < 0.05. All above-mentioned quantitative morphological comparisons were conducted using R v.3.4.1 (R Development Core Team 2014).

Potential distribution model
In order to identify potential distribution areas for Angolan Afroedura, an bioclimatic niche model was conducted using Maxent (Yang et al. 2013). The sampling area included a buffer of two degrees (~250 km) from the most peripheral observations of Afroedura species. Nineteen bioclimatic variables were obtained in WorldClim dataset (Fick and Hijmans 2017; http://www.worldclim.org/) at a spatial resolution of 30 arc-second (~ 1 km 2 ). For those variables, we ran a correlation model to eliminate collinearity between variables in the sampled area and within sample points (Candau and Flemin 2005) and variables, with correlation coefficient ≥ 0.7, were selected in order to capture the bioclimatic range over the distribution of the species (Enriquez-Urzelai et al. 2019). Therefore, the variables included for the analysis were: mean diurnal temperature range (BIO1); maximum temperature of the warmest month (BIO5), minimum temperature of the coldest month (BIO6); annual precipitation (BIO12); precipitation seasonality (BIO15), precipitation of wettest quarter (BIO16). Given the small sample size for some species, we ran a cross-validation model using all the samples but leaving one out in each run, and hinge features with the regularisation parameters set to 2.5, to produce smoother response curves and reduce overfitting (Bittencourt-Silva et al. 2016;Briscoe et al. 2016). The final maps were generated selecting areas with more than 90% of habitat suitability for each species.

Molecular analyses
Molecular analyses confirm that all Angolan Afroedura form a monophyletic clade, distinct from other members of the 'transvaalica' group (Branch et al. 2017a). The Angolan clade is resolved into eight genetically well-supported clades (Fig. 2), most of which show consistently large genetic differences between them (5.9-12%, Table 2) to a level that can be considered representative of species-level divergence ( Table 2). While these values are lower than observed between other species of the 'transvaalica' group (Branch et al. 2017a), they are still surprisingly high for a local monophyletic radiation and indicative of recent and ongoing speciation in the genus Afroedura in Angola.
Two sister clades occur exclusively in the coastal region (except for the first subclade, see below) with a third related clade (Afroedura sp. 2) occurring a little further inland in the Serra da Neve mountain range (Fig. 1). The southernmost taxon within the identified coastal region (Afroedura sp. 1) is currently restricted to the arid Namib Desert region of south-western Angola and clusters with the species from Serra da Neve. Both clades are genetically well differentiated and show minimal internal genetic variation (> 0.8%, Table 2). The sister taxon to those two taxa (Afroedura sp. 3) is much more widely distributed across the semi-arid 'Pro-Namib' habitats of the coastal littoral north of Moçâmedes (= Namibe) (Fig. 1). Congruent with its much larger distribution, it also contains some clear, but consistent, genetic substructure (Fig. 2, Table  2). There are two well-supported major subclades within this taxon and the second, larger subclade shows some additional substructure. The first subclade contains two specimens from Bimbe (NB 745 -6) on the Humpata plateau which are genetically quite distant from the rest of the coastal clade (~ 4.1%, Table 2). However, a third sample from the same locality (NB 743) groups well within the coastal clade. Follow-up analyses will be needed to resolve the status of this population which may prove to be not conspecific with Afroedura sp. 3.
Five further, strongly supported, monophyletic clades occur above the Angolan escarpment and within the region of the putative type locality. The Namba population (here referred to as A. bogerti sensu stricto) is genetically well-differentiated from four southern highland clades (Afroedura sp. 4-7) and shows negligible internal genetic variation (> 0.15%, Table 2). The four remaining southern highland clades , each of which is strongly supported, are genetically closely related (Table 2). This complex is very similar to the monophyletic clade Afroedura sp. 3 in having a comparatively-wide distribution ( Fig. 1) and containing a surprising, but consistent, genetic substructure. The geographic and taxonomic implications of these groupings require further analysis. The clades of Afroedura sp. 4-7 all show very low internal variation versus higher inter-clade variation. Genetic distances between Afroedura sp. 4-7 are about, on average, half as great as the shortest distances identified between the major clades identified in Angola (3.3-4.2% vs. 5.9-12.0%; Table 2).

Morphology
Results for the morphological analysis are summarised in Table 3 and Figure 3 and are discussed in more detail in  the species descriptions below. The Angolan material can be broadly divided into two main morphological groups: 1) south-western (mostly) coastal group: Afroedura sp. 1 and Afroedura sp. 3, defined by higher precloacal pores in males (10-12), higher numbers of scale rows on ventral and dorsal surface per tail verticil (5 and 6, respectively), nasals always in contact and ventre immaculate, without pigmentation; 2) west-central inland/highland group: comprising species Afroedura sp. 4-7 and A. bogerti sensu stricto, defined by lower precloacal pores in males (8-10), lower numbers of scale rows on ventral and dorsal surface per tail verticil (4 and 5, respectively), nasals 8 9-12 (9.7) 11-12 (11.5) 9-11 (10.2) 8 (8.0) Ventral rows per tail verticil 4 (4.0) 4-5 (4.0) 5-6 (5.5) 5-7 (5.0) 4 (4.0) Dorsal rows per tail verticil 5 (5.0) 5-6 (5.1) 6-7 (6.6) 6-7 (6.1) 5 (5.0) Scales below 4 th toe 6-9 (6.9) 6-9 (7.4) 6-8 (7.7) 6-10 (8.0) 9-11 (  in contact (68% of specimens) or separated by smaller granules (33% of specimens) and ventre pigmented with fine black specks. Morphologically, we could not separate members of the Afroedura sp. 4-7 clades, which are supported by close phylogenetic relationships. For that reason, we grouped the morphological data of the four clades together. The south-western coastal group can be further divided into a southern and northern group, based on average adult SVL (57.6 vs. 51.3 mm) and overall colouration (see below). An isolated inland population on the Humpata plateau agrees with the northern lowland coastal group morphologically, with the exception of the darker pigmented ventre with an immaculate throat. The two west-central inland/highland groups are morphologically very similar and differ only in the average number of mid-body scale rows (73 vs. 79). An apparently isolated population on Serra da Neve (Afroedura sp. 2) agrees morphologically with this inland highland group in most aspects, but differs in that the nasals are always in contact (versus 33-68% of the time). The Tukey multiple comparisons of means showed significance differences amongst most pairwise comparisons and support the above morphological separation (Table 4).

Species descriptions
In summary, both genetics and morphology (particularly diagnostic features, such as the nasal condition [separated or in contact], colouration [ventral pigmentation], snout vent length, number of mid-body scale rows, number of pre-cloacal pores, number of dorsal and ventral scale rows per tail verticil), as well as geographical segregation, suggest that most of the major genetic clades recovered should be considered as separate species. We, therefore, take the opportunity here to provide an updated description for A. bogerti and describe the remaining clades as separate species. We apply the general lineage-based species concept, treating all populations that represent independent historical lineages supported by multiple different lines of evidence as listed above as separate species (de Queiroz 1998 Table 3 Note. The Namba Flat Gecko, Afroedura bogerti was described using a single male specimen collected by Harry and Alan Chapman in 1925 during the Vernay Expedition to Angola (Loveridge 1944;Branch et al. 2017b). According to Loveridge (1944), the specimen was collected at 'Namba (Mombolo)', even though Bogert (1940)  A. bogerti -A. donveae P = 0.004 P = 0.273 P = 0.771 P = 0.262 P = 0.298 P = 0.031 P < 0.001 P < 0.001 P = 0.014 P = 0.988 A. vazpintorum -A. donveae P < 0.001 P < 0.001 P < 0.001 P = 0.671 P < 0.001 P < 0.001 P = 0.035 P < 0.001 P = 0.103 P < 0.001 A. wulfhaackei -A. donveae P < 0.001 P = 0.264 P = 0.004 P = 0.642 P < 0.001 P < 0.001 P < 0.001 P < 0.000 P = 0.012 P < 0.001 A. praedicta -A. donveae P = 0.015 P = 0.396 P = 0840 P = 0.951 P = 0.139 P < 0.001 P < 0.001 P < 0.001 P < 0.001 P = 0.730 A. vazpintorum -A. bogerti P = 0.932 P = 0.927 P = 0.424 P = 0.026 P = 0.992 P = 1.000 P < 0.001 P < 0.001 P = 0.168 P < 0.001 A. wulfhaackei -A. bogerti P = 0.942 P = 0.926 P = 0.767 P = 0.020 P = 1.000 P = 0.868 P =1.000 P = 0.983 P = 0.381 P < 0.001 A. praedicta -A. bogerti P = 1.000 P = 1.000 P = 1.000 P = 0.187 P = 1.000 P = 0.993 P = 1.000 P = 1.000 P = 1.000 P = 0.956 A. wulfhaackei -A. vazpintorum P = 1.000 P = 0.066 P = 0.839 P = 1.000 P = 0.769 P = 0.312 P < 0.001 P < 0.001 P = 0.676 P = 0.647 A. praedicta -A. vazpintorum P = 0.980 P = 0.912 P = 0.462 P = 1.000 P = 0.777 P = 0.868 P < 0.001 P < 0.001 P = 0.007 P = 0.005 A. praedicta -A. wulfhaackei P = 0.984 P = 0.967 P = 0.782 P = 1.000 P = 0.987 P = 0.195 P = 1.000 P = 0.972 P = 0.048 P = 0.043   General description. (description based on the adult material listed above). Head and body dorsoventrally compressed; SVL 46.4-53.5 (mean 50.0) mm, HL 10.9-12.6 (mean 11.6) mm, HW 7.8-11.2 (mean 10.6) mm, broadest at posterior level of eye and 1.1-1.4 (mean 1.3) times longer than wide. Eyes large (2.1-4.3 [mean 2.8] mm wide), pupil vertical with indented margins; circumorbital scales small and smooth, elongate at upper anterior margins, upper posterior scales with small upward pointing spines. Snout rounded, 4.5-5.0 (mean 4.3) mm long, slightly longer than distance between eye and ear openings (3.6-4.3 [mean 3.6] mm). Scales on top of snout smooth, rounded, scales to the side larger than central ones, with no intervening minute granules. Scales on snout slightly subequal in size to those on back of head or nape. Scales on eyelids larger than those on the crown. Circumorbital scales are separated by a row of smaller scales from the large scales on eyelid. Nostril pierced between rostral, two to three nasal scales and the 1 st supralabial; the supranasal being much larger than the subequal postnasals and being separated from each other by one to two smaller scales. Nostrils slightly elevated. Rostral roughly rectangular, but with its upper edges elongated due to extensions into the nostril. Supralabials 8-10 (mean 8.4). Infralabials 8-9 (mean 8.3). A total of 11-14 (mean 12.4) scales across the crown at level of front of eyes. A total of 14-16 (mean 15.4) scales from front of ear to back of eye. Mid-body scale rows 69-77 (mean 73.5). Original tail slightly dorsoventrally flattened and distinctly verticillate, with obvious lateral constrictions that are not that distinct to the tip of tail; each verticil comprising 5 imbricate rows of scales dorsally and 4 imbricate scale rows ventrally and with ventral scales approximately twice the size of those on the dorsal surface. Limbs well developed, hindlimbs slightly longer than forelimbs. All digits with a large pair of distal scansors, separated by a large, curved claw and followed after a large gap by a smaller pair of scansors; infero-median row of digital scales enlarged transversely, particularly towards the scansors, where the ter-minal scale adjoining the first pair of scansors may be medially constricted, swollen and scansor-like, enlarged subdigital lamellae on 4 th toe 6-9 (mean 6.9). Precloacal pores 8.
Greyish above with six irregularly spaced darker grey crossbars from the occiput to the sacrum, central crossbars fused to form an X-shape; head with a dark grey band across the posterior edge of crown encompassing a small central pale spot; dark grey to black bar from nostril to the anterior margins of the ear opening; a vague, thin pale canthal stripe, extends on both sides from the nasal region to anterior margins of eye, continuing posteriorly of the eye to above the ear opening; upper and lower labials grey with diffuse yellow edges; lateral sides of the body with very faintly yellow-mustard scales scattered amongst more prominent darker grey scales; limbs greyish above with scattered darker grey markings with intervening faint yellow-mustard colouration; tail (regenerated) with irregular grey mottling; iris golden with a black narrow elliptic pupil with crenulated edge and black reticulation; ventre uniform greyish with scattered black specks; ventral limbs with scattered black specks, more prominent than on the ventrum. In preservative (based on PEM R24187): dorsum with six irregularly-spaced darker grey crossbars from the occiput to the sacrum with beige intervening blotches, central crossbars fused to from dark grey X-shape; ventre beige with numerous small scattered black specks on each scale, more prominent anteriorly, laterally and posteriorly. Variation: Greyish to brownish above with five to six irregularly-spaced darker grey-brown W-shaped crossbars from the occiput to the sacrum, sometimes fused in the middle to form X-shapes, limbs and tail with grey blotches; ventre uniform greyish with scattered black specks.
Juveniles with more sharply-defined pattern. The original description of the holotype (Fig. 6) was based on a preserved and bleached specimen: 'Above, greyish; back with five or six obsolescent, irregularly W-shaped brown crossbars; limbs and tail immaculate. Below, whitish, uniform' (Loveridge 1944). Examination of images of the holotype confirm the presence of fine brown coloured specks ventrally that indicates that the ventre was not immaculate as originally described.
Natural history and habitat (Fig. 4A). An exclusively rupicolous species sheltering in crevices and under flakes of exfoliating rock amongst large granite boulders during the day. Individuals were observed at night foraging in vertical rock faces of large granite boulders. The surrounding habitat was mostly montane grasslands and some cultivated fields, but also included stunted altitude miombo and some Afromontane forest elements in steeper gorges and between boulders. It should be noted that the southern face of Namba Mountain contains the most extensive and well-preserved Afromontane forest patches in Angola, recognised as an important biodiversity reservoir and regional hotspot (Mills et al. 2013). Distribution and conservation. This species complex was previously considered to be widespread in granite boulders throughout south-western Angola Marques et al. 2018). Here, we show that it is restricted to the Namba Region in the southern parts of Cuanza-Sul Province, Angola (Fig. 1). Namba consists of a 25 km west-east orientated mountain range that climbs to an elevation of 2,420 m a.s.l. and towers southwards above the highland plateau of Mombolo. Our new series was collected at elevations of 1,750-1,850 m a.s.l., at the base of the south-facing slopes of the Namba Mountain chain, but the species probably occurs in suitable habitat higher up and possibly well above 2,000 m a.s.l. (Fig. 1).
Considering the deep genetic divergence between this species and Afroedura present at the southern edge of Mombolo and the recognition of Namba as a biodiversity hotspot, it seems plausible for this species to have evolved in isolation on this mountain 'island'. Nevertheless, A. bogerti may also extend its distribution further north into other poorly-studied mountain ranges in Cuanza-Sul Province (Fig. 1). The species seems to be relatively common in the Namba Mountains, but its biology remains poorly known.

Afroedura wulfhaackei sp. nov.
http: Tables 3, 5 Note. The phylogenetic analysis identified four well-defined clades (Afroedura sp. 4-7) within the southern inland/highland major clade (see Results). Our data show that this southern major clade clearly represents at least one undescribed species of Afroedura. Despite comparatively-low genetic distances between the four clades recovered (Table 2), they were all retrieved separately in our phylogenetic analyses and, as our morphological examination failed to separate the four clades, we cannot assign them as separate species. Therefore, we decided to take a conservative approach and to describe only one of the clades as a new species for now and tentatively assign Afroedura sp. 5-7 to this species pending further work. Thus, we grouped the material according to these clades and selected the clade called here Afroedura sp. 4 (Fig. 2) as the name-bearing clade for our new species. Material from lower-altitude localities near Bocoio in Benguela Province groups morphologically with our new species, but needs to be tested in a phylogenetic framework. , with a small mid-ventral incision for the removal of liver sample. Measurements and meristic characters of holotype presented in Table 5. Head and body dorsoventrally compressed; HL 11.9 mm, HW 9.0 mm, broadest at posterior level of eye and 1.32 times longer than wide. Eye large (2.3 mm wide), pupil vertical with indented margins; circumorbital scales small and smooth, elongate at upper anterior margin, upper three posterior scales with small upward pointing spines. Snout rounded, 4.9 mm long, slightly longer than distance between eye and ear openings (4.0 mm). Scales on top of snout smooth, rounded, scales to the edge larger than central ones, with no intervening minute granules. Scales on snout slightly subequal in size to those on the back of head or the nape. Scales on eyelids larger than those on the crown, five scales deep from circumorbital scale to crown. Circumorbital scales are separated by a row of smaller scales from the larger scales on eyelids. Nostril pierced between rostral and three nasal scales; 1 st supralabial narrowly excluded from nostril; the supranasal being much larger than the subequal postnasals and separated from each other by two smaller scales. Nostrils slightly elevated. Rostral roughly rectangular, but with its upper edges elongated due to extensions into the nostril. Eight supralabials on each side, the labial margin flexing upwards at the rictus (approx. mid-orbital position), with 3-4 minute scales proximal to the flexure. Ten infralabials on either side, with a small scale proximal to the flexure. At the lip, mental slightly narrower than adjacent infralabial, only three quarters the width of rostral and in contact with two distinctly elongate postmen-tal scales. Scales on throat much smaller than those on belly; scales touching infralabials larger. Fifteen scales across the crown at level of front of eyes; 16 scales from ear to eye; 83 scales around mid-body. Ear opening deep, oblique and roughly oval, only half as high as wide (0.3 × 0.5 mm). Scales on dorsum smooth, non-overlapping, largest at mid-body, smaller on nape and tail base. Scales on ventrum flattened, not overlapping, more-or-less ovate at mid-ventrum, twice the size of lateral granules and 1.5 times those along the backbone. Original tail slightly dorsoventrally flattened and distinctly verticillate (11 verticils in total), with obvious lateral constrictions that are not that distinct to tip of tail; each verticil comprising 5 imbricate rows of scales dorsally and 4 imbricate scale rows ventrally and with ventral scales approximately twice the size of those on the dorsal surface. Limbs well developed, hindlimbs slightly longer than forelimbs, mite pockets (dermal crevices inhabited by small ectoparasitic mites) only present at anterior margin of hindlimbs. All digits with a large pair of distal scansors, separated by a large, curved claw and followed by a large gap (twice the length of terminal scansor) by a smaller pair of scansors; infero-median row of digital scales enlarged transversely, particularly towards the scansors, where the terminal scale adjoining the first pair of scansors may be medially constricted, swollen and scansor-like; enlarged subdigital lamellae on 4 th toe 9. Precloacal pores 9.

Synonym. Afroedura bogerti
Paratype variation (see Table 5 for more measurements and scale counts of type series). SVL varied from 49.1-55.7 mm; head length 1.34-1.45 times the head width; snout 1.99 times the diameter of eye. Supranasals always separated by smaller granules, usually with a single large granule in contact with the rostral between the supranasals, followed by 1-2 smaller granules in lateral contact; the first upper labial and rostral always enters the nostril and the width of the rostral at the lip margin is always wider than that of the mental; 2-3 postmental scales; supralabials 7-9, infralabials 8-9; scales between anterior edges of eyes 13-14; scales between nostril and anterior edge of orbit 7-9; scales between anterior edge of ear and rear margin of orbit 14-17; scales around midbody 76-82; subdigital lamellae on 4 th toe 7-8; dorsal scales per tail verticil 5; ventral scales per tail verticil 4. Precloacal pores 9.
Additional material variation. SVL 36.4-59.6 mm; original tail length 36.4-57.8 mm, 0.94 times SVL; head length 1.19-1.75 times head width; snout 1.91 times diameter of eye. The supranasals in contact in 22 specimens and separated by granules in eight specimens, usually with a single large granule in contact with the rostral between the supranasals, followed by 1-2 smaller granules in lateral contact; the first upper labial and rostral always enters the nostril and the width of the rostral at the lip margin is always wider than that of the mental; 2-4 postmental scales; supralabials 8-9; infralabials 8-9; scales between anterior edge of eye 12-16; scales between nostril and anterior edge of orbit 7-11; scales between anterior edge of ear and rear margin of orbit 15-18; scales around mid-body 75-88; subdigital lamellae on 4 th toe 7-9; dorsal scales per tail verticil 5 (TM 45366, TM 46588 and PEM R24743 with 6); ventral scales per tail verticil 4 (TM 46588 with 5). Precloacal pores 9-12.

Colouration.
In life (paratype PEM R24232, Fig. 5B): Greyish above with five irregularly-spaced darker crossbars from the occiput to the sacrum, each crossbar consisting posteriorly out of three to four black scales wide forming a W-shape; anterior to W-shape are 8-10 scales deep with a mix of dark grey and mustard colours; each dark crossbar separated by light grey to beige blotches; head with irregular dark grey-mustard blotches on the crown with intervening light grey colouration; dark mustard to dark grey bar from nostril to the anterior margins of the ear opening; a vague, thin pale grey canthal stripe, extends on both sides from the nasal region to anterior margins of eye; upper and lower labials grey with diffuse mustard edges; lateral sides of the body with a mix of dark grey and yellow-mustard colouration; limbs greyish above with scattered darker grey markings with intervening yellow-mustard colouration; tail (regenerated) with irregular grey-mustard mottling; iris golden with a black narrow elliptic pupil with crenulated edge and black reticulation with light grey intervening blotches; ventrum uniform greyish with scattered black specks; ventral limbs with scattered black specks, more prominent than on the ventrum. In preservative (holotype PEM R24234, Fig. 7): Dorsum with five irregularly-spaced dark grey W-shaped crossbars from the occiput to the sacrum with beige intervening blotches; ventrum is beige with numerous small scattered black specks on each scale, more prominent pos-teriorly. Variation: Greyish to brownish above with five to six irregularly-spaced darker grey-brown W-shaped crossbars from the occiput to the sacrum, limbs and tail with grey blotches; ventrum uniform greyish with scattered black specks. Juveniles have sharper patterns and colours.
Natural history and habitat (Fig. 4B). An exclusively rupicolous species living in crevices between rocks or under flakes of exfoliating rocks in boulders at elevations of 920-2,055 m a.s.l. Specimens were collected in cracks of relatively small-to medium-sized boulders of carbonatitic origin surrounded by montane grassland. Some individuals were also found under flakes in large granite boulders and often in steep vertical rock faces and overhangs.
Distribution and conservation. This species is currently known from southern Cuanza-Sul, central Huambo and northern Benguela Provinces, Angola (Fig. 1). It appears to have a relatively large, but patchy, distribution on the Angolan highlands and may extend its range into neighbouring provinces. Although sometimes locally common, it appears to be absent from vast areas in-between, where the species would be expected to occur. Populations in isolated granite outcrops may be threatened by removal of rock flakes for construction of homes and other buildings.

Serra da Neve Flat Gecko
Osga-achatada da Serra da Neve Etymology. The specific epithet reflects the earlier prediction by WRB of the potential existence of an isolated population of Afroedura at Serra da Neve. We use the specific epithet "praedicta", the Latin participle meaning predicted or anticipated, formed in the feminine genitive to match the gender of Afroedura.

Diagnosis.
A member of the greater 'transvaalica' group as it possesses two pairs of enlarged scansors per digit and a strongly verticillate and flattened tail (Jacobsen et al. 2014 .0] and 5-6 [mean 5.1] in A. wulfhaackei sp. nov.; but lower than the 5-6 (mean 5.5) ventral and 6-7 (mean 6.6) in A. donveae sp. nov. and 5-6 (mean 5.0) and 6-7 (mean 6.1) in A. vazpintorum sp. nov.); ventral surfaces grey with black specks on scales (similar to A. bogerti and A. wulfhaackei sp. nov.; immaculate in A. donveae sp. nov. and A. vazpintorum sp. nov.). Afroedura praedicta sp. nov. also differs from its sister highland species A. bogerti sp. nov. and A. wulfhaackei sp. nov. in that the nasals are always in direct contact (versus mostly separated).
Holotype description. Adult male; SVL 51.6 mm; tail 37.0 mm (regenerated tail, except for the first verticil), with a small mid-ventral horizontal incision for the removal of liver sample. Measurements and meristic characters of holotype presented in Table 6. Head and body dorsoventrally compressed; HL 12.9 mm, HW 7.9 mm, broadest at posterior level of eye and 1.6 times longer than wide. Eye large (2.6 mm wide), pupil vertical with indented margins; circumorbital scales small and smooth, elongated at upper anterior margin, the most upper posterior scale with very small upward pointing spines. Snout rounded, 5.0 mm long, slightly larger than the distance between eye and ear openings (4.2 mm). Scales on top of snout slightly granular and elevated, rounded, mostly equal in size, with no intervening minute granules. Scales on snout slightly subequal in size to those on back of head or nape, which in turn is irregular in size and mostly smooth. Scales on eyelids larger than those on the crown, six scales deep from circumorbital scales to crown. Circumorbital scales are separated by a row of smaller scales from the larger scales on eyelid. Nostril pierced between rostral, three nasal scales; 1 st supralabial in contact with nostril; the supranasal being much larger than the subequal postnasals and are separated from each other by two smaller scales. Nostrils slightly elevated. Rostral roughly rectangular, but with its upper edges elongated due to extensions into the nostril. Eight supralabials on each side, the labial margin flexing upwards at the rictus (approx. mid-orbital position), with 2-3 minute scales proximal to the flexure. Nine infralabials on either side, with a small scale proximal to the flexure. At the lip, mental slightly narrower than adjacent infralabial, only three quarters the width of rostral and in contact with three distinctly elongate postmental scales. Scales on throat much smaller than those on belly; scales touching infralabials larger. Twelve scales across the crown at level of front of eye; 9 scales between nostril and front of eye; 16 scales from ear to eye; 74 scales around mid-body. Ear opening deep, oblique and roughly round, backward pointing, nearly equal as long as wide (0.7 × 0.6 mm). Scales on dorsum smooth, non-overlapping, largest at mid-body, smaller on nape and tail base. Sales on ventre flattened, not overlapping, more-or-less ovate at mid-ventrum and twice the size of lateral granules and 1.5 times those along backbone. Tail regenerated, except for the first verticil, with obvious lateral constriction before regenerated tail start; first verticil comprising 5 imbricate rows of scales dorsally and 4 imbricate scale rows ventrally and with ventral scales approximately twice the size of those on the dorsal surface. Limbs well developed, hindlimbs slightly longer than forelimbs, both without obvious mite pockets at posterior margin of limb insertions, mite pockets present at anterior margin of hindlimbs. All digits with a large pair of distal scansors, sep-arated by a large, curved claw and followed after a large gap (twice the length of terminal scansor) by a smaller pair of scansors; infero-median row of digital scales enlarged transversely, particularly towards the scansors, where the terminal scale adjoining the first pair of scansors may be medially constricted, swollen and scansor-like; 9 enlarged scale rows under 4 th toe. Precloacal pores 8. Table 6 for more measurements and scale counts of type series). SVL varied from 46.0-51.6 mm; head length 1.09-1.46 times head width; snout 1.71 times diameter of eye. The supranasals always in contact; the first supralabial enters the nostril (except for ZMB 91607 and NB 855) and rostral always enters the nostril and the width of the rostral at the lip margin is always wider than that of the mental; 2-3 postmental scales; supralabials 8-10, infralabials 8-9; scales between anterior edge of eye 13-15; scales between nostril and anterior edge of orbit 9-11; scales between ear and eye 13-16; scales around mid-body 73-78; subdigital lamellae on 4 th toe 9-11; dorsal scales per tail verticil 5; ventral scales per tail verticil 4; precloacal pores 8.

Colouration.
In life (holotype NB 854, Fig. 5C). Dark grey above with random yellow-olive scattered scales, with six irregularly-spaced, darker grey to black W-shaped crossbars, first one broken up into three blotches, each crossbar anteriorly bordered by a row of light white to yellow scales; head mostly grey with scattered darker grey and yellow scales, dark black bar from nostril to posterior of eye, continuing anteriorly of the eye to ear opening; no thin pale white canthal stripe; upper and lower labials dark grey with diffuse lighter grey edges; lateral sides of the body with a mix of dark grey and yellow blotches; limbs darker grey above with scattered yellow scales; tail (regenerated) with dark black blotches and irregular white to light grey mottling; iris dark brown to golden with a narrow black elliptic pupil with crenulated edge and black reticulation; ventre uniform greyish with scattered black specks; ventral surface of limbs with scattered black specks, more prominent than on the ventrum. In preservative (holotype NB 854, Fig. 8): dorsum with five distinct, irregularly-spaced, dark grey W-shaped crossbars anteriorly, with beige intervening blotches, the posteriorly crossbar on the nape are broken up dark grey blotches; dorsally, the arms and legs are beige with irregular darker grey blotches; tail (regenerated part) with light grey to white mottling on darker grey to black background; dorsally, the head has mottled dark brown scales, dark grey bar running from the nasals through eye to anterior of the ear opening; supralabials dark brownedged ventrally; infralabials scattered with dark brown markings dorsally; ventrum uniform greyish with scattered black specks; ventrally, limbs with scattered black specks, more prominent than on the ventrum. Variation: Similar colouration and patterning as the holotype (preserved) and paratype (life). Dorsal dark W-shaped crossbars number 5-6. Regenerated tails with fine dark brown to black mottling. Juveniles with more sharply-defined pattern and darker colouration.
Natural history and habitat (Fig. 4C). A rupiculous species found inside deep crevices and fractured rocks in large boulders of extrusive origin. All specimens were found on one site between 1,900-2,000 m a.s.l. The surrounding vegetation included altitude dwarf miombo and montane elements. Found in syntopy with Cordylus phonolithos at Serra da Neve.

Distribution and conservation.
This species is known only from the alkaline mountain complex of Serra da Neve (Fig. 1). Etymology. This gecko is named after Donvé Branch, WRB's wife, with the following personal quote: "This, the most beautiful of all the Angolan flat geckos, is named for my wife, Donvé Branch ('Dove') who bore the long periods I was away on fieldwork, and to whose nest I returned, and surrounded me with love until the end". The name is constructed in the feminine singular genitive.
Holotype description. Adult female: SVL 61.0 mm; tail 59.0 mm (full original tail), with a small mid-ventral incision for the removal of liver sample. Measurements and meristic characters of holotype presented in Table 7.
Head and body dorsoventrally compressed; HL 13.6 mm, HW 11.4 mm, broadest at posterior level of eye and 1.19 times longer than wide. Eyes large (3.4 mm wide), pupil vertical with indented margins; circumorbital scales small and smooth, elongated at upper anterior margin, upper three posterior scales with small upward pointing spines. Snout rounded, 5.6 mm long, longer than distance between eye and ear openings (4.5 mm). Scales on top of snout smooth, rounded, equal in size, with no intervening minute granules. Scales on snout slightly larger than those on back of head or nape. Scales on eyelids larger than those on the crown, 5 scales deep from circumorbital scale to crown. Nostril pierced between rostral, three nasal scales; 1 st supralabial narrowly excluded from nostril; the supranasals are much larger than the subequal smaller postnasals, in broad contact with each other. Nostrils slightly elevated. Rostral roughly rectangular, but with its upper edges elongated due to extensions to the nostril. Nine supralabials on each side, the labial margin flexing upwards at the rictus (approx. mid-orbital position), with 1-2 minute scales proximal to the flexure. Ten infralabials on either side, with a small scale proximal to the flexure. At the lip, mental slightly narrower than adjacent infralabial, 63% of rostral and in contact with three distinctly elongated postmental scales. Scales on throat much smaller than those on belly, scales touching infralabials larger. Thirteen scales across the crown at level of front of eyes; 11 scales between nostril and front of eye; 13 scales from ear to eye; 74 scales around mid-body. Ear opening deep, oblique and roughly oval, only half as high as wide (0.4 × 0.9 mm). Scales on dorsum smooth, non-overlapping, largest at mid-body, smaller on nape and tail base. Scales on ventrum flattened, not overlapping, more-or-less ovate at mid-ventrum, twice the size of lateral granules and 1.5 times the size of scales along the dorsal mid-line. Original tail slightly dorsoventrally flattened and distinctly verticillate (16 whorls in total), with obvious lateral constrictions; each verticil comprising 7 rows of imbricate scales dorsally and 6 rows of imbricate scales ventrally, with ventral scales approximately twice the size of those on the dorsal surface. Limbs well-developed, hindlimbs slightly longer than forelimbs, both without obvious mite pockets at posterior margin of limb insertions; mite pockets present on anterior margin of hindlimbs. All digits with a large pair of distal scansors, separated by a large, curved claw and followed after a large gap (twice length of terminal scansor) by a smaller pair of scansors; infero-median row of digital scales enlarged transversely, particularly towards the scansors, where the terminal scale adjoining the first pair of scansors may be medially constricted, swollen and scansor-like; 8 enlarged scale rows under 4 th toe.

Colouration.
In life (holotype PEM R17937, Fig. 5D): Yellowish above, brighter anteriorly, with nine irregularly-shaped darker brown to black crossbars from the occiput to the sacrum, each separated by thinner (3-4 scales deep), light yellow crossbars; head with dark brownblack blotches on the crown with intervening pale yellow colouration; dark brown bar from nostril across the upper margins of the ear opening, connecting with dark brown lateral bar on the neck; a thin pale yellow canthal stripe extends on both sides from the nasal region to anterior margins of eye, continuing posteriorly from the eye on to the nape; upper and lower labials light grey with diffuse brown edges; lateral sides of the body with a mix of dark grey and yellow blotches; limbs yellowish above with scattered darker grey markings; tail (original) with irregular dark brown bars, separated by yellow bars anteriorly, posteriorly with black and white bars; iris dark black with irregular golden spots, a black, narrow, elliptical, crenulate-edged pupil with black reticulation; ventrum uniform beige with scattered brown specks on lateral edges only; ventrally, limbs with scattered brown spots. In preservative (holotype PEM R17937, Fig. 9): dorsum with nine irregularly spaced dark brown crossbars from the occiput to the sacrum with beige intervening blotches; dorsally, arms and legs darkly barred; tail with nine dark brown bars, bolder towards the tip; dorsally, head with mottled dark brown scales; a light beige canthal stripe from nostril to anterior corner of eyes, continuing from the posterior part of the eye to above the ear opening; dark brown bar running from the nasals above the ear opening to the neck; light beige stripe (one-scale-wide) from above the supralabials to the ear opening (two-scales-wide); supralabials ventrally dark brown-edged; infralabials with scattered dark brown markings dorsally; ventrum is mostly immaculate, with dark brown spots ventrally on the arms and legs; ventrally, tail with nine dark brown bars, bolder posteriorly. Variation. Similar colouration and patterning as to the holotype. Dorsal crossbars are often fused to form 2-3 dark brown, X-shaped crossbars. Original tails with 6-7 broad, dark brown to black bars, separated by light beige to white bars; some specimens with fine onescale-wide black bar separating the lighter bars. Regenerated tails with fine dark brown mottling. Juveniles with more sharply-defined patterns.
Natural history and habitat (Fig. 4D). One or two elongated white eggs were laid on limestone rock, to which they adhere in captivity; eggs measure 12.9-14.7 mm (mean 13.8 mm) × 10.5-12.1 mm (mean 11.6 mm) (n = 5) and hatch after 88-92 days, with hatchlings measuring 30.2-30.9 mm SVL and 26.5-27.9 mm tail length (n = 5) (Barts and Haacke 2010). A rupicolous species living in crevices between rocks and under flakes of exfoliating rock amongst larger granite boulders, at elevations of 340-355 m a.s.l. in the arid Namib Desert. It appears to be associated with large granite outcrops in semi-desert shrublands, along the drier fringes of the Pro-Namib. Vegetation includes Senegalia (= Acacia) mellifera, Senegalia spp., Commiphora sp., Boscia foetida and Salvadora persica.
Distribution and conservation. Currently known only from the south-western parts of Namibe Province in Angola, in granite formations around Tambor and on the right bank of the mid-Curoca River (Fig. 1). The species remains poorly known, but it is probably stable in numbers as the local habitat is currently not threatened. It likely occurs in Iona National Park.

Coastal Flat Gecko
Osga-achatada da planície costeira Etymology. This species is named in honour of father and son, Pedro and Afonso Vaz Pinto, two enthusiastic Angolan naturalists with whom WRB spent a great deal of time in the field, to recognise their contributions in collecting and studying Angolan herpetofauna. The name is constructed in the masculine plural genitive.

Diagnosis.
A member of the greater 'transvaalica' group as it possesses two pairs of enlarged scansors per digit and a strongly verticillate and flattened tail (Jacobsen et al. 2014). Part of the A. bogerti-group which differs from other members of the 'transvaalica' group by having less than 86 mid-body scale rows (vs. 97-102 in A. gorongosa, 113-120 in A. loveridgei, 102-119 in A. transvaalica); by the rostral bordering the nostril (nostril excluded from rostral in A. loveridgei); by the anterior nasals always in contact (separated by 1-3 granules in A. gorongosa; always in broad contact in A. loveridgei; usually in broad contact in A. transvaalica ~ 3-18%); and in having 11-14 scales between the anterior borders of the eyes (19-22 in A. gorongosa; 15-19 in A. loveridgei; 15-20 in A. transvaalica) (comparative data fide Branch et al. 2017a). Afroedura vazpintorum sp. nov. differs from other members of the A. bogerti-group by a combination of the following characters (see Tables 3, 4): 73-86 (mean 80.3) mid-body scale rows (69-77 [mean 73.5] in A. bogerti, 76-88 [mean 79.3] in A. wulfhaackei sp. nov., 64-78 [mean 72.8] in A. donveae sp. nov., 73-78 [mean 74.8] in A. praedicta sp. nov.); by the anterior nasals always in contact (similar to A. donveae and A. praedicta sp. nov.; ~ 33% of the time in contact in A. bogerti; ~ 68% of the time in contact in A. wulfhaackei sp. nov.); in each verticil having 5-6 (mean 5.0) ventral and 6-7 (mean 6.1) dorsal rows of scales (5-6 [mean 5.5] and 6-7 [mean 6.6] in A. donveae sp. nov.; 4 and 5 in A. bogerti and A. praedicta sp. nov., 4-5 [mean 4.0 Holotype description. Adult male; SVL 50.6 mm; tail 44.5 mm (partly regenerated tail); with a small mid-ventral incision for the removal of liver sample. Measurements and meristic characters of holotype presented in Table 8. Head and body dorsoventrally compressed; HL 11.1 mm, HW 7.4 mm, broadest at posterior level of eye and 1.5 times longer than wide. Eyes large (2.5 mm wide), pupil vertical with indented margins; circumorbital scales small and smooth, elongate at upper anterior margin, the most upper posterior scale with minute upward pointing spines. Snout rounded, 4.3 mm long, almost equal to distance between eye and ear openings (4.1 mm). Scales on top of snout slightly granular and elevated, rounded, mostly equal in size, with no intervening minute granules. Scales on snout slightly subequal in size to those on back of head or nape which, in turn, is irregular in size and mostly smooth. Scales on eyelids larger than those on the crown, 6 scales deep from circumorbital scales to crown. Circumorbital scales are separated by a row of smaller scales from the larger scales on eyelid. Nostril pierced between rostral, three nasal scales; 1 st supralabial narrowly excluded from nostril; the supranasal being much larger than the subequal postnasals and being separated from each other by two smaller scales. Nostrils slightly elevated. Rostral roughly rectangular, but with its upper edges elongated due to extensions into the nostril. Nine supralabials on each side, the labial margin flexing upwards at the rictus (approx. mid-orbital position), with 2-3 minute scales proximal to the flexure. Eight infralabials on either side, with a small scale proximal to the flexure. At the lip, mental slightly narrower than adjacent infralabial, only three quarters the width of rostral and in contact with three distinctly elongated postmental scales. Scales on throat much smaller than those on belly; scales touching infralabials larger. Eleven scales across the crown at level of front of eye; 8 scales between nostril and front of eye; 14 scales from ear to eye; 73 scales around midbody. Ear opening deep, oblique and roughly oval, backward pointing, much taller than wide (0.6 × 0.1 mm). Scales on dorsum smooth, non-overlapping, largest at mid-body, smaller on nape and tail base. Scales on ventrum flattened, not overlapping, more-or-less ovate at mid-ventrum, twice the size of lateral granules and 1.5 times those along backbone. Original tail slightly dorsoventrally flattened and distinctly verticillate (7 whorls in total), with obvious lateral constrictions; each verticil comprising 6 imbricate rows of scales dorsally and 5 imbricate scale rows ventrally and ventral scales approximately twice the size of those on the dorsal surface. Limbs well developed, hindlimbs slightly longer than forelimbs, both without obvious mite pockets at posterior margin of limb insertions, mite pockets present at anterior margin of hindlimbs. All digits with a large pair of distal scansors, separated by a large, curved claw and followed after a large gap (twice the length of terminal scansor) by a smaller pair of scansors; infero-median row of digital scales enlarged transversely, particularly towards the scansors, where the terminal scale adjoining the first pair of scansors may be medially constricted, swollen and scansor-like; 9 enlarged scale rows under 4 th toe. Precloacal pores 12. Table 8 for more measurements and scale counts of type series). SVL 43.9-58.8 mm; original tail length 42.3-67.0 mm, 0.94 times SVL; head length 1.07-1.61 times head width; snout 1.75 times diameter of eye. The supranasals are always in contact; the first upper labial and rostral always enter the nostril and the rostral at the lip margin is always wider than the mental; 2-3 postmental scales; supralabials 8-10, infralabials 8-9; scales between anterior edges of eyes 11-15; scales between nostril and anterior edge of orbit 7-11; scales between anterior edge of ear and rear margin of orbit 7-11; scales around mid-body 73-86; subdigital lamellae on 4 th toe 6-10; dorsal scales per tail verticil 6-7(mean 6.1); ventral scales per tail verticil 5-6(mean 5.0); precloacal pores 9-12.
Light grey to yellow-olive above with six irregularly-spaced darker, brown to black, W-shaped crossbars from the occiput to the sacrum, central two broken up to form irregularly-shaped crossbars; each crossbar separated by yellow-olive to light grey blotches; head mostly light grey with scattered darker grey scales; dark brown band from nostril across the upper margins of the ear opening connecting with dark brown lateral bar on to the neck; a thin pale yellow canthal stripe extends on both sides from the nasal region to anterior margins of eye, continuing on to the nape; upper and lower labials light grey with diffuse dark brown edges; lateral sides of the body with a mix of dark grey and yellow blotches; limbs yellowish above with scattered darker grey markings; tail (original) with irregular dark brown bars separated by yellow bars; tail (regenerated) with black mottling on light grey background; iris dark brown with a black narrow elliptic pupil with crenulated edge and black reticulation; ventrum uniform beige with scattered brown specks on lateral edges of ventrum alone; ventrally, limbs with scattered brown spots. In preservative (holotype PEM R24118, Fig. 10): dorsum with three distinct, irregularly-spaced, dark brown W-shaped crossbars anteriorly, with beige intervening blotches, the posterior crossbars are broken up and irregularly spaced; dorsally, arms and legs darkly barred; tail (original part) with dark brown bars that are not well defined; tail (regenerated part) with dark brown mottling on beige background; dorsal head with mottled dark brown scales, light beige canthal stripe from nostril to anterior corner of eyes, continuing posterior to the eye to above the ear opening; dark brown band running from the nasals, above the ear opening, to the neck; light beige stripe (one-scalewide) from above the supralabials to the ear opening (two-scales-wide); supralabials dark brown edged ventrally; infralabials scattered with dark brown markings dorsally; ventrum mostly immaculate with dark brown spots ventrally on arms and legs. Variation: Similar colouration and patterning to the holotype (preserved) and paratype (as above, in life). Dorsal crossbars are often fused to form an X-shape or are irregular. When present, crossbars number 5-6. Coastal material is often very light in colouration and not as boldly patterned or brightly coloured compared to highland material. Original tails with 6-7 broad, dark brown to black bars, separated by light beige to white bars, some specimens with a fine one-scale-wide black bar splitting the lighter bars. Regenerated tails with fine dark brown to black mottling. Juveniles with more sharply defined patterns.
Natural history and habitat (Fig. 4E, F). This species was frequently observed at night in coastal areas hunting in small branches, bushes and small trees growing amongst rock crevices between large granite boulders. When disturbed, it quickly jumps to the rock faces and finds shelter under granite flakes. Mainly rupiculous, it is strongly associated with large granite boulders present throughout the semi-arid ecosystems of the Angolan Kaokoveld. Usually found in rocky habitats with succulent plants and dominated by acacia savannah (Senegalia spp.), preferring more wooded and less arid environments when compared to A. donveae sp. nov. Some specimens, referable to A. vazpintorum sp. nov., were found at high altitude near Bimbe, in crevices in pre-cambrian limestone formations, a rocky habitat framed by stunted Afromontane forest elements and surrounded by extensive montane grassland.

Distribution and conservation.
Known from various localities in the coastal lowlands of Namibe and Benguela Provinces of Angola, stretching for about 250 km along the coast and up to 60 km inland, which is probably a fair reflection of the species' global range (Fig. 1). Occurs from near sea level to 500 m elevation. An apparently isolated population occurs on the Angolan highlands along the Humpata Plateau. This population occurs at a much higher elevation (above 2,000 m a.s.l.) and may warrant further phylogenetic analysis employing nuclear markers. The species seems common and widely distributed, mostly in relatively undisturbed habitat and, therefore, it is likely not threatened.

Discussion
When Loveridge (1947) provided the first syntheses of the genus Afroedura, he documented a total of only 12 species. When Onderstall (1984) reviewed the genus, he divided it into distinct morphological groups and described two new subspecies from South Africa and Eswatini (formerly Swaziland). Since then, Mouton and Mostert (1985) described a new endemic species from the Western Cape Province in South Africa, Jacobsen et al. (2014) formally added nine endemic species from the northern provinces of South Africa and elevated three subspecies to species status and Branch et al. (2017a) described a new endemic species from Gorongosa Mountains in central Mozambique, bringing the total number of species in the genus to 28. In the latter study, cryptic diversity within the Angolan A. bogerti group was documented and addressed in more detail afterwards .
Our study provides morphological evidence to support the genetic results, previously reported by Branch et al. (2017a, b) and describes four additional species in the genus, increasing the number of Angolan Afroedura species from one to five and to a total of 32 in Africa. Recent phylogenetic studies in South Africa have documented additional cryptic diversity in more widespread species (e.g. A. amatolica, A. halli, A. nivaria and A. pondolia) and new taxa are awaiting formal description (Makhubo et al. 2014;Busschau et al. 2019). More studies on the Angolan flat geckos are needed to understand their biology, natural history and true distribution and, thus, assess their conservation status. The radiation within Angolan Afroedura seems to have followed a major split that separated the west-central highlands from a south-western group. Being a strictly rupicolous gekkonid genus that typically displays low dispersal mobility, populations could easily become isolated in suitable remaining habitat, thus leading to speciation events in these "rock islands" or inselbergs (Jacobsen et al. 2014;Branch et al. 2017a). All genetically-assigned specimens from the central Angolan highland group were collected at relatively high altitude in what is generally called the Angolan ancient massif or ancient plateau (Huntley 2019), mostly above 1,500 m a.s.l. So far, A. bogerti, A. wulfhaackei sp. nov. and Afroedura sp. 5-7 were found to be restricted to different ancient mountain chains of heterogeneous geology or in isolated outcrops on the old plateau, which may explain the complex speciation patterns exhibited in a relatively-constricted area. In contrast, most of the landscape of south-western Angola has been actively transformed by tectonic forces in more recent geological times, as the coastal plain and southern escarpment have attained the current profile during the Pleistocene (Feio 1981). Exceptionally, Serra da Neve inselberg is an alkaline extrusion dating from the Cretaceous (Pereira 1977) and, therefore, pre-dating by far the rising of the southern escarpment. Thus, the existence of the new species, A. praedicta sp. nov., from the higher elevation on the Serra da Neve inselberg was half-expected, while A. donveae sp. nov. appears to have evolved in association with arid granitic landscapes of the Pro-Namib. Interestingly, A. vazpintorum sp. nov. is comparatively much more widely distributed across the Angolan Kaokoveld, but also linked to the highlands on the top of the southern escarpment at the Humpata Plateau, where an additional population was found. It is likely that the southern escarpment population represents a yet undescribed species and this will be investigated further in forthcoming efforts using a multi-gene approach with a more comprehensive series. The genetic distance indicated that, using only a 16S marker is, on its own, a strong indication, but not enough to consider this isolated population as a formal species.
The obvious genetic (sub-)structure, identified between A. wulfhaackei sp. nov. and the three morphologically-indistinguishable clades Afroedura sp. 5-7 and within A. vazpintorum sp. nov. (Fig. 2), seems surprising at first glance. However, both these species have a relatively-wide distribution area, when compared to the other known Angolan Afroedura species (Fig. 1) and, therefore, they are likely to occupy a broader range of niches, which may suggest they are undergoing a speciation process. Rupiculous species are known to require very specific suitable habitats and this often leads to a micro-fragmentation within their distribution range. These microhabitats are frequently separated from other suitable microhabitats by comparatively long distances, so these species could be considered as allopatric, which could conceivably lead to rapid speciation. This pattern has been reported from other gecko species (Portik et al. 2013;Travers et al.2014). As genetic changes can establish themselves well before any significant and consistent morphological differences develop between populations (see Branch et al. 2014), the lack of morphological differences between the detected morphologically-identical Afroedura sp. 5-7 clades and A. wulfhaackei sp. nov. as well as in A. vazpintorum sp. nov. can be considered as expected. The exact taxonomic status of Afroedura sp. 5-7, as being either conspecific with A. wulfhaackei sp. nov., representing additional taxa or even being hybrid populations, needs to be clarified using a multi-gene study and the morphological analyses of additional specimens from those populations.
Overall, the speciation pattern observed for Angolan Afroedura is similar to that of sympatric species of Cordylus (also rupicolous). Angolan Cordylus also include a central highland and a south-western group. The former group currently comprises a single species, C. angolensis, but it may, in fact, represent a species complex (Stanley et al. 2016;Bates et al. in prep.). On the other hand, the south-western group has, in recent years, been split into three species: C. phonolithos endemic to the Serra da Neve inselberg (Marques et al. 2019b), C. namakuiyus from the coastal Kaokoveld (Stanley et al. 2016) and C. machadoi from the higher elevations of the southern escarpment (Stanley et al. 2016). This parallel evolution highlights the importance of 'rock islands' as speciation sources for Angolan squamates.
A flat gecko discovered in a rock crevice on the summit of the Otjihipa Mountains, northern Opu wo District, Namibia (Branch 1998;Griffin 2003;Branch et al. 2017b) was referred to as Afroedura cf. bogerti. However, no voucher material is available at the National Museum of Namibia (A.M. Bauer, pers. comm. May 2016), nor have any recent specimens been collected to allow us to assign it to one of the species described here. There is also the possibility that it may represent yet another undescribed species. A specimen of Cordylus was also found in the Otjihepa and Baynes Mountains Region of northern Namibia and was assigned to C. machadoi (Branch 1998;Griffin 2003), but was recently shown to be more closely related to C. namakuiyus from south-western Angola (De-Boer et al. 2019). Recently,  recorded Nucras aff. broadleyi, a recently-described species from south-western Angola (Branch et al. 2019), in Cunene Province of Namibia. This parallel between Angolan and Namibian species, as well as the importance of the Cunene River as a barrier to gene flow, needs to be studied further.
In recent years, we have seen Angola embracing international collaboration that has led to an increase in biodiversity knowledge (Huntley 2009;Huntley and Francisco 2015;Brooks 2012Brooks , 2013NGOWP 2017). This is especially true in the field of herpetology, with numerous new species of squamates being described in the last decade (Conradie et al. 2012Stanley et al. 2016;Marques et al. 2019aMarques et al. , b, 2020Branch et al. 2019;Ceríaco et al. 2020a, b, c;Hallerman et al. 2020; this study). Many of these new species descriptions apply to the more arid regions below the Angolan escarpment, reinforcing this area as a regional biodiversity and endemism hotspot ) that requires effective protection.