On the identity of west Saharan geckos of the Tarentola ephippiata complex (Squamata: Phyllodactylidae), with comments on an extreme case of syntopy with their close relative T. annularis

ABSTRACT Tarentola geckos have a widespread geographic distribution and occur both in the Palearctic and Afrotropical realms, as well as the Neotropical region. Particularly, across North Africa phenotypically similar and cryptic species can be found, like the west Saharan members of the T. ephippiata complex. However, the taxonomic relationships and phylogeographic patterns of these geckos are not fully understood. Here we show that some specimens of Tarentola geckos from Mauritania and southern Morocco previously identified as T. hoggarensis can actually be assigned to T. panousei, a taxon treated as a synonym until now. Because the corresponding type specimen has apparently been lost, we designate a neotype for T. panousei referring to a suitable specimen from the type locality. Based on a morphological examination of the neotype and comparative material we provide a detailed redescription of T. panousei and evaluation of its diagnostic characters. Moreover, we report on a syntopic occurrence of T. annularis and T. panousei on an isolated acacia tree in the Western Sahara and the (micro) habitat use of both species.

Arnold 2003; Lyra and Vences 2018). Similarly, the three trans-Atlantic Tarentola geckos in the Caribbean (Cuba, Bahamas, Jamaica) are doubtlessly old endemics that diverged approximately 23 million years ago from an earlier stage of the Atlantic Ocean and have nothing to do with anthropogenic translocation (Carranza et al. 2000;Gamble et al. 2011;Vasconcelos et al. 2012). In contrast to T. atlantica they are considered to represent a subgenus of their own. A more complex pattern can be found off the West African coast: The Palearctic Canary Islands were colonised twice via long transmarine dispersal, once the eastern islands by the T. mauritanica group, then independently the central and western islands by the subgenus Makariogecko. In contrast, a single species of Makariogecko gave rise to the whole radiation of the Afrotropical Cape Verdian Islands approximately seven million years ago (Carranza et al. 2000), which, however, in contrast to scincids (Chalcides (Carranza et al. 2008) and Chioninia, respectively (Carranza et al. 2001;Miralles et al. 2011)), do not represent different subfamilies, but even the same subgenus. However, for Tarentola remarkable biogeographical patterns are not only to be found on islands, but also across their continental distribution range. Although the Mediterranean T. mauritanica group is widespread from northern Italy and southern France southwards to the Morocco-dominated Western Sahara, it is replaced in the south by the T. annularis/ ephippiata group, which extends from the Western Sahara down to West and Central Africa (Joger 1984;Trape et al. 2012;Carranza et al. 2002;Rato et al. 2012).
Studying this transition between the Mediterranean, western Palearctic, and the Afrotropical biogeographical zones was also the inducement for a three-month mission to the western margin of the Sahara desert conducted by the Zoologisches Forschungsmuseum Alexander Koenig (ZFMK) in Bonn in the winter of 1999/2000. As an exemplary case study, the acquisition of reptile communities along a north-south gradient should shed some light on faunal and landscape history (Böhme et al. 2001), considering the fact that the transition has always been a dynamic oscillating process in geological history.
During this expedition, one night was spent close to a single acacia tree (Vachellia tortilis (formerly known as Acacia raddiana)) ( Figure 1) south of the Mauritanian harbour city of Nouadhibou in a monotonous sandy and nearly treeless desert landscape. The isolated acacia at this site (20°57' N, 16°31' W) was the only tree in a radius of at least 3 km, and when we inspected it after nightfall with torchlights, it turned out to be populated by two Tarentola species in relatively high densities: by T. annularis on the one hand ( Figure  2A) and by a member of the T. ephippiata complex on the other ( Figure 2B). Syntopic occurrence of both species on acacia trunks has only been briefly mentioned once, by Sow et al. (2015) from the Banc d'Arguin National Park, which extends not far south from Nouadhibou, along the Mauritanian coast. These authors referred to the second species as T. ephippiata hoggarensis, pending additional clarification of its taxonomic status. This paper aims to resolve the taxonomic position of this member of the T. ephippiata complex, and to discuss the syntopy between it and T. annularis at this specific site.
Taxonomic assignment of these west Saharan members of the T. ephippiata complex is problematic. According to the distribution maps in Joger (1984), Bons and Geniez (1996), Trape et al. (2012), Colacicco (2015) and Mártinez de Mármol et al. (2019), they should belong to the taxon hoggarensis Werner, 1937. However, the type locality of T. hoggarensis, the south Algerian Hoggar Mountains, is situated approximately 2000km eastwards in the interior of the Sahara. Werner (1937) had described it originally as a subspecies of Tarentola delalandii, a species endemic to the Canary Islands, whereas Pasteur (1960) treated it as a species, even a polytypic one, because he regarded the taxon panousei, which he had described the year before (Pasteur 1959) as a subspecies of T. hoggarensis. Joger (1984), in his great taxonomic revision of the genus Tarentola, again placed hoggarensis as a subspecies of T. ephippiata and considered panousei as its synonym and some subsequent authors (Bons and Geniez 1996;Colacicco 2015) followed this view. In contrast, Trape et al. (2012) again treated hoggarensis as a full species, followed by Martínez de Mármol et al. (2019), but the name panousei was no longer mentioned. However, because Joger (1984) had already given hints about some morphological, geographically correlated interpopulational differences within the huge distribution range of hoggarensis, the possible validity of Tarentola (hoggarensis) panousei again merits some attention. Joger and Bischoff (1989) listed specimens of this complex, which they had collected south of the Lesser Atlas (Icht), not far from the type locality of Pasteur's panousei, simply as T. ephippiata.
Tarentola ephippiata O'Shaugnessy, 1875 (sensu stricto) had been described from 'West Africa', restricted by Joger (1984) to Nigeria. He subsumed the morphological differences between this form and hoggarensis and saw evidence for intergrades between both populations from Niger and the Lake Chad area on the one hand; but on the other hand also infrasubspecific differentiation in various populations living in so-called 'favorable regions' ('Gunstregionen') within the Sahara.

Morphological analysis
To evaluate the possible distinctness and validity of Pasteur's taxon panousei, we performed a morphological analysis of 24 T. hoggarensis (sensu lato) specimens from the collection of the Zoologisches Forschungsmuseum A. Koenig (ZFMK), Bonn (see Appendix I: Material examined). The characters were partially selected and modified, based on  Figure 1. B: Second, syntopic Tarentola species on the same tree assigned here to T. panousei. Photographs by W Böhme previous taxonomic studies of Tarentola by Pasteur (1960), Joger (1984 and Vasconcelos et al. (2012).
Scale counts were made using a stereomicroscope, and morphological measurements were taken with a Vernier calliper to the nearest 0.1 mm. For all specimens we examined the sex and the following 25 characters and six ratios were determined: snout-vent length (SVL); tail length (TaL, measured only in specimens with complete original tails); total length (TL, as the sum of SVL and TaL); trunk length (TrL, from the posterior end of forelimb insertion to anterior end of hind limb insertion); tail width (TW, at the widest point of tail); head width (HW, widest part of head, usually at the temple); head height (HH, from the base of the maxilla to the top of head); nostril-eye distance (NED, posterior margin of nostril to anterior edge of eye); snouteye distance (SED, tip of snout anterior margin of eye); ear-eye distance (EED, anterior margin of ear to posterior margin of eye); number of supralabial sales (SLS, counted from rostral to posterior end of mouth cleft); number of non-divided lamellae under first hind toe (Lam1); number of undivided lamellae under fourth hind toe (Lam4); number of transverse dorsal tubercles (Trow, counted paramedially); gular scales (GS, from mental to gular fold); number of interorbital scales (IOS); shape of dorsal tubercles (DTshape): dorsal tubercles arranged in longitudinal rows, clearly distinct (D) or indistinct (ID) from neighbouring dorsal scales; supranasal scales (SNcontact) in contact (C) or separated (S); shape of internasal (IN shape) similar in size as supranasals (L) or smaller (S); shape of ear opening (Eoserr) anteriorly serrated (Serr) or smooth (Sm); contact between nostril and rostral (NRcontact) pointy (O) or broad (B); shape of lamellae at the base of 4th toe (Lbshape) replaced by scale row (R) or uniform (U); supraocular integument (SOIshape) soft (S) or ossified (O); shape of postocular stripe (POstripe) reaching axilla (A) or elbow (E). The colour pattern of each individual was carefully recorded.

Principal component analysis of the morphological data
A principal component analysis (PCA) of the above listed characters was conducted using the Ade4 package (Dray and Dufour 2007) in R (R Development Core Team 2012). Tal, TL, and ratios were excluded from the PCA, because of the presence of incomplete tails in many specimens. One specimen (ZMFK 84913) was excluded, because of its poor state of preservation that allowed no clear identification of some categorical characters. To control for the effect of body size, continuous variables were regressed on SVL and sex, and the residuals were used in the subsequent PCA. The analysis was performed by using mixed variables. All variables were included in a single PCA, in order to evaluate the meaningfulness of the categorical characters based on Pasteur's original description and compare them with both the quantitative traits used in his original description and additional traits used in the present study. Based on both quantitative and categorical variables, the first two principal components (PCs) are visualised in Figure 3. We applied ANOVAs to PCA results and SVL separately in order to investigate whether the specimens from south-east of Nouadhibou in Mauritania and southern Morocco differ significantly from those of southern Mauritania and Algeria.

Resurrection of Tarentola panousei based on morphological and statistical analyses
There are diagnostic differences between the three Operational Taxonomic Units (OTUs) classified as T. hoggarensis on the one hand and on the other hand, the series from Nouadhibou, plus two specimens from South Morocco used in this analysis (Table 1). Some of them are weaker, but some are strong enough to define Pasteur's (1959) nomen panousei as a distinct taxon at the species level.
According to our PCA analysis based on the morphological examination of Tarentola specimens from various localities in north-western Africa the first three components explained 30.6%, 13%, 10.6% of the total variation, respectively. The first component was clearly discriminative between the specimens from southern Mauritania and Algeria assigned to T. hoggarensis on the one hand, and the other operational taxonomic unit (OUT) (Figure 3). The highest loading for the first component refers to the continuous variable GS, and the highest loadings for the same component referring to categorical variables are DTshape, EOserr, POstripe and NRcontact. Based on the first component, the populations from Nouadhibou and South Morocco show a lower number of gular scales than the other OTUs. They also differ by the presence of smooth dorsal tubercles, a serrated ear opening, a postorbital stripe reaching only to the axilla and often a pointy contact of nostril and rostral scales, whereas none or only single of the other OTUs show the presence of these characters. The first axis did not separate different sexes in any of the OTUs. Based on the ANOVA, the two OTUs from Algeria and southern Mauritania were   1  12  14  15  1  14  17  14  1  12  15  Lam4  16  1  15  17  16  1  15  18  16  1  14  17  Trow  15  1  15  17  16  1  15  17  15  2  12  17  GS  34  3  29  37  35  2  33  37  29  1  27  31  IOS  14  1  12  15  12  1  11  13  12  1  11 15 According to our morphological investigation of both T. hoggarensis from southern Mauritania and the type material from Algeria (Hoggar) and individuals of Tarentola from locations in southern Morocco and Nouadhibou in western Mauritania, and the significant statistical distinctness of the two groups, we consider the latter OTU as distinct at the species level and assign it to the previously described T. panousei.

Neotype designation and redefinition of Tarentola panousei
Facing the complicated situation within the T. ephippiata complex in general, and within the taxon hoggarensis in particular, a clear definition of T. panousei based on type material is warranted. However, Pasteur's type specimen, which was published with an informal (private) catalogue no. (BGP 2.54), was stated by its author to be finally deposited in the Muséum d'Histoire Naturelle (MNHN) in Paris, is definitely not there (I Ineich and N Vidal, pers. comm.). The (unlikely) possibility of recovering it in the Natural History Museum (NHMUK in London), because Grandison (1961) had access to the specimen for her revision of West African Tarentola, where the specimen was included and figured, was investigated by P Campbell (pers. comm.) and can also be ruled out, because we were additionally told by I Ineich, that Georges Pasteur had informed the former curator in Paris, ER Brygoo, in July 1989, that he had given the specimen to the Institut Scientifique Chérifien in Rabat instead. However, our inquiry at this institution (Prof. M Fekhaoui) was also unsuccessful, and we were assured that the specimen was not in Morocco (S Fahd, pers. comm.). In view of the complex taxonomic situation within the T. ephippiata complex in general and in the taxon hoggarensis in particular, designation of a neotype for T. panousei is warranted. A suitable specimen from within the type locality (Bassin du Draa moyen) was available. Its redefinition, preceded by a short synonymy/chresonymy list, is given below: Tarentola panousei Pasteur, 1959 Tarentola panousei Pasteur, 1959  Diagnosis: A relatively large (up to 85.4 mm SVL) member of the Tarentola ephippiata complex with a lumpy habitus, which is distinguished from its former conspecific T. hoggarensis ( Figure 5A) by its greater snout-vent length (max. 85 mm vs. 73 mm), a lower number of gular scales (27-31 vs. 29-37), and often a higher tail width/snoutvent length ratio (0.13-0.16 vs. 0.09-0.16). Ear openings small, sometimes concealed, with a serrated (vs. smooth) anterior margin. Dorsal tubercles irregularly arranged (vs. in longitudinal rows) and only indistinctly differentiated from the surrounding scalation. Colour pattern more or less uniformly greyish beige with, in adults, only four indistinct pale dorsal blotches and a postocular dark stripe reaching the axillary region.
T. panousei can be distinguished from T. ephippiata by the presence of indistinct and flat dorsal tubercles (vs. slightly keeled), and from T. senegambiae by a lower number of interorbital scales (11-15 vs. 14-18) and a shorter snout-vent length (maximum 85 mm vs. 129 mm).
Less marked is a different arrangement of the rostral/nasal contact (less broad and more pointy in some T. panousei), the shape of the basal lamellae at the fourth toe (in some T. panousei replaced by scale row and not uniform), and the degree of ossification of the supraocular lamina (no clear differences). For additional differences in scalation see Table 2.
The half-grown female ( Figure 4, Table 2) has a total length of (52.3 + 43.6) 95.9 mm and is consequently smaller than the lost holotype, the snout-vent length of which was 67 mm (Pasteur 1960). Broad head (HW 10.7 mm; HH 7.3 mm) with blunt snout (SED/ SVL 0.11) ( Figures 4C to 4E) (Figures 4C to 4E). Eight supralabial scales, seven infralabial scales. Supranasal scales separated, internasal scale, large with a similar size like the supranasalia. Point contact between nostril and rostral. Supraocular integument smooth, ear opening serrated. Under the first hind toe, thirteen undivided lamellae, and under the fourth hind toe fourteen. Seventeen longitudinal rows of dorsal tubercles, although irregularly arranged and partially indistinct from the surrounding dorsal scalation. Fifteen interorbital scales, thirty scales from mental scale to the gular fold.

Systematics
Based on our morphological analysis, Tarentola panousei is phenotypically most similar to T. hoggarensis ( Figure 5A), with which it was confused by previous authors, who considered both names to be synonymous (see the synonymy-chresonymy list above). However, adult T. panousei show a higher maximal snout-vent length (up to 85.4 mm) with a lumpy appearance, compared with T. hoggarensis ( Figure 5A) and a more or less uniform greyish beige with only four indistinct, pale dorsal patches from snout to vent. Although some of Pasteur's (1960) qualitative diagnostic characters can be considered as strong and justified traits for the definitive identification of T. panousei based on our comparative examination of, for example, shape of ear opening (Eoserr) (in T. panousei: ear opening serrated), shape of postocular stripe (POstripe) (in T. panousei: postocular stripe up to the axilla) and shape of dorsal tubercles (DTshape) (in T. panousei: dorsal tubercles not in rows and indistinct from surrounding dorsal scales), others do not allow a clear separation from individuals of T. hoggarensis, for example, contact between nostril and rostral (Nrcontact), shape of lamellae at the base of 4th toe (Lbshape) and shape of the supraocular integument (SOIshape).
Overall, we conclude that Pasteur's (1959) recognition of a new species was correct, and that a revalidation of T. panousei as a species is fully justified. If one considers the differences between newly described or revalidated species within the T. mauritanica/ T. deserti complex (Joger 1984;Joger and Bshaenia 2010), the differences between T. hoggarensis and T. panousei are at least equivalent. Nonetheless, a future genetic corroboration of our conclusion would be highly desirable. Moreover, it must still be demonstrated that the subpopulations of T. panousei in southwest Morocco and those from south of Nouadhibou, separated by an air-distance of ca. 1 000 km (Figure 7, map), are identical, despite their strong resemblance regarding mensural and meristic characteristics shown. It must also be determined whether the geographically intermediate populations, in particular the coastal ones, will have to be assigned to T. panousei as well.

Habitat selection
Our nocturnal survey of the single, isolated acacia tree mentioned in the introduction (Figure 1), which harboured two closely coexisting species of Tarentola, yielded 23 specimens, of which seven T. annularis (four adults, three juveniles: ZFMK 79476-482) and eight T. panousei (seven adults and one juvenile: ZFMK 79522-529) were collected, the remaining specimens being released. The old acacia tree (Figure 1) offered two types of shelter or hiding places, larger clefts in the wood and small knotholes (Figure 8), of which the former ( Figure 9A) were occupied by T. annularis; whereas the latter ( Figure 9B) were used even by adult T. panousei specimens, which filled these holes completely with their bodies. Nearly all geckos were found and observed exclusively on the tree trunk. Only one big T. annularis male of 22 cm total length descended from the trunk and caught a migratory locust of c. 8-9 cm length, which was attracted by the headlight of our truck, at <1 m distance from the trunk, the distal parts of its toes were bent upwards in the fine desert sand in order to keep the adhesive lamellae clean. Three T. annularis juveniles ( Figure 10) were also seen on the sandy ground, but escaped immediately to the nearby tree trunk. No T. panousei was observed outside of its arboreal microhabitat.

Close syntopy on one single isolated tree trunk
This extreme syntopic situation of two similarly large gecko species suggested that T. annularis was superior in competition, because it occupied the more spacious shelter types. The smaller T. panousei specimens, in contrast, had to be content with the much smaller knotholes where adult specimens were hardly fitting in (Figures 8 and 9), as already indicated by Böhme et al. (2001). This was surprising, because T. annularis is otherwise a rupicolous species, whereas T. hoggarensis and its close relative T. panousei seem to be specialised tree dwellers. Even if T. annularis was, in contrast to T. panousei, also observed on the sandy ground, but very close to the acacia trunk, it is extremely unlikely for both species to have had genetic exchange with conspecifics in neighbouring trees, because these were virtually absent for a radius of at least several kilometres, a distance impossible to cross for either species. Given the extreme isolation of their habitat, we hypothesise that both micropopulations might have faced a long period of inbreeding, but without breakage of the isolating mechanism between these two congeners. This is important to note, because Grandison (1961) had reported on specimens intermediate between T. annularis and 'T. ephippiata' (currently T. senegambiae) from Senegal, Gambia and Guinea Bissau and speculated on 'a breakdown in whatever mechanism isolates' both species from each other. However, these two species are also more similar to each other Figure 7. Distribution map of the Tarentola ephippiata species complex in NW and W Africa, inspired by the grid maps in Trape et al. (2012). The numbered localities refer to the localities from where specimens were studied. T. panousei: 1 -Bassin du Dra moyen, Morocco (white star: type locality); 2 -S of Aouinet Torkoz, Morocco; 3 -SE of Nouadhibou, Mauritania; T. hoggarensis: 4 -Parc National Diawling; 5 -Lac Aleg, Mauritania; 6 -Kaedi, Mauritania; 7 -Tamourt Bougari, Mauritania; 8 -Tamchekket, Mauritania; 9 -Chlim, Mauritania; 10 -N of Tamanrasset, Algeria (black star: type locality). Map by M Flecks with respect to their body size. In contrast, the two strictly syntopic subpopulations of T. panousei and T. annularis, in the isolated acacia tree described here,`were well separated from each other morphologically, including in respect to size, despite the presumably long duration of their isolation from any other conspecific of either species.
Senegambian representatives of the T. ephippiata complex were separated as a subspecies T. e. senegambiae by Joger (1984) and raised to specific rank by Trape et al. (2012). According to the grid map by these authors, its Sahelian distribution range is separated from that of the Saharan T. hoggarensis, whereas T. annularis, sympatric with T. hoggarensis farther northwards, is missing from the T. senegambiae range, except for two rocky sites near Dakar (Yoff and Cap Manuel) and the rocky offshore islands Gorée, Ngor and Ile Madeleine (Cissé 1974;Böhme 1978;Cissé and Karns 1979;Joger 1982). According to Böhme et al. (2001), these marginal points were relictual occurrences where the likewise large, robust T. senegambiae was unable to follow. Tarentola senegambiae is also a primary, but obviously less specialised tree-dweller, sometimes even referred to as 'fig-tree gecko' (Cissé 1974;Cissé and Karns 1979), because it was found, in contrast to the obviously strictly tree-dwelling T. hoggarensis and T. panousei, next to trees, as well as on buildings in human settlements, for example, by WB (still as 'T. ephippiata') behind shutters of an ORSTOM Institute building at Richard-Toll (Böhme 1978).
The only reference reporting 'T. ephippiata' from coastal Mauritania on house walls is by Ineich (1996). He found these geckos close to and in Nouakchott, on the walls of a veterinary station and of a military post, here in sympatry with Hemidactylus brookii (currently H. angulatus). All other discoveries of this species, made farther northwards, were Figure 8. Sketch of the Vachellia tortilis acacia tree investigated (see Figure 1) to show the different shelter places of both syntopic Tarentola species. Labelled with 'A' for T. annularis, and 'B' for T. panousei. Drawing by T Koppetsch on trees (acacia). To the south, in the Diawling National Park in the Senegal River delta, close to the Senegalese border, the geckos were found on Baobab (Adansonia digitata) trees, with T. annularis already largely missing in this area.
However, the exact taxonomic identity of these geckos, in the light of Joger's (1984) revision and the present paper, should be reinvestigated, because now three taxa of the T. ephippiata species complex, viz. T. panousei, T. hoggarensis and, in the south, Figure 9. A: Details of the Vachellia tortilis acacia trunk with larger clefts as hides of T. annularis. B: Small knotholes as hides for T. panousei. The one Ouguiya coin serving as a scale is 21 mm in diameter. Photographs by W Böhme perhaps also T. senegambiae, might be involved. The grid maps in Trape et al. (2012) suggest sympatry between the latter and T. hoggarensis in the Senegal River delta area.
The assumption of the competitive superiority of T. senegambiae over equal-sized T. annularis, as implied by Böhme (1978) can certainly not be applied to the situation of the syntopic presence of the otherwise strictly rupicolous T. annularis with T. panousei, because in this case it seemed that the latter, because of its smaller size, was the inferior competitor, pushed back to the less favourable shelter and hiding places. T. annularis, otherwise known from Egypt and the Sudan to North Cameroon and Mauritania predominantly as a rock-dweller, has been observed on trees so far only by Sow et al. (2015) and in the present study. However, although Sow et al. (2015) found their 'T. e. hoggarensis' only in bare areas in holes of acacia trees, they only occasionally found T. annularis, which they observed three times more often, in syntopy with the former. At our site near Nouadhibou, T. annularis proved to be superior in competition in a habitat structure absolutely atypical for it, whereas it was the primary microhabitat for its smaller congener.
The obviously less narrow ecological specialisation of T. annularis, compared with the taxa of the T. ephippiata complex, might also explain why it is taxonomically uniform from Egypt and the Sudan to the Atlantic Ocean, whereas the more stenoecious T. ephippiata complex is partitioned to several taxa within the same geographic area. The distribution pattern of the here revalidated T. panousei appears to be biogeographically peculiar, with only patchy and locally isolated subpopulations known so far. However, an overlooked more broad occurrence in this region cannot be ruled out, because other records assigned to T. ephippiata sensu stricto might actually represent T. panousei. These broad gaps in Figure 10. Juvenile T. panousei from the acacia tree investigated. Photograph by W Böhme known distribution ranges are also known for other African geckos, for example, Hemidactylus pseudomuriceus (Koppetsch and Böhme 2017). Nevertheless, it can only be speculated whether the localities where T. panousei had been recorded might rather be the result of a relictual patchy distribution. The underlying determinants, for example, different dispersal abilities, and possible scenarios, for example, transformation relictual refugia in the past to the present fragmented distribution of the T. ephippiata complex as a whole, which had caused or shaped those distinct biogeographic patterns, might be identified by future more comprehensive phylogeographic analyses.