Systematic position and taxonomy of Pipistrellus deserti (Chiroptera: Vespertilionidae)

Morphological analyses

The museum specimens labelled Pipistrellus deserti or P. kuhlii from Africa, Europe, and the Middle East listed in Appendix 1 were examined and used in the morphological analyses (427 specimens in total). This comprehensive material includes part of the type specimens concerning the respective group (deserti Thomas, 1902, ikhwanius Cheesman et Hinton, 1924, marginatus Cretzschmar, 1830). Morphological variations of these specimens were first explored on a multivariate space using principal component analyses of skull and dental characters. Maximum factor loadings of variables were also calculated to identify the best discriminating variables. According to these analyses, the samples were then assigned to both morphotypes for further comparisons. Descriptive statistics of each group and multivariate analyses were performed with the Statistica 6.0 software (StatSoft, Inc., Tulsa, Oklahoma, USA). For these morphometric analyses, we primarily took skull, teeth (taken including cingula), and forearm measurements: FA – forearm length (incl. wrist); GLS – greatest length of skull; CBL – condylobasal length; ZB – zygomatic breadth; IC – breadth of interorbital constriction; RBFI – rostral breadth between foramina infraorbitalia; BB – neurocranium (braincase) breadth; MB – mastoidal breadth; BH – neurocranium (braincase) height; CC – rostral breadth between canines (incl.); M³M³ – rostral breadth between third upper molars (incl.); CM³ – length of upper tooth-row between canine and third molar (incl.); M¹M³ – length of upper molar-row (incl.); CP⁴ – length of upper tooth-row between canine and second premolar (incl.); LI¹ – mesio-distal length of first upper incisor; LI² – mesio-distal length of second upper incisor; ML – condylar length of mandible; CH – height of coronoid process; CM₃ – length of lower tooth-row between canine and third molar (incl.); M₁ M₃ – length of lower molar-row (incl.); CP₄ – length of lower tooth-row between canine and second premolar (incl.). All specimens were measured by the same person (PB) in a standard way using mechanical and optical (for teeth measurements) callipers (see Barlow et al. 1997 and Benda et al. 2004b).

Acronyms of collections housing the specimens examined

BMNH – Natural History Museum, London, United Kingdom; IVB – Institute of Vertebrate Biology, Brno, Czech Republic; MHNG – Natural History Museum, Geneva, Switzerland; MNHN – National Museum of Natural History, Paris, France; MUB – Institute of Botany and Zoology, Masaryk University, Brno, Czech Republic; NMP – National Museum (Natural History), Prague, Czech Republic; NMW – Natural History Museum, Vienna, Austria; SMF – Senckenberg Institute and Museum, Frankfurt am Main, Germany; ZFMK – Zoological Institute and Museum Alexander Koenig, Bonn, Germany.

Genetic analyses

Ethanol-preserved tissue samples of Libyan, Moroccan, Middle Eastern, and European specimens of Pipistrellus kuhlii, as well as Libyan and Moroccan specimens of P. deserti, were used for the genetic analyses (see Appendix 2). Total genomic DNA was extracted with a standard salting-out protocol as described by Miller et al. (1988) and re-diluted into 100 μl of pure water. The initial part of the cytochrome b gene was then amplified in a PCR reaction using the primer pair L14724 (Kocher et al. 1989) and MVZ16 (Smith and Patton 1993), and sequenced with an automated DNA sequencer (Applied Biosystems, Life Technologies, Foster City, CA, USA) following standard methods (e.g., Ruedi and Mayer 2001). Sequences were aligned and edited visually using Sequencher 3.0 (Gene Codes Corp., Ann Arbor, MI, USA). All different haplotypes were deposited in GenBank under accession numbers KM252756–KM25277. To determine the phylogenetic position of these taxa relative to other members of the genus Pipistrellus, we also used homologous sequences of P. abramus (Temminck, 1840), P. hanaki Hulva et Benda, 2004, P. hesperidus (Temminck, 1840), P. cf. javanicus (Gray, 1838), P. maderensis Dobson, 1878, P. nathusii (Keyserling et Blasius, 1839), P. pipistrellus (Schreber, 1774), and P. pygmaeus (Leach, 1825), as well as other P. kuhlii sequences deposited in GenBank (see Appendix 2 for origins and/or references). The two Asian species (P. abramus, P. cf. javanicus) were used as a composite outgroup. To estimate levels of DNA sequence divergence, we used the K2P model of correction (K2P distance) that is commonly mentioned for comparative purpose in bat systematics (Bradley and Baker 2001, Ibáñez et al. 2006, Vallo et al. 2013).

To provide a nuclear perspective to the genetic analyses, we also genotyped 12 specimens of deserti and 10 specimens of kuhlii (see details in Appendix 2) at the following five polymorphic microsatellite loci: EF6 (Vonhof et al. 2002), NN8 (Petri et al. 1997), Paur05 (Burland et al. 1998), L45 (Wei et al. 2009), and Ppip05 (Racey et al. 2007). These loci were amplified in three multiplexed reactions containing a forward primer labelled with fluorescent dyes (see Appendix 3). The amplification was achieved in 10 μl reaction volume, with 1.2 μl H₂ O, 5 μl 2× Qiagen Multiplex PCR Master Mix^® (Qiagen, Hilden, Germany), 1 μl of primer mix (10 μm of each primer), 2 μl 5× Q-Solution, and 0.8 μl extracted DNA. The thermal cycling consisted in a 15 min initial denaturation at 95°C, followed by 32 cycles with 40 s denaturation at 94°C, 90 s annealing with a temperature from 55 to 50°C for reaction 1, and 65–60°C for reactions 2 and 3 (the first five cycles consisted in a touchdown with a pitch of 1°C) and 1 min extension at 72°C, followed by a final extension for 30 min at 60°C. Resulting PCR products were run on a Beckman Coulter GeXP Genetic Analysis System and allele callings analysed with the associate software GenomeLab™ (Beckman Coulter, Inc., Brea, CA, USA). Given that the sample size for each locality was too small, no test for Hardy-Weinberg equilibrium was done, but a larger-scale survey of genotypes indicated that no significant deviations or ghost alleles affected these microsatellite loci in Pipistrellus kuhlii s.l. (Andriollo and Ruedi, unpublished data). The individual, multilocus genotypes of the essayed pipistrelles were then submitted to a principal component analysis with the program PCA-GEN v1.2 (Goudet 1999). Given that we had low sample sizes, we opted to use this program because it is better suited for exploring individual relationships than the other programs, e.g., methods based on Bayesian clustering that rely on equilibrium models in population genetics. Number of significant components and overall Fst values were tested with 15,000 randomisations.

Phylogenetic reconstructions

We used Bayesian (BA) inference and maximum likelihood (ML) methods to reconstruct phylogenetic relationships of mitochondrial sequences. For the BA inference and ML method, the appropriate model of nucleotide substitutions was determined with the program MrModeltest version 2.2 (Nylander 2004). The HKY+I+G model best fitted to the cytochrome b data set (I=0.5488, gamma distribution with shape parameter α=2.448). Bayesian posterior probabilities were calculated using four simultaneous Markov chains run for 1 million generations and trees sampled every 1000 generations, as implemented in the software MrBayes version 3.1.2 (Huelsenbeck and Ronquist 2001). After the log-likelihoods of trees reached stationarity, the initial 10% of trees were discarded as burn-in and posterior probabilities were computed from the consensus of the remaining trees. Adequate sampling (ESS>200) and stationarity of values were checked with TRACER v.1.5 (Rambaut and Drummond 2009). ML analyses were conducted with the program RAxML (Stamatakis 2006) and were done on a fully partitioned model, where each codon partition was allowed to have partition-specific model parameters. Reliability of nodes in the ML analyses was assessed by 1000 bootstraps with RAxML.

Morphological characters and identification

Bivariate comparison (Figure 2) and the multivariate factor analysis of all skull and dental measurements (Figure 3; 1st PC 64.86%; 2nd PC 5.49%) confirmed that the two morphotypes identified by the collectors segregate in two non-overlapping groups among African samples. The general high factor loadings of most morphological variables on the first principal component indicate that this is a size factor. One of these morphogroups represents all smaller desert forms of pipistrelles, including the type specimen of Pipistrellusdeserti Thomas, 1902 from Libya (Table 1), whereas most of the larger bats correspond to the P. kuhlii sampled elsewhere in the more mesic parts of North Africa. The latter group also includes the type specimen of Vespertilio marginatus Cretzschmar, 1830 from Egypt. These North African P.kuhlii specimens show a similar morphology and comparable skull and dental variation as the samples of P. kuhlii from Europe and the Middle East (Table 2). In particular, most external and cranial measurements of the type specimen of P. deserti from Murzuq, Fezzan, Libya (male, BMNH 2.11.4.1.) fit well into the size variation of our recent sample of 16 specimens of P. deserti collected from four localities in the same region (Table 1). Exceptions include the mandible length (which is slightly smaller) and length of the second upper incisor, I² (which is slightly larger). In comparison with samples of typical P. kuhlii from the Mediterranean (North Africa, Europe, Middle East), skull dimensions of all south Libyan samples form a separate group without overlap in the larger skull dimensions (GLS, CBL, ZB, MB, ML; Table 2, Figure 2). Other samples from the Sahara, identified as P. deserti by their collectors (Kock 1969, Gaisler et al. 1972, Kowalski and Rzebik-Kowalska 1991, Benda et al. 2004a; cf. Appendix 1), also fit well into the variation range of dimensions reported here for the Libyan Desert pipistrelles, which confirms their original morphotype assignation (Figures 2 and 3). The examined samples from northern Sudan (Wadi Halfa) lies on the lower margin of a cluster composed of P. deserti from Libya, Egypt, Algeria, and Morocco. As a general observation, most skull and dental dimensions illustrate discrete size differences between the smaller P. deserti and the larger P. kuhlii from the Mediterranean (Figures 2 and 3, and Table 3).

Figure 2

Bivariate plot of Pipistrellus kuhlii from the Sahara, North Africa, and the Mediterranean: the greatest length of skull against the neurocranium breadth (both in millimetres). Squares denote the deserti morphotype, circles the kuhlii morphotype. Legend: Central Sahara=south-western Libya (Fezzan) and south-eastern Algeria (Hoggar Mountains.); East Sahara=Upper Egypt and northern Sudan; North-West Sahara=north-western Algeria and south-eastern Morocco; North Africa=Tunisia and the Mediterranean parts of Morocco, Algeria, Libya, and Egypt; T=holotype specimen of Pipistrellus deserti Thomas, 1902.

Figure 3

Bivariate plot of Pipistrellus kuhlii from the Sahara, North Africa, and Mediterranean on the first two principal components of all cranial and dental measurements. Squares denote the deserti morphotype, circles the kuhlii morphotype. For legend see Figure 2.

Qumsiyeh (1985) proposed the greatest length of skull (GLS) as a discriminant criterion for Egyptian populations, where GLS is longer than 12.0 mm in Pipistrellus kuhlii and shorter than 12.0 mm in P. deserti. Our extensive material (including the Egyptian samples) suggests that this discriminant limit lies rather around 12.4–12.5 mm. According to this single criterion, bats with GLS smaller than 12.4 mm would all come from the drier parts of the Sahara and group with the topotype material of the Desert pipistrelles from Libya, whereas bats with GLS larger than 12.5 mm group with the Mediterranean P. kuhlii (see Figure 2). Qumsiyeh (1985) and Kowalski and Rzebik-Kowalska (1991) mentioned also the length of the upper tooth-row (CM³) as another discriminant character, where specimens with CM³ shorter than 4.5 mm were regarded as typical P. deserti, whereas larger bats were recognised as P. kuhlii. In the material examined here, the largest P. deserti identified in the multivariate analysis showed a CM³ of 4.6 mm (NMP 48302, adult female from Gabrun, Libya, GSL 12.0 mm) and 4.7 mm (NMP 90072, adult female from Gorges du Todra, Morocco, GSL 12.2 mm), whereas the smallest specimens of Mediterranean P. kuhlii could have an upper tooth row length as small as 4.4 mm (SMF 34362, adult male from Sush, Iran, GSL 12.7 mm; see also Table 2). Owing to this large overlap of values, the length of tooth-row alone is not useful to discriminate these morphotypes, unlike the greatest length of skull (Figure 2).

Descriptive statistics of skull and dental dimensions are given in Tables 1 and 2, and univariate analyses indicate that almost all comparisons show highly significant size differences between the two morphotypes (Table 3). However, none of the qualitative characters examined differed in a consistent way. For instance, the skull outline of both morphotypes is almost identical, except for absolute size (Figure 4). The second upper incisor (I²) is very tiny in both forms, with the height of crown about 35%–45% of the height of the crown of the first incisor (I¹); the tip of I² is slightly overlapping the cingulum of I¹ in both forms (Figure 5). The ratio between the mesio-distal length of crown of I² and that of I¹ in Pipistrellus deserti is almost identical as that in P. kuhlii (Tables 2 and 3). The first and second upper incisors thus differ essentially in the same way in both morphotypes, which is in accordance with observations described by Qumsiyeh (1985): Figure 20). Gaisler et al. (1972) found difference between P. deserti and P. kuhlii in terms of the degree of reduction of the first upper premolar (P³), with the former species having a more reduced premolar. The degree of reduction of this premolar in a larger material examined here show that this character is variable, but comparable in both morphotypes (Figure 6).

Figure 4

Outline comparison of skulls of Pipistrellus kuhlii: (above) the kuhlii morphotype (♀, Ain Sharshara, Tripolitania, Libya, NMP 49953), (below) the deserti morphotype (♀, Gabrun, Fezzan, Libya, NMP 48310). Scale bar=5 mm.

Figure 5

Fronto-lateral view on the upper incisors of Pipistrellus kuhlii from North Africa. Legend: a – kuhlii morphotype (Tabouda, Morocco, NMP 90030); b – kuhlii morphotype (Ain Sharshara, Tripolitania, Libya, NMP 49953); c – kuhlii morphotype (Tolmeita, N Cyrenaica, Libya, NMP 49930); d – kuhlii morphotype (Jalu, S Cyrenaica, Libya, NMP 49939); e – deserti morphotype (Oued Drâa, Morocco, NMP 90058); f – deserti morphotype (Gabrun, Fezzan, Libya, NMP 48310). Scale bar=1 mm.

Figure 6

Occlusal view on the rostral part of the upper tooth-row (C–P⁴) of Pipistrellus kuhlii from North Africa. For legend, see Figure 5. Scale bar=1 mm.

Qumsiyeh (1985) observed that Egyptian Pipistrellus deserti had more slender rostrum compared to P. kuhlii. This observation is supported in our results (Tables 1–3) as the ratios between breadths of rostrum (RBFI, CC) and greatest length of skull (GLS) differ significantly (Table 3). However, another relative dimension of the breadth of the rostrum (CC/CM³) did not show significant differences between P. deserti and P. kuhlii (Table 3).

The baculum of P. deserti was described by Gaisler et al. (1972) from two specimens from Egypt and by Hill and Harrison (1987) from a specimen from Algeria. This bone has the typical habitus known in members of the genus Pipistrellus (Lanza 1959, Hill and Harrison 1987) with a thin stick arched dorsally and a bifurcation on both epiphyses. As in other morphological characters, the baculum of P. deserti differs from that of P. kuhlii only in absolute size, but we found no difference in the shape of the published preparations (Lanza 1959, Gaisler et al. 1972, Wassif and Madkour 1972, Hill and Harrison 1987).

Most previous authors (e.g., Gaisler et al. 1972, Qumsiyeh 1985, Kowalski and Rzebik-Kowalska 1991) mentioned substantial differences in colouration between Pipistrellus deserti and P. kuhlii, the former being generally paler than the latter for both wing and pelage characteristics. Indeed, as the deserti morphotype lives in the more arid parts of the Sahara, its colouration is very pale (pale olive brown in dorsal pelage, pale brown skin on face, ears and wing membranes), and the very pale (creamish, whitish, or translucent) posterior wing margin is up to 4.5–5.0 mm wide on plagiopatagium with indistinct transition to darker colour of the wing membrane. However, the same pattern of colouration was observed in bats classified by multivariate analyses as P. kuhlii and caught in the Libyan oases of Sinawan and Jalu (situated in the Sahara, ca. 250 km from the sea coast). This very pale colouration pattern was also found in some populations of the arid regions of the Middle East (Syrian Mesopotamia, Iranian Baluchistan), all of which were classified in our multivariate analyses as typical kuhlii (Figures 2 and 3). The darkest individuals representing the typical pelage colouration of P. kuhlii were observed in Iberia and Morocco (Rif Mts.) specimens. These bats had dark chestnut brown dorsal pelage and very narrow (<0.5 mm) and sharply delimited pale (not white) strip on the posterior wing margins, the remaining wing membrane being dark brown. However, in the same region of northern Morocco we found also relatively pale individuals, resembling in colouration the desert forms.

Thus, P. kuhlii specimens identified as such by their skull and dental dimensions (Figures 2 and 3) can have very variable colouration. As a general trend, we found that the colouration intensity of pipistrelles varied clinally from uniformly paler bats in more arid habitats to darker and more variable colours in populations inhabiting more mesic regions (i.e., in the Mediterranean zone). This trend was described in P. kuhlii from Algeria already by Kowalski and Rzebik-Kowalska (1991) and in populations from the eastern Mediterranean by Lewis and Harrison (1962). Nevertheless, without a proper GIS analysis of these colouration trends along dedicated transects, it is difficult to determine if such local variation is significant and valid across the entire distribution of this species complex. Pelage colouration and width of the pale posterior margin of wing membrane are highly variable characters that bear little taxonomic importance (in agreement with Corbet 1984, contra Panouse 1951, Deleuil and Labbé 1955a,b, Gaisler et al. 1972, Hanák and Elgadi 1984, Qumsiyeh 1985, etc.).

Genetic comparison

We sequenced the initial 620 bases pairs of the cytochrome b gene of 22 individuals identified in the multivariate morphological analyses as typical Pipistrellus kuhlii and 12 individuals from Morocco and Libya identified as P. deserti (Appendix 2). The alignment of these cytochrome b sequences with six further homologous sequences of P. kuhlii taken from the GenBank resulted in 12 distinct haplotypes (C1, C4, D1 to D3, K1 to K5, K8, and PK03; Appendix 2). Two haplotypes (K2 and K5) were found in all specimens from the Middle East (Iran and Syria), three others (K8, C1 and C4) were confined to West European bats, whereas the remaining ones originated from a vast area comprising North Africa and Europe and included both typical P. kuhlii and typical P. deserti individuals (Figure 7). Within the kuhlii complex, haplotypes differed by one (D2 vs. D3 or K2 vs. K5) to 36 mutations (K2 vs. C1), which correspond to a K2P divergence of <1% to up to 6% (Table 4). Outside this group, interspecific divergences were much larger, exceeding 12%, except between P. pygmaeus and P. hanaki (7.5%); the latter species is, however, represented only by a partial sequence of 402 bp, which does not include the more variable, central portion of the cytochrome b gene, and thus is not directly comparable to other distances.

Figure 7

Bayesian consensus tree representing the phylogenetic relationships of haplotypes of the Pipistrellus kuhlii group based on sequences (620 bp) of the mitochondrial cytochrome b gene. A star associated to a node denotes that it is supported by at least 95% posterior probability (BA reconstructions) and 95% bootstrap (ML reconstructions). Other support values are given as percentages. We named the clades according to the recent review of Çoraman et al. (2013) and pruned the tree from all outgroup species (i.e., P. abramus, P. hanaki, P. hesperidus, P. nathusii, P. pipistrellus, P. cf. javanicus, and P. pygmaeus; see Appendix 2) to simplify this representation. Scale bar=5% divergence.

All phylogenetic reconstructions (ML and BA) identified a strongly supported (>95% bootstrap or posterior probability; Figure 7) monophyletic group comprising all haplotypes of Pipistrelluskuhlii, P. deserti, and P.maderensis, to the exclusion of any other species (including a Southern African P. hesperidus). As already documented in all reconstructions based on distinct mitochondrial genes (e.g., Ibáñez et al. 2006, Mayer et al. 2007, Evin et al. 2011, Veith et al. 2011, Çoraman et al. 2013), sequences issued from this species complex form more or less well-separated clades, but relationships among them lack resolution, which is fully consistent with the reconstructions presented in Figure 7.

Representatives of Clade 1 are widespread across most of Europe, North Africa, the Canary Islands, the Balkans, and the Levant (see Çoraman et al. 2013 for a larger geographic sampling). This clade is well supported and also includes typical Pipistrellus kuhlii, as well as all sequences of P. deserti from Morocco and Libya (Figure 7). More specifically, all nine pipistrelles sampled near the type locality of P. deserti in Fezzan (Libya) shared a single haplotype (D1), which is most closely related (1.1% K2P divergence) to the widespread haplotype K1 in Clade 1 (Table 4). Haplotypes from Clade 3 are restricted to Western European P. kuhlii, whereas those of Clade 2 include all sequences from the Middle East (Figure 7 and Çoraman et al. 2013). Partial cytochrome b haplotypes from the desert region of Arabia reported in Bray et al. (2013) also pertain to this Clade 2 and differs only by two point mutations from the haplotype K5 (result not shown). The haplotypes of P. maderensis form a sister group close to the widespread Clade 1, albeit with moderate support (Figure 7), rendering this taxon paraphyletic, as shown by Pestano et al. (2003). Thus, according to all molecular reconstructions based on mitochondrial DNA, P. deserti and P. kuhlii do not appear in distinct units (Figure 7; Mayer et al. 2007).

However, these conclusions on phylogenetic relationships are based on mitochondrial markers, which retain the history of the females only. As such, female lineages may underlie a different history than the organisms themselves (Ballard and Whitlock 2004). For instance, at least two pairs of biological species of bats (Myotis myotis vs. M. blythii and/or Eptesicus serotinus vs. E. nilssonii; Berthier et al. 2006, Artyushin et al. 2009) show striking cytonuclear discordance, supposedly due to ancient but massive episodes of mtDNA introgression. In these introgressed species the unusually low divergence measured at mtDNA genes thus does not reflect their true organismal relationships, as shown by their divergent external morphology or nuclear genes (Berthier et al. 2006, Juste et al. 2013).

To exclude the possibility that the strong similarities in mtDNA genes of both morphotypes are due to introgression, we also report the nuclear genetic relationships of a subsample of 22 pipistrelles (mostly collected in Libya, Morocco, and Switzerland; see Appendix 2) in Figure 8. This subsample represents both typical kuhlii and deserti identified in the previous multivariate, morphological analysis. The PCA-GEN output of these multilocus, nuclear genotypes suggests that samples are grouped according to their geographic origin rather than morphotypes (Figure 8). Most of the inertia of the first (and only significant) component is indeed due to the separation of the African versus European samples, regardless of the morphotype tested. If deserti would represent a distinct species, kuhlii genotypes from both sides of the Mediterranean should be more closely related to each other than either is to desert genotypes. If the same analyses are repeated without the European pipistrelles, the results are similar (not shown), with all Moroccan and all Libyan samples being grouped together, regardless of morphotypes. This data set, albeit limited, therefore clearly confirms that pipistrelles of the two morphotypes in North Africa not only share very similar mitochondrial cytochrome b genes (Figure 7), but also share similar allelic composition at five independent, nuclear loci (Figure 8). The hypothesis that the morphologically identified Pipistrellus deserti samples from Morocco and Libya (and by extension all those from the Sahara) would share a single common ancestor that is distinct from other kuhlii morphotypes is thus falsified by all current genetic evidence.

Figure 8

Projection on the first two principal components (each representing 28% and 12% of total inertia, respectively) of the multilocus genotypes of 10 kuhlii morphotypes (in bold face) and 12 deserti morphotypes (greyed) analysed with PCA-GEN (Goudet 1999). Samples come from Morocco (MO), Libya (LY), and Switzerland (SW). Only the first axis is significant according to the broken-stick distribution.

Taxonomic and biogeographical conclusions

Our phylogenetic reconstructions clearly suggest that the mitochondrial DNA of bats representing the deserti morphotype are very closely related and are imbedded within the broader radiation of other mitochondrial DNA lineages belonging to the kuhlii morphotype sampled around the Mediterranean (Figure 7). The same conclusion can be drawn from a limited number of samples genotyped at five nuclear loci (Figure 8), which excludes the possibility that the results revealed from the mtDNA analysis could be biased by recent or past events of introgression. These genetic results rather suggest that the bats of the deserti morphotype are issued from multiple, independent kuhlii ancestors, but evolved a convergent, desert-adapted morphology in different parts of the Sahara. According to the genetic or biological species concepts (de Queiroz 2007) and despite the clear morphologic differences observed between the deserti and kuhlii populations in North Africa (Figures 2 and 3), the small-sized and pale-coloured Saharan populations of the deserti morphotype do not seem to represent a distinct biological species. Hence, we suggest to consider the name Pipistrellus deserti Thomas, 1902 a junior synonym of Vespertilio kuhlii Kuhl, 1817=Pipistrellus kuhlii (Kuhl, 1817).

The significant meristic differences between the two morphotypes of Pipistrellus kuhlii in North Africa, which is not reflected in their genetic characters, may be due to the contrasting environments found in this region. The deserti morphotype is clearly a desert inhabitant of the Sahara, which is composed of a complex of relatively young habitats – the current state being approximately 5500 years old (Foley et al. 2003, Bray et al. 2013). Thus, the form deserti is living in this harsh environment of the Sahara since a short period on an evolutionary time scale. This short period did not lead to major genetic differences but was sufficient for morphologic (and supposedly physiologic) adaptations to evolve in response to such harsh desert habitats.

The two morphotypes of P. kuhlii coincident with mesic and arid habitats is a general phenomenon that is found also in other organisms living in such contrasted environments (Heim de Balsac 1936, Lewis and Harrison 1962, Guillaumet et al. 2008). Indeed, several bat species are represented by two distinct morphotypes in North Africa, in which a larger form (and often also darker morph) inhabits coastal regions of the Mediterranean Sea and a smaller (and paler) form occurs in a belt of the central Sahara. Examples include Rhinopoma cystops Thomas, 1903, Rhinolophus clivosus Cretzschmar, 1828 and/or Asellia tridens (Geoffroy, 1818). Similarly as in the kuhlii complex, these pairs of morphotypes were originally described and, for a long time, treated as pairs of separate taxa (see Gaisler et al. 1972, Hill 1977, Qumsiyeh 1985, Owen and Qumsiyeh 1987, Van Cakenberghe and De Vree 1994). All are now considered as morphologically distinct populations of a single species or even subspecies because of close similarities in genetic traits found between morphotypes (Hulva et al. 2007, Benda et al. 2011, Benda and Vallo 2012).

Small and pale individuals of Pipistrellus kuhlii were also found in other desert regions of the western Palaearctic, e.g., in Arabia (Gaisler et al. 1972, Bray et al. 2013) or in the Iranian Baluchistan (Benda et al. 2012). Whereas in Arabia the degree of size reduction in Desert pipistrelles does not exceed the variation extremes of Mediterranean P. kuhlii, the Baluchistani bats differ significantly in size and represent a dimensional transition between the African kuhlii and deserti morphotypes (see Benda et al. 2012 for details).

Most records of the deserti morphotype come from the belt of the central Sahara from southern Algeria via southern Libya and southern Egypt to northern Sudan, from 28°N to the south (see Figure 1), i.e., an area with the lowest annual precipitation in the Sahara (≤20 mm). In this region, this species was captured in oases, where it uses petrophilous or synathropic shelters (rocky fissures and fissures between beams in abandoned houses) or was netted among palm trees (Gaisler et al. 1972, Kowalski and Rzebik-Kowalska 1991; our records). However, some records came also from the north-Saharan region of north-western Algeria and south-eastern Morocco (Qumsiyeh 1985, Gaisler and Kowalski 1986, Kowalski and Rzebik-Kowalska 1991, Benda et al. 2004a, 2010), where these bats were netted in oases. In the latter region, the distribution range of the deserti morphotype was found to be in close parapatry to connect with the kuhlii morphotype in the High Atlas and Anti-Atlas Mountains in Morocco (Benda et al. 2004a) and probably also in the Saharan Atlas Mountains in Algeria (Kowalski and Rzebik-Kowalska 1991, our own observations).

In general, we observed that the border between the ranges of both morphotypes extends to a distance of ca. 250–350 km from the sea coast throughout North Africa (Figure 9). However, in the region of the north-western Sahara in Morocco, the distribution of both morphotypes is more mosaic-like and the exact limits probably depend on the mesic/arid character of each site of occurrence. Thus, the reported sympatric occurrence of both forms in nearby Algeria (Gaisler et al. 1972, Kowalski and Rzebik-Kowalska 1991) may probably result from sparse sampling in such mosaics, where annual precipitation (50–100 mm) could represent a transition zone for intermediate populations.

Figure 9

Greatest length of skull (GLS, in millimetres) against relative aridity of a record (expressed as a geographical distance of the respective site from the Mediterranean Sea and/or Atlantic Ocean coasts; in kilometres) in North African samples of Pipistrellus kuhlii group. Legend: North-West Sahara=Morocco, Algeria and Tunisia; North-East Sahara=Libya, Egypt and the Sudan.

The deserti morphotype is a desert form of Pipistrellus kuhlii, which most probably developed in the most arid habitats of the northern and/or central Sahara after several independent invasions from the Mediterranean or, perhaps less probably, from relic populations that persisted in oasis islets in the Sahara. Therefore, we consider as highly unlikely the possibility that these desert forms may have a sub-Saharan origin, i.e., in the Afro-tropic region. Thus, we hypothesise that the published records assigned to P. deserti from sub-Saharan Africa (see Van Cakenberghe and Benda 2013 actually belong to another species. One such candidate is P. aero Heller, 1912 a bat described from central Kenya and distributed in the north-western part of this country (Aggundey and Schlitter 1984, Van Cakenberghe and Happold 2013). Koopman (1975) differentiated this species from P. deserti only on the basis of its slightly larger size. However, new sampling coupled with proper phylogeographic analyses are needed to validate this taxonomic and biogeographic hypothesis.