Shrews (Mammalia, Eulipotyphla) from a biodiversity hotspot, Mount Nimba (West Africa), with a field identification key to species

ABSTRACT In this study, we collected 226 shrew specimens originating from 16 localities on the Guinean and Liberian sides of Mount Nimba. We surveyed all major vegetation zones from 400 to 1600 m above sea level (asl), including forest and savannah habitats. We recorded 11 species, whose identifications were confirmed by genetic analyses and classical morphometrics. Furthermore, we provide cytogenetic data for five of these species. The shrew community at Mount Nimba is composed of a mix of both savannah- and forest-dependent species, which is related to the peculiar position of Mount Nimba situated at the transition between lowland rainforest to the south and Guinean woodlands to the north. We recorded 11 species of shrews in syntopy in lowland rainforest, seven in edaphic savannah and mountain forest, and five in high-altitude savannah at 1600 m asl. Based on morphometric analyses, we show that these syntopic species separate along a size axis, allowing species to occupy different ecological niches, which we speculate allows them to access different food resources. We also highlight that Crocidura theresae Heim de Balsac, 1968 from Mount Nimba has a different karyotype from that described in Côte d'Ivoire. Finally, we develop a novel identification key for shrews from Mount Nimba using external characters and standard body measurements, allowing it to be used in the field on live specimens. In total 12 shrew species are now known from Mount Nimba, which highlights its exceptional position as a tropical African biodiversity hotspot.

The diversity of small mammals, and especially of shrews, is promoted by factors acting as biogeographical barriers such as different vegetation zones (Jacquet et al. 2014), rivers (Stanley & Esselstyn 2010;Jacquet et al. 2013), or mountains (Stanley & Olson 2005;Stanley & Esselstyn 2010). Mount Nimba, which is situated at the triple frontier point between Guinea, Liberia and Côte d'Ivoire, constitutes one of the three highest West African mountains (1752 m above sea level). Mount Nimba has been subject to several surveys aimed at understanding the structuring of its plant and animal communities (see Coe 1975 and references therein). Stratification of the vegetation is observed along the elevational gradient, with lowland forest up to 600 m, mid-elevation forest between 600 and 1200 m and edaphic altitude savannah above 1200 m. This latter habitat is known to offer isolation conditions for both plant and animal species owing to its distinct climate and vegetation characteristics (Coe 1975;Lamotte & Roy 2003). For this reason, Mount Nimba may act as a biodiversity promoter (White 1981). Moreover, it is situated in the ecotone zone between rainforest and savannah, which is high in diversity (Fahr & Kalko 2011), and is known to play an important role in speciation processes (Schilthuizen 2000).
Cytogenetic analyses have also been successfully employed to assess the validity of shrew species and to explore intraand inter-specific variability (Meylan & Vogel 1982;Schlitter et al. 1999). Several cytogenetic studies have been conducted in West Africa, especially in Côte d' Ivoire andBurkina Faso (Meylan 1967, 1971;de Hondt 1974;Meylan & Vogel 1982;Maddalena & Ruedi 1994;Baskevitch et al. 1995;Lavrenchenko et al. 1997;Schlitter et al. 1999), however many gaps remain in our knowledge of the intra-and inter-specific variability in the karyotypes of shrews in West Africa, and no karyotypes have been published for shrews at Mount Nimba.
Despite progress made with resolving the taxonomy and systematic relationships of West African shrews, little is known about their ecological requirements and morphological variability. Morphometric analyses are useful in describing the variability within and between species, and have been used to demonstrate that in species-rich shrew communities such as those in African rainforests, species are typically ordered along a size gradient (Hutterer et al. 1987;Brosset 1988;Churchfield et al. 1999). It has been suggested that this assists partitioning access to food resources, allowing these species to co-exist sympatrically or even syntopically (Hutterer et al. 1987).
A serious challenge to research on West African shrews is the difficulty with species identification. Indeed, tools allowing rapid and accurate identification of species in the field are totally lacking; even the most recent and comprehensive synthesis of African mammals 'Mammals of Africa' does not present characters for distinguishing different species of the super-diverse genus Crocidura (Happold & Happold 2013). To complicate matters, the morphology of shrews may vary with environmental factors (Rácz & Demeter 1998;Wójcik et al. 2000Wójcik et al. , 2003. As a result, the identification of species in the field is challenging even for specialist taxonomists, especially owing to the paucity of available external characters (e.g. fur and skin colour, tail length and thickness, quantity and length of tail vibrissae) for the discrimination of species (Heim de Balsac 1956, 1958, 1968Hutterer & Happold 1983). An identification key was presented by Hutterer & Happold (1983) for Nigerian shrews but is mostly inapplicable to the Upper Guinea rainforest, since species composition is different. Given the important status of Mount Nimba (Monadjem et al. 2016), the development of an identification key based on external and morphometric characters would be highly useful.
The aim of this paper is to synthesize our recent work on shrew diversity at Mount Nimba by: 1) providing a faunal Denys C. et al. list with information on habitat and ecological requirements; 2) describing karyotypes of Mount Nimba shrews and comparing them with those available for other West African localities in the literature; 3) exploring the morphological variability within and between species in order to understand how diverse communities are structured; and 4) providing a new identification key easily usable in the field. We have at our disposal a valuable collection of genotyped specimens that represents a unique opportunity to update our knowledge on the distribution, ecology, and morphology of shrew taxa from this biodiversity hotspot. In 2003, Liberia proclaimed the East Nimba Nature Reserve (ENNR) covering most of the Liberian part of the mountain, whose management costs are currently subsidized by ArcelorMittal that views it as a possible offset site for its operations. In contrast, the entire Guinean and Ivorian parts of the mountain were protected from 1944 onwards. The Mount Nimba Strict Nature Reserve (La Réserve naturelle intégrale du Mont Nimba), was initially established by the former French colonial government in 1943 by Order no. 4190 SE/F in Côte d'Ivoire and subsequently by decree in Guinea in 1944(JO-RF, 1944. In 1980 the Guinean portion of the Strict Nature Reserve was proclaimed a Biosphere Reserve. The Mount Nimba Strict Nature has been recognised as a single transboundary 'Natural World Heritage Site' (NWHS) by UNESCO since 1981 (Cote d'Ivoire section) and 1982 (Guinea section) (Granier & Martinez 2011). Nevertheless, this ecosystem is currently threatened by intensive human activities like cacao, coffee, rubber and oil palm cultures, forestry and iron ore exploitation.

Study area and biological material
During biodiversity inventories performed from 2008 to 2013, we trapped shrews at several localities on the Guinean and Liberian sides of Mount Nimba. During the dry (February-March 2008, December 2009-January 2010, December 2010-January 2011 and wet (October 2008) seasons, 153 shrew specimens were trapped using Sherman and pitfalls traps at five localities in Guinea (Table 1). In 2011 and 2013 we collected 73 additional shrews at 11 localities in Liberia, in the ArcelorMittal concession including the ENNR. We surveyed all habitats encountered on this mountain, i.e., lowland and montane forests, secondary forests and agroecosystems from 400 to 1350 m, as well as edaphic savannahs at altitudes of 500-600, 1200 and 1600 m. (Figs 1; 2).
We used 180 Sherman traps baited with a mixture of wheat, oil, peanuts, dryfish and palm nuts to capture small mammals. Each line consisted of 20 to 40 traps at 5 m intervals. In addition, we used pitfall trap lines that consisted of 20 buckets (10 l) 5 m apart and linked with a plastic sheet drift fence 40 cm tall. Our total sampling effort was 4532 bucket-nights for the pitfalls and 9515 trap-nights for Shermans and tomahawks.
Animals were autopsied immediately after capture, following the guidelines of the American Society of Mammalogists (Sikes et al. 2016). We took the following standard external  (30) measurements: head and body length (HB), tail length (T), ear length (E) and hindfoot length without claw (HF). We also recorded weight, sex, and reproductive state. Skulls were either prepared in the field, at the Muséum national d'Histoire naturelle (MNHN) in Paris, France, or in the Durban Natural Science Museum, Durban (DM), South Africa. Eleven craniodental measurements were taken adapted from Dippenaar (1977) and Hutterer & Kock (2002) (Fig. 2) using a Mitutoyo caliper (Mitutoyo, Kawasaki, Japan) with a precision of 0.01 mm. All specimens are housed in the mammal collections of the MNHN under catalogue numbers MNHN-ZM-2012-1051to 2012-1205, 2014-899 to 2014 and at the DM under the numbers 13176-13204 (Appendices 1; 2; 3). Guinean species identification was based on an integrative taxonomic approach using morphological and molecular data (Jacquet et al. 2012). Based on genetic data the morphospecies C. muricauda could represent a complex of species (Jacquet et al. 2012), but pending revision of this species we take the conservative approach and consider it as a single species. The Liberian specimens were barcoded by two of us (A.L., J.G.B.) using cytochrome b partial sequences and identified using BLAST ("Basic Local Alignement Search Tool") on the NCBI website (http://blast. ncbi.nlm.nih.gov/Blast.cgi; option "megablast").  Lee & Elder (1980). We compared our data with those from other cytogenetic surveys in West Africa (Meylan 1967(Meylan , 1971de Hondt 1974;Meylan & Vogel 1982;Lavrenchenko et al. 1997;Baskevitch et al. 1995;Schlitter et al. 1999).

morphometric analySeS
We performed univariate statistics on the external and craniodental measurements on all the newly collected specimens from Mount Nimba ( Fig. 3; Appendices 1, 2). Means, standard deviations and ratios were calculated for weight and external measurements (HB, T, E and HF lengths) for 141 adult specimens in order to provide diagnostic characters between species and to quantify the morphometric variability within them. Similarly, we calculated standard statistics using eleven craniodental measurements (Fig. 3) (a: condyle-incisive length, b: nasal width, c: interorbital width, d: occipital greatest width, e: greatest maxillary width, f: upper tooth row length, g; height of the skull at M2 level, h: greatest braincase height, i: mandibular length, j: lower tooth row length, k: greatest length between extremities of the coronoid and angular processes) for 122 specimens, which had intact skulls.
We then tested for sex-and species-related differences using ANOVAs (Analysis of Variance) and MANOVAs (Multiple

conStruction of an identification key
To assist in the identification of shrew species occurring on Mount Nimba we developed a dichotomous key that could be usable in the field and could work on live animals. Therefore, we used, as far as possible, external characters visible or measurable in the field. However, in a few instances reference to craniodental features was necessary to ensure accurate identifications. To take intraspecific variability into account, especially when few specimens of a species, like C. douceti, C. eburnea and C. nimbasilvanus, were collected, we used additional shrew specimens of the same species housed in the collections of the Muséum national d'Histoire naturelle, among which were 10 type specimens described from Mount Nimba and neighbouring regions (Appendix 3).

RESULTS
trapping reSultS and ecological data Trapping success per locality and device for both the dry and wet seasons are presented in Tables 2 and 3. In Guinea, with pitfall traps, the overall trapping success was better in the wet than in the dry season. In the case of the edaphic and high-altitude savannahs (600 and 1600 m; Gbié and Mare d'hivernage), the trapping success was more than doubled between the dry and wet season (Table 2). There was great variability in capture success rate between pitfall sites, with the best results from primary lowland forest at Bentor, which was alongside a medium-sized river, and in the Mare d'hivernage (high-altitude savannah). With Sherman traps, capture success was always low and similar in the dry and wet seasons, in the different habitats of Guinea (Table 3). The best results were obtained in gallery forests at 600 m (Ziela) and in the high-altitude savannah at 1600 m (Mare d'hivernage). No specimens were trapped in houses in the Guinean side. In Liberia best trapping success was obtained in Tailings (lowland forest) and Grassfield (edaphic savannah). A single individual of C. grandiceps was found in a house in Zolowee village in Liberia in 2013.
Of the 226 specimens captured in this study, 209 (92%) belong to the genus Crocidura and 17 (8%) to the genus Suncus Ehrenberg, 1832 (Table 4). The genus Crocidura was represented by 10 species, and Suncus by one species. We collected mostly the same species on both sides of Mount Nimba with the exceptions of C. eburnea (Liberian side only) and C. douceti (Guinean side only). We did not trap C. nimbae, which was previously recorded from Mount Nimba by Heim de Balsac (1958). Suncus megalura (Jentink, 1888) was trapped in both Liberia and Guinea. In the Guinean side, the most abundant species in our sampling was C. obscurior, which represented 23.5% of all captured specimens. Crocidura buettikoferi, C. theresae and C. jouvenetae each represented about 15% of all captures.

karyotypeS
The three males and two females of C. buettikoferi displayed the same karyotype with diploid number 2n = 52 and fundamental number of autosomes NFa = 66 (Fig. 4A). The autosomal set comprised three pairs of small meta-submetacentric (1 st -3 rd ), five pairs of subtelocentric (4 th -8 th ) and 17 pairs of acrocentric chromosomes (9 th -25 th ) decreasing in size. The X and Y chromosomes were constituted respectively of the largest metacentric and an acrocentric element similar in size to the 19 th pair of autosomes.
The two males of C. grandiceps had 2n = 46 and NFa = 64 (Fig. 4B). The karyotype comprised one pair of small metacentric (1 st ), nine pairs of subtelocentric (2 nd -10 th ) and 12 pairs of acrocentric chromosomes (11 th -22 nd ) decreasing in size. The X was a large-sized metacentric chromosome and the Y was acrocentric and similar in size to the 17 th pair of autosomes.
The three C. jouvenetae specimens displayed identical karyotypes with 2n = 44 and NFa = 68 (Fig. 4C). The autosomal set comprises nine pairs of large to small metacentric and submetacentric chromosomes (1 st -9 th ), four subtelocentric pairs (10 th -13 th ) and eight acrocentric pairs (14 th -21 st ). The X and Y chromosomes were easily recognizable and represented a large-sized metacentric and an acrocentric slightly higher in size than the 21 st pair of autosomes, respectively.
The three males and the female of C. olivieri were characterized by 2n = 50 and NFa = 60 (Fig. 4D). The karyotypes consisted of four small meta-submetacentric pairs (1 st -4 th ), two medium-sized subtelocentric pairs (5 th -6 th ) and 18 acrocentric pairs of chromosomes (7 th -24 th ) decreasing in size. The X and Y chromosomes were metacentric and small acrocentric, respectively.
The male and the female of C. theresae displayed the same karyotype with 2n = 50 and NFa = 70 (Fig. 4E). The karyotype contained three pairs of small bi-armed autosomes similar in size (1 st -3 rd ), eight subtelocentric pairs (4 th -11 th ) and 13 acrocentric pairs (12 th -24 th ) progressively decreasing in size. The X chromosome was a large-sized metacentric and the Y was acrocentric and similar in size to the 21 st pair of autosomes (data not shown).
morphometric data Shrew species occurring on Mount Nimba differed in external body measurements (F = 15.596, P < 0.001), in body weight (F = 135.8, P < 0.001) and in craniodental measurements (F = 3.7812, P < 0.001). Some sexual dimorphism was detected (F = 1.8896, P = 0.120, F = 0.661, P = 0.120 and F = 1.9245, P = 0.070, respectively), but we preferred to keep male and female specimens pooled for the remaining analyses due to relatively low sample sizes.
The CVA performed on the four body measurements (Fig. 5) showed that most of the variability was recovered on axis 1, representing 70.5 % of the total variability (Appendices 4; 5). The highest loadings on axis 1 were HB, HF, E. Axis 2 represented 26% of the variability and the highest loading was TL (Appendix 4). The CVA on external measurements allowed discrimination of C. muricauda and S. megalura from all other species along the second axis with regards to their long tail length (Fig. 5). Crocidura douceti was also partially differentiated along axis 2. Along axis 1 we observed an opposition between the smallest species (negative side of the axis) C. obscurior, C. eburnea, C. douceti and the largest (positive side) C. nimbasilvanus, relatively well differentiated. Mediumsized species like C. jouvenetae, C. theresae, C. buettikoferi, C. nimbae were situated in the centre of axis 1. Crocidura olivieri and C. grandiceps were two large-sized species, the former being larger than the latter. Calculation of means and standard deviations for external body measurements (Table 5) also showed that most values did not overlap between species, which appeared to be size calibrated when considering only the Mount Nimba specimens. A continuum from the smallest to the largest species was observed, each being replaced in turn by another slightly larger species. The smallest species were C. obscurior, C. eburnea and C. douceti that were followed by five slightly larger species C. jouvenetae, C. muricauda, S. megalura, C. theresae and C. buettikoferi. Next were C. grandiceps, C. olivieri and C. nimbasilvanus, which were the largest species. Crocidura muricauda and S. megalura are further differentiated from all other species by their long tails, where tail to head-body ratio is greater than 1.0 ( Table 5).
The plot between the two first axes of the CVAs and on the 11 cranio-dental measurements (Fig. 5) allowed good discrimination between the 11 species of our data set (Fig. 6, Appendices 6; 7). Most of the variability was recovered on axis 1 (86.5 %). All variables were highly correlated with this axis, indicating a size relationship (Appendix 6). On axis 2 (5.5 %), the best correlated variables were nasal width (b), occipital height (h) and molar height (g). Crocidura nimbae could be distinguished from all other species on this. The three smallest shrews (C. obscurior, C. eburnea, and C. douceti) can be distinguished on the negative side of axis 1, with the mean-sized species C. megalura + C. muricauda + C. jouvenetae, followed by the medium-large species C. theresae + C. grandiceps + C. buettikoferi and finally  on the opposite positive side, the two largest species, C. olivieri and C. nimbasilvanus. We obtained a total of 83.6% of correctly classified specimens with the highest scores (100%) for three species C. olivieri, C. nimbae and C. nimbasilvanus. The worst scores were obtained for C. grandiceps (52.6%) and C. eburnea (40%) for which some specimens were incorrectly classified and regrouped by the analysis inside C. theresae, C. buettikoferi and C. obscurior variability ranges (Appendix 7).
The same observations about size classes were made for craniodental measurements for only those species present at Mount Nimba (Table 6) with C. obscurior being the smallest shrew and C. nimbasilvanus the largest. Crocidura nimbae, which was not captured during this study, is a relatively large shrew with skull proportions close to other medium-to large-sized species such as C. grandiceps, C. buettikoferi and C. theresae (Appendices 4, 5).

Species
We confirm here for the first time that C. eburnea occurs on the Liberian side of Mount Nimba and that it is sympatric with C. obscurior in that region. Its absence on the Guinean side of Mount Nimba (Jacquet et al. 2014) and more generally in western parts of Eastern Guinea, despite extensive sampling, needs to be verified by further studies, but may be related to the relatively drier conditions there.
Of the shrews that occur at Mount Nimba, three species are currently classified as "Near Threatened": Crocidura buettikoferi, C. grandiceps (13.7 and 7.1% of all captures, respectively) and C. nimbae (no captures during our study). We captured a single specimen of the rare C. douceti, whose conservation status is currently "Least Concern" (IUCN, 2019). This species has been rarely captured since its original description in 1984. In Ziama forest Nicolas et al. (2009) reported a low relative abundance at 5.9%, and Mamba et al. (2021) recorded a single individual from Wologizi forest in Liberia. Based on these observations, we suggest that the conservation status of this species be re-appraised. To date, there is no IUCN assessment for the two most recently recognised species, C. eburnea and C. nimbasilvanus. We captured two specimens of C. nimbasilvanus and 12 specimens of C. eburnea.
The 12 species of shrews currently known at Mount Nimba may be an underestimate because the high genetic divergence between the two lineages of C. muricauda (Jacquet et al. 2012) may represent cryptic species and requires further investigation.
When compared with the shrew diversity of the Upper Guinea rainforest zone, which harbours 16 species (Burgin & He 2018) we did not record the following species in our study: C. nimbae, C. crossei, C. lamottei, C. nigeriae and C. wimmeri. Crocidura nimbae was recovered in S Sierra Leone, N Liberia and SW Côte d'Ivoire in Taï National Park (Churchfield et al. 2004, Hutterer 2005 and Dodo-Cavally forest reserve (Decher et al. 2005)  with C. jouvenetae, which is now considered as a distinct valid species. Crocidura lamottei was described from Lamto (Côte d'Ivoire) and does not occur in forested environments. Crocidura wimmeri is classified as Critically Endangered and was collected only in Adiopodoume and Banco National Park near Abidjan (Côte d'Ivoire) in a secondary forest (Kadjo et al. 2013, Vogel et al. 2014). Crocidura nigeriae, described from Nigeria, is supposed to be also present in SE Côte d'Ivoire and NE Ghana (Burgin & He 2019) but a recent phylogenetic study concluded that West African specimens attributed to this species should be renamed (Nicolas et al. 2009(Nicolas et al. , 2020. According to Heim de Balsac (1971) C. denti could be present on Mount Loma (Sierra Leone) but these specimens most probably represent misidentifications (Dambry et al. 2016).
The species associations we report here are like those previously published, even though their abundances vary. The three forest-dwelling species C. obscurior (excluding C. eburnea), C. jouvenetae and C. buettikoferi were also trapped in the highelevation savannah (at 1600 m asl) during both wet and dry seasons. Furthermore, our study revealed that most species from Mount Nimba live in syntopy, sharing the same habitat. The lowland primary and secondary forests at 600 m asl harboured 11 shrew species representing two genera (Table 7). This species richness is similar to that reported in other West African (Churchfield et al. 2004;Nicolas et al. 2009;Mamba et al. 2021) and Central African (Brosset 1988;Nicolas et al. 2005) rainforests. Nevertheless, some more species rich communities have been recorded in the rainforests of the Central African Republic. For example, Ray & Hutterer (1996)  We could compare these species abundance data to those of other surveys led in Tai forest (Churchfield et al. 2004), in Haute Dodo and Cavally forests (Decher et al. 2005) and in Ziama forest (Nicolas et al. 2009;Mamba et al. 2021) (Fig. 6). In Taï and Ziama forests, C. obscurior + C. eburnea, C. jouvenetae and C. buettikoferi were abundant, while C. douceti, C. nimbae and C. nimbasilvanus were rare. All were typical forest-dwelling species and were trapped in gallery or secondary forests on Mount Nimba. The relative abundance of the species C. olivieri, C. grandiceps and S. megalura are higher in Mount Nimba than in other studies (Fig. 7). Crocidura obscurior + C. eburnea, C. muricauda dominate in Taï and in Ziama. In Dodo/Cavally it is C. obscurior + eburnea, C. jouvenetae and C. theresae that dominate the community. Crocidura theresae has been recorded from Guinean savannahs, mixed forest and savannah areas and rice fields (Verschuren & Meester 1977; Hutterer 2005; Wilson & Mittermeier 2018), while C. olivieri and S. megalura were encountered in a wide variety of forest and savannah habitats (Hutterer 2005). Crocidura grandiceps is associated with primary rain forests (Hutterer 1983;Grubb et al. 1998), but has also been recorded from small scattered forest fragments surrounded by savannahs in Ghana (Hutterer 2005), and in our study a single individual was trapped in a house. Several authors have reported a synanthropic behavior of C. olivieri (Heim de Balsac 1968Heim de Balsac & Barloy 1966;Verschuren & Meester 1977, Nicolas et al. 2020), but we did not trap this species in houses. Crocidura eburnea is found in lowland forests up to 600 m. We did not capture it in montane forest or in high-elevation savannah, which is in contrast to C. obscurior that was found at all elevations and in all vegetation types. The peculiar composition of the shrew community at Mount Nimba compared with other West African forests can be explained by the geographic location of this mountain at the ecotone zone between rainforest and Guinean woodland. Mount Nimba thus harbours a mixed shrew community composed of both savannah and forest-dwelling taxa.
Crocidura muricauda is less abundant on Mount Nimba (5.3 % of all captures) than in Ziama and Tai National Parks forests (respectively 19.35 & 21.5 %) (Fig. 7). Heim de Balsac (1958) had already noticed the scarcity of this forest-dwelling species on Mount Nimba (four specimens out of more than 100 shrew specimens). Nicolas et al (2009) observed that in Ziama C. muricauda is a typical forest species: it was much more abundant at sites with high understorey density, high canopy height and cover, and high density of stems and trees, than in logged or agricultural landscapes. Both lowland and montane forests in the Nimba Mountains are of variable states of integrity, ranging from relatively intact to highly degraded by anthropogenic activities (seasonal fires across the entire range, mining in two localised areas, and logging and clearance for agriculture in the lowlands). This may explain the relative rarity of C. mauricauda on Mount Nimba. None of the forest-dwelling shrew species on Mount Nimba appeared to be restricted to primary habitat. These results are congruent with those of the long-term study of Nicolas et al. (2009) in Ziama Man Biosphere Reserve forest who demonstrated that shrew communities were not significantly affected by agricultural activities and that their diversity was similar in primary and secondary forests.
At Mount Nimba, the shrew community in montane forest (1000-1350 m asl) was not different from that in lowland forest except in terms of diversity with seven species instead of 11; demonstrating that the shrew community in montane forest may be a subset of that in lowland forest (Table 7). However, for logistics reasons our trapping efforts concentrated mostly on lowland forest (6750 Trap nights) compared to montane forest (2573 trap nights). The edaphic savannah (from 1200 to 1600 m asl), which is peculiar to the Guinean side of Mount Nimba, harboured five species, C. obscurior, C. jouvenetae, C. theresae, C. buettikoferi and S. megalurus. Only two species, C. obscurior and C. theresae, were recorded from this habitat by Heim de Balsac (1958), Verschuren & Meester (1977) and Lamotte & Roy (2003). Crocidura theresae was found in grassland savannahs, bush between 600 and 1600 m by Verchuren & Meester (1977). The altitude savannah biotope does not seem to promote isolation for shrews, as no endemic taxon was identified, suggesting that this vegetation is relatively recent in origin. Our results suggest that shrews demonstrate habitat plasticity at Mount Nimba, as suggested by Verschuren & Meester (1977). Some species like C. eburnea and C. douceti or C. nimbasilvanus were found only in lowland forests. They all corresponded to species trapped in low abundance despite relatively intense trapping on the Guinean side of the Mount in this environment. Crocidura douceti is known in relict and gallery forest as well as anthropogenic environments in Taï National Park (Churchfield et al. 2004) but was considered as a forest shrew of low density by Heim de Balsac (1958). The holotype of C. eburnea comes from Mt Tonkoui situated at 1200 m asl, which could indicate that the species is also living in montane forest of Nimba range region. Crocidura nimbasilvanus is more common in closed environments according to Nicolas et al. (2009) but was captured in low abundance in Ziama forests. According to Burgin & He (2018), C. nimbae is found in submontane and lowland forest but was known by a few specimens in Zouguepo Nimba lowland forest (Heim de Balsac 1958).

cytogenetic characterization of mount nimba ShrewS
Cytogenetic analyses in mammals showed that there is extensive karyotypic diversity among extant species and that many closely related species or even populations possess different karyotypes indicating that chromosomal differentiation often occurs during, or shortly after cladogenesis (Dobigny et al. 2017). Thus, descriptions of mammalian karyotypes serve an important role for characterizing chromosomal rearrangements, which provide information on genetic barriers to gene flow and ultimately on the processes involved in speciation. In this paper we provide cytogenetic data for five of the 11 captured species. For each species several individuals were studied in order to test for intraspecific variability but a remarkable result of this study is that karyotypes of all species are stable. In the following paragraphs we compare our data with previously published data from the same or closely related species.
The standard karyotype of C. buettikoferi is characterized by 2n = 52 and NFa = 66. Our results are in good agreement with the report of Meylan (1971) and Meylan & Vogel (1982) from Côte d'Ivoire (as C. poensis pamela Dollman, 1915). The authors identified 17 pairs of acrocentric decreasing in size, five pairs of subtelocentric and three pairs of small meta/submetacentric chromosomes. The X and Y chromosomes are a large-sized metacentric and a small acrocentric, respectively. It is important to note that C. buettikoferi from Mount Nimba has very small short arms at the 9 th pair of autosomes, the same as C. olivieri (7 th ) and C. theresae (12 th ). With a significant chromosome contraction, those arms are barely noticeable. In addition, they are C-positive and vary in size, so we consider this pair of autosomes as acrocentric.
The karyotype of C. grandiceps is 2n = 46 and NFa = 64. This is in agreement with data previously published by Meylan & Vogel (1982) in Côte d'Ivoire, as C. cf. nimbae.

C . o b s c u r i o r + e b u r n e a C . j o u v e n e t a e C . t h e r e s a e C . b u e t t i k o f e r i C . o l i v i e r i C . g r a n d i c e p s C . m u r i c a u d a C . d o u c e t i C . n i m b a s i l v a n u s C . n i m b a e C . m e g a l u r a
Taï Ziama Dodo/Cavally  Table 5.
Denys C. et al.
Our specimens of C. jouvenetae are characterized by 2n = 44 and NFa = 68. The first karyotype of C. jouvenetae was described from Côte d'Ivoire by Meylan (1971) with 2n = 44 and NFa = 62. Variation in the short arms of several pairs of chromosomes can result from difficulties in categorizing chromosomes and homologous pairs, which can be considered as acrocentric or subtelocentric (see Fig. 2 in Meylan, 1971). The chromosome formula 2n = 44 and NFa = 62 was described for C. crossei (jouvenetae) ebriensis from Côte d'Ivoire (Meylan & Vogel 1982), but the taxonomic status of these specimens still needs to be assessed and they could be a synonym of C. jouvenetae (Hutterer 2005). In the same work, Meylan & Vogel (1982) described the karyotype of C. cf. planiceps from Côte d'Ivoire, which also displayed 2n = 44, but differed by NFa = 68. The five C. jouvenetae investigated in our study display the same standard formula as C. cf. planiceps. According to Wilson & Mittermeier (2018) the species C. planiceps Heller, 1910 is known from Nigeria and Uganda. Thus the taxonomic status of the C. cf. planiceps specimens from Côte d'Ivoire needs to be re-evaluated. However, an integrative study combining morphological, molecular and cytogenetic data is now required to review the phylogeography and taxonomy of the species C. jouvenetae.
Cytogenetic (Meylan 1967(Meylan , 1971de Hondt 1974;Meylan & Vogel 1982;Baskevitch et al. 1995;Lavrenchenko et al. 1997) and molecular studies (Quérouil et al. 2005;Dubey et al. 2007Dubey et al. , 2008Jacquet et al. 2015) have led to the unification of a high number of large-sized shrews under the name C. olivieri (Hutterer 2005; Burgin &He 2018). A summary of the various available karyotypic data for C. olivieri was presented by Schlitter et al. (1999). The first karyotype of C. olivieri was described by Meylan (1967) under the name C. occidentalis kivu Osgood, 1910 from Democratic Republic of Congo with 2n = 50 and NFa = 62. Later surveys from various parts of Africa revealed the same karyotype for all forms that are currently considered conspecific to C. olivieri (Meylan 1971 from Côte d'Ivoire;de Hondt, 1974 from Egypt;Meylan & Vogel 1982 from Mali, Cameroon, Nigeria and Burkina Faso;Baskevitch et al. 1995 from Ethiopia). The formula we describe here from Mount Nimba (2n = 50, NFa = 60) is in agreement with these previous surveys. However, some differences exist in categorizing acrocentric/subtelocentric largest pair of autosomes. In the first studies, this pair is considered as acrocentric (Meylan 1967(Meylan , 1971. In later papers, it is attributed to the subtelocentric (De Hondt 1974;Meylan & Vogel 1982;Maddalena et al. 1987;Lavrenchenko et al. 1997;Schlitter et al. 1999). In our study we adhere to the terms of Meylan (1967Meylan ( , 1971. We consider this pair of autosomes as acrocentric because its very small short arms are C-positive and can vary in size, while the short arms of the other two pairs of subtelocentrics are C-negative and dimensionally stable (Aniskin, unpubl. data).
The diploid number of C. theresae from Mount Nimba is the same as C. olivieri (2n = 50), but the karyotype differs by the fundamental number of autosomes (NFa). In our study the karyotype of C. theresae comprises 11 pairs of bi-armed autosomes and 13 acrocentric pairs, with NFa = 70. Our results show some differences with the karyotype of C. theresae from Côte d'Ivoire by Meylan (1971), who identified 15 pairs biarmed autosomes and nine acrocentric pairs, with NFa = 78. Additional analyses combining morphological, molecular and cytogenetic data is now required to review the phylogeography and taxonomy of this species. morphometric characteriSticS of ShrewS from mount nimba Some species could not readly be discriminated with certainty based on morphometric analyses. This is particularly the case for C. eburnea and C. obscurior for which external and craniodental measurements overlapped significantly (Tables 5, 6), demonstrating their status as sibling species. The two holotypes have distinct sizes (Appendix 3), with C. eburnea slightly smaller, but this difference disappears when additional specimens are included in the analysis. In the original description, Heim de Balsac (1958) highlighted some dental differences between these taxa like a more developed P3 and paracone of P4. These characters were not validated as diagnostic by Jacquet et al (2014), who instead demonstrated that both species are small sized ones (body length 50-65 mm) with a uniform dark pigmentation of the skin and the fur (Jacquet et al. 2014).  Balsac, 1958C. eburnea Heim de Balsac, 1958C. jouvenetae Heim de Balsac, 1958C. theresae Heim de Balsac, 1968C. buettikoferi Jentink, 1888C. grandiceps Hutterer, 1983C. muricauda (Miller, 1900 C. douceti Heim de Balsac, 1958C. olivieri (Lesson, 1827) C. nimbasilvanus Hutterer, 2003S. megalura (Jentink, 1888 The morphometric analyses based on weight, external and craniodental measurements show that sympatric shrew species from Mount Nimba can be classified along a size gradient from the tiny C. obscurior-C. eburnea to the giant C. nimbasilvanus. This is congruent with the size range theory of Brosset (1988), who demonstrated that within a shrew community composed of 11 species from Makokou forest (Gabon), each species was replaced by another along a size gradient. All species were syntopic and showed the same activity periods. This suggests that species-rich shrew communities in African rain forests partition resources in this way, decreasing interspecific competition and allowing coexistence of multiple species (Hutterer et al. 1987;Brosset 1988;Churchfield et al. 1999;Goodman et al. 2001).
The morphometric analyses also reveal an important intraspecific uniformity for most species. External measures are thus often used to define diagnostic characters and to build identification keys (Hutterer & Happold 1983) (Appendices 4; 5). A notable exception is found between the smallest species C. eburnea and C. obscurior which are impossible to separate by external size characters (Jacquet et al. 2014).

morphological identification key
We developed an identification key based as much as possible, on both body measurements (head & body, tail and hindfoot lengths, tail / head + body length ratio) and external characters (fur and skin color, tail hairiness, quantity and length of tail vibrissae). The same characters were used by Hutterer & Happold (1983) to build a key for Nigerian shrews.
Our key needs to be tested in the future on other west African sites and be amended by the missing taxa once they have been well described using an integrative taxonomic approach sensu Taylor et al. (2019).

CONCLUSION
Our surveys on the Guinean and Liberian slopes of Mount Nimba has allowed us to better understand the biodiversity and ecological requirements of shrews at this unique mountain. We collected 11 species of shrews in total, which confirmed the high diversity of shrews on Mount Nimba, especially in lowland forests (11 species) compared to the edaphic savannah from 1200 to 1600 m (five species). Only three species known in the western parts of the Upper Guinea rainforest zone were not captured in our study: C. nimbae, C. wimmeri and C. lamottei. The shrew community at Mount Nimba combines a mixture of both savannah (Crocidura theresae, C. olivieri, S. megalurus) and forest-dwelling species (C. buettikoferi, C. obscurior, C. eburnea, C. jouvenetae and C. nimbasilvanus). This highlights the importance of Mount Nimba for the conservation of shrews, as has been demonstrated for bats (Monadjem et al. 2016). We also described for the first time karyological data of shrews from Mount Nimba. Our results were congruent with previous data, except for C. theresae, for which differences with a karyotype described from Côte d'Ivoire were highlighted, maybe owing to cryptic vari-ability. Morphometric analyses based on external body and craniodental measurements revealed that shrew species from Mount Nimba can be classified along an increasing size gradient, congruently with the size range theory of Brosset (1988). Based on body measurements and external characters, we developed an identification key of shrews from Mount Nimba, which can be easily used in the field and on live animals. the field study. The research and exportation authorisations were provided by the Guinean Ministry "Eaux et Forêts" and we are grateful to Dr C. Sagno and O. Diallo. We dedicate this work to the late Fodé Kourouma who participated to the 2008, 2010 and 2013 fieldworks and was a great field assistant. Fodé Kourouma also prepared the skulls. M. Mfonkumun prepared skulls under supervision of A. Delapré in SPOT (Paris, MNHN). Peter Farnloe and Moses Darpey provided fantastic assistance in the field; their effort and company was greatly appreciated. Tanya Romanenko was always quick to provide support wherever it was needed during earlier trips to Liberian Nimba. Martyn Gest is thanked for help in the field during March 2013. Wing-Yunn Crawley and John Howell, ArcelorMittal, have provided critical support for continued biodiversity studies on the ArcelorMittal concession. AM and CD were employed by ArcelorMittal and SMFG and had signed non-disclosure agreements with these mining companies. Molecular analyses were performed at the Service de Systématique moléculaire (UMS 2700 Acquisition et Analyse de Données pour l'Histoire naturelle (2AD), MNHN, Paris). Guy Rabache must be thanked for his help on figures design.