Taxonomic revision of the pampas cat Leopardus colocola complex (Carnivora: Felidae): an integrative approach

Abstract

The pampas cat Leopardus colocola has been subject to conflicting classifications over the years. Currently, one polytypic species with seven subspecies is recognized, but integrative taxonomic study for this debated group has never been done. Here, we combine the broadest morphological coverage of the pampas cat to date with molecular data and ecological niche models to clarify its species composition and test the validity of recently proposed subspecies. The multiple lines of evidence derived from morphology, molecular, biogeography and climatic niche datasets converged on the recognition of five monotypic species: L. braccatus, L. colocola, L. garleppi (including thomasi, budini, steinbachi, crespoi and wolffsohni as synonyms), L. munoai and L. pajeros (including crucina as synonym). These five species are morphologically diagnosable based on skin and skull traits, have evolved in distinct climatic niche spaces and were recovered in molecular species delimitation. Contrary to previous taxonomic arrangements, we do not recognize subspecies in pampas cats. To objectively define the two most controversial species, we designate neotypes for L. colocola and L. pajeros. The diversification of pampas cats is associated with Middle Pleistocene glaciations, but additional genetic samples from the central Andean region are still needed to conclusively reconstruct its evolutionary history.

INTRODUCTION

The pampas cat, Leopardus colocola (Molina, 1782), is a South American small-sized felid (± 3 kg) characterized by long hairs on the body, an erectile spinal crest slightly darker than ground colour, transverse dark stripes on the throat, markings on the flanks, legs with transverse dark stripes in the proximal portion, ears more pointed and tail relatively shorter than other South American felids and a lingual cavity between lower canine teeth (Pocock, 1941; Cabrera, 1958, 1961; Guggisberg, 1975; Redford & Eisenberg, 1992; Salles, 1992; García-Perea, 1994; Eisenberg & Redford, 1999; Sunquist & Sunquist, 2002, 2009; Prevosti, 2006). Pampas cats are typical of open areas and show a broad elevational distribution from sea level up to 5000 m (Redford & Eisenberg, 1992; García-Perea, 1994; Nowak, 1999; Sunquist & Sunquist, 2002, 2009). Phylogenetic studies show that pampas cats are part of the group informally known as the ‘ocelot lineage’, which corresponds to small- and medium-sized Neotropical spotted cats of the genus LeopardusGray, 1842 (Johnson et al., 1999, 2006; Mattern & McLennan, 2000; Li et al., 2016; Kitchener et al., 2017). Pampas cats are closely related to the Andean mountain cat, Leopardus jacobita (Cornalia, 1865), the oncilla or tigrina, Leopardus tigrinus (von Schreber, 1775) species group, Geoffroy’s cat, Leopardus geoffroyi (d’Orbigny & Gervais, 1844) and the kodkod, Leopardus guigna (Molina, 1782) (Johnson & O’Brien, 1997; Johnson et al., 1999, 2006; Mattern & McLennan, 2000; Li et al., 2016).

The taxonomic history of pampas cats started with Molina (1782) describing Felis colocola, based on a cat from the forests of Chile (‘Boschi del Chili’). Since then, numerous related species and subspecies have been described, usually based on a single or a few specimens: Felis pajerosDesmarest, 1816 from Argentina, Felis braccataCope, 1889 from central Brazil, Felis pajeros crucinaThomas, 1901 from southern Argentina, Lynchailurus pajeros garleppiMatschie, 1912 from Peru, Felis pajeros thomasiLönnberg, 1913 from Ecuador, Lynchailurus pajeros budiniPocock, 1941 from north-western Argentina, Lynchailurus pajeros steinbachiPocock, 1941 from Bolivia, Felis (Lynchailurus) colocolo crespoiCabrera, 1957 from north-western Argentina, Felis colocola muñoaiXiménez, 1961 from Uruguay and Lynchailurus colocolo wolffsohniGarcía-Perea, 1994 from northern Chile.

The use of the specific name colocola and its derivations (colocolo, colorolla, colicollo, etc.) (Molina, 1782, 1786, 1788, 1789, 1808, 1809, 1810; Bechstein, 1800; Shaw, 1800; Fischer, 1829; Wagner, 1841) has been historically debated since the late 18^th century (Cabrera, 1940, 1958, 1961; Osgood, 1943; Kitchener et al., 2017). The main problem started when Griffith (1821) and Griffith et al. (1827) associated Molina’s F. colocola with an unidentified specimen from Guyana based on information and plates provided by Charles Hamilton Smith, an opinion largely followed in subsequent works (e.g. Fischer, 1829; Jardine, 1834; Reichenbach, 1834; Gay, 1847; Giebel, 1855; Gray, 1867; Fitzinger, 1869; Mivart, 1881; Trousseart, 1885, 1897, 1904; Elliot, 1883; Matschie, 1894, 1895), although rejected by others (e.g. Swainson, 1838; Wagner, 1841; Gray, 1874). The confusion became even greater when Felis jacobitaCornalia, 1865 (= Leopardus jacobita) was erroneously associated with the name colocola (Philippi, 1869, 1870, 1873; Burmeister, 1879; Lydekker 1896; Allen, 1919; Yepes, 1929; Pocock, 1941; Schwangart, 1941).

Throughout the 20^th and early 21^st centuries, a series of taxonomic studies involving pampas cats was carried out, resulting in contrasting taxonomic schemes (Table 1). In the most recent revision of the group, García-Perea (1994), based on skull and external morphological characters of 86 specimens, recognized three allopatric polytypic species: Lynchailurus colocolo (including Ly. c. colocolo and Ly. c. wolffsohni) distributed on the western slope of the southern Andes (northern and central Chile), Ly. pajeros (including Ly. p. pajeros, Ly. p. crucinus, Ly. p. budini, Ly. p. steinbachi, Ly. p. crespoi, Ly. p. garleppi and Ly. p. thomasi) distributed along the Andes, from Ecuador to southern Chile and Argentina and Ly. braccatus (including Ly. b. braccatus and Ly. b. munoai) found in Brazil, Paraguay and Uruguay. This arrangement was largely followed by Wozencraft (2005), but he placed the group in the genus Leopardus and considered crespoiCabrera, 1957 and crucinaThomas, 1901 as junior synonyms of L. p. budini (Pocock, 1941). Later, Nascimento (2010), in his taxonomic revision of genus Leopardus, recognized six species for the pampas cat species complex: Le. colocolo, Le. pajeros, Le. braccatus, Le. munoai, Le. garleppi and Le. budini. Recently, Kitchener et al. (2017), evaluating previous studies, recognized only one species with seven subspecies: Leopardus colocola colocola, Le. c. wolffsohni, Le. c. pajeros (including crucina), Le. c. budini (including steinbachi), Le. c. garleppi (including thomasi), Le. c. braccatus and Le. c. munoai.

In the last decades, molecular studies have revealed a complex and recent evolutionary and biogeographic history of the pampas cats, including cases of hybridization with Snethlage’s tigrinas Leopardus emiliae (Thomas, 1914) in central and north-eastern Brazil (Johnson et al., 1999; Napolitano et al., 2008; Trigo et al., 2008, 2013, 2014; Cossíos et al., 2009; Ruiz-García et al., 2013; Li et al., 2016; Santos et al., 2018). Glaciation events during the Pleistocene caused repeated changes in population sizes and distributions, resulting in the establishment of structured phylogenetic clades that have been linked to some of the previously morphology-based taxa (Johnson et al., 1999; Cossíos et al., 2009; Santos et al., 2018). Nevertheless, comprehensive systematic analyses of the pampas cats complex throughout its entire distribution have never been done. In this study, integrating morphology, molecular and ecological datasets, we aim to revise the taxonomy of the pampas cats complex to properly define its taxa and test the validity of species and subspecies currently recognized.

MATERIAL AND METHODS

Samples and collections

We studied skins and skulls of L. colocola group specimens (N = 142) (Fig. 1) housed in the following museums and institutional collections: American Museum of Natural History, New York, USA (AMNH); Colección Boliviana de Fauna, La Paz, Bolivia (CBF); Colección Mamíferos Lillo, Tucuman, Argentina (CML); Field Museum of Natural History, Chicago, USA (FMNH); Instituto Alexander von Humboldt, Villa de Leyva, Colombia (IAVH); Museo Argentino de Ciencias Naturales ‘Bernardino Rivadavia’, Buenos Aires, Argentina (MACN); Museo de Historia Natural Javier Prado, Universidad Nacional Mayor de San Marcos, Lima, Peru (MUSM); Museo Nacional de Historia Natural, Asuncion, Paraguay (MNHNP); Museo Nacional de Historia Natural, Montevideo, Uruguay (MNHN-M); Museu de Zoologia da Universidade de São Paulo, São Paulo, Brazil (MZUSP); Museu Nacional da Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil (MNRJ); National Museum of Natural History, Smithsonian Institution, Washington DC, USA (USNM); and Zoologisches Museum Berlin (= Museum für Naturkunde), Berlin, Germany (ZMB_MAM). We also examined photographs of specimens housed in the following collections: Coleção de Mamíferos, Universidade de Brasília, Brasília, Brazil (UnB); Estación Biológica de Doñana, Sevilla, Spain (EBD); Senckenberg Gesellschaft für Naturforschung, Frankfurt, Germany (SMF); Muséum National d’Histoire Naturelle, Paris, France (MNHN); Museum of Comparative Zoology, Harvard University, Cambridge, USA (MCZ); Museum of Vertebrate Zoology, University of California, Berkeley, USA (MVZ); and the Natural History Museum, London, UK (NHM). A list of all specimens examined, including their museum collection numbers and localities, is presented in the Supporting Information, Appendix S1.

Examined material included the type specimens of Felis braccataCope, 1889 housed at AMNH; Felis (Lynchailurus) colocolo crespoiCabrera, 1957 housed at MACN; Felis colocola muñoaiXiménez, 1961 housed at MNHN-M; and Lynchailurus colocolo wolffsohniGarcía-Perea, 1994 housed at USNM. We also examined photographs of the type specimens of Felis pajeros crucinaThomas, 1901, Lynchailurus pajeros budiniPocock, 1941, Lynchailurus pajeros huinaPocock, 1941 and Lynchailurus pajeros steinbachiPocock, 1941 housed at NHM and Lynchailurus pajeros garleppiMatschie, 1912 housed at ZMB_MAM.

We provide updated distribution maps based on records from museum specimens and literature. Some literature records were not included due to dubious identifications.

Morphological dataset and statistical analyses

Analyses of characters

All specimens were examined in qualitative and quantitative terms based on external and cranial features. For a detailed analysis of geographic variation, individuals from nearby localities were grouped according to the following criteria: morphological homogeneity of the sample, geographical proximity, ecological similarity of the localities and the absence of clear geographical barriers between collecting localities. In some cases, it was not possible to group the exact same location for the analysis of the skull and skin because the distribution of the available sample of each type of material was different.

The external qualitative characters refer to the coloration and markings pattern of the pelage: (1) colour of cheek lines, (2) colour of the cheek region, (3) colour of the mandible, (4) colour of the medial infranasal region, (5) colour of forehead, crown and nape, (6) colour of the dorsal surface of ears, (7) width of gular stripes, (8) colour of gular stripes, (9) colour of the sides of the body, (10) colour of throat, chest, abdomen and ventral parts of the legs, (11) colour of the spinal crest, (12) markings on the sides of the body, (13) shape of chest and abdomen markings, (14) colour of chest and abdomen markings, (15) colour of leg stripes, (16) colour of the feet and (17) tail markings. Craniodental qualitative characters were largely based on García-Perea (1994): (1) degree of development of ectotympanic and entotympanic, (2) morphology of mastoid process, (3) shape of notch for postpalatine vein, (4) shape of posterior margin of the palate, (5) development of the sagittal crest, (6) presence of upper second premolar (P2), (7) paracone of the upper third premolar, (8) parastyle of the upper third premolar and (9) protocone of the upper fourth premolar. A detailed description and the distribution of states of each character analysed in this study are presented in Supporting Information, Appendix S2.

The quantitative external characters are the measurements taken from the specimen labels as follows: (1) total length (TL), (2) head–body length (HB), (3) tail length (T), (4) hindfoot length (HF) and (5) ear length (E). When only total length was provided, we subtracted the recorded tail length from total length to obtain the values of the head and body length. Quantitative craniodental characters comprises 16 measurements taken with digital callipers to the nearest 0.01 mm from adult specimens of both sexes (Fig. 2): (1) greatest length of the skull (GLS), (2) condylobasal length (CBL), (3) rostral length (RL), (4) zygomatic breadth (ZB), (5) greatest breadth of braincase (GBB), (6) maximum rostral breadth across outer margins of upper canines (RBC), (7) breadth between the infraorbital foramina (IFB), (8) greatest palatal breadth (GPB), (9) greatest palatal length (GPL), (10) C‑M1 length (CM1L), (11) anteroposterior diameter of the auditory bulla (DAB), (12) maximum breadth across mastoid processes (MST), (13) anteroposterior length of temporal fossa (ALT), (14) p3-m1 length (p3m1L), (15) mandible height (MH) and (16) mandible length (ML).

Statistical analyses

In all statistical analyses, we only used adult individuals defined based on dental morphology (permanent teeth fully erupted) and fusion of the cranial sutures (especially the spheno-occipital suture) (Ximénez, 1974; García-Perea, 2002). We first applied the Kolmogorov–Smirnov test to assess the normality of craniodental variables and Student’s t-test to evaluate the existence of sexual dimorphism (P < 0.05). For the multivariate analyses, all craniodental variables were log₁₀ transformed and missing values (< 2% of the dataset) were estimated using the Amelia package (Honaker et al., 2011) in R software (R Core Team, 2019). We evaluated the reliability of the estimation by comparing the distribution of the inferred and observed values. These two sets of values showed similar distribution patterns indicating an accurate estimation (Honaker et al., 2011). Principal component analysis (PCA) was extracted from the correlation matrix and used as an exploratory tool for investigating the major patterns of variation and to evaluate the degree of separation among pampas cats. Only principal components with eigenvalues > 1 were extracted. We performed a discriminant function analysis (DFA) to investigate whether the groups of pampas cats studied could be distinguished based on craniodental morphology. Lastly, as used in previous taxonomic studies of Carnivora (e.g. Christiansen, 2008), we assessed the diagnosability of groups using the ratio of cranial measurements. Statistical analyses were performed using SPSS 17.0 and R software.

Molecular sampling and analyses

To investigate the phylogenetic relationships among pampas cat populations we combined all sequences from previous studies (Johnson et al., 2006; Cossíos et al., 2009; Trigo et al., 2013; Santos et al., 2018). In total, the dataset encompassed 2317 bp (134 bp for ATP8, 588 bp for CR, 1029 bp for Cytb and 566 bp for ND5) of sequences obtained from 7, 38, 12 and 24 individuals, respectively. We also included sequences from other extant species of Leopardus as outgroups (Le. geoffroyi, Le. guigna, Le. pardalis, Le. tigrinus and Le. wiedii). The sequence alignments were performed in MEGA 6 (Tamura et al., 2013) using the MUSCLE algorithm (Edgar, 2004).

Phylogenetic and divergence time analyses

To reconstruct the phylogenetic tree of pampas cats, we used a Bayesian inference (BI) approach, in which all four loci were partitioned. To select the best model of molecular evolution of each locus, we used the Bayesian information criterion (BIC) in jModelTest 2.1.7 (Darriba et al., 2012). The following substitution models were selected: HKY+I for ATP8, HKY+I+G for CR, HKY+G for Cytb and HKY+I for ND5. One cold and three heated Markov chain Monte Carlo (MCMC) analyses were performed twice for at least 20 million generations, with trees sampled every 1000 generations until convergence (SD < 0.01). The first 25% of the Markov chain samples (N = 20 000) were discarded as burn-in, and the remaining samples were used to generate majority rule consensus trees and summary statistics.

To estimate the divergence times of the Le. colocola complex, we used a Bayesian phylogenetic approach in BEAST 1.8.2 (Drummond et al., 2012). The analysis was performed using the concatenated dataset. An HKY+I+G model of base pair substitution and a prior coalescent tree of constant size were used. Two calibration points (CP) were used in credibility intervals following Johnson et al. (2006) and the parameters were set to log-normal distributions with a 95% interval boundary. CP1: the basal diversification of Leopardus (2.02–4.25 Myr), we assigned a minimum age of 2.02 Myr [offset = 2.02; mean = 0.50; log (SD) = 1.00]. CP2 represents the basal divergence in the clade (Le. jacobita + Le. colocola + Le. tigrinus + Le. geoffroyi + Le. guigna) (1.68–3.56 Myr), we assigned a minimum age of 1.68 Myr [offset = 1.68; mean = 0.50; log (SD) = 1.00]. The clock model was selected based on the marginal likelihood estimated (MLE) from stepping stone and path sampling (Baele et al., 2012) with 50 path steps, 500 000 iterations and samples every 500 generations. The procedure was run twice to ensure convergence and an uncorrelated relaxed log-normal molecular clock was chosen. The MCMC chains were run for 20 million generations, with sampling every 1000 generations. The convergence of the MCMC chains was examined in TRACER 1.6 (Rambaut et al., 2014) and the first 25% of runs were discarded as burn-in.

Species delimitation

To test species boundaries of the Leopardus colocola complex, we used four different methods of species discovery and species validation. The Poisson tree process (PTP) method (Zhang et al., 2013) was implemented in the majority consensus tree from the above-concatenated Bayesian analysis without a priori assignment regarding putative species. The bPTP web server (http://species.h-its.org/ptp) runs the original maximum likelihood (ML) version of the PTP and an updated version that adds Bayesian support to delimited species in the input tree. We ran for 500 000 generations with thinning set to 100 and burn-in to 10%. The mean interspecific Kimura two-parameter (K2P) distance was assessed for each candidate species in MEGA 6.04 (Tamura et al., 2013). The species delimitation plugin (Masters et al., 2011) in GENEIOUS 9.1.4 (Kearse et al., 2012) was used to obtain PID (Liberal) statistical values, which are based on the putative species resulting from BI trees and used to calculate the mean probability of the interspecific genetic distance ratios for these candidate species. The putative priori ‘species’ was set according to the phylogenetic topologies from the gene trees (clades A–E recovered in BI tree).

To validate the candidate species discovered from the previous steps, BP&P v.3.2 (Yang & Rannala, 2010) and *BEAST (Heled & Drummond, 2010) in BEAST 1.8.2 were used and the topology was obtained from BI tree. In BP&P, we used the A11 and A01 algorithms to explore different species delimitation models and species phylogenies. We ran reversible‐jump MCMC (rjMCMC) analyses for 500 000 generations (with a five-generation sampling interval) with a burn‐in phase of 100 000. The priors were set to θ ~ G (20, 2, 000) and τ ~ G (20, 1, 000), and the other divergence time parameters were assigned to the Dirichlet prior (Yang & Rannala, 2014). Each analysis was run twice with different starting seeds to ensure consistency. In *BEAST, the substitution models were linked and the substitution parameters were the same as in the divergence time analysis. We chose the Yule process species tree priors in a piecewise linear fashion and a constant root population size model. The uncorrelated relaxed log-normal clock was set for all loci. The MCMC chains were run for 100 million generations with sampling every 2000 generations. The convergence of the MCMC chains was examined in TRACER 1.6 and the first 25% were discarded as burn-in.

Climatic niche comparison

In a complementary approach, we investigated climatic niche segregation among groups of pampas cats. Specifically, we tested whether the groups defined on morphological grounds also showed ecological differences and occupied distinct niches. The use of ecological niche analyses as a tool to help set species boundaries has been largely recognized (Sites & Marshall, 2003; Wiens, 2004; Wiens & Graham, 2005; Raxworthy et al., 2007). According to Wiens & Graham (2005: 522), ‘if a set of populations of uncertain taxonomic status [as in the Le. colocola complex] is geographically separated from closely related species by areas that are outside of the climatic niche envelope of all of these species, then gene flow within these species is unlikely because it would involve crossing unsuitable habitat. This pattern would support the hypothesis that the populations of uncertain status represent a distinct species.’ Therefore, by testing the climatic niche segregation among groups of pampas cats we aim to provide additional evidence on their taxonomic rank.

For each individual collection locality (Fig. 1), we extracted 22 environmental parameters via the raster R package (Hijmans, 2017). Nineteen environmental variables (11 based on temperature and eight on precipitation) and elevation were extracted from WordClim v.2, at 30 s spatial resolution (Fick & Hijmans, 2017). Net primary productivity (NPP) and potential evapotranspiration (PET) were obtained from the NASA Socioeconomic Data and Applications Center (Imhoff et al., 2004; Imhoff & Bounoua, 2006) and the CGIAR Consortium for Spatial Information (Trabucco & Zomer, 2009).

We first evaluated the breadth of the environmental space occupied by pampas cats and assessed which set of environmental variables were most closely associated with their distribution via PCA. The niche segregation was tested via multivariate analysis of variance (MANOVA), followed up by one-way analysis of variance (ANOVA) to test the isolated effect of each parameter. All environmental traits were standardized prior to analyses.

In addition, we built ecological niche models (ENM) for each group of pampas cats to estimate their potential distribution. To avoid collinearity, we excluded highly correlated environmental variables (r > 0.80). The remaining variables are temperature diurnal range, isothermality, temperature annual range, temperature dry quarter, temperature warmest quarter, temperature coldest quarter, annual precipitation, precipitation seasonality, precipitation wettest month, precipitation driest quarter, precipitation warmest quarter and precipitation coldest quarter. ENMs were constructed using the MaxEnt 3.4.1 algorithm implemented in the ENMenv R package (Muscarella et al., 2014). To test the performance of our models, we partitioned the localities into testing and training bins using the ‘checkerboard2’ method for Groups II, III, IV and V (see Results for group definition) and the ‘jackknife’ method for Group I due to its small sample size (Muscarella et al., 2014). One thousand background points were randomly selected for model training from a buffer area that extended five decimal degrees from the most marginal records. We compared models with distinct complexities using regularization multipliers ranging from 0 to 5 with five combinations of feature classes (L, LQ, LQP, H, LGH; L = linear, Q = quadratic, H = hinge, P = product) and ranked via second‐order corrected Akaike information criterion (AICc). The performance of the best model (smallest AICc) was evaluated using the area under the curve (AUC) (Mason & Graham, 2002). AUC values can range from 0 to 1, where values closer to 1 indicate higher performance in discriminating occurrence data of species from random background points (Phillips et al., 2004).

RESULTS

Definition of groups

Based on a consistent combination of external and cranial qualitative traits (for a full description of characters, see: Supporting Information, Appendix S2, Figs. S1–4), we sorted our sample into five groups that were geographically structured and without intermediate forms (Fig. 3). Below we describe each group with their respective geographical distribution. Subsequent statistical analyses were performed using these groups as baselines.

Group I includes individuals found in Central Chile on the western slope of the Andes. They are characterized by an ash grey background colour of forehead, crown and nape, speckled with rusty cinnamon; the medial infranasal area, cheeks, throat and areas around the lips and eyes are white; the mandible is white or pale yellowish brown. Two rusty cinnamon lines cross the cheeks in parallel and rusty cinnamon, transversal gular stripes are present, of which at least one is markedly wider than others. The ears are triangular with tips and margins blackish, with the remaining portion rusty or cinnamon. The spinal crest is dark grey with a few cinnamon hairs. The general colour of the body is ash grey as on the forehead, with distinct rusty cinnamon oblique lines on the sides of the body. The throat, chest, abdomen and ventral parts of the legs are white, and there are irregular transversal rusty cinnamon stripes on the chest and abdomen. Dark rusty stripes forming complete or incomplete rings around the legs between elbow and wrist, and between knee and ankle; stripes on the forelegs are more intensely coloured than on the hind legs. Feet are entirely light coloured cinnamon. The tail has an ash-grey ground colour and displays dark grey rings, more evident in the distal portion. The skull has a well-developed sagittal crest, occupying the total length of the parietal suture. The notch of the postpalatine vein is broad and relatively shallow in most of the specimens and the posterior margin of the palate has a U-shaped edge. The zygomatic arch has its anterior part more robust and expanded laterally, giving its external side a more convex curvature when observed in the ventral or dorsal view. The mastoid processes are well developed in the posterior portion and separated from paraoccipital processes by a notch, covering the surface of the auditory bulla, which is large and oval. The ectotympanic is equivalent to 40–50% of entotympanic size. P2 is present. P3 has a long and narrow paracone and an absent parastyle. P4 paracone is present in most specimens.

Group II is distributed along and on both slopes of the Andes from Ecuador, Peru and western Bolivia to northern Chile and north-west Argentina (province of Catamarca). The following features define it: forehead, crown and nape brownish grey with speckled orange background colour; mandible, medial infranasal area, cheeks, areas around the lips and eyes, throat, chest, abdomen and ventral parts of legs white; two dark brown or reddish brown cheek lines; black, dark brown, yellowish brown or dark yellowish brown transversal gular stripes, with at least one of them markedly wider than others; ears triangular with distal half blackish and basal half greyish or pale brownish; spinal crest dark brownish grey with some orangish hairs; sides of body pale brownish grey (most specimens) or pale yellowish brown; well-marked rosettes with reddish brown border and orangish brown interior forming small oblique bands on the sides of body; chest and abdomen with dark brown to black rounded/oval markings or longitudinal stripes; dark brown to black stripes around the legs present between elbow and wrist, and between knee and ankles; feet entirely light coloured, similar or slightly lighter than the ground colour of sides of body; tail with reddish brown rings present from the base to the tip; sagittal crest poorly developed and restricted to interparietal region or moderately developed, occupying the posterior half of the suture between the parietal bones; notch of postpalatine vein broad and comparatively shallow or narrow and deep; posterior margin of the palate with U-shaped edge and it may or may not have a medial notch; anterior part of the zygomatic arches more robust and expanded laterally, giving their external side a more convex curvature when observed in the ventral or dorsal views; mastoid process poorly developed posteriorly and not covering the surface of the auditory bulla; medium- to large-sized ectotympanic, which varies from 25 to 50% of the entotympanic size; P2 absent; P3 paracone long and narrow or short and wide; P3 parastyle and P4 paracone may be absent or present.

Group III includes specimens from north-west Argentina (province of Catamarca) to the Strait of Magellan, on the east slope of the Andes. They are characterized by the forehead, crown and nape brownish grey or dark brownish grey colour; mandible, medial infranasal area, cheeks, areas around the lips and eyes white; two cheek lines that may be dark brown, dark greyish brown, yellowish brown or orangish brown; black, dark brown or yellowish brown transversal gular stripes, with a width that varies among specimens; ears triangular with distal half blackish and basal half greyish or pale brownish; spinal crest brownish grey or dark brownish grey; sides of body brownish grey with markings absent or, when present, with dark brown or dark yellowish brown indistinct oblique lines; throat, chest, abdomen and ventral parts of legs white or creamy white colour; dark yellowish brown to dark brown and black rounded/oval markings or longitudinal stripes present in chest and abdomen; dark brown to black stripes around the legs; feet entirely light coloured; tail brownish grey without rings and black tips; sagittal crest poorly developed and restricted to interparietal region or moderately developed, occupying posterior half of suture between parietal bones; notch of postpalatine vein broad and comparatively shallow or narrow and deep; posterior margin of the palate with or without medial notch; anterior part of the zygomatic arches, when observed in ventral or dorsal views, may be more robust and expanded laterally, giving their external side a more convex curvature, or little robust and not laterally expanded, giving their external surface a slightly convex or an almost rectilinear shape; mastoid process may be poorly developed posteriorly and not covering the surface of the auditory bulla, or well-developed posteriorly and covering the bulla; medium- to large-sized ectotympanic, which varies from 25 to 50% of entotympanic size; P2 absent in most specimens; P3 paracone long and narrow or short and wide; P3 parastyle and P4 paracone may be absent or present.

Group IV occurs in central Brazil and Paraguay and the specimens are characterized by the brown forehead, crown and nape; mandible and medial infranasal area white or pale yellowish brown; two dark brown cheek lines; yellowish brown cheeks; narrow dark brown or black gular stripes; ears triangular with distal half blackish and basal half greyish or pale brownish; spinal crest dark brown to black; sides of body brown; indistinct oblique brown lines, slightly darker than the sides of body; throat, chest, abdomen and ventral parts of legs light yellowish; dark brown to black rounded/oval or some stripe-like markings in chest and abdomen; dark brown to black stripes around the legs; feet entirely blackish; tail brown without rings and black tip present; sagittal crest poorly developed and restricted to interparietal region, or moderately developed, occupying posterior half of suture between parietals; notch of postpalatine vein broad and comparatively shallow or narrow and deep; posterior margin of the palate with or without medial notch; anterior part of the zygomatic arches, when observed in ventral or dorsal views, not laterally expanded, giving their external surface a slightly convex or an almost rectilinear shape; mastoid process poorly developed posteriorly and not covering the auditory bulla; small- to medium-sized ectotympanic, which varies from 20 to 35% of the entotympanic size; P2 present in most specimens; P3 paracone long and narrow or short and wide; P3 parastyle and P4 paracone may be absent or present.

Group V includes samples from southern Brazil (Rio Grande do Sul state), Uruguay and north-eastern Argentina. They are characterized by the yellowish grey forehead, crown and nape; mandible, cheeks and medial infranasal area white; two dark yellowish brown cheek lines; narrow gular stripes in dark yellowish brown colour; ears triangular with distal half blackish and basal half greyish or pale brownish; spinal crest dark yellowish grey; sides of body yellowish grey with distinct or indistinct dark yellowish grey oblique lines; throat, chest, abdomen and ventral parts of legs yellowish brown; dark brown to black longitudinal stripes in chest and abdomen; dark brown to black stripes around the legs; feet with dorsal surface light coloured and palmar/plantar surfaces blackish; tail with few discontinuous rings near the distal end and reduced black tip; sagittal crest poorly developed and restricted to interparietal region; notch of postpalatine vein broad and comparatively shallow or narrow and deep; posterior margin of the palate with or without medial notch; anterior part of the zygomatic arches when observed in ventral or dorsal view not laterally expanded giving their external surface a slightly convex or almost rectilinear shape, but some specimens may show a more convex curvature; mastoid process poorly developed posteriorly and not covering the auditory bulla; small- to medium-sized ectotympanic, which varies from 20 to 35% of entotympanic size; P2 present in most specimens; P3 paracone long and narrow or short and wide; P3 parastyle and P4 paracone may be absent or present.

Morphometric comparison among groups

The descriptive statistics of the skull and external morphology are available in Tables 2 and 3. All 16 cranial measurements show no sexual dimorphism and no interaction with groups (P > 0.05). Therefore, in the following analyses, we pooled males, females and specimens of undetermined sex. The first principal component (PC1) of the linear skull measurements explained 68.1% of overall variation, while the second principal component (PC2) explained 7.9% (Supporting Information, Table S1). The PC1 axis shows a considerable overlap among groups, whereas, on the PC2 axis, Groups IV and V are clearly differentiated from Groups I and II by their positive values (Fig. 4). The first discriminant function is responsible for 70.7% of the total variance, whereas the second and third functions account for 16.9% and 7.9% (Supporting Information, Table S2). Plots of the first, second and third discriminant functions depict a clear separation among groups (Fig. 4). Groups IV and V are clearly differentiated from all others along DF1. Specimens of the Group III are differentiated from Groups I and II with highly negative scores on DF2. Along the DF3 axis, Group I cluster apart from Groups II and III. By comparing the ratios of the cranial measurements, we found that IFB/CBL, ZB/CBL and GPL/CBL ratios are the most important to differentiate pampas cat groups (Fig. 5).

Climatic niche comparison

MANOVA revealed a marked climatic niche differentiation among our five groups of Le. colocola complex (Pillai’s Trace = 2.23, F_4,180 = 37.67, P < 0.000). Similarly, ANOVA showed significant differences (P < 0.000) among all five groups in each of the 22 climatic parameters (Supporting Information, Table S3). PCA plot depicts a clear separation among most of the groups in the climatic space, except Group I that overlaps with Group II. Individuals of Groups I and II occur in high-elevation areas, with pronounced diurnal temperature ranges and precipitation seasonality. Group III inhabits areas with a high annual temperature range and marked temperature seasonality, and Group V occurs in regions with strong precipitation. Animals of Group IV are affected by a set of multiple climatic parameters (Fig. 6).

The best ecological niche models (smallest AICc) show high AUC values, ranging from 0.904 for Group IV to 0.998 for Group I (Supporting Information, Table S4), indicating good performance. Overall, the predicted distribution for each group shows high congruence with the geographic range shown in Figure 3, with virtually no overlap between groups (Fig. 7). The main differences refer to the northern range limit of Group III, estimated at the border of Argentina, Bolivia and Paraguay, the eastward extension of Group IV into the central area of the Brazilian Atlantic Forest and the presence of Group II in the Colombian Andes. Taken together, niche analyses reveal that each group of pampas cats defined on morphological grounds exhibit marked climatic niche segregation and are associated with distinct ecological conditions.

Phylogeny and species delimitation

Phylogenetic and divergence time analyses

The monophyly of the Le. colocola complex is well supported. BI tree recovers five main clades with strong support. The first clade (clade A) includes two individuals from central Chile of the west slope of the Andes and represents our Group I (Fig. 8). The second clade (clade B) is mostly represented by sequences from central Argentina (our Group III), plus one sequence from central Chile and two sequences from southern Bolivia and northern Chile. Clade C presents animals from Peru and Bolivia attributed to Group II. These three clades together form a monophyletic lineage and separate from clade D and clade E, representing pampas cats from central Brazil (our Group IV) and from Uruguay and southern Brazil (our Group V), respectively (Fig. 8).

The divergence time indicated that the most recent common ancestor of the extant species of the Le. colocola complex diverged at approximately 0.54 Myr during the Middle Pleistocene [0.31–0.82 Myr, 95% highest posterior density (HPD)]. Clade C first diverged at approximately 0.43 Myr (0.25–0.65 Myr, 95% HPD) from clade A+B, which then diverged at approximately 0.37 Myr (0.22–0.56 Myr, 95% HPD). Clade D and E diverged at 0.26 Myr (0.11–0.50 Myr, 95% HPD).

Species delimitation

The ML and Bayesian solution PTPs identified clades A, C, D and E as ‘species’, while recognizing four ‘species’ within clade B (Fig. 8B). The smallest Kimura two-parameter genetic distance between analysed samples is 4.0% between clades D and E, whereas the highest distance is 10.5% between clades C and E (Table 4). The results based on the BI tree show higher P ID (Liberal) values (≥ 0.85) among clades. It is noteworthy that clades B and C showed high intra/inter values representing large intraspecific genetic variation (Table 5). The BPP and *BEAST species tree present congruent topology as the gene tree, recovering five species in pampas cats (Fig. 8C). *BEAST showed relatively higher support than BPP species tree for most of the nodes. Consonantly, five putative species were supported using BPP A11 algorithm with high PP values (> 0.95).

DISCUSSION

Taxonomic assessment

By using the broadest morphological coverage of Le. colocola complex to date, combined with multilocus phylogeny, species-delimitation techniques and ecological niche analyses, we show that treating pampas cats as a single species underestimates its actual diversity. When analysing the distribution and frequency of characters along the entire pampas cat range in order to detect clines and sharp discontinuities and to account for intra- and interpopulation variation – key factors for defining taxa and searching for diagnostic traits (Chiquito et al., 2014; Feijó & Cordeiro-Estrela, 2016; Nascimento & Feijó, 2017; Feijó et al., 2018) – we recognized five allopatric groups, each with clear diagnostic traits and well-defined geographic distribution. These phenetic units were then consistently recovered from each of the approaches applied here. Thus, given that the five groups are morphologically diagnosable (even in nearby areas), have evolved distinct niche spaces, occupy distinct geographic distributions mirroring major biogeographical units in the Neotropical region (Morrone, 2014), have moderate to high genetic distances and were recovered in species delimitation analyses, we hypothesize that they constitute valid species. The oldest available epithets for our groups are: Group I = colocolaMolina, 1782; Group II = garleppiMatschie, 1912; Group III = pajerosDesmarest, 1816; Group IV = braccataCope, 1889; and Group V = muñoaiXiménez, 1961.

Our polytypic classification with five living species within the Le. colocola complex largely accords with prior phylogenetic studies and solves some of the mismatches between previous morphology (e.g. García-Perea, 1994) and molecular datasets. Johnson et al. (1999) recovered a monophyletic clade in pampas cats with samples from Uruguay and southern Brazil that represents Le. munoai.Napolitano et al. (2008) found monophyletic groups from central Brazil (= Le. braccatus). Cossíos et al. (2009) recovered clades of pampas cats from Peru and Bolivia (= Le. garleppi) and from central and north Argentina (= Le. pajeros). Ruiz-García et al. (2013) pointed out that the northern Chilean population of pampas cats represents an extension of the Peruvian and the western Bolivian group, which compares with the distribution of Le. garleppi. Recently, Santos et al. (2018) also recovered five main clades in pampas cats, which is not surprising given that we used a similar molecular dataset. They reported limited dispersal and low gene flow among these five clades, reflecting their high genetic differentiation.

Despite the overall congruence among multiple methods, it is noteworthy to discuss some disagreements between the morphology and molecular datasets. Specimens from Ecuador, Peru, Bolivia and northern Chile are grouped into a single species (Le. garleppi) based on skin and cranial traits. While the two sequences available from northern Chile and southern Bolivia cluster with samples from Argentina (Le. pajeros). Additionally, one specimen from central Chile (Le. colocola) also clusters with Argentinian individuals (Fig. 8). Previous genetic studies have recognized that the central Andean region (i.e. northern Chile and Argentina and southern Bolivia) harbours a complex evolutionary history for pampas cats (Napolitano et al., 2008; Cossios et al., 2009; Santos et al., 2018), possibly representing the centre of diversification of the group (Ruiz-García et al., 2013). Pleistocene glaciations appear to have prompted events of population expansion and isolation, resulting in a mixed genetic signature there (Ruiz-García et al., 2013; Santos et al., 2018). For example, Napolitano et al. (2008) and Cossios et al. (2009) found samples from north Chile split into two clades. Posterior expansions of divergent lineages were also suggested to explain this admixture in the central Andes (Santos et al., 2018). Indeed, this region shows the highest phenotypic diversity for some of the characters analysed here (see Supporting Information, Appendix S2).

Unfortunately, most of the sequences available were not linked to a museum specimen, thereby we could not determine their skin pattern. All the 16 specimens examined by us from northern Chile and northern Argentina show the typical pattern of Le. garleppi shared by the Ecuadorian, Peruvian and Bolivian populations. Ruiz-García et al. (2013), who have assessed both skins and DNA sequences of pampas cat individuals, found that north Chilean populations have a similar skin pattern, mitochondrial haplotypes and microsatellites alleles as those found in animals from Peru and northern Bolivia. This finding supports our classification. Based on the morphological evidence, the contact zone between Le. garleppi and Le. pajeros lies at the Province of Catamarca, Argentina. Future efforts to generate new sequences from the central Andean region associated with museum vouchers are needed to clarify this conflict and to determine whether Le. pajeros extends northward up to southern Bolivia and northern Argentina (as predicted in the ecological niche model; Fig 7C), in sympatry with Le. garleppi.

The second mismatch refers to one individual from central Argentina nested in the central Brazilian clade. This can be explained by Pleistocene connections between the faunas of the Chaco and Cerrado (Werneck, 2011). Santos et al. (2018) found that the strongest migration events in pampas cats occurred between south-western (Argentina) and central (Brazil) populations, likely via the open vegetation corridor connecting the Chaco and savannas of the Cerrado.

Furthermore, we cannot rule out that these conflicts might be due to shorter gene sequences (< 567 bp) available from the complex central Andean region. Kumar et al. (2017) show that postspeciation gene flow between recently diverged taxa – as reported between Leopardus species (Trigo et al., 2008, 2013; Santos et al., 2018) – might lead to inaccurate individual gene phylogenies, even for well-recognized species with marked differences in morphology, habitat and ecology. It is thus clear that the available genetic data are insufficient to provide a clear picture of the evolutionary history and to conclusively reconstruct the relationships of pampas cats. Nevertheless, from a taxonomic viewpoint, the complementary lines of evidence employed here (discrete characters of skin and skull, cranial and body measurements, niche modelling, molecular species discovery and validation) converge on the recognition of five allopatric species in the Le. colocola complex. This finding shows that, even under gene flow, these five species maintain their identity (Kumar et al., 2017), which strengthens our classification.

Unlike previous taxonomic schemes, we do not recognize subspecies in the Le. colocola complex. Subspecies is a long-term controversial topic in taxonomy and, due to a myriad of concepts employed over the years, has misled its primary idea of representing geographical subdivisions, rather than well-defined allopatric taxa (Mayr, 1942; Willson & Brown, 1953; Mayr, 1977). According to Mayr (1942), subspecies are not ‘clear-cut units which can easily be separated from one another’ and their distinctness ‘can often be proven only by the careful comparison or biometric analysis of large series from various parts of the range of the species’. However, most of the subspecies of pampas cats were defined based on one or a few individuals without considering the intra- and interpopulation variation (see species account for more details). The five species here recognized show clear diagnosable phenotypes that are consistent along their distributions, even in proximate areas as in north-western Argentina, supporting their species rank (Feijó & Cordeiro-Estrela, 2016; Nascimento & Feijó, 2017; Feijó et al., 2018).

Speciation events and current geographic barriers

The recent diversification of pampas cats follows the general pattern in felids, where about half of modern species originated during the Pleistocene (Johnson et al., 2006). Pleistocene glaciations resulted in alternate cycles of expansion and retraction of open and forested biomes (Vuilleumier, 1971; Van der Hammen, 1974; Webb, 1978). Glacial periods are marked by the expansion of open areas and the spread of puna grasslands on both slopes of the Andes, allowing the open dwellers to migrate along the Andes, as well as to invade lowland areas via savanna-like corridors (Vuilleumier, 1971; Webb, 1978; Marshall, 1988). Interglacial times, on the other hand, represented the expansion of forests and the retraction of open biomes. It is thus reasonable to assume that the late Early and Middle Pleistocene glacial conditions in the tropical Andes played an important role in the evolutionary history of the pampas cat (Santos et al., 2018). The expansion of open areas might have allowed the ancestral pampas cats to increase their distribution both in a north–south axis along the Andes, as well as eastward into central Brazil via the diagonal open belt (Webb, 1978). In line with this explanation, Santos et al. (2018) showed evidence of colonization of pampas cats from western South America towards the central region of Brazil and a posterior colonization of the north-western part of its range, coinciding with two distinct glacial Pleistocene periods.

The modern distribution of the five species are well-defined by geographic barriers (Fig. 9). Pampas cats are typical of open areas, thereby the Amazonian and the Atlantic forests limiting their distribution. Leopardus garleppi is strongly associated with high elevations (Fig. 6) along the Andes, while Le. pajeros prefers the low-elevation areas of Argentina. Leopardus colocola is confined to the west coast of the Andes in central Chile, where its northern and southern limits coincide with the Atacama Desert and Valdivian Temperate forest (Fig. 9). Leopardus munoai occurs in grassland areas with marked precipitation in the dry season in southern Brazil, Uruguay and north-eastern Argentina (Figs 6, 9). Leopardus braccatus is associated with savanna-like habitats in central Brazil, eastern Paraguay and Bolivia. The seasonal dry forests of the Caatinga and Chaco seem to act as barriers to dispersal of pampas cats (Fig. 9).

Species accounts

In this section, we provide a list of synonyms, type locality, type material, diagnosis, detailed geographic distribution and remarks about morphological variation and taxonomic notes of each species. External and cranial measurements of the five species are in Tables 2 and 3. The list of specimens examined and localities are provided in the Supporting Information, Appendix S1.

Leopardus colocola (Molina, 1782) (Fig. 10)

Colocolo or Central Chilean pampas cat

Felis colocolaMolina, 1782: 295; type locality: ‘boschi del Chili,’ restricted to ‘Province of Valparaiso’ (Chile) by Osgood (1943: 79). The neotype was collected in Santiago, Chile.

feliscolocolla: M. Gruvel, 1789: 275 [French translation of Molina (1782)]; incorrect subsequent spelling of Felis colocola Molina.

Felis colorolla: Bechstein, 1800: 699; incorrect subsequent spelling of Felis colocola Molina.

Felis Corololo: Shaw, 1800: 369; incorrect subsequent spelling of Felis colocola Molina.

Felis colocolo: Anonymous [‘an American Gentleman’] 1808: 206 [English translation of Molina (1782)]; incorrect subsequent spelling of Felis colocola Molina.

F[elis]. Colocollo: Fischer, 1829: 204; incorrect subsequent spelling of Felis colocola Molina.

F[elis]. Colicollo: Wagner, 1841: 546; incorrect subsequent spelling of Felis colocola Molina.

Felis pajeros: Gay, 1847: lámina nº3; not Felis pajeros Desmarest.

Catus Pajeros: Fitzinger, 1861: 391; name combination; not Felis pajeros Desmarest.

Panthera Maracaya, albescens: Fitzinger, 1869: 22; name combination.

Felis passerum: Sclater, 1871: 700; part; name combination.

Oncifelis colocolo: J. A. Allen, 1919: 374; part; name combination.

Lynchailurus colocolus colocolus: Cabrera, 1940: 12; name combination; unjustified emendation of colocola Molina.

Lynchailurus pajeros huinaPocock, 1941: 261; type locality: ‘The range of mountains over Lake Catapilco, near Aconcagua, 900 m. alt.’.

Felis pajeros colocolo: Osgood, 1943: 79; name combination.

Felis [(Lynchailurus)] colocolo colocolo: Cabrera, 1958: 276; name combination.

F[elis]. colocola colocola: Ximénez, 1961:5; name combination.

Oncifelis (Lynchailurus) pajeros: Hemmer, 1978: 77; part; name combination.

Lynchailurus colocolo: García-Perea, 1994: 24; part.

L[ynchailurus]. c[olocolo]. colocolo: García-Perea, 1994: 31.

Leopardus colocolo: Wozencraft, 2005: 538; part; name combination.

[Leopardus colocolo] colocolo: Wozencraft, 2005: 538; name combination.

Leopardus colocola: Kitchener, Breitenmoser-Würsten, Eizirik, Gentry, Werdelin, Wilting, Yamaguchi, Abramov, Christiansen, Driscoll, Duckworth, Johnson, Luo, Meijaard, O’Donoghue, Sanderson, Seymour, Bruford, Groves, Hoffmann, Nowell, Timmons & Tobe, 2017: 51; part; name combination.

Leopardus colocola colocola: Kitchener, Breitenmoser-Würsten, Eizirik, Gentry, Werdelin, Wilting, Yamaguchi, Abramov, Christiansen, Driscoll, Duckworth, Johnson, Luo, Meijaard, O’Donoghue, Sanderson, Seymour, Bruford, Groves, Hoffmann, Nowell, Timmons & Tobe, 2017: 53; name combination.

Type material

Molina (1782) did not mention any specimen on which he based his Felis colocola. According to Osgood (1943: 81), ‘Molina’s descriptions were usually colored by hearsay, that such specimens as he may have seen were not in his hands at the time of writing’. Therefore, no original material is available, leaving the species without any name-bearing type specimen. The taxonomic history of Leopardus colocola is dynamic and debated (see: Wolffsohn, 1908; Cabrera, 1940; Osgood, 1943; and taxonomic notes below). Numerous interpretations of Molina’s Felis colocola have led to contrasting classifications and only a designation of a neotype can define this nominal taxon objectively. Thus, to clarify the taxonomic status of Leopardus colocola, we designate the specimen USNM391852 housed at the National Museum of Natural History, Smithsonian Institution, Washington DC, USA as the neotype of Felis colocolaMolina, 1782 (Fig. 10). The specimen consists of a skin of unknow sex and agrees with the current concept of the species.

Type locality

The neotype was collected in Santiago, Chile.

Diagnosis

Leopardus colocola is distinguished by ash grey background colour speckled with rusty cinnamon hairs on forehead, crown, nape, sides of body and tail. The spinal crest is dark grey with few cinnamon hairs. The sides of the body show a distinct rusty cinnamon colour with dark grey oblique lines. Two rusty cinnamon cheek stripes are present. The ear has a cinnamon colour with blackish tip and margins. The gular stripes are rusty cinnamon. Chest and abdomen exhibit irregular transversal rusty cinnamon stripes and legs present marked dark rusty rings. The tail is marked by dark grey rings. The skull has a well-developed sagittal crest, occupying the total length of the parietal suture.

Geographical distribution

Leopardus colocola is restricted to Central Chile on the western slope of the Andes (from sea level to 1800 m). It is found in Mediterranean forests, woodlands and shrublands of the Chilean Matorral. The geographical distribution is limited to the north by the Atacama Desert (apparently with a marginal distribution to the south of this region, in the coastal range), to the east by the Andes and to the south by the Valdivian temperate forests, where it seems to be replaced by the kodkod, Le. guigna (Fig. 9; Supporting Information, Appendix S3, Fig. S5).

Taxonomic notes

The use of colocola and its derivations, such as colocolo and colocolus, have been traditionally problematic and have generated numerous interpretations among different authors about the correct spelling (Cabrera, 1940, 1958, 1961; Osgood, 1943; Kitchener et al., 2017). Further details on this subject can be found in Kitchener et al. (2017).

Cabrera (1940), in his review of the identity of Molina’s Felis colocola, treats pampas cats as a polytypic species with six subspecies: Lynchailurus colocolus colocolus, Ly. c. pajeros, Ly. c. braccatus, Ly. c. crucinus, Ly. c. garleppi, and Ly. c. thomasi. Furthermore, Cabrera (1940) proposed the genus Oreailurus for Felis jacobitaCornalia, 1865, a taxon that has historically been confused with Le. colocola. Probably unaware of Cabrera’s article, Pocock (1941) and Schwangart (1941) (both authors were also probably unaware of each other’s articles) mistakenly recognized Felis jacobitaCornalia, 1865 as a synonym of Felis colocolaMolina, 1782, and classified the Andean mountain cat in different genera. Pocock (1941) classified the Andean mountain cat as Colocolo colocola, while Schwangart (1941) named it Montifelis colocola. In addition, the true Leopardus colocola of Molina from Central Chile was named Lynchailurus pajeros huina by Pocock, whereas Schwangart considered it as the same species of the Andean mountain cat. Notably, Pocock’s and Schwangart’s taxonomic arrangements were shared by the following authors: Burmeister, 1854; Philippi, 1869, 1870, 1873; Lydekker, 1896; J. A. Allen, 1919; Yepes, 1929.

García-Perea (1994) recognized two subspecies: the nominal from central Chile and a new taxon, Lynchailurus colocolo wolffsohni, from northern Chile (type locality: ‘Camarones, Tarapacá, Chile’). However, our results indicate that wolffsohni is a junior synonym of Le. garleppi (see the account of this species for more details).

Remarks

Wolffsohn (1908) commented that there are two morphs of background colour, one ash grey and the other orangish, but we did not observe the latter pattern among the examined specimens.

Leopardus pajeros (Desmarest, 1816) (Fig. 11)

Southern pampas cat.

Felis pajerosDesmarest, 1816: 114; based on ‘Le Pajeros de d’Azara’ (Azara, 1802: 160; 1801: 179); the neotype was collected near the city of Santa Rosa, department Capital, province of La Pampa, Argentina.

L[eo]. brunneusOken, 1816: 1070; part; unavailable name (ICZN, 1956: Opinion 417).

Felis PampaSchinz, 1821: 237; based on Azara (1801) and Oken (1816); type locality: ‘Provinz Buenos-Ayres’ (Province of Buenos Aires, Argentina).

Felis pageros: Lesson, 1827: 175; incorrect subsequent spelling of Felis pajeros Desmarest.

P[uma]. pajeros: Jardine, 1834: 267; name combination.

Lynx pageros:Boitard, 1843: 427; name combination and incorrect subsequent spelling of Felis pajeros Desmarest.

[Felis (Lynchailurus)] pajeros: Severtzov, 1858: 386; name combination.

Pajeros pampanus: Gray, 1867: 270; name combination.

Panthera Pajeros: Fitzinger, 1869: 66; name combination.

Felis passerum: Sclater, 1871: 700; part; name combination.

Felis pajero: Burmeister, 1879: 128; part; incorrect subsequent spelling of Felis pajeros Desmarest.

[Catus] pajeros: Matschie, 1895: 199; name combination; not Catus Pajeros Fitzinger.

[Felis (Felis)] pajeros: Trouessart, 1897: 364; name combination.

Felis pajeros crucinaThomas, 1901: 247; based on the specimen collected by Darwin in 1834 (BM 55.12.24.261) and described and figured as Felis pajeros by Waterhouse (1839: 18, plate IX); type locality: ‘Santa Cruz’ (Argentina).

Lynchailurus pajeros crucina: Allen, 1905: 183; name combination.

[Dendrailurus] pajeros: Pocock, 1917: 348; name combination.

Lynchailurus pajeros pajeros: Allen, 1919: 375; name combination.

Felis fasciatusLarrañaga, 1923: 345; based on ‘Le Chat Pajeros’ of Azara.

Lynchailurus pajeros: Cabrera, 1925: 93; name combination.

L[ynchailurus]. colocolus pajeros: Cabrera, 1940: 12; name combination; unjustified emendation of colocola Molina.

L[ynchailurus]. colocolus crucinus: Cabrera, 1940: 12; name combination; unjustified emendation of colocola Molina.

Lynchailurus pajeros pajeros: Pocock, 1941: 259; part (Felis pajeros crucina Thomas treated as a synonym); name combination.

Lynchailurus pajeros pajeros: Schwangart, 1941: 26; part (Felis pajeros crucina Thomas treated as a synonym); name combination.

[Lynchailurus (Lynch[ailurus].) pajeros braccatus] Phase A: Schwangart, 1941: 31; name combination.

Felis [(Lynchailurus)] colocolo pajeros: Cabrera, 1958: 277; part (Felis pajeros crucina Thomas treated as a synonym); name combination.

Felis (Lynchailurus) colocolo pajeros: Cabrera, 1961: 199; part (Felis pajeros crucina Thomas treated as a synonym); name combination.

F[elis]. colocola pajeros: Ximénez, 1961: 6; name combination.

Felis (Lynchailurus) colacola pajeros: Daciuk, 1974: 35; incorrect subsequent spelling of colocola of Molina.

Oncifelis (Lynchailurus) pajeros: Hemmer, 1978: 77; part; name combination.

Lynchailurus pajeros: García-Perea, 1994: 31; part; name combination.

L[ynchailurus]. p[ajeros]. crucinus: García-Perea, 1994: 32; name combination.

L[ynchailurus]. p[ajeros]. pajeros: García-Perea, 1994: 32; name combination.

Leopardus pajeros: Wozencraft, 2005: 538; part; name combination.

Leopardus colocola: Kitchener, Breitenmoser-Würsten, Eizirik, Gentry, Werdelin, Wilting, Yamaguchi, Abramov, Christiansen, Driscoll, Duckworth, Johnson, Luo, Meijaard, O’Donoghue, Sanderson, Seymour, Bruford, Groves, Hoffmann, Nowell, Timmons & Tobe, 2017: 51; part; name combination.

Leopardus colocola pajeros: Kitchener, Breitenmoser-Würsten, Eizirik, Gentry, Werdelin, Wilting, Yamaguchi, Abramov, Christiansen, Driscoll, Duckworth, Johnson, Luo, Meijaard, O’Donoghue, Sanderson, Seymour, Bruford, Groves, Hoffmann, Nowell, Timmons & Tobe, 2017: 53; part (Felis pajeros crucina Thomas treated as a synonym); name combination.

Type material

Desmarest (1816: 114) named ‘Le Pajeros de d’Azara’ as Felis pajeros. The Pajero (Azara, 1802: 160) or the Chat Pampa (Azara, 1801: 179) of Azara refer to a medium felid (86 cm) with small tail (27 cm), faint stripes along the body and pale hands and feet. Although few specimens from Azara’s expedition are known to have survived to the present day (Hershkovitz, 1987; Voss et al., 2009), we have failed to find any existing felid among them, and assume that none exist, leaving the species without any original material. Because the long and debated taxonomy of L. pajeros, with conflicting classifications proposed over the years, including seven subspecies (see taxonomic notes below), we find it justified to designate a neotype to fix the application of Felis pajerosDesmarest, 1816. The neotype is an adult female deposited in the Museo Argentino de Ciencias Naturales ‘Bernardino Rivadavia’, Buenos Aires, Argentina (MACN25843), consisting of a skin, skull and postcranial skeleton (Fig. 11). The specimen was collected by Javier Pereira on 9 April 2013 and agrees with Azara’s (1801, 1802) external description of Le. pajeros. The cranial measurements of the neotype are: GLS 96.64; CBL 83.40; RL 33.53; ZB 66.67; GBB 46.30; IFB 28.14; GPB 40.54; GPL 36.53; CM1L 29.32; ALT 58.52; p3m1L 22.56; MH 26.79; ML 60.87; postorbital breadth 28.72; interorbital length 19.91; greatest length of P4 12.44; temporal fossa height 29.44.

Type locality

The neotype was collected near the city of Santa Rosa, department Capital, province of La Pampa, Argentina. This locality is part of the original distribution of the species as reported by Azara, who stated that it occurs between 35° and 36° S latitude in the pampas south of Buenos Aires (Azara, 1801: 179).

Diagnosis

Forehead, crown, nape and spinal crest are brownish grey or dark brownish grey colour; black, dark brown, yellowish brown or dark yellowish brown gular stripes, with at least one of them markedly wider than others; sides of body brownish grey with markings absent or, when present, with dark brown or dark yellowish brown indistinct oblique lines; tail brownish grey without rings and black tips.

Geographical distribution

Confirmed records range from north-western Argentina (Province of Catamarca) to the Strait of Magellan, Chile (Fig. 9; Supporting Information, Appendix S3, Fig. S6). Its presence in the northern portion of Argentina pending confirmation and may represent a contact zone with Le. garleppi.

Taxonomic notes

The taxonomic history of the southern pampas cat began with the description of Felis pajeros by Desmarest in 1816, which was based on ‘Le Chat Pampa’ of Azara (1801). In the same year, Oken described his ‘Leo brunneus’, which part of it was also based on the work of Azara. However, this name proposed by Oken is not Linnean and, therefore, unavailable for Zoological Nomenclature (see: ICZN, 1956; Opinion 417). For a long time, the name pajeros was used for the pampas cats, mainly due to uncertainties about the true identity of colocola and whether this name would be valid or not. Since the name colocola was well established to designate the pampas cats, the name pajeros was relegated to subspecies level (e.g. Cabrera, 1958, 1961; Ximénez, 1961, 1970).

García-Perea (1994) recognized Le. pajeros as a polytypic species with seven subspecies (Lynchailurus p. pajeros, Ly. p. budini, Ly. p. crespoi, Ly. p. crucinus, Ly. p. garleppi, Ly. p. steinbachi and Ly. p. thomasi), which apparently represented a clinal variation. However, we did not detect any clinal variation along the populations distributed from Ecuador to the south of Argentina. In fact, we found a clear discontinuity in the pelage patterns in the Catamarca region, north-western Argentina, where two patterns are present: one represented by Le. garleppi (e.g. MACN50.446, La Atravesada, Andalgalá, Catamarca) and other by Le. pajeros (MACN37.39 and 37.40, both from Catamarca, without precise locality). The first pattern occurs from Catamarca northwards to Ecuador, while the second is distributed southwards to the Strait of Magellan. Thus, we treat budini, crespoi, steinbachi, thomasi and wolffsohni as junior synonyms of Le. garleppi (see the taxonomic notes of that species). We also found no evidence that crucinaThomas, 1901 and pajerosDesmarest, 1816 are different taxa, thus we treat the former as a junior synonym of the latter, following Schwangart (1941) and Kitchener et al. (2017).

Remarks

In adults and subadults the markings on the side of the body may be faint or evident; in the latter case, they form indistinct oblique lines that may be dark brown or dark yellowish brown. In young individuals, conspicuous dark markings arranged in the scapuloinguinal direction along the flanks are visible. We did not find melanistic specimens. One specimen (MACN 1.23) from the central Pampa (precise locality unknown) shows a distinct coloration, the background colour of the throat, chest, venter, limbs and fore and hind feet were ochraceous, while the markings in these areas were cinnamon (spots and stripes in the throat, chest and venter) and dark brown (the stripes in limbs). Cabrera (1961), who also examined this specimen, commented that it may be a case of partial erythrism.

Leopardus braccatus (Cope, 1889) (Fig. 12)

Pantanal cat or Brazilian pampas cat.

Felis braccataCope, 1889: 146; type locality: ‘province of Rio Grande do Sul, or Matto Grosso’, restricted to ‘Chapada dos Guimarães’ (State of Mato Grosso, Brazil) by J. A. Allen (1919).

[Felis (Catopuma)] braccata: Trouessart, 1897: 364; name combination.

Lynchailurus pajeros braccatus: Allen, 1919; name combination.

[Felis (]Mungofelis[) braccata]: Antonius, 1933: 13; name combination.

L[ynchailurus]. colocolus braccatus: Cabrera, 1940: 12; name combination; unjustified emendation of colocola of Molina.

Lynchailurus (Lynch[ailurus].) pajeros braccatus: Schwangart, 1941: 29; part; name combination.

[Lynchailurus (Lynch[ailurus].) pajeros braccatus] Phase C: Schwangart, 1941: 32; name combination.

Felis [(Lynchailurus)] colocolo braccata: Cabrera, 1958: 275; name combination.

F[elis]. colocola braccata: Ximénez, 1961: 6; name combination.

Oncifelis (Lynchailurus) pajeros: Hemmer, 1978: 77; part; name combination.

Lynchailurus braccatus: García-Perea, 1994: 25; part.

L[ynchailurus]. b[raccatus]. braccatus: García-Perea, 1994: 25; name combination.

Leopardus braccatus: Wozencraft, 2005: 537; part; name combination.

Leopardus colocola: Kitchener, Breitenmoser-Würsten, Eizirik, Gentry, Werdelin, Wilting, Yamaguchi, Abramov, Christiansen, Driscoll, Duckworth, Johnson, Luo, Meijaard, O’Donoghue, Sanderson, Seymour, Bruford, Groves, Hoffmann, Nowell, Timmons & Tobe, 2017: 51; part; name combination.

Leopardus colocola braccatus: Kitchener, Breitenmoser-Würsten, Eizirik, Gentry, Werdelin, Wilting, Yamaguchi, Abramov, Christiansen, Driscoll, Duckworth, Johnson, Luo, Meijaard, O’Donoghue, Sanderson, Seymour, Bruford, Groves, Hoffmann, Nowell, Timmons & Tobe, 2017: 53; part; name combination.

Type material

AMNH354, skull and skin of an adult male collected by Herbert H. Smith in November 1884.

Type locality

‘province of Rio Grande do Sul, or Matto Grosso’ (Brazil) (Cope 1889: 146). Later restricted as ‘Chapada dos Guimarães’ by J. A. Allen (1919).

Diagnosis

Leopardus braccatus is distinguished by a brown background colour of the forehead, crown and nape; mandible and medial infranasal area white or pale yellowish brown; spinal crest dark brown to black. The sides of the body have a brown colour with indistinct oblique slightly darker brown lines. Feet are entirely blackish. The tail is uniform brownish without rings and with a black tip. P2 is present in most specimens.

Geographical distribution

The species occurs in open habitats of central Brazil (from Pantanal to south-western Piauí in north-eastern Brazil), Paraguay, Bolivian lowlands (records from the Department of Beni) and northern Argentina (record from the Province of Formosa) (García-Perea, 1994; Silveira, 1995; Bagno et al., 2004; Chebez et al., 2008; Luque et al., 2012; Nascimento et al., 2016) (Fig. 9; Supporting Information, Appendix S3, Fig. S7).

Taxonomic notes

Leopardus braccatus has been regarded either as a subspecies of Le. colocola (Cabrera, 1940, 1958; Ximénez, 1961; Ximénez, 1970; Pecon-Slattery et al., 1994, 2000, 2004; Masuda et al., 1996; Johnson & O’Brien, 1997; Eizirik et al., 1998; Pecon-Slattery & O’Brien, 1998; Johnson et al., 1999; Napolitano et al., 2008; Cossíos et al., 2009; Sunquist & Sunquist, 2009; Kitchener et al., 2017) or of L. pajeros (Allen, 1919; Pocock, 1941; Schwangart, 1941). García-Perea (1994) recognized it as a polytypic species with two subspecies, Ly. b. braccatus from central Brazil and Paraguay and Ly. b. munoai from southern Brazil and Uruguay. Our study shows that these two subspecies are warranted recognition at the species level. Schwangart (1941) recognized three phases for his Lynchailurus (Lynchailurus) pajeros braccata, which distinguished themselves from each other mainly by the pattern of darkening of the feet: Phase A, ‘without blackening of the feet’, Phase B, ‘intermediate stage for the blackening of the feet’ and Phase C, ‘perfect blackening of the feet’. However, only Schwangart’s Phase C actually represents Le. braccatus, whereas the two former correspond, respectively, to Le. pajeros and Le. munoai.

Remarks

Despite young individuals having the overall brown body colour and entirely black feet similar to adults, they show conspicuous dark markings arranged in the scapuloinguinal direction along the flanks, which fades in adults. This pattern was observed by us in the specimen MNRJ4868 from Maracaju, Mato Grosso do Sul, Brazil (Supporting Information, Appendix S4) and by Silveira (1995), who observed conspicuous irregular black bands over the body of a 75 days old individual, that disappeared completely when it reached 8 months old. Beviglieri et al. (2018) provided photos of a live late juvenile, in which it is possible to observe some visible oblique bands on the side of the animal. Melanistic specimens of Le. braccatus have been reported in the wild (Silveira et al., 2005; Parque Nacional das Emas, Goiás, Brazil), in captivity (Monticelli & Marques, 2015) and we also observed one in a museum collection (MZUSP7670, from Três Lagoas, Mato Grosso do Sul, Brazil).

Leopardus garleppi (Matschie, 1912), comb. nov. (Fig. 13)

Garlepp’s pampas cat or northern pampas cat.

Felis pajeros: Matschie, 1894: 60; part; not Felis pajeros Desmarest.

Felis (Lynchailurus) pajeros garleppiMatschie, 1912: 259; type locality: ‘von Cuzco in Südost-Peru, im Gebiet des Apurimac, der durch den Ucayali zum oberen Amazonas abwässert’.

Felispajeros thomasiLönnberg, 1913: 7; type locality: ‘near Quito, Ecuador’.

Felis pajeros garleppi: Lönnberg, 1913: 8; name combination.

Lynchailurus pajeros thomasi: Allen, 1919: 376; name combination.

Lynchailurus pajeros garleppi: Allen, 1919: 376; name combination.

Felis (Lynchailurus) colocolo garleppi:Pearson, 1951: 130, 137; name combination.

Lynchaylurus pajeros pajeros: Yepes, 1936: 36; incorrect spelling; not Felis pajeros Desmarest.

L[ynchailurus]. colocolus garleppi: Cabrera, 1940: 12; name combination; unjustified emendation of colocola Molina.

L[ynchailurus]. colocolus thomasi: Cabrera, 1940: 12; name combination; unjustified emendation of colocola Molina.

Lynchailurus pajeros budiniPocock, 1941: 263; type locality: ‘Mount Sola, 2500 m., in Salta, northern Argentine’.

Lynchailurus pajeros steinbachiPocock, 1941: 264; type locality: ‘Tiraque, Cochabamba, western Bolivia, 4000 m. alt.’.

Lynchailurus pajeros garleppi: Pocock, 1941: 266; part (Felis pajeros thomasi Lönnberg treated as a synonym).

Lynchailurus (Lynchailurus) garleppi: Schwangart, 1941: 33; part (Felis pajeros thomasi Lönnberg treated as a synonym); name combination.

Felis (Lynchailurus) colocolo crespoiCabrera, 1957: 71; type locality: ‘Aguaray, província de Salta [Argentina]’.

Felis [(Lynchailurus)] colocolo budini: Cabrera, 1958: 276; name combination.

Felis [(Lynchailurus)] colocolo crespoi: Cabrera, 1958: 277; name combination.

Felis [(Lynchailurus)] colocolo garleppi: Cabrera, 1958: 277; part (Lynchailurus pajeros steinbachi Pocock treated as a synonym); name combination.

Felis [(Lynchailurus)] colocolo thomasi: Cabrera, 1958: 278; name combination.

F[elis]. colocola budini: Ximénez, 1961: 6; name combination.

F[elis]. colocola crespoi: Ximénez, 1961: 6; name combination.

F[elis]. colocola garleppi: Ximénez, 1961: 6; name combination.

F[elis]. colocola thomasi: Ximénez, 1961: 6; name combination.

Lynchailurus colocolo wolffsohniGarcía-Perea, 1994: 31; type locality: ‘río Camarones, provincia Tarapacá, between 2000 and 4000 m, Chile’.

Lynchailurus pajeros: García-Perea, 1994: 31; part; name combination.

L[ynchailurus]. p[ajeros]. budini: García-Perea, 1994: 31; name combination.

L[ynchailurus]. p[ajeros]. crespoi: García-Perea, 1994: 32; name combination.

L[ynchailurus]. p[ajeros]. garleppi: García-Perea, 1994: 32; name combination.

L[ynchailurus]. p[ajeros]. steinbachi: García-Perea, 1994: 32; name combination.

L[ynchailurus]. p[ajeros]. thomasi: García-Perea, 1994: 32; name combination.

[Leopardus colocolo] wolffsohni: Wozencraft, 2005: 538; name combination.

Leopardus pajeros: Wozencraft, 2005: 538; part; name combination.

[Leopardus pajeros] budini: Wozencraft, 2005: 539; name combination.

[Leopardus pajeros] garleppi: Wozencraft, 2005: 539; name combination.

[Leopardus pajeros] steinbachi: Wozencraft, 2005: 539; name combination.

[Leopardus pajeros] thomasi: Wozencraft, 2005: 539; name combination.

Leopardus colocola: Kitchener, Breitenmoser-Würsten, Eizirik, Gentry, Werdelin, Wilting, Yamaguchi, Abramov, Christiansen, Driscoll, Duckworth, Johnson, Luo, Meijaard, O’Donoghue, Sanderson, Seymour, Bruford, Groves, Hoffmann, Nowell, Timmons & Tobe, 2017: 51; part; name combination.

Leopardus colocola wolffsohni: Kitchener, Breitenmoser-Würsten, Eizirik, Gentry, Werdelin, Wilting, Yamaguchi, Abramov, Christiansen, Driscoll, Duckworth, Johnson, Luo, Meijaard, O’Donoghue, Sanderson, Seymour, Bruford, Groves, Hoffmann, Nowell, Timmons & Tobe, 2017: 53; part; name combination.

Leoparduscolocola budini: Kitchener, Breitenmoser-Würsten, Eizirik, Gentry, Werdelin, Wilting, Yamaguchi, Abramov, Christiansen, Driscoll, Duckworth, Johnson, Luo, Meijaard, O’Donoghue, Sanderson, Seymour, Bruford, Groves, Hoffmann, Nowell, Timmons & Tobe, 2017: 53; part (Lynchailurus pajeros steinbachi Pocock treated as a synonym); name combination.

Leopardus colocola garleppi: Kitchener, Breitenmoser-Würsten, Eizirik, Gentry, Werdelin, Wilting, Yamaguchi, Abramov, Christiansen, Driscoll, Duckworth, Johnson, Luo, Meijaard, O’Donoghue, Sanderson, Seymour, Bruford, Groves, Hoffmann, Nowell, Timmons & Tobe, 2017: 53; part (Felis (Lynchailurus) pajeros garleppi Matschie treated as a synonym); name combination.

Type locality

Cuzco, Peru: ‘von Cuzco in Südost-Peru, im Gebiet des Apurimac, der durch den Ucayali zum oberen Amazonas abwässert’ [‘from Cuzco in south-eastern Peru, in the region of Apurímac, which drains through the Ucayali to the upper Amazon’] (Matschie, 1912: 259).

Type material

ZMB_MAM_21244, skin and skull of unknown sex adult specimen collected by Otto Garlepp.

Diagnosis

Leopardus garleppi is distinguished by the brownish grey forehead, crown and nape speckled with orange; black, dark brown, yellowish brown or dark yellowish brown transversal gular stripes, with at least one of them markedly wider than others; spinal crest dark brownish grey with some orangish hairs; background colour on the sides of the body pale brownish grey in most specimens or pale yellowish brown; well-marked rosettes with reddish brown border and orangish brown interior forming small oblique bands on the sides of body; tail with reddish brown rings present from the base to tip.

Geographical distribution

Along and on both slopes of the Andes from Ecuador to north-western Argentina (Provinces of Catamarca and Córdoba) and northern Chile (Tarapacá Region) (Fig. 9; Supporting Information, Appendix S3, Fig. S8). Ruíz-García et al. (2013) suggested that this species possibly occurs marginally in Colombia, based on a specimen collected in the Parque Natural Nacional Volcano Galeras, Department of Nariño, south-western Colombia, in August 1989 (IAVH5857). However, the specimen identified by them as a pampas cat is, in fact, a northern tigrina (Le. tigrinus) (see: Nascimento & Feijó, 2017). Nevertheless, it is likely that Le. garleppi also occurs in the Andean region of Colombia, as predicted in our niche model (Fig. 7B). Future studies in this areas may confirm this hypothesis.

Taxonomic notes

Depending on the author, the Garlepp’s pampas cat was regarded as either a subspecies of L. colocola (Cabrera, 1940, 1958; Pearson, 1951; Ximénez, 1961; Pecon-Slattery et al., 1994, 2000, 2004; Masuda et al., 1996; Johnson & O’Brien, 1997; Eizirik et al., 1998; Pecon-Slattery & O’Brien, 1998; Johnson et al., 1999; Cossios et al., 2007; Napolitano et al., 2008; Cossíos et al., 2009; Sunquist & Sunquist, 2009; Li et al., 2016; Kitchener et al., 2017) or of L. pajeros (Matschie, 1912; Lönnberg, 1913; Allen, 1919; Pocock, 1941; García-Perea, 1994; Wozencraft, 2005).

Mastchie (1912) described a specimen from Peru as Felis (Lynchailurus) pajeros garleppi and, a year later, Lönnberg (1913) described in detail another individual from Ecuador with similar pelage characters and named it Felis pajeros thomasi. Lönnberg also examined the type of garleppi and albeit indicated a close relationship between both taxa, he pointed out some characters that would be sufficient to differentiate these two species. However, Pocock (1941), examining the material from Ecuador and Peru available at the Museum of Natural History in London, did not detect significant differences between the populations of these countries that would justify the recognition of two taxa and, therefore, treated thomasi as a junior synonym of garleppi. In addition, Pocock described two new subspecies, Lynchailurus pajeros budini (Salta, NW Argentina) and Ly. p. steinbachi (W Bolivia), which showed characters similar to garleppi. In the same year, Schwangart (1941) had independently reached the same conclusion in treating garleppi and thomasi as synonyms.

Cabrera (1957) described a new subspecies from the province of Salta, north-western Argentina, named by him as Felis colocolo crespoi, which shared similarities with thomasi in the pattern of coat markings, but they differed from each other in background colour, and crespoi mainly differed from budini and pajeros by longer fur during the summer. Cabrera (1958), in his taxonomic catalogue of South American mammals, recognized garleppi and thomasi as distinct subspecies and considered steinbachi as the junior synonym of the former, and recognized crespoi and budini as valid subspecies. Later, Cabrera (1961) commented that budini would be an intermediate geographic form between garleppi from southern Peru and south-western Bolivia and colocola from Chile, resembling the former more than the latter, and the difference between budini and garleppi did not seem so evident and it is possible that a comparative study with good sampling of specimens could prove budini to be a junior synonym of garleppi.

Decades later, García-Perea (1994) recognized garleppi, thomasi, steinbachi, crespoi and budini as subspecies of L. pajeros. She examined the populations from Ecuador (thomasi) (N = 2) and Peru (garleppi) (N = 7) and realized that the skull and pelage characters, as well as habitat preferences (high-elevation steppes), are similar, but they only differ in body size, with garleppi being larger than thomasi. García-Perea also affirmed that budini and crespoi showed the same colour pattern, but they differ from each other in the colour shades of the markings, and they inhabit a transitional zone between the higher steppes and dry forests. Based on the similarities in the pelage characters and habitat preferences, García-Perea suggested that crespoi should be a synonym of budini, but due to the small sample size and the poor condition of the type specimen (confirmed by one of us, FON, who examined the specimen), she provisionally treated them as distinct subspecies.

Moreover, García-Perea (1994) recognized the populations in northern Chile as a new subspecies and named these Lynchailurus colocolo wolffsohni. According to García-Perea’s description, the skull characters are close to those found in Ly. c. colocolo, but her Ly. c. wolffsohni differed from the nominal subspecies mainly by the pelage pattern, which in turn was similar to the pattern of the Peruvian and Bolivian populations. However, our results indicate that wolffsohni is a junior synonym of Le. garleppi, in agreement with previous authors. Osgood (1943) noted that the northern Chilean specimen could not ‘be identified satisfactorily as to subspecies, but perhaps will prove to be nearer to garleppi than colocolo’. Furthermore, Osgood pointed out that the position of the central Chilean populations (= colocola) in relation to the northern ‘varieties’ (garleppi, thomasi, budini and steinbachi) was uncertain. In addition, Pine et al. (1979), comparing specimens of colocola from Central Chile and garleppi from Peru with the specimen from Tarapacá, north Chile, concluded ‘that the form in northern Chile is F. c. garleppi’. Consonantly, molecular data suggested that samples from northern Chile, Bolivia and Peru form a well-supported group (Johnson, 1999; Napolitano et al., 2008; Cossíos et al., 2009; Ruiz et al., 2013; Santos et al., 2018). Therefore, contradicting García-Perea (1994), the pampas cats from the western Andean slope of Peru and northern Chile belong to the same taxon.

Based on our results and on previous studies, we conclude that all forms – budini, crespoi, garleppi, steinbachi, thomasi and wolffsohni – are synonyms and should be united in a single species, Le. garleppi. The colour and spot patterns are similar and geographically consistent. The variation in colour shades, fur length and the skull dimensions, pointed out by previous authors are, in fact, of an individual nature.

Remarks

Most of the specimens showed the overall ground colour of body pale brownish grey, but some were pale yellowish brown. We did not find melanistic specimens in museum specimens, but Giordano et al. (2012) reported a melanistic individual in south-west of Parque Nacional Yanachaga-Chemillén, Pusapno, Peru (10°43’26”S, 75°27’43”W).

Leopardus munoai (Ximénez, 1961), comb. nov. (Fig. 14)

Muñoa’s or Uruguayan pampas cat.

Felis pajero: Burmeister, 1879: 128; part; incorrect subsequent spelling of Felis pajeros Desmarest.

Felis pajeros: Arechavaleta, 1882: 43; not Felis pajeros Desmarest.

Felis passerum: Aplin, 1894: 298; not Felis passerum Sclater.

Lynchailurus (Lynch[ailurus].) pajeros braccatus: Schwangart, 1941: 29; part; name combination.

[Lynchailurus (Lynch[ailurus].) pajeros braccatus] Phase B: Schwangart, 1941: 32; name combination.

Felis colocola muñoaiXiménez, 1961: 3; type locality: ‘Arroyo Perdido, Departamento de Soriano’ (Uruguay).

Felis colocola munoai: Ximénez, 1970: 1; corrected spelling of subspecific name (ICZN, 1999: Art. 32.5.2).

Felis (Lynchailurus) colocola munoai: Ximénez, Langguth & Praderi, 1972: 17; name combination.

Lynchailurus braccatus: García-Perea, 1994: 25; part.

L[ynchailurus]. braccatus munoai: García-Perea, 1994: 32; name combination.

Leopardus braccatus: Wozencraft, 2005: 537; part; name combination.

[Leopardus braccatus] munoai: Wozencraft, 2005: 538; name combination.

L[eopardus]. b[raccatus]. fasciatus: González & Martínez-Lanfranco, 2010: 188; name combination; not Felis fasciatus Larrañaga.

Leopardus colocola: Kitchener, Breitenmoser-Würsten, Eizirik, Gentry, Werdelin, Wilting, Yamaguchi, Abramov, Christiansen, Driscoll, Duckworth, Johnson, Luo, Meijaard, O’Donoghue, Sanderson, Seymour, Bruford, Groves, Hoffmann, Nowell, Timmons & Tobe, 2017: 51; part; name combination.

Leopardus colocola munoai: Kitchener, Breitenmoser-Würsten, Eizirik, Gentry, Werdelin, Wilting, Yamaguchi, Abramov, Christiansen, Driscoll, Duckworth, Johnson, Luo, Meijaard, O’Donoghue, Sanderson, Seymour, Bruford, Groves, Hoffmann, Nowell, Timmons & Tobe, 2017: 53; name combination.

Type locality

‘Arroyo Perdido, Departamento de Soriano’ (Uruguay) (Ximénez, 1961: 5).

Type material

Holotype: MNHN-M884, skin and skull of an adult female collected by Florencio Lezica in the type locality on 4 August 1959; skull damaged. Paratypes: MNHN-M875, a skin of unknown gender collected in May, 1959 by Juan Escoto in Estancia Juan Escoto, Tarariras, Departamento de Cerro Largo, Uruguay; MNHN-M879, a skin and skull of a juvenile male, collected in July, 1959 by Orosmán Constantino in Chamizo, Departamento de San José, Uruguay; MNHN-M971, skin and skull of an adult male collected in February, 1960 by Juan Fernandez in Estancia Bella Vista, Zapicán, Departamento de Lavalleja, Uruguay.

Diagnosis

Leopardus munoai is distinguished by the forehead, crown and nape having a yellowish grey colour; two dark yellowish dark brown cheek lines; narrow gular stripes in dark yellowish brown colour; spinal crest dark yellowish grey; sides of body yellowish grey with distinct or indistinct dark yellowish grey oblique lines; dark brown to black longitudinal stripes in chest and abdomen; feet with dorsal surface light coloured and palmar/plantar surfaces blackish; tail with few discontinuous rings near the distal end and reduced black tip; sagittal crest poorly developed and restricted to interparietal region; P2 is present in most specimens.

Geographical distribution

Open areas of southern Brazil (the southern portion of the Rio Grande do Sul state), Uruguay and north-eastern Argentina (Province of Corrientes) (Ximénez, 1961; García-Perea, 1994; Queirolo, 2009, 2016) (Fig. 9; Supporting Information, Appendix S3, Fig. S9). Leopardus munoai is probably separated geographically from Le. braccatus and Le. pajeros by the Paraná River in the west and from Le. braccatus by forested regions in southern and south-western Brazil in the north (Fig. 9).

Taxonomic notes

Leopardus munoai has been regarded for a long time either as a subspecies of Le. colocola (Ximénez, 1961; Ximénez, 1970; Pecon-Slattery et al., 1994, 2000, 2004; Masuda et al., 1996; Johnson & O’Brien, 1997; Eizirik et al., 1998; Pecon-Slattery & O’Brien, 1998; Johnson et al., 1999; Napolitano et al., 2008; Cossíos et al., 2009; Sunquist & Sunquist, 2009; Kitchener et al., 2017) or of Le. braccatus (García-Perea, 1994; Wozencraft, 2005; Chebez et al., 2008; Barstow & Leslie Jr., 2012). García-Perea (1994) treated it as a subspecies of Lynchailurus braccatus based on morphological similarities of the skull, on the existence of variation in the pelage pattern and body size, and on the distribution ranges, which are separated from each other by a forested region.

González & Martínez-Lanfranco (2010) commented that the subspecies present in Uruguay is called Le. b. fasciatus (Larrañaga, 1923), without justifying the use of this name in place of munoai, ignoring the discussion of Ximénez et al. (1970). The latter authors have argued that Felis fasciatusLarrañaga, 1923 is a junior synonym of Le. pajeros (Desmarest, 1816), since the description and the body measures coincide with those of ‘Le Pajeros’ of Azara (1801, 1802), which was the basis of the description of Desmarest’s Felis pajeros.

Remarks

Schwangart (1941) commented on two skins that represent an ‘intermediate stage for the blackening of the feet’ [his Lynchailurus (Lynchailurus) pajeros braccatus Phase B], which is a characteristic present in Le. munoai. These two skins were deposited in the Munich Museum at the time Schwangart studied them (before 1941). One of them (‘Barbieux 1931, No. 81’) has an unknown locality, whereas the other (‘H. Krieg 1931/32’) was purchased in Asuncion, Paraguay during the Hans Krieg Expedition. Regarding the locality of this last skin, care must be taken, because there is no evidence that animals with this feature are present in Paraguay. We only found individuals with completely blackened feet (Le. braccatus) in this country. Possibly the animal may have been collected on the other bank (right) of the Paraná River (NE Argentina), where the partially dark feet specimens (i.e. Le. munoai pattern) are found, and then it was transported to Asuncion. In addition, Burmeister (1879: 129) described a specimen collected in the province of Entre Rios, north-eastern Argentina, that shows the sole of the foot almost black.

CONCLUSIONS

In this study, we propose a new taxonomic classification for the pampas cats. Based on multiple lines of evidence derived from morphology, molecular phylogeny, biogeography and climatic niche datasets, we recognize five living species in the Le. colocola complex: Leopardus colocola (Molina, 1782), Le. pajeros (Desmarest, 1816), Le. braccatus (Cope, 1889), Le. garleppi (Matschie, 1912) and Le. munoai (Ximénez, 1961). This classification solves some of the mismatches between prior morphology-based and molecular-based studies. We further show that previous subspecies classifications were largely defined without considering interpopulation variation. Our study shows that pelage variation or skull traits used to diagnose subspecies are more of an individual nature. Future efforts to generate new sequences from the central Andean region is needed to clarify the evolutionary history and to conclusively reconstruct the phylogenetic relationships of pampas cats from this complex area.

SUPPORTING INFORMATION

Additional Supporting Information may be found in the online version of this article at the publisher's web-site.

Appendix S1. List of all specimens examined including their museum collection numbers and records obtained from literature with their localities. Geographic data on each taxa were obtained directly from the labels of the specimens when available, from published gazetteers (Paynter Jr., 1985, 1988, 1989, 1992, 1993, 1994; Stephens & Traylor Jr., 1983; Paynter Jr. & Traylor Jr., 1991; Vanzolini & Traylor Jr., 1992) or online databases (Global Gazetteer 2.3, www.fallingrain.com/world/index.html; GeoNames, www.geonames.org). When the exact locality was not available or not found in the methodologies employed, so we used the coordinates of the nearest county.

Appendix S2. Description of states of each character analysed in this study. External and cranial qualitative characters used in this study to assess the morphological variation in Leopardus colocola complex. Characters were selected based on our direct observation of specimens in museums and on their use as diagnostic traits in previous taxonomic works. The distribution of each state of characters is shown in Figures S1–S4.

Appendix S3. Maps of geographic distribution of the species and their respective gazetteers.

Appendix S4. Figure of juvenile specimen of Leopardus braccatus.

Figure S1. Distribution of the states of external characters (1–8).

Figure S2. Distribution of the states of external characters (9–17).

Figure S3. Distribution of the states of cranial characters (1–6)

Figure S4. Distribution of the states of cranial characters (7–9).

Figure S5. Geographic distribution of Leopardus colocola. Red dots correspond to specimens housed in scientific collections and black dots refer to the localities obtained from literature. Numbers correspond to collection localities listed in the Gazetteer and the red star refers to the type locality.

Figure S6. Geographic distribution of Leopardus pajeros. Red dots correspond to specimens housed in scientific collections and black dots refer to the localities obtained from literature. Numbers correspond to collection localities listed in the Gazetteer and the red star refers to the type locality.

Figure S7. Geographic distribution of Leopardus braccatus. Red dots correspond to specimens housed in scientific collections and black dots refer to the localities obtained from literature. Numbers correspond to collection localities listed in the Gazetteer (see and the red star refers to the type locality.

Figure S8. Geographic distribution of Leopardus garleppi. Red dots correspond to specimens housed in scientific collections and black dots refer to the localities obtained from literature. Numbers correspond to collection localities listed in the Gazetteer and the red star refers to the type locality.

Figure S9. Geographic distribution of Leopardus munoai. Red dots correspond to specimens housed in scientific collections and black dots refer to the localities obtained from literature. Numbers correspond to collection localities listed in the Gazetteer and the red star refers to the type locality.

Table S1. Factor loadings and percentage of variance of Principal Component Analysis (PCA) for groups of pampas cats using 16 craniodental variables.

Table S2. Function loadings and percentage of variance of Discriminant Function Analysis (DFA) for groups of pampas cats using 16 craniodental variables.

Table S3. ANOVA results of ecologic variables among groups of L. colocola complex.

Table S4. Evaluation metrics of ecological niche models generated by ENMeval R package for five groups of L. colocola species complex. Results are shown for models with lowest AICc values. Number of occurrence records used for each species is given by N.

ACKNOWLEDGEMENTS

We are grateful to the following curators and collection managers for permission to examine specimens in their respective collections: Mario de Vivo, Luís Fábio Silveira and Juliana Gualda-Barros (MZUSP); João Alves de Oliveira and Sérgio Maia Vaz (MNRJ); Jader Marinho-Filho (UNB); Robert Voss and Eileen Westwig (AMNH); Bruce D. Patterson (FMNH); David Flores, Pablo Teta and Sergio Lucero (MACN); Isabel Gamarra de Fox (MNHNP); Enrique M. González (MNHN-Uruguay); Kristofer M. Helgen, Craig A. Ludwig, Darrin P. Lunde, Esther M. Langan and Nicole R. Edmison (USNM); Mónica Díaz and Rubén M. Barquez (CML); Isabel Dias (CBF); Víctor Pacheco (MUSM); Claudia Medina and Fernando Forero (IAVH); Christiane Funk (ZMB_Mam); Judith M. Chupasko (MCZ); and Roberto Portela Miguez (NHM). We also thank Luis A. Ruedas (Portland State University, Portland, United States) and Deyan Ge (Chinese Academy of Science, China) for kindly providing photos of specimens housed at the NHM, Thiago B. F. Semedo (Universidade Federal de Mato Grosso, Mato Grosso, Brazil) for photos of the types housed in NHM and some species in MNHN-Paris, Pablo Teta and Sergio Lucero (MACN) for photos and information of the neotype of Leopardus pajeros Carla Aquino (MZUSP), Teresa Garcia (EBD), Katrin Krohmann (SMF) and Hwa Ja Götz (ZMB_MAM) for the photos sent to us. James Patton (MVZ), who gave permission for the photographs of the specimens of pampas cats housed in the collection, and Daniela Rossoni, Bárbara Andrade Costa and Ana Paula Aprígio Assis, who kindly took the photographs. Eduardo Eizirik kindly provided numerous sequences of pampas cat individuals. Guilherme Siniciato Terra Garbino (UFMG) José Serrano-Villavicencio (MZUSP) and Luís Fábio Silveira (MZUSP) for valuable suggestions and comments. We want to give special thanks to Roy Baethe for his amazing illustrations of pampas cat species. We also give special thanks to Dione Seripierri (Biblioteca do Museu de Zoologia da Universidade de São Paulo) and Juliane Diller (ZSM) for their valuable effort in obtaining some references needed for the conception of this manuscript. We wish to thank Maarten Christenhusz for his carefully comments on this manuscript. The authors declare that there is no conflict of interest.

FUNDING

FON was financed by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior – Brasil (CAPES) - Finance Code 001 and the AMNH Grants Program (Collection Study). AF is supported by the Chinese Academy of Sciences President’s International Fellowship Initiative (Grant number 2018PB0040).

REFERENCES

Author notes