Morphological evidence for the taxonomic status of the Bridge’s Guan, Penelope bridgesi, with comments on the validity of P. obscura bronzina (Aves: Cracidae)

Penelope obscura Temminck, 1815 is a forest guan found in Brazil, Paraguay, Uruguay, Argentina and Bolivia. Three subspecies are currently recognized: Penelope o. obscura Temminck, 1815, P. o. bridgesi Gray, 1860, and P. o. bronzina Hellmayr, 1914. The limits between Penelope taxa are poorly understood since few studies have evaluated their differences in plumage, distribution and taxonomy. Based on 104 specimens deposited in ornithological collections we studied the variations in the plumage of P. obscura, including all characters that have been used to describe the included subspecies. Our results show that the plumage of these birds is extremely variable in southern and southeastern Brazil. Without any morphological and morphometric characters to support P. o. bronzina as a valid taxon, we synonymized it with P. obscura. Conversely, P. o. bridgesi, which occurs in the Yungas and the Chaco, is a distinct taxon and should be treated as a separate species from P. obscura.


INTRODUCTION
Penelope obscura Temminck, 1815 is distributed from eastern Minas Gerais and northern Espírito Santo in Brazil, to southeastern Paraguay, northeastern Argentina (extending to central Bolivia) and Uruguay (Del Hoyo and Kirwan 2016).As it is often the case among members of the genus, P. obscura is syntopic with other species throughout its distribution: on the eastern slopes of the Andes of Bolivia and northwestern Argentina, this species shares its habitat with P. montagnii and P. dabbenei, and in southeastern Brazil, eastern Paraguay and northeastern Argentina, it is usually found with P. superciliaris (Vaurie 1968).
The body plumage of P. obscura is usually dark brown; neck, mantle, upper wing coverts and breast are edged with dull white, as the forehead and eyebrow.The tarsus is blackish brown or black in life (Vaurie 1968).It presents sexual dimorphism in the color of the iris, red to orange in males and brown or dark brown in females, which is already well differentiated in the first year of life (Sick 1997, Chalukian 1997, Nacinovic 2012).
Based on the "Yacuhú" of Azara (1805), Temminck (1815) described P. obscura.Paraguay is considered as its type locality.Gray (1860) described Penelope bridgesi, type locality in Bolivia (restricted to Villa Montes, Tarija by Vaurie 1966).Ogilvie-Grant (1893) suggested that P. bridgesi should be considered a synonym of P. obscura; however, Hartert and Venturi (1909) considered P. bridgesi as a subspecies of P. obscura, a placement that is currently accepted.Finally, Hellmayr (1914) described P. obscura bronzina based on a specimen collected in Santa Catarina (Colonia Hansa, currently Corupá).The author also delimited the geographic distribution of the three subspecies.Hellmayr's hypotheses (1914) have been adopted up to now, and no revisions of Penelope have been published since the taxonomic study of Vaurie (1968).
Penelope obscura bronzina is an endemic subspecies of the Brazilian Atlantic Forest and has a wide distribution, occurring from east of Minas Gerais to Santa Catarina through Espírito Santo, Rio de Janeiro, São Paulo and Paraná (Sick and Teixeira 1979).The nominate form is distributed in southern Brazil, in the states of Santa Catarina and Rio Grande do Sul, also including the forested regions in Paraguay, Uruguay and northeastern Argentina (in the provinces of Misiones, Chaco, Corrientes, Entre Ríos and Santa Fe); the exact range of this subspecies in Brazil is unclear, but it is most likely restricted to the forests of the Paraná and Uruguay Rivers.Lastly, P. o. bridgesi occurs in western South America, being found in the Bolivian Yungas in the provinces of Cochabamba, Santa Cruz, Chuquisaca and Tarija, and towards the south in Argentina in the forests of eastern slopes of the Andes in the provinces of Jujuy, Salta, Tucumán and Catamarca.The gap between the ranges of this and the two other subspecies may have been caused by an increase of the aridity in the Chaco, which has isolated this taxon in the Andean foothills (Vaurie 1968).Vaurie (1966) recognized that the nominate subspecies differs from P. o. bronzina by having the feathers of the head entirely blackish with only a suggestion of light edgings on the forehead and brows.Penelope o. bronzina, according to this author, has the head with long and bordered feathers, especially on the anterior half of the crown, with silvery-white.Also, the race obscura would be smaller in size and lighter, with plumage browner and throughout duller.Other authors such as Delacour and Amadon (2004) documented that P. o. bridgesi is similar to P. o. bronzina, being larger and having the crest feathers less edged with white and little or no trace of eyebrow stripe, with the plumage duller and browner.
Del Hoyo and Kirwan (2016) mentioned that plumage differences among races suggest that more than one species may be involved in the P. obscura complex.They also pointed out that the possibility of contact and intergradation of the two eastern races, P. o. bronzina and P. o. obscura, required detailed study.Vaurie (1968) suggested that these races do not intergrade, and Del Hoyo and Motis (2004) noted a gap between the two forms in the state of Santa Catarina, Brazil, and also recognized the need of studies to evaluate the taxonomy of this complex.

MATERIAL AND METHODS
We analyzed and measured 80 museum specimens of all taxa currently considered as belonging to the P. obscura complex (36 males, 29 females and 15 of unknown sex), all of which are distributed in Brazil, Argentina, Paraguay and Bolivia.These specimens are housed in the following institutions: Museu de Zoologia da Universidade de São Paulo (MZUSP, including the topotype of P. o. bronzina), São Paulo, Brazil; Museu Nacional da Universidade Federal do Rio de Janeiro (MNRJ), Rio de Janeiro, Brazil; Museu de História Natural do Capão da Imbuia (MHNCI), Curitiba, Brazil; Museu de Ciências Naturais da Fundação Zoobotânica do Rio Grande do Sul (MCNRS), Porto Alegre, Brazil; and Museo Argentino de Ciencias Naturales "Bernardino Rivadavia" (MACN), Buenos Aires, Argentina.Photographs of the specimens were taken against a white background using 10 watt LED spotlights, always in the same place to standardize the lightening.We also examined photographs of 24 specimens, including the holotypes of P. o. bronzina and P. o. bridgesi, deposited in the following institutions: Museum für Naturkunde (MFN), Berlin, Germany; Field Museum of Natural History (FMNH), Chicago, USA; Rijksmuseum van Natuurlijke Historie (RNHL), Leiden, Netherlands; Zoologische Staatssammlung München (ZSM), Munich, Germany; and the Natural History Museum (BMNH), Tring, UK (Fig. 1).
Localities and geographic coordinates were obtained from the labels attached to the specimens.When the coordinates were not available they were obtained by consulting ornithological gazetteers (Paynter and Traylor 1991, Paynter 1992, 1995, Vanzolini 1992).The maps were designed using the software QGis 2.6.0,showing the divisions of countries, Brazilian states and the biogeographic provinces proposed by Morrone (2009) to differentiate the areas where the taxa occur.In order to perform detailed analysis of geographic variations in body size and plumage coloration we defined nine geographic groups considering morphological homogeneity, geographic proximity and ecological similarity.Specimens identified as juveniles by distinct plumage and smaller sizes were excluded from the morphometric and plumage analyses, and were considered only to compile the distribution of the taxa.
We measured the length of the exposed culmen, bill from nostril, tarsus, middle toe, wing and tail using calipers with precision of 0.01 mm following Baldwin et al. (1931).Morphometric data were analyzed to test the sexual dimorphism and intra-and inter-geographic group variations.Shapiro-Wilk and Lilliefors (Kolmogorov-Smirnov) tests were performed to test the normality of data.Sexual dimorphism in each group was evaluated through the Student's t test, testing the differences between sexes for each selected character.For multivariate analyses, the data requires log-transformed variables as input so that all of them have the same relevance for the analysis.Multivariate analyses are based on multiple variables that are measured in the same individual and are not independent from each other, and we tested the general morphology of P. obscura with all the morphometric variables taken together.To corroborate the sexual dimorphism we performed the Hotelling's T square test, the multivariate counterpart of Student's t, testing the differences between the multivariate means.A Lineal Discriminant Analysis (LDA) was also used; this test identifies all the elements that belong to each geographic group (including each measure for each variable) and generates a linear combination of variables that can find the pattern within the data and effectively classify it.It represents in percentages the possibility of individuals belonging to the pre-determined group or meeting the pattern of another group.The Canonical Variates Analysis (CVA) was performed to check whether the groups form significantly distinct clusters morphometrically.The univariate part of the tests was conducted in R software statistical package v. 3.2.1 and the multivariate analysis was carried out in Past v. 2.15.
The plumage was evaluated according to the Munsell Color Guide (1994), based on traits usually considered as diagnostic for the three subspecies: supercilium, crown, throat, breast, abdomen, back, tail and wing coverts.We also analyzed photographs of living individuals available in the online databases Wiki Aves (http://www.wikiaves.com.br) and Ecoregistros (http://www.ecoregistros.org).From WikiAves were reviewed 4.342 images from the states of Espírito Santo (122), Minas Gerais (976), Rio de Janeiro (711), São Paulo (1.582), Paraná (333), Santa Catarina (274) and Rio Grande do Sul (344) in Brazil; and 100 images from Ecoregistros, for the records in Uruguay (18) and northeast (31) and northwest (51) of Argentina (Fig. 1).Some of the examined photographs were excluded from our analysis because they had been taken from a long distance, were very dark, or portrayed partially or wrongly identified taxa.

RESULTS
Statistical analyzes: nine geographic groups were defined: three from the Atlantic forest, two from the Araucaria forest, and one from the Parana forest, Pampa, Chaco and Yungas (Table 1).For statistical analyses, we examined 77 adult specimens distributed in the aforementioned biogeographic provinces (Fig. 2).Table 2 presents the descriptive statistical analysis including mean, standard deviation and minimum and maximum values.
Univariate analysis of normality suggests that data of the six variables meet the criteria of the parametric tests.Student's t-test was not performed for the groups with small sample number; it was used to test the null hypothesis that the mean value of each variable does not differ between the two sexes for each group.Based on the test, the vast majority of the means of the two sexes do not differ significantly; only in five specific cases was there a significant difference between the sexes, with males being larger than females (Table 3).The T 2 Hotelling test has as null hypothesis that the vectors of the means of the sexes are equal, and shows that there is no sexual dimorphism in the same geographic groups tested for Student's t-test (p > 0.05).
Table 4 presents the results of Lineal Discriminant Analysis and shows that everything in the off-diagonal are the frequencies of individuals that also meet the pattern of other groups.Only 33 (44 %) of the individuals were correctly classified within their pre-determined group.The MANOVA/CVA explained  the variations between the geographic groups in the first two axes, and most of them could not be differentiated by any of the morphometric variables.Groups 6 and 7 from southern Brazil and northeastern Argentina (Araucaria forest and Pampa, respectively) differed slightly from the others in the length of the tail (Fig. 3).Specimens collected in eastern South America showed continuous distribution with no gap in the state of Santa Catarina in Brazil.Their plumage showed no sexual dimorphism in any geographic group.The characters used to diagnose the subspecies P. o. obscura were found in specimens from various parts of its wide range.For instance, the whitish edges of head feathers are wide (MZUSP 31577, from Varjão de Guaratuba, São Paulo, BR), thin (MZUSP 34130, Maromba, Rio de Janeiro, BR) or almost imperceptible (MACN 44036, Chantecler, Misiones, ARG), without any apparent geographic trend.The overall color of the crown feathers affects the appearance of the head, which can be black, gray, or brown.In addition, the prominence of the white  BR) to the south often have less noticeable whitish edges and tend to have a blackish general coloration, but this pattern is also recorded in birds from the northern portion of the range, which present general blackish coloration and feathers with white edges reduced or absent (WA700594, WA823706).On the other hand, specimens in the south of the range may present feathers with noticeable white edges and general grayish coloration (WA914588, WA1763524, WA1782826; ER10072, ER10141).
The plumage of the back also showed considerable variation, without any apparent geographical pattern.Across the entire range of the species we found individuals with the back greenish-brown (MZUSP 32906, Teresópolis, Rio de Janeiro, BR), brownish (MZUSP 7021, Fazenda Monte Alegre, Paraná, BR) or blackish (MZUSP 49384, Rocha, São Paulo, BR).Apparently, specimens from the northern part of the range have the white streaks on the back more noticeable when compared to the southern specimens.However, we also found specimens in the south with wide white streaks (MCNRS 1765, Bagé, Rio Grande do Sul), and in the north the white streaks in some specimens are almost absent (MN 45542, Simão Pereira, Minas Gerais, BR) (Fig. 4).
It is common to find individuals with different overall coloration in the same place, demonstrating the existence of a large individual variation (e. g.WA1681568, Pouso Alegre, Minas Gerais; WA450514, Nova Friburgo, Rio de Janeiro; WA1518511, Itatiba, São Paulo; WA1084879 and WA1090972, Tibagi, Paraná, BR).As mentioned before we located specimens with plumage characters supposedly diagnostic for the nominate subspecies in specimens found in the range of P. o. bronzina (WA1765142, Bocaina de Minas, Minas Gerais, BR) and specimens with characters considered diagnostic for P. o. bronzina in the geographic range of nominate subspecies (ER10141; Ribera del Norte, Buenos Aires, Argentina).
On the other hand, all specimens attributed to P. o. bridgesi lacked olive brown coloration with a green tinge to the upperparts (unlike P. o. obscura and P. o. bronzina) (Fig. 4).Instead, specimens of P. o. bridgesi showed chestnut upperparts contrasting with the black head.The underparts were consis-tently uniform and slightly paler than the back, whereas in P. o. obscura and P. o. bronzina the coloration of the upperparts was highly variable.The wing coverts of P. o. bridgesi show broad white margins, a character state not found in P. o. obscura and P. o. bronzina, which have indistinct white margins to the wing coverts (Fig. 5).Overall, P. o. bridgesi shows less pronounced individual variation in plumage coloration than P. o. obscura and P. o. bronzina.

DISCUSSION
Misidentifications of Penelope obscura are still common, especially where it is sympatric with P. superciliaris.The range of P. obscura in eastern South America is continuous, without any gap between the distribution of the nominate race and P. o. bronzina in southern Brazil (Fig. 1).We suspect that the absence of specimens collected between the states of Santa Catarina (from north, called P. o. bronzina) and Rio Grande do Sul (to south, P. o. obscura), deposited in North American and European museums may have led to the assumption that there was a large geographic gap between the ranges of P. o. obscura and P. o. bronzina.As no detailed specimen-based study has been conducted to date, including analyses of individual and geographic variation, the validity of P. obscura bronzina remained uncontested for many decades in the ornithological literature.The same applies to the taxonomic rank of P. o. bridgesi, an isolated taxon occurring in the biogeographic provinces of Yungas and Chaco on the eastern slopes of the Andes in Bolivia and Argentina, which also was never studied in detail.
The eastern population (P.o. obscura and P. o. bronzina) showed considerable variation in plumage coloration, with many individuals presenting characters previously considered as diagnostic for one of the subspecies in the range of the other.The plumage in P. o. bridgesi was much less variable throughout its range than that of P. o. obscura/bronzina.Statistical analyses showed no morphological differences between males and females within the groups, and the few differences among some geographic groups do not represent a pattern of biogeographic variation in the eastern or western populations.We cannot be sure that there is a marked difference in size between the aforementioned subspecies of P. obscura, as was previously mentioned by Vaurie (1968).Total size measurements attached to some specimens' labels suggest that males are larger than females, but the color of the iris is the only character that effectively differentiates the two sexes.
The occurrence of P. obscura is still questionable in some regions.Sousa (1999) mentioned the presence of this species in Elísio Medrado (13°S, 39°45'W) in the state of Bahia (Brazil), being the north limit of its distribution.However, as this record was based on a personal communication from a third part, and there are no records from naturalists in the 19 th and 20 th centuries, when the habitat was undisturbed, we consider unlikely that P. obscura occurs in the state of Bahia since there is no reliable evidence for it.In Paraguay, Hayes (1995) mentioned Penelope obscura as a hypothetical species.It was visually reported in the Chaco (President Hayes, Boquerón) and Orient (Central, Alto Paraná and Ñeembucú), but unfortunately those reports had no other details.The only specimens mentioned by the author are deposited at the National Museum of Natural History, Smithsonian Institution (USNM 256875, 58994), and they lack precise information on collecting dates and localities, and were obtained at a time when Paraguay included parts of what is now Brazil and Argentina.Smith and Derna (2015) recently reviewed the records in the country and showed that museum skins attributed to P. obscura actually represent members of P. superciliaris.Recent records from Paraguay are photographs taken in the Refugio de Vida Silvestre Atinguy, in Misiones department (27°20'S, 56°41'W), claimed by the authors to be wild-origin birds.
After the review of a large series of specimens of P. o. bridgesi, and by comparing it with P. o. obscura and P. o. bron zina, we find that what is currently defined in the literature as Penelope obscura does not include the characters observed in the individuals of P. o. bridgesi.When Ogilvie-Grant (1893) proposed bridgesi as a synonym of Penelope obscura, he only reviewed seven specimens (four from Brazil, one from Uruguay, one from Para-guay, and the holotype from Bolivia), excluding the characters of bridgesi cited by Gray (1860) that are diagnosable to considered it as a species; the same situation was also observed in Hartert and Venturi (1909), which considered it as a subspecies of P. obscura.Our results clearly show that Penelope bridgesi is a taxon that can be differentiated from all others without overlapping with any other Penelope.It is geographically isolated and should be treated as a valid species (see below).The populations in eastern and southern Brazil are continuously distributed and show considerable individual variation in plumage, without any clear geographic trend.There are any morphological or morphometric characters to support the continued existence of more than one taxon in the region.As P. o. bronzina cannot be distinguished from the nominate subspecies, we here treat Penelope obscura bronzina Hellmayr, 1914 as a synonym of Penelope obscura Temminck, 1815.The treatment proposed here is consistent with the current proposals found for the other species of Penelope.

TAXONOMY
Penelope obscura Temminck, 1815 Holotype.The description of the species is based on the 'Yacuhú', made by Azara (1805).Type locality: Paraguay.
Diagnosis.Distinguished from P. bridgesi by olive brown coloration with green tinge in upperparts, and generally darker overall coloration.Underparts very variable, unlike P. bridgesi.White margins of wing coverts markedly reduced.Head, hindneck, and back coloration more uniform than in P. bridgesi.
Re-description.No sexual dimorphism found in plumage coloration.Whitish edges extension of head feathers varies from wide to almost imperceptible.Some individuals show a moderately distinct superciliary stripe formed by short feathers with white edges.Overall color of crown feathers variable, ranging from black, gray to brown.White hue of the edge of the feathers varies individually.Brown back may show greenish, brownish or blackish tinge.Neck, breast, upper back, and shoulders conspicuously edged with white.Birds in the northern parts of the species' range tend to have more noticeable white streaks on the back, but streaks also may be lacking.The coloration of the underparts is variable and may show a contrastingly darker chest than belly, but underparts may also be uniform dark or lighter than the breast.The belly feathers have contrasting horizontal scaling, which may range from brown to orangish, and may vary in intensity.Tail feathers are dark brown with bronze or purplish sheen.Bare skin of face slaty, legs dusky gray to blackish, and wattle red.
Distribution.Southeastern and southern Brazil (from the east of Minas Gerais and north of Espírito Santo to Rio Grande do Sul), northeastern Argentina (Misiones, Chaco, Corrientes, Entre Ríos, Santa Fe and Buenos Aires), and the forested regions grants received and for the financial support; Idea Wild for the equipment donated to DEV; the photographers who posted the photos in EcoRegistros and WikiAves websites.The curators and staff of the following institutions kindly welcomed us and allowed us to study the specimens under their care: MFN and ZSM (Germany); FMNH (USA); RNHL (Netherlands); BMNH (UK); MACN (Argentina); MNRJ, MHNCI and MCNRS (Brazil).We also thank Instituto Chico Mendes de Conservação da Biodiversidade (ICMBio/SISBIO) for issuing collecting permits.LFS is supported by a CNPq research productivity fellowship.George Sangster kindly reviewed the manuscript and offered important contributions to this text.

Table 1 .
Geographic groups and respective number of skins used for statistical analysis.

Table 2 .
Mean, standard deviation, minimum and maximum values of body measurements (mm) from the geographic groups of Penelope obscura.

Table 3 .
Student t-test comparing males and females of each geographic group with more than four individuals for each sex.Asterisks indicates significant differences between both sexes at p < 0.05.

Table 4 .
Frequency of percentages of predictive performance from the Lineal Discriminant Analysis (LDA) for all the geographic groups.