CRANIOMETRIC VARIATION IN JAGUAR SUBSPECIES (PANTHERA ONCA) FROM COLOMBIA

Transcription

1 Chapter 11 CRANIOMETRIC VARIATION IN JAGUAR SUBSPECIES (PANTHERA ONCA) FROM COLOMBIA Manuel Ruiz-García 1, and Esteban Payan 2 1 Unidad de Genética (Grupo de Genética de Poblaciones Molecular-Biología Evolutiva). Departamento de Biología. Facultad de Ciencias. Pontificia Universidad Javeriana. Bogotá D.C. Colombia. 2 Panthera, Colombia. Bogotá, Colombia ABSTRACT Forty one jaguar skulls from Colombia were measured for 46 quantitative morphometric variables. We detected no differences between two supposed subspecies of jaguars or their gender. However, size was more important in determining individual differences than shape. Therefore, morphometric data seem to be less powerful in detecting differences among different gene pools than DNA microsatellites or other molecular markers. Keywords: Colombia, jaguar, morphometric skull traits, Panthera onca INTRODUCTION Jaguars are solitary big cats, living in Neotropical rainforests. They are nocturnal and are extremely shy of humans. Although the pelt trade has significantly decreased, jaguars are still threatened due to habitat loss by logging, cattle ranching and hunting by humans due to domestic livestock attacks (Payán and Almeida, 2002). They are classified in the CITES in Appendix I and are considered to be vulnerable by the IUCN. Consequently, there is great difficulty in obtaining samples and tissues from wild animals and the effort is not entirely risk free, making systematic, jaguar evolutionary and population genetics studies extremely rare. Therefore, jaguar skeletons are relatively scarce in the world s Natural History Museums mruiz@javeriana.edu.co

2 2 Manuel Ruiz-García and Esteban Payan compared to those from other mammalian species. Futhermore, very few Colombian jaguar skulls have been analyzed in previous studies (Nelson and Goldman, 1933; Pocock, 1939; Larson, 1997) and no paleontological analysis has been completed of this species in the current Colombian territory. The good news is that the paleontological evolution of the jaguar is partly known. Jaguar fossils in North America date back to the mid-pleistocene (Kurten, 1973; Kurten and Anderson, 1980; Schultz et al., 1985). Kurten (1973) suggested that the earliest jaguars may have been conspecific with the Euroasiatic P. gombaszoegensis which migrated across the Bering land bridge to current North America. Fossil findings of pre-wisconsinan jaguars are distributed as far north as Washington, Nebraska and Maryland. This fossil form was classified as P. onca augusta and was 20% larger than any of the supposed extant subspecies. In North America 73 fossils have been recorded; however only 18 have been recovered in South America (Seymour, 1983). Morphological studies have intended to establish systematic and taxonomical differences among individuals from diverse geographical origins using craniometrical characters, body size, color and coat patterns. The first anatomical studies intending to determine geographical relationships among individuals from different regions were those from Martin (1832), Mearns (1901) and Hollister (1914). Goldman (1932) and Nelson and Goldman (1933) were the first authors to analyze a series of 101 specimens (35 skins and skulls, 57 skulls without skins and 9 skins without skulls) and defined 16 subspecies (see also Cabrera, 1957). Here we describe the type locality and distribution range of two of these subspecies. The first one, P. onca onca, (type locality Pernambuco; now Recife, fixed by Thomas 1910, in Nelson and Goldman 1933) presents a distribution ranging from extreme eastern Brazil to west and north of the lower Amazon, including the Colombian Amazon and the Eastern Llanos. The second one, P. onca centralis (type locality: Talamanca, Costa Rica), is from the valley of the Sicsola River (near Sipurio, Costa Rica) to the north of El Salvador and along the Pacific coast covering the Isthmus of Tehuantepec and all the Pacific- north of Colombia to Guaduas (Cundinamarca Department) at the center of Colombia. Pocock (1939) studying 59 skulls with their respective geographic origins, determined the existence of only 8 different subspecies: P. onca onca, P. onca palustris = paragueyensis, P. onca peruviana, P. onca centralis, P. onca goldmani, P. onca hernandesii, P. onca veraecrucis and P. onca arizonensis. This author suggested that several of the subspecies defined by Nelson and Goldman (1933) were synonyms. Therefore, P. onca onca now includes P. o. major, P. o. mexianae, P. o. coxi, P. o. boliviensis, P. o. ucayalae and P. o. madeirae and P. onca paragueyensis includes P. o. milleri and P. o. paulensis. Cabrera (1940) and Cabrera and Yepes (1960) studied in detail specific characters which clearly determined P. onca paragueyensis as a different population from other geographical jaguar populations. The work of Larson (1997) is based on eleven skull characters and considers the existence of the 8 subspecies of jaguars as defined by Pocock (1939). The above are the subspecies currently recognized by Weigel (1961), Seymour (1989) and Swank and Teer (1989). Larson (1997) studied 170 specimens in natural history museums in the United States belonging to these 8 subspecies using diverse multivariate analysis and a multigroup discriminant analysis among them and did not find significant differences among the supposed subspecies. He concluded that jaguar populations tend to be highly variable within taxa, with low variation between populations and therefore, following Templeton et al., (1986), jaguars have high variation between individuals, regardless of their subspecific status,

3 Craniometric Variation in Jaguar Subspecies (Panthera Onca) from Colombia 3 implying that the multiple populations comprise a single taxon. Members of a subspecies share a unique geographical range, a number of phylogenetically concordant characters, and an evolutionary history and natural history which is specific to each (Avise and Ball 1990; O Brien and Mayr, 1991). Although subspecies are not reproductively isolated, they will normally be allopatric and exhibit recognizable phylogenetic partitioning (O Brien and Mayr, 1991). During the last century, geographic variants of species were named as subspecies without clear definition. When this indeterminate taxonomy is overlaid with precise wildlife legislation, it is sometimes detrimental to conservation efforts (Avise and Nelson 1989; O Brien and Mayr, 1991; Wayne and Jenks, 1991). Thus, there is an imminent need for conservation biologists to define taxonomic units explicitly, particularly those that are endangered and threatened. Colombian jaguar populations are considered to be in immediate need of research and are in priority areas of the species range. Having all this in mind, a craniometric analysis was conducted on jaguar skulls from the geographic range of the two supposed Colombian subspecies (Panthera onca onca and Panthera onca centralis). Several of the specimens analyzed herein were also genotyped for 18 DNA microsatellites (FCA01, FCA08, FCA24, FCA43, FCA45, FCA70, FCA94, FCA96, FCA126, FCA136, FCA176, FCA200, FCA225, FCA251, FCA290, FCA294, FCA391 and FCA506; Genotyped samples are presented with an asterisk in Appendix 1). The microsatellite results are presented elsewhere (Ruiz-Garcia et al., 2003; 2006). Previous scientists have also based jaguar taxonomy on skull characteristics. For example, Pocock (1939) described the Panthera onca onca as having great individual size variation and characterized this subspecies based on a differential condylobasal length. The second subspecies Panthera onca centralis also showed high skull variability but was similar to P. onca hernandesii (the West Mexican jaguar, see Nelson and Goldman 1933), but with less depressed nasals. But a rigorous analysis of morphological differences with a considerable number of samples has never been conducted. Furthermore, actual jaguar taxonomy has been derived from such general characterizations as above and the Colombian subspecies are in need of a detailed revision. This chapter examines the cranial variation between skulls of jaguar belonging to the two traditionally proposed subspecies in Colombia (Panthera onca onca and Panthera onca centralis); in order to search for evidence to support the existence of these subspecies by means of more powerful and more recently developed methods of morphological analysis. Furthermore, this chapter provides some comparisons of craniometric data with our molecular studies to determine if there is a correlation between morphological and molecular evolution in the jaguar. Multivariate Craniometric Analysis MATERIAL AND METHODS Appendix I shows the collection number and the geographic origins of the 41 adult skulls measured for this study. Maturity was determined by closure of the basisphenoid-basioccipital suture (Smuts et al., 1978). Thirty-eight of these skulls are from the Mammalogy collection belonging to the Instituto Alexander Von Humboldt in Boyacá Department, Colombia.

4 4 Manuel Ruiz-García and Esteban Payan Figure 1. Graphical description of several skull variables measured in 41 Colombian jaguars. The remaining 3 skulls belong to private collectors. Only 5 of these skulls have no precise geographical origin. The 46 craniometric variables used are shown in Table 1 and in Figure 1, and some of them are discussed by Fernández et al., (1992) and García-Perea (1994). Different multivariate techniques were applied to analyze the relationships among the 41 Colombian individuals and the relationships among variables and the discriminatory capacity of the morphometric variables employed. In the first analysis carried out, the craniometric distances were obtained without using any type of standardization or transformation to determine the simultaneous impact of size and shape among the individuals analyzed, keeping in mind that the differences separating the alleged jaguar subspecies deal with size (Nelson and Goldman, 1933). Different distance matrices (correlation, variancecovariance, Euclidean and Manhattan distances; Sneath and Sokal, 1973; Marcus, 1990) were calculated among the individuals analyzed. Each one of these procedures has different mathematical properties, which must be evaluated to see the effects on the results obtained. The UPGMA algorithm was applied to each one of the matrices obtained in order to construct a phenogram which shows the relationships among the individuals analyzed. The cophenetic correlation coefficient was calculated for each one of these trees. In order to analyze the degree of similarity among these diverse phenograms, a strict consensus tree was obtained (Rohlf, 1982). Table 1. List of variables used and their abbreviations. The majority of measurements are taken from Garcia-Perea, (1994) and Garcia-Perea et al., (1985) Variable Abbreviations 1 Greatest length of skull along the medial plane GLS 2 Condylobasal length CBL 3 Rostral width across canines RWC 4 Mastoid width MW 5 Interorbital width IOW

5 Craniometric Variation in Jaguar Subspecies (Panthera Onca) from Colombia 5 Variable Abbreviations 6 Postorbital width POW 7 Zygomatic width ZW 8 Length of P 4 P 4 L 9 Mandible length ML 10 Length of sagittal crest SCL 11 Basal length LBA 12 Palatine length PAL 13 Length of rostrum LRO 14 Length of the neurocranium NCL 15 Length of the temporalis lineae LLT 16 Condilomolar length CML 17 Molar separation MS 18 Distance between lineae temporalis measured transversally at the intersection of the sutura parietofrontalis with the sutura sagittalis DLTPS 19 Length from the incisure between the condylar and angular processess to the infradentale LCPIA 20 Length from the aboral extreme of the angular process to the infradentale LAAPI 21 Length from the aboral extreme of the coronoid process to the infradentale LACP 22 Height of the mandibular arch HMA 23 Distance between C 1 and P 4 C 1 -P 4 24 Distance between P 3 and P 4 P 3 -P 4 25 Width of the P 4 P 4 W Distance between C 1 and M 1 C 1 -M 1 27 Length of M 1 LM 1 28 Maximum width between molars 1. AMM 1 29 Craneal width CW 30 Distance between the foramen lacerum and the foramen ovale DFLFO 31 Maxilla-Palatine index 1 MPI1 32 Maxilla-Palatine index 2 MPI2 33 Nasal width NW 34 Nasal Height NH 35 Occipital condyle width OCW 36 Intermaxilar width IW 37 Greatest diameter of superior canine MDSC 38 Cranium depth CD 39 Greatest diameter of inferior canine MDIC 40 Width of incisors Win 41 Maxilar width MW 42 Length of bulla LB 43 Width of bulla WB 44 Height of the cranium CH 45 Length of upper tooth row + incisors LCST 46 Length of lower tooth row LIM

6 6 Manuel Ruiz-García and Esteban Payan To establish other possible relationships among the 41 jaguars studied, a Principal Coordinate Analysis (PcorA) was applied. The PcorA procedure employed was that used by Gower (1966). A graphic matrix ( Minimum Spanning tree ) was superimposed (Gower and Ross, 1969; Rohlf, 1970) in order to see the probable local distortion generated by the process of dimensional reduction. To establish the relationships among the individuals studied and the influence of size and shape, a Principal Component Analysis (PCA) with the standardized data was carried out. Therefore, the same weight was given to all the morphometric variables employed. A high and positive correlation of all variables with the first component usually denotes differences of size among the individuals. Otherwise, the following components mainly describe the shape of the individuals. The following tests were applied to analyze if there are differences among the variance percentage explained for each component: the Bartlett (1950) expected variance proportion (employing the broken-stick model), the Anderson (1963) test, the Lebart and Fenelon s (1973) test and the Ibanez test (1973). The relationships among the second and third components were analyzed to determine the relationships among the individuals exclusively by shape. Burnaby s size adjustment method was also applied (Burnaby, 1966). The major part of these analyses were undertaken with the program NTSYS 2.02g. Other analyses were conducted with a program created by the first author in In order to detect differences among individuals belonging to different Colombian geographical regions and belonging to the subspecies Panthera onca onca and P. onca centralis, a Canonical Population analysis (PCA) was performed. This method separates groups along axes with high discrimination power (canonic axes) using the Mahalanobis distance (Mahalanobis, 1936). This analysis is based on two hypotheses: there is homogeneity between all covariance matrices corresponding to the populations groups (this was verified with a maximum-likelihood test) and the means of the k groups must be significantly different. To contrast this hypotheses, the Wilks s and the Fisher-Snedecor F associate value test by means of the Rao (1951) approximation was used. Subsequently, a canonical transformation was made and the eigenvalues and the significance of the first canonical axes with Barlett s test were calculated. Additionally, the factorial structure of the canonic variables, the canonical representation and the confidence region radius at 90% level, were determined. The expression for the radius is R /N 1/2, where R 2 = F (N k) n / (N k- n + 1), with P (F > F ) = 1 - for the Fisher-Snedecor distribution with n and N k n 1 degrees of freedom (N = total population number, k = number of population groups, and n = number of variables). In order to perform the first canonical analysis, 4 geographical groups were determined: the jaguars from the western Pacific area of Colombia, including Chocó and parts of the Antioquia Department (Panthera onca centralis), jaguars from northern Colombia (Northern Santander and Bolivar Departments) (Panthera onca centralis), jaguars from the Eastern Llanos (Panthera onca onca) and jaguars from the Amazon (Panthera onca onca). In the second analysis, individuals were grouped by subspecies: Panthera onca centralis vs. Panthera onca onca. In both analyses, three skulls with unknown origin were included to identify their possible origins. Both analyses were carried out in two sets, the first one contained cranial data and the second dental data. This analysis was performed using the MULTICUA software created by Cuadras (1991). Additionally, two discriminant analyses were performed. One included the cranial variables and the other included the dental and mandible variables for both supposed Colombian jaguar subspecies samples using the theory

7 Craniometric Variation in Jaguar Subspecies (Panthera Onca) from Colombia 7 proposed by Marks and Dunn (1974), Lachenbruch and Goldstein (1979) and Cuadras (1989) by means of the SAS and the BMDP programs. We used three different procedures: 1) A linear discriminator with the cross-validation and the leave-one out techniques, 2) a Euclidean linear discriminator and 3) distance discriminators with three different distances, the absolute value one, the Minkowski (exponential = 3) one and the Mahalanobis one. A quadratic discriminator was not employed because its use was not efficient for the current data. RESULTS Results without Transformation or Standardization Several studies have claimed that the main differences among some jaguar subspecies are related to the size of the animals. For this reason, the first multivariate craniometrical analyses were undertaken without any transformation or standardization. Figure 2a. shows a UPGMA tree using a correlation matrix (cophenetic correlation, r = , t = 6.221, P = ) with the 41 jaguar skulls studied. This tree did not reveal any conspicuous geographic association among the animals studied. For instance, we found an association of jaguar skulls from the Chocó region (Colombian Pacific area, Panthera onca centralis), two from the Colombian Amazon (P. onca onca) and one from the Santander Department (P. onca centralis), with these areas in opposite parts of Colombia. Association among skulls of the same gender was also absent. Male and female skulls were mixed across all trees obtained. Thus, no geographical or sexual associations were discovered with the first multivariate analysis of the 46 quantitative variables and incorporating size as a factor. The resulting strict consensus tree, using simultaneously the Euclidean and the Manhattan distances is shown in Figure 2b. Two main clusters were present. One of them was small and composed of three skulls with origins from Riosucio (Chocó, Pacific area), Itilla River in the Vaupés Department (north of Colombian Amazon) and Leticia (south of Colombian Amazon). These skulls representing both Colombian jaguar subspecies, were characterized for their small sizes, although they were adults. Here, size seems to be related with individual variation rather than with differences between the two subspecies studied. In the other main cluster, the most divergent individual was an extremely big jaguar from Sao Paolo do Olivença (number 10), a Brazilian Amazonian locality near the Colombian border. Within this cluster, two sub-clusters were detected. In one of them, we found animals from diverse Colombian regions (Riosucio, Chocó; Eastern Llanos; Amazon region and Pto. Berrio in the Antioquia Department, north of the country), which shows a lack of discrimination between the two Colombian subspecies at this craniometrical level. Nevertheless, the second sub-cluster was composed of nine jaguars, six of which belonged to animals from the Amazon and the Eastern Llanos (i.e. P. onca onca) and three were of unknown origin. Therefore, these three animals together with the other six jaguars could possibly constitute the only sub-cluster of the same subspecies detected by the craniometric data. The PcorA analyses (Figure 3) were undertaken with correlation, variancecovariance, Manhattan and Euclidean matrices.

8 8 Manuel Ruiz-García and Esteban Payan Figure 2. UPGMA phenograms of 41 Colombian jaguar skulls simultaneously considering the influence of size and shape with the correlation matrix (A). Strict Consensus tree of 41 Colombian jaguar skulls simultaneously considering the influence of size and shape with the Euclidean and Manhattan distance matrices (B). Geographic origins and subspecies of skulls: 1.Chocó (P. o. centralis), 2. Vaupés (P. o. onca), 3. Brazilian-Peruvian Amazon (P. o. onca), 4. Colombian subspecies? 5. Colombian subspecies? 6. Vaupés (P. o. onca), 7. Leticia (P. o. onca), 8. Vaupés (P. o. onca), 9. Vaupés (P. o. onca), 10. Brazilian Amazon (P. o. onca), 11. Vaupés (P. o. onca), 12. Guaviare (P. o. onca), 13. Colombian subspecies? 14. Colombian subspecies? 15. Colombian subspecies? 16. Colombian subspecies? 17. Santander (P. o. centralis), 18. Antioquia (P. o. centralis), 19. Guaviare (P. o. onca), 20. Guaviare (P. o. onca), 21. Guaviare (P. o. onca), 22. Guaviare (P. o. onca), 23. Colombian Amazon, Leticia (P. o. onca), 24. Guainía (P. o. onca), 25. Colombian Amazon, Leticia (P. o. onca), 26. Vaupés (P. o. onca), 28. Arauca (P. o. onca), 29. Vaupés (P. o. onca), 30. Bolivar (P. o. centralis), 31. Guainía (P. o. onca), 32. Chocó (P. o. centralis), 33. Guainía (P. o. onca), 34. Caquetá (P. o. onca), 35. Colombian Amazon, Leticia (P. o. onca), 36. Colombian Amazon, Leticia (P. o. onca), 37. Meta (P. o. onca), 38. Vaupés (P. o. onca), 39. Guainía (P. o. onca), 40. Colombian subspecies? 41. Chocó (P. o. centralis).

9 Craniometric Variation in Jaguar Subspecies (Panthera Onca) from Colombia 9 Figure 3. Principal Coordinates analyses (PCorA) of 41 Colombian jaguar skulls simultaneously considering the influence of size and shape. PCorA with the correlation matrix (A). PCorA with the Manhattan distance matrix (B). It is interesting to note that PcorA with the correlation and the variance-covariance matrices showed four significant principal coordinates, using the broken-stick model, the Lebart and Fenelon s (1973) and the Ibanez (1973) tests. The variance-covariance method showed a first principal coordinate explaining 53.24% of the total variance and the first principal coordinate with the correlation matrix of only 39.60%. Contrarily, the Euclidean and Manhattan distances, had only the first principal coordinate as significant and the amount of variance explained by it was substantially higher than those observed with the correlation and variance-covariance methods (Manhattan distance: 79.53% and Euclidean distance: 66.62%, respectively). The PcorA with the correlation matrix (Figure 3a) showed that the majority of the jaguar skulls were very similar and that the first principal coordinate differentiated only three skulls with the following origins: one from Sao Paolo de Olivença, another from the Guayabero River (Guaviare, Colombian Amazon) and a last one from Miraflores (Vaupés,

10 10 Manuel Ruiz-García and Esteban Payan Colombian Amazon), all Amazonian jaguars. Towards the other side of the coordinate (right, positive) a jaguar is separated from Pto. Inirida (Guainía). The second principal coordinate differentiated only three individuals from the clustered nucleus set, one from Itilla River (Vaupés, Colombian Amazon), another from Colorado (Bolivar, north of Colombia) and one animal from La Macarena (Meta, Colombia). The PcorA with the Manhattan distance (Figure 3b) showed as well a compact nucleus containing the majority of the skulls analyzed. Three small individuals diverge from the nucleus in the first principal coordinate (one animal from Leticia, Colombian Amazon, one from Riosucio, Chocó, Pacific Colombian coast, and one from Itilla River, Vaupés, Colombian Amazon). Figure 4. Canonical Population Analysis of 41 Colombian jaguar skulls simultaneously considering the influence of size and shape. Two sets of variables were considered. One set was mainly composed of cranial and face variables and the other one was composed of maxilla, mandible and teeth variables. Considering four different jaguar populations (Pacific Colombia, North Colombia, Western Llanos and Amazon) with cranial and facial variables (A); Considering four different Colombian jaguar populations with dental and mandible variables (B).

11 Craniometric Variation in Jaguar Subspecies (Panthera Onca) from Colombia 11 The second principal coordinate strongly differentiates from the central nucleus downward the skull from Sao Paolo de Olivença, and upward, but only slightly three individuals: one from Miraflores (Vaupés, Colombian Amazon), the other from the Javarí River (Peruvian-Brazil Amazon frontier near to the Colombian Amazon) and a last one of unknown origin. In sum, there were no strong craniometrical differences between Colombian subspecies or populations by regional or ecological areas. The differences found seem to be more related with individual characteristics than with population trends. Two PCA analyses of the entire sample size divided into four geographic groups (Colombian Pacific area and Antioquia Department, Western Colombia; Santander and Bolivar Departments, North of Colombia; Colombian Eastern Llanos; Colombian Amazon) showed different discriminatory capacity depending on the morphological variables chosen (Figure 4). When the variables were principally cranial and facial ones, a certain geographic difference was detected. Wilks was 0 with F equal to with 150 and 37 degrees of freedom. This was significant (F = with = 0.05). Therefore, the mean values of the diverse geographical groups were different. The Amazon group was clearly differentiated from the other groups. Curiously, the Pacific and the North Colombian groups (supposedly belonging to Panthera onca centralis) were separated although both were linked to the Eastern Llanos group (supposedly P. onca onca). Additionally, three skulls with unknown origin were introduced in this analysis. One was associated to the Pacific group and the other two were not associated with either geographical group. The first two canonic axes explained % of the total variance in this analysis. In the second PCA analysis, with the same geographic groups, but using dental and mandible variables, Wilks was with a F = with 120 and 65 degree of freedom. This means that there were not significant differences for these variables among the four geographical groups considered (F = with = 0.05). The two canonic axes explained 93.15% of the total variance. Thus, the dental and mandible variables discriminated less than the cranial and facial variables among the different Colombian jaguar population considered. A second PCA with two groups (P. onca onca and P. onca centralis) clearly showed greater discriminative power for the cranial and facial variables than for the teeth and mandible variables although the first variables did not reach a complete separation of the subspecies (not shown here). The results of the discriminant analysis could be seen in Table 2. The analyses carried out with only the cranial variables showed one procedure (linear cross-validation method) which perfectly differentiated the skulls of P. onca centralis from those of P. onca. onca (number of wrong classified individuals = 0; Probability of wrong classification = ), but the other procedures always misclassified some individuals (probability of wrong classification = %). A similar result was obtained with the dental and the mandible characters. Two methods offered perfect classifications (linear cross validation and Mahalanobis distance procedures). However, all the other methods showed significant probabilities of wrong classification (from 8.33 to 27.77%). Therefore, these discriminant analyses with the main exception of the linear cross-validation method were unable to completely differentiate the skulls of both supposed Colombian jaguar subspecies. Nevertheless in this case, there were no excessive differences among the cranial and facial variables and the teeth and mandible ones, which is opposite to that discovered with the PCA.

12 12 Manuel Ruiz-García and Esteban Payan Table 2. Discriminant Analyses for both supposed Colombian jaguar subspecies with 2 sets of data (cranial and facial variables and dental and mandible variables) Cranial and facial variables Linear cross-validation discriminator Number of wrongly classified individuals = 0 Probability of wrong classification = 0 Linear leave-one-out discriminator Number of wrongly classified individuals = 8 Probability of wrong classification = P. o. centralis were assigned to P. onca onca 4 P. o. onca were assigned to P. onca centralis Euclidean linear discriminator Not possible to analyze with the current data Absolute value distance discriminator Number of wrongly classified individuals = 2 Probability of wrong classification = P. onca onca were assigned to P. o. centralis Minkowski distance discriminator Number of wrongly classified individuals = 8 Probability of wrong classification = P. o. centralis were assigned to P. onca onca 5 P. o. onca were assigned to P. o. centralis Mahalanobis distance discriminator Not possible to analyze with the current data Dental and mandible variables Linear cross-validation discriminator Number of wrongly classified individuals = 0 Probability of wrong classification = 0 Linear leave-one-out discriminator Number of wrongly classified individuals = 10 Probability of wrong classification = P. o. centralis were assigned to P. onca onca 6 P. o. onca were assigned to P. onca centralis

13 Craniometric Variation in Jaguar Subspecies (Panthera Onca) from Colombia 13 Euclidean linear discriminator Number of wrongly classified individuals = 5 Probability of wrong classification = P. o. centralis were assigned to P. onca onca 1 P. o. onca were assigned to P. onca centralis Absolute value distance discriminator Number of wrongly classified individuals = 3 Probability of wrong classification = P. onca onca were assigned to P. o. centralis Minkowski distance discriminator Number of wrongly classified individuals = 8 Probability of wrong classification = P. o. centralis were assigned to P. onca onca 5 P. o. onca were assigned to P. o. centralis Mahalanobis distance discriminator Number of wrongly classified individuals = 0 Probability of wrong classification = Results with Transformation or Standardization of Original Data The Principal Component analysis with standardized data from the correlation matrix showed a first principal component, which explained an extraordinarily high variance of the total variance of the system (96.81%). This first principal component was the only significant one in this analysis using diverse methodologies. Nearly all variables showed positive correlations with this first component. This could mean that this first component, which practically comprised all the variance of the model, represents differences in size among the jaguars studied, whereas only a very little fraction of the differences is determined by the shape of the jaguar skulls. Then the individual sizes of the animals seem to be the main factor of differentiation. In this analysis, comparing the first and the second components, we observed a very homogeneous nucleus from which several individuals diverged. When we compared the second and the third components (only shape), we also observed an important nucleus of individuals from which two lines of individuals diverged. One of these lines was composed of five individuals, one from La Macarena (Meta), one from Riosucio (Chocó) and three of unknown origin. The second line was integrated by seven exemplars whose origins were Yuruparí (Vaupés), Arauca, Itilla River (Vaupés), Colorado (Bolivar), Pto.Inirida (Guainía) and two of unknown origin. No geographic or gender trends

14 14 Manuel Ruiz-García and Esteban Payan were detected. A second Principal Component Analysis from the variance-covariance matrix with the log-data also revealed that the first component explained 93.58% of the total variance. Therefore, size again is the fundamental cause of the differences found. The comparison of the second and third components revealed three lineages from a homogeneous nucleus. One of them was integrated only by two geographic unknown origin individuals. One second group was constituted by 11 individuals, with one individual from Bahía Solano (Chocó) the center of this group. The last line was integrated by skulls from Guayabero River (Guaviare), two from Miraflores (Vaupés) and two of unknown origin. Figure 5. UPGMA (A) and Neighbor-joining trees (B) of 41 Colombian jaguar skulls only considering the influence of shape by means of the Burnaby (1966) procedure.

15 Craniometric Variation in Jaguar Subspecies (Panthera Onca) from Colombia 15 The last method employed was the Burnaby (1966) s procedure which extracts the first eigen vector representing size. The UPGMA and the Neighbor-joining phenograms with the Manhattan distance are shown in Figure 5. The P. onca centralis skulls were mixed among the P. onca onca skulls and did not conform a separated cluster. We only found a sub-cluster of animals belonging to the distribution of P. onca onca in the NJ tree. DISCUSSION We didn t detect conspicuous geographic skull associations or differential classification by gender of jaguars when we employed different distances and different procedures (PCorA and phenetic trees). The major fraction of the distance variables analyzed did not show significant dimorphism (ANOVAs and MANOVAs not shown here). Nevertheless, Larson (1997) showed that a discriminant analysis of the sexes led to 75% correct allocation by gender. This author applied several Principal Component Analyses with size-constrained and non-size constrained procedures. The latter PCA data points tend to be more separated, with males toward one side and females toward the other side of the graphics, although showing considerable overlap in the middle. With the former PCA, sexes were intermixed. Thus, limited differentiation among males and females is due to size and to shape. Larson (1997) states that sexual dimorphism exists in jaguars, but that it is very weak and unlikely to obscure taxonomic differences. Furthermore, the condylobasal length of males and females showed a similar pattern. The main differences found herein were basically among the individuals, irrespective of subspecies. Only the strict consensus tree using the Euclidean and the Manhattan distances, showed a consistent sub-cluster with all the individuals belonging to Panthera onca onca and three skulls of unknown origin. Although the capacity of the skull variables to assign individuals to some geographic origin seems low, the position of the three above skulls suggests a possible origin from the Eastern Llanos or from the Amazon region. Individual characteristics rather than population or subspecies ones seem to play a key role as differentiating agents. Larson (1997) based on skull morphology, did not find support for the traditionally proposed jaguar subspecies either. Pocock (1939) had demonstrated that several skull features, such as shape of the nasals and frontals, size of the bullae, width of the mesopterygoid fossa, height of the sagittal crest, height of the mastoid crest, the postorbital area and the zygomatic arches had no systematic value in jaguars. Herein, we demonstrate the low power of a set of 46 quantitative biometric variables to detect differences among supposed jaguar subspecies and therefore suggest the use of molecular data. Additionally, Larson (1997), using the condylobasal length alone, the most commonly employed variable to discriminate jaguar subspecies, showed that it could assign individuals to subspecies correctly only 14% of the time. A correct assignment of the jaguar skulls to their geographical groups was only 40-50%. A comparison of within-group and between group variations discloses cases of higher within-group variation than between-group variation. For instance, several P. onca hernandesii skulls presented more craniometric resemblances to P. onca veraecrucis, P. onca goldmani, P. onca centralis, P. onca onca, and P. onca peruviana than to those within its own subspecies. Apparently microsatellite molecular evolution is highly useful in distinguishing significant genetic heterogeneity in the loci studied, and seems to be of greater differentiating power than the skull morphological measurements which were unable to

16 16 Manuel Ruiz-García and Esteban Payan separate animals by geographical subspecies in Colombia (Ruiz-García et al., 2006). The results of several molecular studies suggest the value and power of using molecular tools in taxonomy of mammals. Some kinds of assignment analyses (frequency and Bayesian methods) correctly classified more than 97% jaguars belonging to the corresponded geographical area of P. onca centralis and P. onca onca with only 8 DNA microsatellites. Furthermore, the FCA96, FCA45, FCA391 and the global microsatellite set employed showed significant heterogeneity among the two supposed Colombian jaguar subspecies. This fact does not mean that the DNA microsatellites markers supported the two subspecies of jaguars named by Nelson and Goldman (1933) or Pocock (1939). Rather that the molecular markers indicated significant genetic heterogeneity among jaguars from western and northern Colombia with respect to those from the Eastern and Amazon areas of Colombia. The possibility of genetic heterogeneity inside subspecies remains and differences between jaguars from northern Colombia and Panamá and Costa Rica (supposedly P. onca centralis) should not be ruled out. Differences could even exist between the Llanos and Amazon animals with respect to other Central Brazilian animals when microsatellites are employed (Ruiz-García et al., 2006). Indeed, the detection of Wahlund effect inside the two jaguar subspecies indicates a structure composed of finer genetic units than previously believed. In sum, morphological analysis did not reveal evidence for the two supposed subspecies, whereas the molecular procedures did detect slight but significant differentiation between and within geographical areas, indicating small population units widely connected by gene flow and/or very recent divergence times. The above mentioned studies and data clearly shows different evolution rates among the cranial (morphologic) and molecular characters, with higher variation in the latter. Other aspects on the morphological analysis were as follows. The PcorA with the correlation and the variance-covariance matrices showed four significant coordinates; while the same analysis with the Euclidean and the Manhattan matrices only detected one significant coordinate, which emphasizes the importance of using several different mathematical procedures to obtain a clearer view of the question. The CPA analysis using cranial and facial distance variables separately from the dental, maxilla and mandible variables, yielded different discriminative power. The neurocranial and facial variables showed certain discriminative capacity among different Colombian geographical areas, although not in the expected way (subspecies wise). Contrarily, the teeth, maxilla and mandible variables did not differentiate jaguars from different geographical areas, although the variance explained by the first two canonical axes were relatively similar. Gay and Best (1996) found that in pumas, teeth reach their full-grown size at 2 years of age, so characters without significant variation among age groups are those related to dentition. Dentition characters probably did not discriminate among our jaguars for this reason. Larson (1997) found a considerable overlap of individuals but distinguished a pattern of clinal variation between northern and southern populations of the Americas. We did not detect any significant spatial pattern for the Colombian jaguar skulls because our scale was comparatively smaller than the continent. The Principal Component Analysis showed that the size component is extremely more important that the shape component. In fact, comparisons among different cat species revealed that the first factor is correlated intensively with the skull size of the species (Werdelin, 1983). This same author emphasized that when all skull and dental variables were included, the separation between large and small cats was more intensely marked than when

17 Craniometric Variation in Jaguar Subspecies (Panthera Onca) from Colombia 17 only the lower dentition was employed. This agrees quite well with our dentition variable analysis and its low power of discrimination among individuals. Gay and Best (1996) showed that for pumas 15 out of 19 characters exhibited significant variation among male age classes and 16 out of 19 for female age classes. This explanation could be consistent with our jaguar skull results, implying classification by age (size) rather than by origin. However, in African lions (Panthera leo), skulls dimensions overlap those of older lions shortly after 3 years of age (Smuts et al., 1978). Subspecies revisions and clarifications are also extremely relevant for creating useful conservation units. This is urgently needed for the future preservation of this species and the creation of stronger population areas in countries such as Colombia that lack these types of studies. Molecular population genetics studies are to be undertaken in parallel to morphological biometric ones in order to correctly determine differences. ACKNOWLEDGMENTS The authors strongly thank the Vicerectory of the Pontificia Universidad Javeriana for the approval of grant 905 submitted by the first author to study the molecular genetic structure of the six wild cat species present in Colombia. Thanks also go to the Instituto von Humboldt (IVH) at Villa de Leyva for the use of its facilities, to measure the jaguar skulls and providing the opportunity to take skin samples from its jaguar pelt collection for extracting DNA. These acknowledgments are mainly directed to the curator of the mammalogy collection, Ms. Yaneth Muñoz-Sabas and the directors of IVH, Dr. Cristian Samper and Dr. Fernando Gast. Similarly, the authors express their gratitude to Professor E. Villarraga for permitting access to several skulls from his personal collection and to the Huitoto, Ticuna and Jaguas Indian communities for providing skulls of jaguars across the Colombian Amazon. APPENDIX 1 Subspecies, localities, date of collection and institutional collection number of jaguar skulls measured. Subspecies Department Region Collection Collection Institution/contact number date P. o. centralis Chocó Riosucio 3997* 1974 IAvH P. o. onca Vaupés Miraflores 844* 1970 IAvH P. o. onca Brazil-Peru Javari River 3973* 1975 IAvH P. o. onca Colombia Eastern Llanos Prof. E. Villarraga P. o. onca Colombia Eastern Llanos Prof. E. Villarraga P. o. onca Vaupés Miraflores 1621* 1970 IAvH P. o. onca Amazonas Leticia 1321* 1975 IAvH P. o. onca Vaupés Miraflores 836* 1973 IAvH P. o. onca Vaupés Miraflores 1171* 1970 IAvH P. o. onca Brazil Sao Paulo de Olivenca 3994* 1975 IAvH P. o. onca Vaupés Miraflores 833* * IAvH P. o. onca Guaviare Itilla river 1170* 1971 IAvH P. onca?? IAvH

18 18 Manuel Ruiz-García and Esteban Payan Appendix 1. (Continued) Subspecies, localities, date of collection and institutional collection number of jaguar skulls measured. Subspecies Department Region Collection Collection Institution/contact number date P. onca?? IAvH P. onca?? IAvH P. onca?? IAvH P. o. onca Santander Santander 1999 Dr. M. Ruiz-Garcia P. o. centralis Antioquia Puerto Berrío 830* 1971 IAvH P. o. onca Guaviare Apaporis River 830* IAvH P. o. onca Guaviare Guayabero River 3995* 1971 IAvH P. o. onca Guaviare Tacunema 3996* 1972 IAvH P. o. onca Guaviare Guaviare 4001* 1970 IAvH P. o. onca Amazonas Leticia 854* 1971* IAvH P. o. onca Guainía Puerto Inírida 3999* * IAvH P. o. onca Amazonas Araracuara 843* 1971 IAvH P. o. onca Amazonas Leticia 854* 1971 IAvH P. o. onca Vaupés Yuruparí 1595* 1975 IAvH P. o. onca Arauca Arauca 1873* 1976 IAvH P. o. onca Vaupés Itilla River 837* 1972 IAvH P. o. centralis Bolivar Colorado 3068* 1972 IAvH P. o. onca Guainía Puerto Inírida 4000* * IAvH P. o. centralis Chocó Solano bay IAvH P. o. onca Guainía Puerto Inírida 1172* IAvH P. o. onca Caquetá Caguan River 3998* * IAvH P. o. onca Amazonas Leticia 1320* * IAvH P. o. onca Amazonas Leticia 1319* 1975 IAvH P. o. onca Meta La Macarena 1179* * IAvH P. o. onca Vaupés Miraflores 1620* 1972 IAvH P. o. onca Guainía Isana River 5845* 1975 IAvH P. onca?? IAvH P. o. centralis Chocó Riosucio 1907* 1971 IAvH REFERENCES [1] Anderson, T. W. (1963). Asymptotic theory for principal component analysis. Annals of Mathematic Statistics, 34, [2] Avise, J. C., and R. M. Ball. (1990). Principles of genealogical concordance in species concepts and biological taxonomy. Oxford Surveys of Evolutionary Biology, 7, [3] Bartlett, M. S. (1950). Tests of significance in factor analysis. British Journal of Psychometry, 28, [4] Burnaby, T. P. (1966). Growth-invariant discriminant functions and generalized distances. Biometrics, 22, [5] Cabrera, A. (1940). Notas sobre Carnivoros sudamericanos. Notas Museo de La Plata, Zool. Buenos Aires, 29, 1-22.

19 Craniometric Variation in Jaguar Subspecies (Panthera Onca) from Colombia 19 [6] Cabrera, A. (1957). Catalogo de los mamíferos de América del Sur. I (Metatheria, Unguiculata, Carnívora). Revista del Museo Argentino de Ciencias Naturales Bernardo Rivadavia, Zoología, 4, [7] Cabrera, A., and J. Yepes Mamíferos Sud Americanos. Second 2 nd Ed. Ediar, Buenos Aires 1: [8] Cuadras, C. M. (1989). Distance analysis in discrimination and classification using both continuous and categorical variables. In Statistical data analysis and inference, North Holland, (pp ). [9] Cuadras, C. M. (1991). Métodos de Análisis multivariantes. Promociones y Publicaciones Universitarias, Barcelona. [10] Fernández, E., F. de Lope, and C. de la Cruz. (1992). Morphologie cranienne du chat sauvage (Felis silvestris) dans le sud de la Péninsule ibérique: importance de l introgression par le chat domestique (F. catus). Mammalia, 56, [11] Garcia-Perea, R. (1994). The Pampas cat group (Genus Lynchailurus Severtzov, 1858) (Carnivora: Felidae), a systematic and biogeographic review. American Museum Novitates, 3096, [12] Gay, S. W., and T. L. Best. (1996). Age-Related variation in skulls of the puma (Puma concolor). Journal of Mammalogy, 77, [13] Goldman, E. A. (1932). The jaguars of North America. Proceedings Biological Society Washington, 45, [14] Gower, J. C. (1966). Some distance properties of latent root and vector methods used in multivariate analysis. Biometrika,, 53, [15] Gower, J.C., and G. J. S. Ross. (1969). Minimum spanning trees and single cluster analysis. Applicative Statistics, 18, [16] Hollister, N Two New South American jaguars. Proceedings of the U. S. National Museum, 48, [17] Ibanez, F. (1973). Un test de significativité des composants principales. Annales Institute Oceanografique, 49, [18] Kurten, B. (1973). Pleistocene jaguars in North America. Comment. Biologic. Sci Fennica, 62, [19] Kurten, B., and E. Anderson. (1980). Pleistocene mammals of North America. Columbia University Press, New York. [20] Larson, S. E. (1997). Taxonomic re-evaluation of the jaguar. Zoo Biology, 16, [21] Lachenbruch, P. A. and M. Goldstein. (1979). Discriminant Analysis. Biometrics, 35, [22] Lebart, L., and J. P. Fenelon. (1973). Statistique et Informatique Appliquées. Dunod, Paris. [23] Mahalanobis, P. C. (1936). On the generalized distance in statistics. Proceeding of National Institute of Science India, 2, [24] Marcus, L. F. (1990). Traditional Morphometrics. Pp In Proceedings of the Michigan Morphometrics Workshop. F. J. Rohlf and F. L. Bookstein (Eds.). Special Publication Number 2. The University of Michigan Museum of Zoology. [25] Marks, S. and O. J. Dunn. (1974). Discriminant functions when covariance matrices are unequal. Journal of the American Statistic Association, 69, [26] Martin, W. (1832). On the anatomy of the jaguar (Felis onca L.) Proceedings of the Comm. Science and Corres. Zoological Society of London, Pt., 2, 7-9.

20 20 Manuel Ruiz-García and Esteban Payan [27] Mearns, E. A. (1901). The American jaguars. Proceedings of the Biological Society of Washington, 14, [28] Nelson, E. W., and E. A. Goldman. (1933). Revision of the jaguars. Journal of Mammalogy, 14, [29] O Brien, S. J., and E. Mayr. (1991). Bureaucratic mischief: recognizing endangered species and subspecies. Science, 251, [30] Payán, C. E. and R. Almeida. (2002). Jaguar predation and jaguar hunting in Costa Rica. Cat News, 36, Bougy Villars, Switzerland. [31] Pocock, R. I. (1939). The races of the jaguar (Panthera onca). Novitates Zoologicae, 41, [32] Rao, C. R. (1951). Advanced statistical methods in biometric research. Hafner Publishing Company, Darien. [33] Rohlf, F. J. (1970). Adaptive hierarchical clustering schemes. Systematic Zoology, 19, [34] Rohlf, F. J. (1982). Consensus indices for comparing classifications. Mathematical Biosciences, 59, [35] Ruiz-García, M., Orozco-terWengel, P., Payán, E., and Castellanos, A. (2003). Genética de Poblaciones molecular aplicada al estudio de dos grandes carnívoros (Tremarctos ornatus Oso andino, Panthera onca- jaguar): lecciones de conservación. Boletin de la Real Sociedad Española de Historia Natural. Sección Biológica,, 98 (1-4), [36] Ruiz-García, M., Payán, E., Murillo, A., and Alvarez, D. (2006). DNA Microsatellite characterization of the Jaguar (Panthera onca) in Colombia. Genes and Genetics Systems, 81, [37] Schultz, C. B., L. D. Martin, and M. R. Schultz. (1985). A Pleistocene jaguar from north-central Nebraska. Transactions Nebraska Academy of Sciences, 8, [38] Seymour, K. L. (1983). The Felinae (Mammalia: Felidae) from the late Pleistocene tar seeps at Talara, Peru, with a critical examination of the fossil and recent felines of North and South America. Unpublished M.S. thesis, University of Toronto, Toronto, Ontario. [39] Seymour, K. L. (1989). Panthera onca. Mammalian Species, 340, 1-9. [40] Smuts, G. L., J. L. Anderson, and J. C. Austin. (1978). Age determination of the African lion (Panthera leo). Journal Zoological Society of London, 185, [41] Sneath, P. H., and R. R. Sokal. (1973). Numerical Taxonomy. Freeman. San Francisco [42] Swank, W. G., and J. G. Teer. (1989). Report. National Fish and Wildlife Foundation, Washington, D.C. [43] Templeton, A. R., H. Hemmer, G. Mace, U. S. Seal, W. M. Shields, and D. S. Woodruff. (1986). Local adaptation, coadaptation, and population boundaries. Zoo Biology, 5, [44] Wayne, R. K., and S. M. Jenks. (1991). Mitochondrial DNA analysis implying extensive hybridization of the endangered red wolf Canis rufus. Nature, 351, [45] Weigel, I. (1961). Das fellmuster der wildlebenden katzenarten un der hasuskatze in vergleichender und stammesgeschichtlicher hinsicht. Saugetierkundlichen Mitteilungen, 9, [46] Werdelin, L. (1983). Morphological patterns in the skulls of cats. Biological Journal of the Linnean Society, 19,