Fossil remains of South American tapirs are often fragmentary and scarce compared with those of other mammals that entered South America during the “Great American Biotic Interchange”. Here, we review and add to the Pleistocene tapir remains from the Tarija Valley (Bolivia), and provide a taxonomic re-evaluation of Tapirus tarijensis. T. tarijensis was a large-sized animal, approximating the size of the living Malaysian T. indicus and the extinct North American T. haysii. The geographical distribution of Pleistocene records of Tapirus in South America indicates that T. tarijensis was the only known species inhabiting the Tarija Valley during this time.
Introduction
Fossil tapirs (Perissodactyla: Tapiridae) are known from Europe, North America, South America and Southeast Asia, including China (e.g., Cerdeño and Ginsburg 1988; Hulbert 2005; Tong 2005; Ferrero and Noriega 2007; Holanda et al. 2011, Medici 2011, Scherler et al. 2011). In South America, the family is only represented by Tapirus Brisson, 1762, with the oldest reliable records of the genus dating to the Early-Middle Pleistocene of the Pampean region of Argentina (see Tonni 1992; Soibelzon 2005; Soibelzon et al. 2008). All of the living tapirs occurring in South America today (T. bairdii [Gill, 1865]; T. pinchaque [Roulin, 1829] and T. terrestris [Linnaeus, 1758]) also belong to this genus.
Although remains of Tapirus are often fragmentary and scarce compared with other mammalian migrants that took part in the “Great American Biotic Interchange” (Woodburne et al. 2006; Webb 2006), the fossil record of South American tapirs has substantially improved in recent years, and now comprises material from Argentina, Brazil, Peru, Uruguay, and Venezuela (e.g., Ubilla 1983, 2004; Marshall et al. 1984; Hoffstetter 1986; Tonni 1992; Ferrero and Noriega 2003, 2007; Holanda and Cozzuol 2006; Ubilla and Rinderknecht 2007; Holanda et al. 2007, 2011; Ferrero et al. 2007, 2009; Holanda and Rincón 2012).
In Bolivia, finds of fossil tapirs are restricted to the Tarija Valley (Fig. 1). The first tapirid specimen from this locality was described by Ameghino (1902), who erected the new species Tapirus tarijensis based on a poorly preserved mandibular fragment bearing pm2-pm4 in situ. Subsequent studies generally followed this assignment (e.g., Hoffstetter 1963; Marshall and Sempere 1991; Takai et al. 1982, 1984), but did not discuss the diagnostic features of this species, with the exception of its relatively large size (Boule and Thevenin 1920).
Here, we provide a detailed morphological description of the holotype and new referred material of Tapirus tarijensis, and discuss its diagnostic features and taxonomy. In addition, we describe some additional material referable to Tapirus sp. from the Tarija Valley, and briefly comment on the geographical distribution of the genus during the Pleistocene-Holocene of South America.
Institutional abbreviations.—AMNH, American Museum of Natural History, New York, USA; CICYTTP, Colección Paleontológica, Centro de Investigaciones Científicas y Transferencia de Tecnología a la Producción, Diamante, Argentina; CRK, Coleção de Referência Zoológica Renato Kipnis, Laboratório de Estudos Evolutivos e Humanos, Universidade de São Paulo, São Paulo, Brazil; F:AM, Frick Fossil Mammal Collection, American Museum of Natural History, New York, USA; MACN-M, Colección de Mastozoología del Museo Argentino de Ciencias Naturales “Bernardino Rivadavia”, Buenos Aires, Argentina; MACN-PV, Colección Paleontología Vertebrados del Museo Argentino de Ciencias Naturales “Bernardino Rivadavia”, Buenos Aires, Argentina; MCN, Museu de Ciências Naturais, Fundação Zoobotânica do Estado do Rio Grande do Sul; MHN, Museu de História Natural, Universidade Federal de Minas Gerais, Belo Horizonte, Brazil; MLP, Museo de La Plata, La Plata, Argentina; MNHN, Museo Nacional de Historia Natural de Montevideo, Uruguay; MNHN-TAR, Museum nationnational d'Histoire naturelle de Paris, France, Tarija Collection; MNPA-V, Museo Nacional Paleontológico-Arqueológico, Universidad Autónoma Juan Misael Saracho, Tarija, Bolivia; MNRJ, Museu Nacional do Rio de Janeiro, Rio de Janeiro, Brazil; MZUSP, Museu de Zoologia, Universidade de São Paulo, São Paulo, Brazil; UNIR, Universidade Federal de Rondônia, Porto Velho, Brazil; UF, Florida Museum of Natural History, Gainesville, USA; UFMG, Universidade Federal de Minas Gerais, Belo Horizonte, Brazil.
Other abbreviations.—PCA, Principal Component Analysis.
Material and methods
The present analysis is mainly based on specimens collected from 1978 to 1980 by researchers of The Research Institute of Evolutionary Biology located in Tokyo, Japan (see Takai et al. 1982, 1984), as well as material that arose from expeditions organized by staff of the University of Florida, USA (Mac-Fadden et al. 1983, 1994; MacFadden and Shockey 1997; MacFadden 2000). Measurements of teeth and metapodial elements were obtained using calipers, following Simpson (1945) and Hulbert (2005), and are reported in Tables 1-4 and the Supplementary Online Material (SOM available at http://app.pan.pl/SOM/app59-Ferrero_etal_SOM.pdf). Where sample size permitted, the material from Tarija was compared with T. terrestris using Student's t-tests, performed in PAST v. 2.16 (Hammer et al. 2001). In addition, we used Principal Component Analysis (PCA) to quantify the observed variation in the dental measurements, and compare T. tarijensis with a range of other living and fossil tapirs, including T. indicus, T. haysii, T. johnsoni, T. rondoniensis, T. simpsoni, T. terrestris, T. veroensis, and T. webbi (Fig. 6, Table 5). T. cristatellus, T. greslebini, T. mesopotamicus, and T. rioplatensis could not be included in our PCA owing to their relatively poor or incomplete state of preservation.
Geological setting
The age of the sediments from the Tarija Valley is controversial (Tonni et al. 2009; Soibelzon et al. 2011). Despite discrepancies regarding the chronology of the fossil-bearing sediments, and especially the disparity between fossil-bearing localities and dated stratigraphic columns, some authors (e.g., Takai et al. 1982, 1984; Alberdi and Prado 2004) assigned the entire fauna exclusively to the Ensenadan (Early Pleistocene) of the Pampean regional time scale (see Fig. 1 in Soibelzon et al. 2009) or to the “Ensenadan Land Mammal Age” (Marshall et al. 1984; MacFadden 2000), although Marshall et al. (1984: 33) mentioned the possibility “that some fossils from this locality are younger”. By contrast, radiocarbon dating of organic levels from several different localities suggests a Late Pleistocene age (between 27180 and 39000 14C BP) for the fossiliferous strata (Coltorti et al. 2007). Following Tonni et al. (2009), we consider the mammalian fauna of Tarija to represent neither exclusively the Ensenadan Stage/Age (Early to Middle Pleistocene) nor the Lujanian Stage/Age (Late Pleistocene to early Holocene), as it contains several taxa typical of either: at least 3 species only occurring in the Ensenadan Mesotherium cristatum Biozone and 10 species typical of the Lujanian Equus (Amerhippus) neogaeus Biozone (see Tonni et al. 2009 and bibliography cited therein).
Systematic paleontology
Class Mammalia Linnaeus, 1758
Order Perissodactyla Owen, 1848
Family Tapiridae Burnett, 1830
Genus Tapirus Brisson, 1762
Type species: Hippopotamus terrestris Linnaeus, 1758; Pernambuco, Brazil, Oligocene-Recent.
Tapirus tarijensis Ameghino, 1902
Fig. 3A.
Lectotype: MACN-PV-1523, left mandible with pm2-pm4 (Ameghino 1902: 247–248, pl. 5: 22a, b).
Type locality: Tarija Valley, Bolivia (Fig. 1).
Type horizon: Pleistocene, “Pampean formation” (sensu Carles 1888).
Referred material.—MNPA-V 006038, palatal fragment bearing PM1-M2 and an unerupted M3 (Fig. 3B); MNHN-TAR 843, partial left maxilla bearing PM2-M2 (Fig. 3C); MNHN-TAR 842, left mandible with dpm2-dpm4, m1 (Fig. 3D); MNHN-TAR 847, partial right mandible with fragmentary m1 and complete m2-m3 (Fig. 3E); MNHN-TAR 846 partial left mandible with m1-m2 and broken m3 (root) (Fig. 3F); MNPA-V485, partial right mandible with pm2-pm3 (Fig. 4A); MNPA-V1437, partial right mandible with pm2-m2 (Fig. 4B); MNPA-V1448, partial left mandible with m2?-m3? (Fig. 4C); MNPA-V1446, partial right mandible with pm3-m1 (Fig. 4D); MNPA-V1447, partial left mandible with pm4-m2 (Fig. 4E); MNPA-V1445 partial right mandible with m2? (Fig. 4F); MNPA-V 5942, partial mandible of a juvenile specimen with some molariform teeth and an unerupted left m1?, right m2?, left pm3-pm4, right pm3, mandibular symphysis with right i1-i2 and left i1-i3 (Fig. 4G); MNHN-TAR 844, partial left mandible with m3 (Fig. 5A); MNHN-TAR 845, right mandible with pm4-m3 (Fig. 5B); UF 91589, partial left mandible with partial pm4, m1-m3, mandibular symphysis, and partial right mandible bearing m2-m3 (Fig. 5C); AMNH 55999, left mandible with pm2-m3. Tarija Valley (Bolivia), Pleistocene.
Emended diagnosis.-Differs from all extant tapirs in having an accessory cusp on the lingual side of the paraconid of dpm2, as well as a pm2 characterized by a metalophid (or oblique cristid) originating at the base of the hypoconid and extending anteriorly to the base of the protoconid in an almost straight line, as well as a trapezoidal lingual interlophid valley; differs from T. mesopotamicus, T. rondoniensis, and extant South American tapirs in its larger size, and from T. rioplatensis, T. mesopotamicus, and extant South American tapirs in its more robust and dorsoventrally higher mandible; shares a well-developed cingulid on pm3 with T. rioplatensis, T. veroensis, T. haysii, and T. lundeliusi, but not T. webbi, T. jonhsoni, and T. polkensis; resembles T. indicus, T. haysii, T. oliverasi, T. cristatellus, T. greslebini, and T. rioplatensis in terms of size; shares with all other tapirs a mesial cingulid better developed on pm3 than on pm4; resembles all other tapirs except T. greslebini in the morphology of the upper molars.
Description.—The lectotype comprises a partial left mandible bearing pm2-pm4 (Figs. 2A, 3A; SOM: table S1). The mandible is robust and every tooth is worn. The pm2 is broken anterior to the protoconid (the paralophid and paraconid are not preserved). In occlusal view, the metaconid and the protoconid are transversely aligned (Fig. 2A), unlike in other tapirs, in which the protoconid is located more mesially (Fig. 2B). The metaconid is developed as a strong, high cusp. The metalophid (or oblique cristid) originates at the base of the hypoconid and extends anteriorly to the base of the protoconid in an almost straight line, thus resulting in a trapezoidal lingual interlophid valley (Fig. 2A). By contrast, the metalophid follows an oblique route and terminates between the protoconid and the metaconid in other tapirs (Fig. 2B). The pm3 is worn and broken at the level of the protolophid. The hypoconid is not preserved. The mesial cingulid is prominent and better developed than in extant South American tapirs. The pm4 is the best preserved tooth, and bears a mesial cingulid less developed than that of pm3, as also seen in T. terrestris. The posterior border of pm4 bears a poorly developed distal cingulid, showing a subtle slope where its dorsalmost portion would have approximated the base of the hypoconid.
Comparative description.—Most of the studied specimens represent incomplete cranial and mandibular fragments. Judging from the dental measurements (Tables 1–3; Fig. 6; SOM), Tapirus tarijensis is larger than T. mesopotamicus, T. rondoniensis and all of the extant South American tapirs, slightly smaller than T. rioplatensis, and at least partially overlaps with the size range of T. cristatellus, T. greslebini (see Holanda and Ferrero 2013), T. oliverasi, and T. indicus. Its PM1 resembles that of other species of Tapirus (except T. pinchaque) in being subtriangular in occlusal view and in bearing a well-developed hypocone. The PM2 is molariform and characterized by a well-developed protocone and hypocone. As in all living and fossil American tapirs except T. pinchaque and T. rondoniensis, the protoloph reaches the base of the ectoloph. The parastyle is poorly developed on PM2, better developed on PM3-PM4, and strongly developed on M1 and especially M2, similar to all other American tapirs. It is presently unclear whether a well-developed parastyle also occurs on M3, since the latter is so far only known from unerupted teeth (e.g., MNPA-V 006038); however, it is likely that M3 resembles M2 in this regard. As in other tapirs, there are no lingual cingula on any of the cheek teeth. The morphology of the upper molars resembles that of other species of Tapirus except T. greslebini, which is uniquely distinguished by a mesial subcingulum at the base of the parastyle, a well-marked labial cingulum between the distal wall of the paracone and the labiodistal portion of the metacone, and a small lingual cingulum between the protocone and the hypocone (Holanda and Ferrero 2013).
The lower border of the mandibular body is somewhat concave (MACN-PV 1523, MNPA-V 1447, 1437) or straight (AMNH 55999, MNPA-V 1446). The mandible of the juvenile specimen (MNHN-TAR 842) is robust and approximately 20% higher (at the level of m1) than that of T. terrestris at a comparable ontogenetic stage (Table 3). The dpm2 is rectangular in occlusal view, relatively robust and, unlike in any other species of Tapirus, bears an accessory cusp on the lingual side of the paraconid, with the two being connected by a relatively low, oblique crest. An accessory crest on dpm2 also occurs in other perissodatyles, such as rhinoceroses, in which it, however, corresponds to a bifurcation of the paralophid (Cerdeño and Sánchez 2000: 284). The dpm3 and dpm4 of T. tarijensis bear well-developed mesial cingulids, as also seen in deciduous teeth of T. terrestris. Only two specimens (MNPA-PV-485 and AMNH 55999) preserve a complete pm2, which is characterised by a more robust talonid than in the extant South American species. No pm2 has been found for any of the other South American fossil tapirs (T. mesopotamicus, T. rondoniensis, T. cristatellus, T. oliverasi, T. greslebini, and T. rioplatensis).
The pm3 of T. tarijensis (MNPA-V485, 5942, 1437, 1446) resembles that of T. rioplatensis, T. veroensis, T. haysii, and T. lundeliusi in having a well-developed, mesiolingually extending cingulid originating at the level of the protoconid. By contrast, the mesial cingulid is less developed and originates at a more ventral position relative to the protoconid in all of the living and some fossil North American species, such as T. webbi, T. jonhsoni, and T. polkensis. No pm3 has yet been found for T. mesopotamicus, T. oliverasi, T. greslebini, T. rondoniensis, and T. cristatellus.
The pm4 is rectangular in occlusal outline and bears a well-developed cingulid, although the latter is less sharp and lower than that of pm3. This tooth is nearly indistinguishable in its morphology from that of other fossil and extant South American tapirs. Similarly, the lower molars of T. tarijensis generally resemble those of other members of the genus. Stratigraphic and geographic range.—Pleistocene, Tarija Valley, Bolivia.
Table 1.
Summary statistics for the upper teeth of Tapirus terrestris (see Supplementary Online Material) and measurement values for Tapirus tarijensis. Abbreviations: AW, greatest anterior width; CV, coefficient of variation; L, length; MAX, maximum; MIN, minimum; N, sample size; PW, greatest posterior width; s, standard deviation; x, sample mean.
Table 2.
Summary statistics for the permanent lower teeth of Tapirus (see Supplementary Online Material). The two rightmost columns in Tapirus tarijensis show the results of a two-tailed t-test for significant differences in relation to Tapirus terrestris. Abbreviations: AW, greatest anterior width; CV, coefficient of variation; (D), deciduous tooth; L, length; MAX, maximum; MIN, minimum; N, sample size; ns, not significant (p >0.05); PW, greatest posterior width; s, standard deviation; x, sample mean. Values marked with an asterisk are from Perini et al. (2011).
Table 3.
Summary statistics for the deciduous premolars and the mandibular height of Tapirus terrestris (see Supplementary Online Material), and measurement values for Tapirus tarijensis. Abbreviations: AW, greatest anterior width; CV, coefficient of variation; Hm, dorsoventral height of the body of the mandible, measured at the level of m1; L, length; MAX, maximum; MIN, minimum; N, sample size; PW, greatest posterior width; s, standard deviation; W, width; x, sample mean.
Table 4.
Summary statistics for the metapodials of Tapirus terrestris (see Supplementary Online Material), and measurement values for Tapirus sp. from Tarija and Tapirus mesopotamicus. Abbreviations: CV, coefficient of variation; DDP, dorsopalmar diameter of proximal portion of metacarpal and dorsoplantar diameter of metatarsals; L, length; MAX, maximum; MDB, mediolateral diameter of the body; MDD, mediolateral diameter of distal portion; MDP, mediolateral diameter of proximal portion; MIN, minimum; N, sample size; s, standard deviation; x, sample mean.
Tapirus sp.
Fig. 5D-F.
Referred material.—MACN-PV-604, left metacarpal III (Fig. 5D); MNHN-TAR 849, left metacarpal III (Fig. 5F); MNHN-TAR 848, left metatarsal IV (Fig. 5G); MNHN-TAR 850 right metatarsal IV (Fig. 5E) from the Tarija Valley, Bolivia, “Pampean formation” (sensu Carles, 1888).
Remarks.—All of referred elements are robustly built, but too worn to discern details of their morphology. Because they were not clearly associated with any dental material, these remains cannot be confidently assigned to Tapirus tarijensis.
Morphometric analyses
In terms of the lower tooth measurements, the material assigned to T. tarijensis is on average significantly larger than T. terrestris (Student's t-test, p <0.001), with the exception of the width of the anterior and posterior lophids of p3 and p4 (Table 2). Small sample sizes prevented us from conducting statistical tests on the lower deciduous and upper tooth data, but the latter still generally seem to corroborate the size difference between the two species. Thus, although MNPA-V 006038 overlaps in its size with the largest specimens of T. terrestris, MNHN TAR-843 is 10–20% larger according to all measurements (Table 1). Similarly, MNHN-TAR 842 generally exceeds the size of the largest specimens of T. terrestris by 10–21% (Table 3).
In all of the Principal Component analyses, the first two principal components account for more than 80% of the total variation, with PC1 reflecting size in all cases (Fig. 6). The PCA of the upper teeth (Fig. 6A) revealed three groupings, comprising (i) T. terrestris, T. rondoniensis, T. johnsoni, and some specimens of T. veroensis; (ii) MNPA-V 006038 (referred to T. tarijensis), T. veroensis, T. simpsoni, T. webbi, and T. indicus; and (iii) the largest specimens of T. indicus, T. tarijensis (MNHN-Tar 843), and T. haysii.
The lower tooth data were analyzed using three separate PCAs. In the first analysis (Fig. 6B), which includes all of the measurements preserved in the lectotype of T. tarijensis (MACN-PV-1523), T. terrestris, T. lundeliusi, T. johnsoni, and T. webbi broadly overlap, with MACN-PV-1523 occupying an intermediate position between them and the larger species T. veroensis, T. indicus, and T. haysii. Although the lectotype falls close to the larger specimens of T. terrestris specimens, the two species do not overlap. Another specimen assigned to T. tarijensis (AMNH 55999) clusters with T. haysii, and indeed represents one of the largest specimens in the analysis; however, it should be noted that the data for this specimen were obtained from a cast. The second PCA of the lower teeth (Fig. 6C) is based on m1 and m2 only, and distinguishes a group comprising T. terrestris, T. lundeliusi and T. johnsoni from a second cluster including T. tarijensis along with large specimens of T. veroensis, T. webbi, T. indicus, T. oliverasi, and T. haysii. A similar pattern is evident in the third analysis (Fig. 6D), which focuses only on m3 and clusters T. terrestris, T. mesopotamicus, T. lundeliusi, some specimens of T. veroensis and some of T. webbi on the one hand, and T. tarijensis, T. haysii, and T. oliverasi on the other.
Table 5.
Factor loadings for the first two principal components arising from the Principal Component analyses of the lower and upper teeth. Abbreviations as in Table 1.
In terms of the postcranial remains, 8 of our 17 metapodial measurements fall outside the size range of T. terrestris (Table 4). Thus, MNHN-TAR 849 is about 18% longer and 30% wider than the largest metatarsal III measured for T. terrestris. Similarly, MNHN-TAR 848 exceeds T. terrestris by 10% in three of the four measurements of metatarsal IV. Moreover, MNHN-TAR 849 and 848 are about 10% and 25% larger, respectively, than the equivalent elements of T. mesopotamicus (Table 4).
Discussion and conclusions
Morphology and taxonomy .—The remarkably small degree of variation in dental morphology across different species of Tapirus is well known (Simpson 1945; Hershkovitz 1954; Ray and Sanders 1984; Hulbert 1995, Holanda and Cozzuol 2006; Ferrero and Noriega 2007), and has posed a considerable challenge to the definition of diagnostic characters and, thus, the specific assignment of often fragmentary fossil material. The extant tapirs (T. terrestis, T. pinchaque, T. bairdii, and T. indicus) clearly differ in their cranial morphology from South American fossil species, such as T. mesopotamicus and T. rondoniensis, but, with the exception of the large T. indicus, overlap with the latter in terms of their tooth size. In his description of T. tarijensis, Ameghino (1902) listed high lophids on the lower molars and a high mandibular body with a concave lower border as diagnostic characters of this species. He furthermore commented that T. tarijensis might be slightly larger than T. americanum (= T. terrestris), but did not provide any data on differences in tooth size. Later, Boule and Thevenin (1920) interpreted all of the remains from Tarija as those of a large tapir resembling T. terrestris, and assigned them to T. cf. americanum (= T. cf. terrestris), rather than T. tarijensis.
Based on our comparisons of the lectotype of T. tarijensis (MACN-PV-1523) with the other specimens from Tarija, we consider a concave lower border of the mandibular body to be the result of individual variation, rather than a species-diagnostic character. However, the existence of T. tarijensis as a separate species is supported by several unique traits, as detailed above in the emended diagnosis. This conclusion is further corroborated by our morphometric analyses of the teeth, which show T. tarijensis to be significantly larger than T. terrestris, T. mesopotamicus, and T. rondoniensis, and similar in size to other large living and fossil tapirs, such as T. indicus, T. oliverasi, and T. haysii. Equally, the metapodials of Tapirus sp. from Tarija are clearly more robust that those of T. terrestris and T. mesopotamicus.
Ray and Sanders (1984) suggested that North American tapirs usually fall into two size groups, including (i) those species similar in size to T. terrestris and (ii) those species larger than T. terrestris. A similar pattern occurs among South American tapirs, with T. tarijensis belonging to the larger size group including T. greslebini, T. oliverasi, T. rioplatensis, and, in some cases, T. cristatellus (Holanda and Ferrero 2013).
Biogeographic distribution.—Based on available data, T. tarijensis seems to have been endemic to the Tarija Valley, Bolivia. However, it is possible that the species has escaped detection in other regions of South America owing to the difficulty of distinguishing species based on their teeth and the fragmentary nature of much of the described material. Thus, for example, poorly preserved material currently makes it difficult to distinguish T. oliverasi from both T. tarijensis and T. rioplatensis (Holanda and Ferrero 2013). Several other taxa previously proposed to be endemic to Tarija were later also found in the Pampean region of Argentina (e.g., Scelidodon tarijensis, Arctotherium tarijense, A. wingei; Tonni et al. 2009; Soibelzon et al. 2011), thus adding to an extensive list of species with a wide geographic distribution across several parts of South America (e.g., Alberdi and Prado 1992, 1993; Soibelzon 2004; Prevosti 2007; Gasparini et al. 2009; Tonni et al. 2009, Zurita et al. 2009; Soibelzon et al. 2011). In addition, it is possible that T. terrestris may have been distributed alongside T. tarijensis in Bolivia, resembling the sympatric occurrence of the three extant South American species in Colombia, or the co-occurrence of T. cristatellus and T. terrestris in Brazil (Cartelle 1999). Testing such ideas must await the discovery of more informative material.
Acknowledgements
We thank Jorge I. Noriega (CICYTTP-Conicet) for his valuable contribution and comments, Noelia Nuñez Otaño (CICYTTP-Conicet) for making the drawings, and Alejandro G. Kramarz, David Flores (both MACN), Christian de Muizon (MNHN), Cástor Cartelle (Pontifícia Universidade Católica de Minas Gerais, Belo Horizonte, Brazil), Mariano Merino (MLP), Freddy P. Rios (MNPA) and Richard C. Hulbert (UF) for access to material under their care. We furthermore thank to Alessandra D. Boos (Universidade Federal do Rio Grande do Sul, Porto Alegre, Brazil) for language revision, Esperanza Cerdeño (Instituto Argentino de Nivología, Glaciología y Ciencias Ambientales-Conicet, Mendoza, Argentina) for providing useful bibliography, and Daniel Perea and Martín Ubilla (Universidad de la República, Montevideo, Uruguay) for comments which helped to improve the paper. This study was partially funded by the Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET), the Universidad Nacional de La Plata, (both Argentina), and the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq), Brazil. This paper is a contribution to the following projects: PIP 886; PICT-ANPCYT 804 and 392; and PICTO 164.