First fossil small-clawed otter , Amblonyx , with a note on some specimens of Lutra , from the Upper Siwaliks , India

Otter-like mammals have been reported from the Indian sub-continent for more than a century and a half. Enhydriodontini are far more common and have a longer fossil record than either Lutrini or Aonychini. This paper describes and interprets two fossils from the Pleistocene of Panchkula and Pipalwala, India. The Panchkula specimen represents a large species of Amblonyx (Aonychini) whereas the Pipalwala specimen is likely to represent a large individual of Lutra palaeindica (Lutrini).


Introduction
The recovery of lutrine otter fossils is a rare event in the Indian subcontinent in contrast to the remains of enhydriodont otters, which have been found on a more regular basis.More than a century passed between the discovery of the skull and mandible of Lutra palaeindica Falconer (1868) and the Pipalwala specimen (Verma et al., 2003) which probably represents the same taxon.We here report the discovery of an aonychine otter in the Pinjor Formation of the Siwalik Group, the first of this tribe known from the subcontinent.

Geological context and faunal association
The Siwalik Group of freshwater molasse sediments (~6000 m) spanning the period ~18-0.5 Ma, developed in response to the rising Himalayas to the north (fig. 1) and sinking Foreland Basin to the south (Tandon, 1991) (fig.1).The Siwalik Group is subdivided into the Lower, Middle and Upper Siwalik Subgroups based on faunal assemblages and lithofacies criteria (Pilgrim, 1913;Prakash et al., 1980).The Upper Siwalik Subgroup is divided into Tatrot (~5.26 to ~2.5 Ma), Pinjor (~2.5 to ~1.7 Ma) and Boulder Conglomerate Formations (~1.7 to First fossil small-clawed otter, Amblonyx, with a note on some specimens of Lutra, from the Upper Siwaliks, India ~0.5 Ma).The Pinjor and Boulder Conglomerate Formations are well exposed along the Ghaggar River Section which is located northeast of Chandigarh near the town of Panchkula (fig.2).Tandon & Kumar (1985) reported a volcanic ash bed in the Ghaggar river section (fig.2b), which was dated to around 2.14 ± 0.5 Ma on the basis of fission track dating (Mehta et al., 1988).Recently, Kumaravel et al., (2005) carried out a comprehensive palaeomagnetic stratigraphy of the Ghaggar River section spanning a thickness of ~ 1000 m aged between 2.7 Ma and 0.5 Ma (fig.2b, c).The Upper Siwalik sequence in this section includes a small outcrop of the Tatrot Formation (slightly over 60 m thick) near Khetpurali, characterized by the presence of fine, medium and coarse-grained, grey sandstones and variegated mudstones deposited in low sinuosity streams, mainly of a trunk river system (Kumaravel et al., 2005).The Tatrot fauna of this area comprises Stegodon bombifrons, Pentalophodon khetpu-ralensis, Hipparion antelopinum, Hexaprotodon sivalensis, Hipparion theobaldi, and Camelus sivalensis (Sahni & Mitra, 1980;Nanda, 1978).
Pinjor is the only type locality of the Siwalik Group in India: the rest of the type localities are in Pakistan.In the Ghaggar River Section, the Pinjor Formation is about 380 m thick (basal part of section in fig.2b), and consists of brown to greyish-brown, fine, medium and coarse-grained sandstones, multistorey sandstones, pebbly sandstone, pedogenic and non-pedogenic overbank facies deposited in high gradient, low sinuosity streams, mainly of piedmont drainage (Kumaravel et al., 2005).Pilgrim (1913) characterized the Pinjor zone by the first appearance of Equus and Camelus.Other taxa known from this region are Dilatomys sp., Tatera pinjoricus, Mus linnaeusi, Stegodon insignis, Archidiskodon planifrons, Elephas hysudricus, Equus sivalensis, Leptobos and Bos (Patnaik, 2003;Nanda, 1978).The overlying (~575 m thick) Boulder Conglomerate Formation is largely unfossiliferous and was deposited mainly in braided river channels and proximal alluvial fans (upper part of section in fig.2b).
Age: Late Pliocene ca 2.14 Ma on the basis of stratigraphy, fission track dating and magnetostratigraphy (fig.2b).

Description
The mandible (fig.3) contains the right m/1 in light wear.In the trigonid, the metaconid is slightly behind the level of the protoconid.The crests joining the metaconid and the protoconid are separated by a notch in the midline of the tooth.The paraconid is far from the metaconid, leaving a V-shaped lingual opening in the trigonid basin.There is a deep notch between the paraconid and protoconid.The post-protoconid crest is low but sharp and is separated from the hypoconid by a narrow transverse incision.This crest, and the hypoconid behind it, are positioned some distance lingually of the buccal margin of the crown.Behind the hypoconid, there is a small cusp, now confluent with the hypoconid due to wear, which swings buccally at its distal extremity.The buccal cingulum is continuous from anterior to posterior and anteriorly it extends onto the lingual aspect of the paraconid.The talonid basin slopes bucco-lingually and is bordered posterolingually by a low rim forming the lingual talonid cusp.Distally this cusp is separated from the posterior hypoconid crest by a shallow valley.The trigo-   Tandon, 1985).B-D) Lithostratigraphic column, magnetostratigraphy and correlation of Ghaggar River Section with the GPTS (modified from Kumaravel et al., 2005).G1-G3 fossil sites.nid is marginally narrower (9.2 mm) than the talonid (10.0 mm).
The remnant of p/4 in the right mandible shows a buccal cingulum, but the tooth is too poorly preserved to reveal details of the crown morphology although its breadth can be estimated (table 1).The m/2 is uniradiculate, and its crown would have been higher than the talonid of the m/1.The masseteric fossa is deep and extends as far forwards as the m/2.
The fragment containing the canine and a premolar are poorly preserved, but reveal that the anteriormost premolar lies behind the canine and not lingual to it.However, the jaw is poorly preserved in this part and difficult to interpret.
The overall shape of the crown of M1/ appears to have been losenge-shaped, but it is so poorly preserved that it is not possible to provide measurements or details of the morphology of the crown except that it shows a wedge-or dune-shaped protocone behind which is a broad hypocone (table 1).

Discussion
The specimen from Panchkula differs from Lutra lutra by the relatively broad talonid of its m/1 in which the post-protoconid crest and hypoconid are positioned some distance from the buccal cingulum, leaving a narrow shelf between these cusps and the buccal edge of the tooth.The post-hypoconid crest swings buccally at its distal extremity, which is unusual among otters (fig.3) many of which have the crest swinging lingually.This morphology of the hypoconid complex is important to understand, since this part of the tooth occludes with the metacone-metastyle complex of the P4/, which consequently ought to be displaced more buccally in this species than it is in other otters.The buccal cingulum is also complete.
In all these features and in the details of crown shape and disposition of cusps, the m/1 in the Panchkula otter resembles those of the extant Asian small-clawed otter, Amblonyx cinereus and the Clawless otter of Africa, Aonyx capensis.Bivariate plots of the tooth and those of Lutra lutra, Aonyx capensis and Amblonyx cinereus show that the Panchkula fossil plots within the range of metric variation of Aonyx capensis (fig.5) confirming the broad outline of the lower first molar relative to its length.The m/1 of Lutra aonychoides from Yushe, China, (Teilhard de Chardin & Leroy (1945) also falls within the range of metric variation of Aonyx capensis but it is a bigger specimen than the Panchkula example.
Because the Panchkula tooth is closer morphologically to those of Amblonyx cinereus than to those of Aonyx capensis we conclude that it belongs to this genus, but since it clearly represents a much larger animal, we create a new species for it, Amblonyx indicus.The Panchkula otter is smaller than the specimen of Lutra aonychoides from Yushe, China (Teilhard de Chardin & Leroy, 1945).
Lutra aonychoides from Yushe and Locality 49, China (Teilhard de Chardin & Leroy, 1945;Zdansky, 1924) is significantly larger (m/1 = 19 x 11.5 mm) than the Panchkula otter (m/1 = 17.2 x 9.9 mm), but the proportions of the talonid to trigonid in the m/1, and the detailed morphology of the cusps indicate that the two fossils are closely related.Neither of the specimens falls near the extrapolated regression line of Lutra, but close to the Aonyx -Amblonyx regression and it is considered likely that both these specimens belong to Amblonyx rather than to Lutra.
The fossil species Lutrogale palaeoleptonyx (Dubois, 1908) has a relatively broad m/1 (fig.5) which is close in size to that of Amblonyx indicus described here.Von Koenigswald (1933Koenigswald ( , 1940) ) did not illustrate the m/1 from Sangiran, Java, so a suitable comparison cannot be made with the Indian specimen.Willemsen (1986) examined the question, and concluded that the Javan fossil belonged instead to a species close to Lutrogale perspicillata.Given that teeth of Lutrogale perspicillata, the extant Smooth-coated otter, tend to be narrower relative to length than those of Lutra lutra (fig.5) (Pohle, 1919;Willemsen, 1980Willemsen, , 1986)), then it is more likely that the Javan specimen belongs to Amblonyx or another Aonychini, rather than to Lutrogale, a Lutrini.
Other fossil otters that have a relatively broad talonid in the m/1 are Tyrrhenolutra helbingi (Hürzeler, 1987) from Italy, Cyrnaonyx antiqua (Helbing, 1935) from various sites in Europe, and Limnonyx species from Spain and other parts of Europe (Crusafont Pairo, 1950;Nordmann, 1858) (fig.5).None of them are as large as the Indian fossil, all plotting close to Amblonyx cinereus.Megalenhydris barbaricina from Sardinia was interpreted by Willemsen & Malatesta (1987) to be an Aonychini, but the length-breadth proportions of the m/1 are close to those of Lutra lutra and widely divergent from the regression of the aonychine group (fig.5) and in terms of m/1 proportions, it seems to be an enlarged version of Lutra.
Lutra bravardi has been approached to Aonyx by some authors (Kurten, 1968;Pohle, 1919) on account of the relatively broad P4/, but Willemsen (1992) considered that it fell comfortably within the range of morphometric variation of the genus Lutra.Direct comparisons cannot be made with Amblonyx indicus as the latter species shares no parts in common with Lutra bravardi, which is known only by upper teeth, and possibly a tibia.
The body weight of Amblonyx indicus would have been appreciably greater than that of Amblonyx cinereus which weighs about 3.2 kg (Larivière, 2003;Timmis, 1971), and judging from the dimensions of the m/1 it would have fallen near the lower limit of the range of body weight variation of Aonyx capensis, that is ca 11-12 kg (Larivière, 2001).
SFP 238 is the snout of a large otter (Verma et al., 2002) retaining damaged left P3/ and P4/, and the roots or partial to complete alveoli of the other teeth (fig.4).The intermaxillary suture is not yet fused, despite the fact that the individual had the M1/ erupted at the time of death, judging by the sediment-filled alveolus behind the P4/ (fig.4).There are differences in the lingual half of the P4/, from that of the holotype of Lutra palaeindica, the protocone being slightly more distally positioned relative to the anterior margin of the tooth.The parastyle is strongly developed, but the paracone and metacone are broken.Measurements are provided in table 2.
The P3/ in SFP 238 is a robust tooth, lacking the apex due to damage.It is narrower anteriorly than posteriorly, and is completely surrounded by a cingulum.The protocone has a prominent post-protocone crest and the tooth is positioned in the palate such that this crest is in line with the inner edge of the parastyle of the P4/.The same disposition seems to have occurred in the holotype in which the P3/ itself is missing, but in which the alveoli are well preserved.
The P1/ alveolus and that of the anterior root of the P2/ are located lingually to the canine alveolus as in the holotype of L. palaeindica (table 2).

Discussion
The Pipalwala otter snout SFP 238 (20833) was briefly described and illustrated by Verma et al., (2002).It is appreciably larger than the holotype of Lutra palaeindica, but its P4/ plots out close to the large end of the range of variation of Lutra lutra (fig.5).Since the P4/ in the holotype of Lutra palaeindica plots close to the centre of the scatter of points of Lutra lutra, it is not improbable that the two Siwalik specimens belong to a single species.An alternative identification would be with the genus Lutrogale (fig.6), but the specimen is too poorly preserved to provide satisfactory evidence for this suggestion.
On the basis of m/1 dimensions the body weight of Lutra palaeindica would have been similar to that of extant Lutra lutra, or about 5-10 kg (Yom-Tov et al., 2006).
At present, niche separation plays an important role in otter diversity in South-East Asia (Sivasothi & Nor, 1994) and the same seems to have been the case in the past.The Pleistocene deposits of the Siwalik Group have yielded remains of true otters encompassing two different dietary adaptations, a small piscivore (Lutra palaeindica) and a large shellfish predator (Amblonyx indicus sp.nov.).If we add to this the well known giant otter-like mammal, Enhydriodon sivalensis, the diversity of otters in these deposits is quite high.Enhydriodon probably preyed predominantly on bivalves rather than crabs.The Siwalik otters thus present an interesting case of niche separation among Pleistocene riverine otters of the Indian subcontinent.
The disproportionate representation of otter tribes in the Siwaliks Group remains to be explained: Enhydriodontini (several dozen specimens), Lutrini (three fossils) and Aonychini (one fossil).Fish and molluscs are present throughout the Siwalik sequence, so it is probably not a question of availability of prey categories.Otters spend much of their lives in or near water (rivers, swamps, ponds, lakes) so one would expect them to be better represented in the fossil record than they are.Perhaps the skeletal remains of Lutrini and Aonychini, being quite small otters, were subject to destruction by predators (crocodiles, large snakes, carnivores) whereas the more solidly constructed and large snouts of Enhydriodon had better chances of becoming fossilized, indicating the operation of a taphonomic filter.For the moment, the question remains unanswered.The adaptations of the Indian fossil otters to different niches are reflected in the dental morphology and body size, the piscivore being small (ca 5-10 kg) with sectorial carnassials, the cancrivore being medium sized (ca 11-12 kg) with semi-bunodont carnassials whereas the molluscivore was large (ca 22-25 kg) with mastoidized carnassial cusps.

Conclusions
Lutrini and Aonychini are extremely rare in Indian fossil deposits and are confined to Plio-Pleistocene deposits, unlike the remains of Enhydriodontini which are reasonably well represented and span the period from Middle Miocene to Pleistocene.We here describe and interpret some fossils from the Pleistocene of India attributed to the otter tribes Aonychini and Lutrini.A large shellfish predator is identified as a new species of Amblonyx, A. indicus.The dimensions of the teeth indicate that this species was similar in body size (estimated at ca 11-12 kg) to the extant Cape clawless otter (Aonyx capensis) (ca 12-18 kg (Larivière, 2001)), from which it differs by the morphology of the talonid basin of the m/1, which, in contrast, approaches the Indian specimen to the genus Amblonyx.
Other large otters from the Plio-Pleistocene of Asia have been described, including Lutra aonychoides from China (Zdansky, 1924) and Lutra palaeoleptonyx from Java (Dubois, 1908;Von Fig. 6.-Aonychini and Lutrini P4/ length and breadth dimensions and comparisons with Indian Plio-Pleistocene otters attributed to Lutra palaeindica (data for extant Lutrini and Aonychini from Pohle, 1919, andWillemsen, 1992).Koenigswald, 1933Koenigswald, , 1940)).Both of these taxa possess broad m/1s and are likely to represent Aonychini rather than Lutrini, and were thus probably predominantly shellfish predators.Lutra robusta (Von Koenigswald, 1933) is based on a P4/ from Java, which Willemsen (1986) considered to belong to Lutrogale palaeoleptonyx.It cannot be compared with the Panchkula otter due to lack of parts in common.
Lutrini from the Pleistocene of India have been known for more than a century (Falconer, 1868) but their remains are rare.The new specimen from Pipalwala is larger than the type specimen of Lutra palaeindica, but probably belongs to this species.This is because the teeth of the type specimens plot near the centre of metric variation of Lutra lutra, while the Pipalwala specimen plots near the large end of the range of variation.The range of variation of the teeth of Lutra sumatrana is based on only four specimens, all of which fall into the lower half of the range of variation Lutra lutra.Unfortunately the teeth in the Pipalwala specimen are damaged, making it difficult to assess the meaning of the slight differences in morphology that can be made out between it and the holotype of Lutra palaeindica and there is a faint possibility that it belongs to Lutrogale rather than to Lutra.