faunal and environmental change in the late miocene siwaliks of...

q 2002 The Paleontological Society. All rights reserved. 0094-8373/02/2802-0001/$1.00

Copyright ( 2002, The Paleontological Society

Faunal and environmental change in the late Miocene Siwaliksof northern Pakistan

John C. Barry, Michele E. Morgan, Lawrence J. Flynn, David Pilbeam,Anna K. Behrensmeyer, S. Mahmood Raza, Imran A. Khan, Catherine Badgley,Jason Hicks, and Jay Kelley

Abstract.—The Siwalik formations of northern Pakistan consist of deposits of ancient rivers thatexisted throughout the early Miocene through the late Pliocene. The formations are highly fossil-iferous with a diverse array of terrestrial and freshwater vertebrates, which in combination withexceptional lateral exposure and good chronostratigraphic control allows a more detailed and tem-porally resolved study of the sediments and faunas than is typical in terrestrial deposits. Conse-quently the Siwaliks provide an opportunity to document temporal differences in species richness,turnover, and ecological structure in a terrestrial setting, and to investigate how such differencesare related to changes in the fluvial system, vegetation, and climate. Here we focus on the intervalbetween 10.7 and 5.7 Ma, a time of significant local tectonic and global climatic change. It is alsothe interval with the best temporal calibration of Siwalik faunas and most comprehensive data onspecies occurrences. A methodological focus of this paper is on controlling sampling biases thatconfound biological and ecological signals. Such biases include uneven sampling through time,differential preservation of larger animals and more durable skeletal elements, errors in age-datingimposed by uncertainties in correlation and paleomagnetic timescale calibrations, and uneven tax-onomic treatment across groups. We attempt to control for them primarily by using a relative-abun-dance model to estimate limits for the first and last appearances from the occurrence data. Thismodel also incorporates uncertainties in age estimates. Because of sampling limitations inherentin the terrestrial fossil record, our 100-Kyr temporal resolution may approach the finest possiblelevel of resolution for studies of vertebrate faunal changes over periods of millions of years.

Approximately 40,000 specimens from surface and screenwash collections made at 555 localitiesform the basis of our study. Sixty percent of the localities have maximum and minimum age esti-mates differing by 100 Kyr or less, 82% by 200 Kyr or less. The fossils represent 115 mammalianspecies or lineages of ten orders: Insectivora, Scandentia, Primates, Tubulidentata, Proboscidea,Pholidota, Lagomorpha, Perissodactyla, Artiodactyla, and Rodentia. Important taxa omitted fromthis study include Carnivora, Elephantoidea, and Rhinocerotidae. Because different collectingmethods were used for large and small species, they are treated separately in analyses. Small spe-cies include insectivores, tree shrews, rodents, lagomorphs, and small primates. They generallyweigh less than 5 kg.

The sediments of the study interval were deposited by coexisting fluvial systems, with the largeremergent Nagri system being displaced between 10.1 and 9.0 Ma by an interfan Dhok Pathan sys-tem. In comparison to Nagri floodplains, Dhok Pathan floodplains were less well drained, withsmaller rivers having more seasonally variable flow and more frequent avulsions. Paleosol se-quences indicate reorganization of topography and drainage accompanying a transition to a moreseasonal climate. A few paleosols may have formed under waterlogged, grassy woodlands, butmost formed under drier conditions and more closed vegetation.

The oxygen isotopic record also indicates significant change in the patterns of precipitation be-ginning at 9.2 Ma, in what may have been a shift to a drier and more seasonal climate. The carbonisotope record demonstrates that after 8.1 Ma significant amounts of C4 grasses began to appearand that by 6.8 Ma floodplain habitats included extensive C4 grasslands. Plant communities withpredominantly C3 plants were greatly diminished after 7.0 Ma, and those with predominantly C4

plants, which would have been open woodlands or grassy woodlands, appeared as early as 7.4 Ma.Inferred first and last appearances show a constant, low level of faunal turnover throughout the

interval 10.7–5.7-Ma, with three short periods of elevated turnover at 10.3, 7.8, and 7.3–7.0 Ma. Thethree pulses account for nearly 44% of all turnover. Throughout the late Miocene, species richnessdeclined steadily, and diversity and richness indices together with data on body size imply thatcommunity ecological structure changed abruptly just after 10 Ma, and then again at 7.8 Ma. Be-tween 10 and 7.8 Ma the large-mammal assemblages were strongly dominated by equids, withmore balanced faunas before and after. The pattern of appearance and disappearance is selectivewith respect to inferred habits of the animals. Species appearing after 9.0 Ma are grazers or typicalof more open habitats, whereas many species that disappear can be linked to more closed vege-tation. We presume exceptions to this pattern were animals of the mixed C3/C4 communities or thewetter parts of the floodplain that did not persist into the latest Miocene. The pace of extinctionaccelerates once there is C4 vegetation on the floodplain.

The 10.3 Ma event primarily comprises disappearance of taxa that were both common and oflong duration. The event does not correlate to any obvious local environmental or climatic event,

2 JOHN C. BARRY ET AL.

and the pattern of species disappearance and appearance suggests that biotic interactions may havebeen more important than environmental change.

The 7.8 Ma event is characterized solely by appearances, and that at 7.3 Ma by a combination ofappearances and disappearances. These two latest Miocene events include more taxa that wereshorter ranging and less common, a difference of mode that developed between approximately 9.0and 8.5 Ma when many short-ranging and rare species began to make appearances. Both eventsalso show a close temporal correlation to changes in floodplain deposition and vegetation. The 7.8Ma event follows the widespread appearance of C4 vegetation and is coincident with the shift fromequid-dominated to more evenly balanced large-mammal assemblages. The 7.3 to 7.0 Ma eventstarts with the first occurrence of C4-dominated floras and ends with the last occurrence of C3-dominated vegetation. Absence of a consistent relationship between depositional facies and thecomposition of faunal assemblages leads us to reject fluvial system dynamics as a major cause offaunal change. The close correlation of latest Miocene species turnover and ecological change toexpansion of C4 plants on the floodplain, in association with oxygen isotopic and sedimentologicalevidence for increasingly drier and more seasonal climates, causes us to favor explanations basedon climatic change for both latest Miocene pulses.

The Siwalik record supports neither ‘‘coordinated stasis’’ nor ‘‘turnover pulse’’ evolutionarymodels. The brief, irregularly spaced pulses of high turnover are characteristic of both the stasisand pulse models, but the high level of background turnover that eliminates 65–70% of the initialspecies shows there is no stasis in the Siwalik record. In addition, the steadily declining speciesrichness and abrupt, uncoordinated changes in diversity do not fit either model.

John C. Barry, Michele E. Morgan, Lawrence J. Flynn, and David Pilbeam. Department of Anthropologyand Peabody Museum, Harvard University, Cambridge, Massachusetts 02138

Anna K. Behrensmeyer. Department of Paleobiology, MRC NHB 121, Smithsonian Institution, Washing-ton, D.C. 20560

S. Mahmood Raza and Imran A. Khan. Geological Survey of Pakistan, Sariab Road, Quetta, PakistanCatherine Badgley. Museum of Paleontology, University of Michigan, Ann Arbor, Michigan 48109Jason Hicks. Denver Museum of Natural History, 2001 Colorado Boulevard, Denver, Colorado 80205Jay Kelley. Department of Oral Biology, College of Dentistry, University of Illinois at Chicago, Chicago,

Illinois 60612

Accepted: 12 October 2001

Introduction

One research goal of paleobiology is inves-tigation of how external environmental, andespecially climatic, events shape species andecosystems over geologic timescales (e.g.,Vrba 1980). Such investigations seek to relatetemporal patterns in species richness, com-position, and ecological structure to concur-rent environmental change. Previous investi-gations in both marine and nonmarine set-tings have found that a recurrent pattern inthe preserved record is alternation betweenbrief intervals of concentrated biotic changeand much longer periods with little or nochange, that is, ‘‘coordinated stasis’’ (Olson1980; Webb 1984; Brett et al. 1996). Theoreticalconsiderations suggest such patterns shouldbe closely linked to environmental events andevolutionary responses to them (Vrba 1985,1995). Authors have debated, however, the rel-ative importance of environmental versus bi-otic factors, the causal relationships with en-vironmental change, and even whether theterrestrial vertebrate record does document

such patterns (e.g., Alroy 1996; McKinney etal. 1996; Behrensmeyer et al. 1997; Prothero1998, 1999). In addition, the appropriate tem-poral and geographic scales for such investi-gations are uncertain, because significant cli-matic/environmental changes may be local-ized and can occur in just thousands of years,a level of resolution that is rarely found in fos-sil sequences of long duration.

The Siwalik formations of northern Pakistanconsist of deposits of ancient rivers that ac-cumulated throughout the early Miocenethrough the late Pliocene. The preserved sed-iments are highly fossiliferous and have beena focus of paleontological research for over100 years (see summaries in Barry et al. 1985;Flynn et al. 1990; Badgley and Behrensmeyer1995b). Since 1973, collaborative research be-tween the Geological Survey of Pakistan (GSP)and several American universities has result-ed in the compilation of a large body of newinformation on the age, sedimentology, andvertebrate paleontology of these sediments.Consequently the Siwaliks now comprise one

3LATE MIOCENE FAUNAL AND ENVIRONMENTAL CHANGE

of the best-known Cenozoic nonmarine re-cords of both environmental and bioticchange, and they are an ideal site for exploringthe problems and implications of temporalpatterns of faunal change.

Previous analyses of Miocene Siwalik fau-nal change found evidence for short periods ofhigh faunal turnover between longer periodswith low turnover, as well as indications of alate Miocene decline in species richness. Thestrongest peaks of both appearances and dis-appearances were in the middle Miocene,with turnover in the late Miocene being moreconstant than in the middle Miocene, but therewas no evidence that increased extinction co-incided with increased first appearances (Bar-ry et al. 1990, 1995; Barry 1995). Related stud-ies of species longevity, trophic structure, andbody size (Flynn et al. 1995; Gunnell et al.1995; Morgan et al. 1995) have also docu-mented differences between the middle andlate Miocene. These earlier analyses all used500-Kyr intervals as the temporal unit of anal-ysis.

Considerable evidence for significant lateMiocene Siwalik environmental change alsoexists. This evidence comes from both the sed-iments (Retallack 1991; Willis 1993a,b; Willisand Behrensmeyer 1994, 1995; Zaleha 1997a,b)and the stable isotopic content of paleosolsand fossils (Quade et al. 1989, 1992; Morgan etal. 1994; Stern et al. 1994; Quade and Cerling1995). Quade and others (1989, 1992, 1995)first reported changes in stable carbon and ox-ygen isotope ratios in both d13C and d18O val-ues in the late Miocene, which they interpret-ed as resulting from long-term changes in cli-mate and vegetation that led to drier, moreseasonal climate and extensive grasslands.Both mammalian and avian isotopic valuesalso show a transition from C3-dominated di-ets to mixed and C4-dominated diets (Morganet al. 1994; Stern et al. 1994), but the dietaryevidence implies an earlier change than that ofthe soil carbonates.

In the following, we summarize our workon faunal and environmental change in theMiddle Siwaliks of northern Pakistan. In do-ing this we extend our earlier inquiries by in-cluding more species level taxa and improvingthe time resolution fivefold, while trying to

present as much primary data as practicable.In contrast to our previous analyses using 500-Kyr intervals (Barry et al. 1990, 1991, 1995), inthis paper we use a temporal resolution of 100Kyr. The use of intervals of shorter duration isa compromise between analytic resolutionand having adequate data in each unit. It mayalso be a lower limit imposed by the rock re-cord. In addition, we restrict our analysis tothe interval between 10.7 and 5.7 Ma (inclu-sive). This 5-Myr-long period approximatesthe late Miocene, a time of very significantglobal climatic change as well as more region-al and local tectonic and habitat modifications.It is also the interval for which we currentlyhave the best temporal calibration of Siwalikfaunas and most comprehensive data on spe-cies occurrences. Equids first appear in the Si-waliks at 10.7 Ma as immigrants from NorthAmerica, so this important faunal event marksan appropriate starting point. We occasionallyinclude ‘‘pre-hipparion’’ intervals in our dis-cussions to place trends into a larger context.

Our primary focus is on documenting pat-terns of faunal turnover and ecologicalchange, and investigating the relationships be-tween observed faunal patterns and changesin the fluvial system, climate, and local habi-tats. Methodological issues of dating and anal-ysis are also significant concerns. We thereforeattempt to incorporate uncertainties in thecorrelation and ages of localities in our anal-yses, and seek to control for the diverse sam-pling biases confounding the biological andecological signals. We begin with discussionof the occurrence of the fossils and correlationof the fossil localities to the Geomagnetic Po-larity Time Scale, and then assess the suitabil-ity and limitations of the vertebrate record forour analyses. We next describe the Siwalik flu-vial system and the changes it underwent dur-ing the late Miocene, as well as the paleosoland stable isotope evidence for late Mioceneenvironmental change. Finally, we analyze thefaunal data and discuss the relationships be-tween patterns of observed faunal change andindicators of environmental change. In doingthis, we assume that if ecological and evolu-tionary change primarily results from envi-ronmental and especially climate change, thenturnover events should temporally correlate to


FIGURE 1. Map of the Potwar Plateau in northern Pakistan, showing the locations of regional networks of strati-graphic sections along the Soan River (A), on the southern limb of the Soan Synclinorium near Chinji (B), and atthe eastern end of the Potwar Plateau near Hasnot (C). The location of sections at Jalalpur and Rohtas are also shown.Barbed lines denote the boundaries of thrust sheets. Gray stippled area to east is a reservoir. Figure based in parton Burbank and Beck 1991: Figure 2b.

environmental events (Vrba 1985, 1995). Giv-en a sufficiently well sampled, temporally re-solved, unbiased record, we would expectturnover and other faunal events to (1) be con-centrated in brief intervals separated by lon-ger periods with little or no change and (2)show a close temporal correlation to environ-mental changes. Such correlations can be usedto infer causal links between faunal and en-vironmental change.

Stratigraphic and Fossil Data

Occurrence of the Fossils

The Siwalik formations comprise Neogeneage fluvial deposits exposed widely through-out the Indian subcontinent along the north-ern and western mountains paralleling thecollision zone between India and Asia. In Pak-istan, Siwalik sediments are especially exten-sive on the Potwar Plateau, where they are ex-posed in a folded belt extending from the SaltRange in the south to the Margala Hills in thenorth, and from the Jhelum River in the east

to the Indus on the west (Fig. 1). The depositsfill basins with sediment eroded off adjacentuplifted areas, and in some places single sec-tions exceed 3000 m in thickness and incor-porate 10 Myr of time. Taken together, the sed-iments span the time interval from about 18Ma to younger than 2 Ma. The sediments arewell exposed throughout the Plateau alongmajor streams and rivers, where the Pleisto-cene Potwar Loess has been removed by re-cent erosion. The combination of minimal de-formation and faulting, great thickness, andlateral continuity of exposure allows a moredetailed and temporally resolved study of thestratigraphic succession of lithologies and fau-nas than is usual in terrestrial deposits.

Siwalik lithologies include sandstones, silt-stones, mudstones, and rare marls and clays.The Potwar sequence is frequently dividedinto six or more formations (Table 1) (Cheemaet al. 1977). Primary importance in most clas-sification schemes is given to the relative pro-portions of sandstones and fine-grained sed-


TABLE 1. Siwalik formations of the Potwar Plateau.

Formation Age range (Ma)

TatrotSamwalDhok PathanNagriChinjiKamlial/Murree*Murree

3.5 to 3.3ca. 3.6 to ca. 1.510.1 to ca. 3.511.2 to 9.014.2 to 11.218.3 to 14.2? to 18.3

* In the southern Potwar, the Kamlial and Murree Formations are notdifferentiated. In the north, the Murree is considerably older than 18 Ma.

iments, so that Siwalik formations are distin-guished on the basis of the size and frequencyof large sandstone bodies and the relative pro-portions of silt and mudstones. Correlationsbetween outcrop exposures and sectionsthroughout the Potwar Plateau and elsewhereare based on both magnetostratigraphy andthe lateral tracing of distinctive isochronouslithological horizons (Behrensmeyer andTauxe 1982; Johnson et al. 1988; Badgley andTauxe 1990; McRae 1990; Kappelman et al.1991; S. M. Raza unpublished data; K. A.Sheikh unpublished data). In Table 1 the over-lap in ages between formations represents theapproximate degree of known time transgres-sion. All the formations are likely to be timetransgressive to some degree, even if notshown.

Siwalik depositional environments includesmall to very large channels, levees, paleosols,and rare pond or swamp deposits. These de-positional environments are present in all theSiwalik formations but differ in their frequen-cy of occurrence. Vertebrate fossils are foundin all lithologies, but are very rare in the thinmarls deposited in shallow ponds. The fossilsare thought to have accumulated as attritionalassemblages derived from nearby surround-ings, but habitat specific associations of fossilsseem not to have been preserved (Badgley andBehrensmeyer 1980, 1995; Badgley et al. 1995).Fossil vertebrates are most common in thesmall floodplain channels, where they typi-cally occur as concentrations of disarticulated,often fragmentary bones. In the small flood-plain channels, a very few concentrations con-tain 1000 or more specimens, but the majorityof concentrations have only 5–200 fossils. Be-cause the large channel, levee, and paleosol fa-cies contain only isolated specimens or small,

low-density scatters, our collections of fossilsfrom the different facies are biased towardoverrepresentation of small channel flood-plain sites. This bias, in addition to those in-troduced by taphonomic processes commonto the different facies, places strict limits onour ability to reconstruct the living commu-nities from the fossil assemblages (Badgley1986a,b), and draw meaningful conclusionsfrom the occurrences of individual taxa. Someof these limitations are considered in subse-quent sections, where we discuss uncertain-ties in our analyses. Specific concerns includeuneven sampling through time, differentialpreservation of larger-bodied animals and du-rable parts such as teeth, errors in age-datingimposed by uncertainties in correlation andpaleomagnetic timescale calibrations, and un-even taxonomic treatment across groups.

We use two terms to designate the prove-nience of our fossils; locality and survey block.A locality, as we use the term here, is limitedtemporally and spatially. Most localities en-compass a few tens to several hundred squaremeters of outcrop and have a single sedimen-tary layer as the source of the fossil bone. Theduration of accumulation of the fossil materialin a given locality is thought to be brief, andgenerally between a few tens to, rarely, at most50 Kyr (Behrensmeyer 1982; Badgley 1986b;Badgley et al. 1995; Behrensmeyer et al. 1995).

A survey block or collecting level is an al-ternative to designating and documenting in-dividual localities for the many isolated fossilsof biostratigraphic interest, such as singleequid or bovid molars. Survey blocks differfrom localities in that they are more extensivein area, span a greater stratigraphic thickness,and usually include fossils from multiple sed-imentary layers. Survey blocks included inthis paper span between 30 and 350 Kyr andextend laterally 1 to 2 km. Many of these sur-vey blocks have been systematically searched,with all identifiable surface fossils recorded;others represent less-coordinated collecting ofwell-preserved material. Fossils collectedfrom both survey blocks and localities areused in the analyses of this paper.

The Siwalik fossil collections have beenmade using a variety of techniques. Most ofthe larger specimens, and some of the smaller,


TABLE 2. Biostratigraphic survey results.

Survey block Age (Ma)Duration

(Myr)

Collect-ing

effort(hours)

Scrapfound

IDFossilsfound

IDMam-mals

foundEq-uids

Bo-vids

Gir-affes

Rhi-nos

Pro-bosc.

KL07KL09 1 10

KL08KL11 1 21

KL12ML04 1 05

KL16ML06

KL02 1 14DK01 1 ML07 1 KL13

KL01KL03

KL05 1 06KL04KL20RK01RK02

7.2897.5617.7197.9498.4858.6588.7338.7878.8409.0019.1849.2709.5439.7039.784

10.30710.385

0.3240.2920.1250.3440.1490.1230.1370.2460.1260.1550.0430.0620.1260.1090.1390.1140.124

10.37.8

13.318.310.410.610.5

8.310.715.1

9.513.812.812.711.3

9.56.5

10094

181—

210206154

76179176180125169154—

14848

422895

164139148104

46123108130100140147

428037

322463

14898

10778398180

10278

108129

396629

125

2210

815

8141424151417

750

44

142415

74586445

14266

00064040343634335

213

12608055478

11161

1330

1414

84

159

122

132051

were collected from the surface of naturallyeroded outcrops. Quarrying is mostly unpro-ductive, because even in the richest concentra-tions the density of fossils in the rock is low.Only four sites in the 5-Myr interval have beenexcavated to any extent (localities Y182, Y260,Y270, and Y311). On the other hand, screen-washing bulk samples has been successful inproviding remains of very small species, suchas rodents and insectivores. Over 60 sites havebeen screenwashed to date, and most of theselocalities also have remains of large species.We have also made controlled biostratigraphicsurveys to document the relative number offossils and the relative abundance of taxa atdifferent levels (Table 2). These biostratigraph-ic surveys involved carefully searching andrecording all occurrences of fossil bone alongextensive, strike-parallel belts of outcropwithin the well-defined stratigraphic intervalsthat comprise the survey blocks. During eachsurvey we kept a record of the number ofhours spent searching, and all finds wereidentified as closely as possible and tabulated,although most fossils were not collected. Webelieve the resulting data give less biased es-timates of the abundance of fossils and the rel-ative proportions of taxa than other collectingprocedures.

Harvard-GSP localities and survey blocksare assigned unique letter prefixes and num-

bers, and their geographical positions aremarked on topographic maps or overlays ofaerial photographs. We also have a localitycard file that includes a brief lithological de-scription of each locality, sketch maps andcross-sections, Polaroid ground photographs,and other information (Behrensmeyer andRaza 1984). Collected fossils are catalogedwith a field number, coded description of thespecimen, taxonomic identification, notationof the locality or survey block, and a score forpredepositional damage (Barry 1984). A verysmall number of fossils collected by BarnumBrown and G. Edward Lewis and now in theAmerican Museum of Natural History and theYale Peabody Museum are well enough doc-umented as to provenience and stratigraphichorizon to include in our fossil and localitydatabases. Unfortunately the extensive collec-tions made during the colonial era by the Geo-logical Survey of India and its predecessorscannot be included, because the geographicand stratigraphic data are inadequate. Thismeans that type specimens and other Siwalikfossils described by Falconer, Lydekker, andPilgrim cannot be incorporated in our bio-stratigraphic analyses.

Both the GSP-Harvard Siwalik specimenand locality catalogs are configured as dBasefiles. In addition we have a species file, whichcontains information on each fossil species,


and related auxiliary files. The entire speci-men database presently includes over 40,000records of vertebrate and nonvertebrate fossilsfrom 1291 localities and survey blocks, 86% ofwhich are well-documented localities foundsince 1973. As noted, our project database alsoincludes a few selected records for fossils inthe American Museum of Natural History andYale Peabody Museum collections that can beconfidently placed in our stratigraphic frame-work. The data presented in this paper are de-rived from these databases.

Chronostratigraphic Framework

The drainage patterns of the Potwar Plateauare characterized by a latticelike arrangementof intermittent streams and smaller river trib-utaries (‘‘kas’’ or ‘‘nala’’ in local vernaculars),which have incised shallow ravines in thegently dipping Miocene-age strata. The majorstreams run perpendicular to strike, are typi-cally spaced every 3–5 km, and are connectedalong strike by lateral valleys.

The chronostratigraphic framework of thispaper is based on 20 measured sections thatrange between 250 and 3200 m thick, as wellas 27 shorter sections. Of the 47 sections, 19 ofthe 20 long sections have determined magnet-ic-polarity stratigraphies, and 16 of the short-er have at least some magnetic determina-tions. Apart from the two sections at Rohtasand Jalalpur (Opdyke et al. 1979; Johnson et al.1982), the included sections form three region-al networks (Fig. 1): one on the northern limbof the Soan Synclinorium at Khaur, another onthe southern limb near Chinji, and a third atthe eastern end of the Potwar Plateau nearHasnot. Because exposures are continuouswithin each region, we are able to correlatesections in each regional network by means oflithological horizons, as well as by magneticpolarity. However, because of the distance andabsence of continuous linking exposures be-tween regions, chronostratigraphic correla-tions between the three areas depend on themagnetic-polarity stratigraphy. No sectionswere correlated on the basis of biostratigra-phy.

In the Siwaliks individual stratigraphic sec-tions of sufficient thickness to contain six orseven magnetic transitions (on the Potwar

typically about 250 m) can usually be corre-lated to the Geomagnetic Polarity Time Scale(GPTS) (Johnson and McGee 1983). Althoughsections recording fewer transitions normallycannot be independently correlated to theGPTS, they can still be confidently placed inthe chronostratigraphic framework by deter-mining their stratigraphic relationships to thelonger and better-constrained sections. Thishas been done by laterally tracing isochronouslithological horizons, such as sandstone bod-ies or paleosols (Behrensmeyer and Tauxe1982; Johnson et al. 1988; Badgley and Tauxe1990; McRae 1990; Kappelman et al. 1991; K.A. Sheikh unpublished data). In making cor-relations to the GPTS, it is not always possibleto match every observed magnetic transitionto a GPTS magnetic boundary, either becauseof sampling problems (Johnson and McGee1983) or because of the presence of very shortmagnetic intervals not documented in theGPTS (Kappelman et al. 1991).

Appendix 1 lists the thicknesses and refer-ences for the sections used to construct thechronostratigraphic framework for the PotwarSiwaliks. In most cases we use the same po-larity zonation and correlation to the GPTS asthe references, or have made only trivialchanges. In a few cases subsequent work hasled us to make substantial reinterpretations ofthe polarity zonations or the correlations, andwe comment on these in the remainder of thissection. We also present results of new workdone in the Hasnot area and provide more de-tails for already published sections in theKhaur region.

Jarmaghal Kas. Johnson et al. (1982) madeno specific correlations for the middle third ofthis section. New details in the GPTS of Candeand Kent (1995) dictate different preferredcorrelations for the intervals between the baseof 3An and the top of 4An. We now correlatethe N5 interval of Johnson et al. (1982) to3An.1n, the N4 interval to 3An.2n, N3 to 3Bn,N2 to 3Br.2n through 4N.2n, and N1 to 4An.

Andar Kas. Johnson et al. (1982) againmade few specific correlations for this section,but we have correlated more of the observedtransitions to the GPTS. Specifically we cor-relate the N7 interval of Johnson et al. (1982)to 3An.1n, N6 to 3An.2n, the top of N4 to the


3Br.3r/4n.1n transition and its bottom to4n.2n/4r.1r, the top of N0 to the top of 5n.1nand the bottom to the base of 5n.2n, N1 to5r.2n, and the top of N2 to 5r.3r/5An.1n andits bottom to the base of 5An.2n. We make nocorrelations between the top of zone R4 andthe base of R1, nor below the base of N2.

Dhok Saira. This recently measured sectionparallels the upper part of the Dhala Nala sec-tion of Johnson et al. (1982). At its base the sec-tion is continuous with the lower Dhala Nalasection, and its uppermost 20 m can be cor-related by means of the Pillar Sand to the ‘‘Ta-trot’’ and ‘‘Pillar’’ I–VI sections of McMurtry(1980). The stratigraphic and magnetic polar-ity data, together with our preferred correla-tion to the GPTS, are shown in Figure 2. Thebase of the uppermost N4 polarity zone is atthe same stratigraphic level as a polarity tran-sition in all of McMurtry’s (1980) Pillar sec-tions, and we follow his correlation of thetransition to the 3An.2n/3Ar transition. Thelong reversed interval between 132 and 53 mis supported by only a few reversed sites,most of which are in the upper portion (R3).The lower part (R2) is in a single, thick sandin which our sampling efforts failed becausethe magnetics were unfavorable. If our corre-lation is correct, future sampling at this hori-zon should show normal polarity represent-ing 4n.1n. A single low-quality site 92 m abovethe base suggests the presence of a short nor-mal zone (N2). It could be the 3Br.2n zone. Theupper Dhala Nala section of Johnson et al.(1982) also had a normal zone at this horizon.We have correlated the base of N1, which issparsely sampled, to the base of 4r.1n, ratherthan the base of 4n.2n, because this correlationpreserves the proportionality of the thicknessof the zones.

Lower Dhala Nala. Johnson et al. (1982) didnot discuss this section in detail, but they didpresent lithologic and magnetic logs in theirFigure 11. They correlated the section bymeans of marker beds to Kotal Kund, but theymade no specific correlations to the GPTS oth-er than to note it was very similar to their Ko-tal Kund section. Subsequently the sectionwas extended downward an additional 329 m,and polarity determinations made for the up-per 150 m (unpublished data from the late

Noye Johnson personal communication 1981).Although the virtual geomagnetic pole lati-tudes and data quality are not available to us,we present the stratigraphic and polarity datain Figure 3. Our preferred correlation to theGPTS is the same as that of Johnson et al.(1982). In drafting Figure 3 we have assumedthat the lowest normal sample at the base ofthe measured section is near the base of the5n.2n interval. However, because that corre-lation is not confirmed by an underlying re-versed site, we do not use it to determine theages of any fossil sites in the section. Thesparse sampling in the upper quarter of theN1 interval makes correlation of the N1/R1transition particularly difficult. It is likely oneor more reversed zones were missed in the in-terval between 300 and 400 m. The correlationof N4 is from the Dhok Saira section discussedabove. New magnetic samples have recentlybeen collected from the upper 200 m and froman additional 100 m below the base of this crit-ical section, but the results are not yet avail-able.

Dhok Pathan Rest House. Polarities for thissection were shown in Barry et al. (1980), butno details or correlations were given. Themagnetic polarity and stratigraphic data areshown in Figure 4 (unpublished data from thelate Noye Johnson personal communication1980), together with a preferred correlation tothe GPTS. Unfortunately, the position of theSoan River and structural complexities makeit difficult to measure the complete section ac-curately. We think that the gap shown in Fig-ure 4 below the 300-m level may be underes-timated. The section includes the ‘‘U’’ Sand-stone interval (Behrensmeyer and Tauxe 1982),which allows the base of the section to beplaced in magnetozone 4Ar.1r. Nearby sec-tions (Barry et al. 1980) indicate that the R1 in-terval should contain a substantial normal in-terval correlating to 4An. The N3 magneto-zone is unbounded at its top, but rather thanextrapolating locality ages from rates of thesubjacent zones, we have used sediment ac-cumulation rates from the same interval in theKaulial Kas section. We do this because thetwo subjacent Dhok Pathan magnetozones arevery short and their inconsistent rates give un-realistic estimates.


FIGURE 2. Dhok Saira lithologic log, magnetic polarity zonation, and correlation to geomagnetic polarity timescaleof Cande and Kent (1995), with upper 70 m of Lower Dhala Nala section. Black shaded areas indicate paleosols;light-gray areas, fine-grained overbank; stippled areas, sandstone; cross-hatched areas, covered intervals; double-headed arrows, sandstones used to correlate lithologic sections; and dotted lines, correlations to geomagnetic po-larity timescale. R, r 5 reversed site, N, n 5 normal site, X 5 indeterminate site. Uppercase letters indicate siteswith samples having consistent polarity; lowercase letters indicate sites with questionable polarity.

The Dhok Pathan Rest House section is par-ticularly important because much of the DhokPathan fauna of Pilgrim, Brown, and Lewisapparently came from the vicinity of the DhokPathan Rest House. The most fossiliferous

beds are above the Hilltop Sand in the N2, R3,and N3 magnetozones. It is from these levelsthat the older collections presumably came.

Hasal Kas. Tauxe and Opdyke (1982) ther-mally demagnetized and reanalyzed the sam-


FIGURE 3. Lower Dhala Nala lithologic log, magnetic polarity zonation, and correlation to geomagnetic polaritytimescale of Cande and Kent (1995), with lower 70 m of Dhok Saira section. Symbols as in Figure 2.

ples of Barndt et al. (1978), and then relatedthem to a remeasured lithological section. Wefollow Tauxe and Opdyke’s (1982) polarity as-signments and correlations to the GPTS, butwe make four additional assignments by cor-relating their R11 polarity zone to 4Ar.2r andthe R10 polarity zone to 4Ar.3r. Nevertheless

the polarity zonation of the upper third is notwell defined, and many of the correlations ofpolarity reversals are based on lithologicalcorrelations to the nearby Mahluwala Kas andRatha Kas sections.

Kaulial Kas. Tauxe and Opdyke (1982) dis-cussed this section, and we follow their cor-


FIGURE 4. Dhok Pathan Rest House lithologic log, magnetic polarity zonation, and correlation to geomagnetic po-larity timescale of Cande and Kent (1995). Symbols as in Figure 2. The double-headed dotted arrow marks thecorrelation of the U-level to the GPTS as established by Behrensmeyer and Tauxe (1982).

relations with two minor differences. First, wecorrelate the uppermost N7 zone to 3Br.1n,rather than 3Bn. This results from recognitionof two new small polarity zones in the GPTS.Zone 3Br.2n is then missing, with zone N6 cor-related to 4N.1n, in agreement with Tauxe andOpdyke (1982). It follows then that zone N5 is

a new normal zone in 4n.1r, not recorded inthe standard GPTS. We note that it is based ona single sample of normal polarity and that itis the correlation implied by Tauxe and Op-dyke (1982). Second, we treat the N3 zone as anew normal interval, not as 4r.1n as implied byTauxe and Opdyke (1982). It is noteworthy


that N3 is based on a single sample having aVGP latitude near zero (Tauxe and Opdyke1982: Fig. 4i). Zone 4r.1n is unrecorded in theKaulial Kas section.

In the Chinji region there are many addi-tional long and short sections, including thoseof Kappelman et al. (1991). The focus of thisstudy, however, is limited to the interval be-tween 10.7 and 5.7 Ma. We have therefore in-cluded only two sections (Gabhir and Chita-parwala) that extend down into the early andmiddle Miocene to provide a baseline for ourstudy.

Ages of Localities and Assignment toIntervals

Most localities are in exposures near thestreams along which we have measured localstratigraphic sections and are typically lessthan 1 km off the line of a section. To establishstratigraphic relationships for individual lo-calities, we have then traced lithological mark-ers back into the nearest section. These hori-zons are usually paleosols or small sandstonebodies, types of lithologies that have beendemonstrated to be isochronous over the dis-tances involved (Behrensmeyer and Tauxe1982). Any such correlation has a degree ofuncertainty. To preserve this uncertainty, wehave recorded the stratigraphic positions oflocalities as being bracketed by uppermostand lowermost bounds in the sections. We areconfident that localities are not above or belowthese bounds, but we assume that the localitycould be at any level within the bracketed in-terval with equal likelihood.

To order faunal events in time, we had to as-sign each locality an absolute age, which canbe interpolated from its stratigraphic positionrelative to known marker horizons or timelines. Ages can be estimated to varying de-grees of accuracy. Estimates for the ages of 555localities and survey blocks that lie between10.7 and 5.7 Ma are presented in Appendix 2.The algorithm used to determine these ages isbased on interpolation between two points ofknown age. In this paper all points of knownage are magnetic polarity transitions, but inprinciple, radiometric dates and biostrati-graphic or isotopic events of known age could

also be used. The formula for finding an ageis

Age 5 UAge 1 ((LAge 2 UAge)/T)SD,

where UAge and LAge are, respectively, theages of the top and bottom of the geomagneticzone, T is the stratigraphic thickness of themagnetozone in the local section, and SD isthe stratigraphic distance between the localityand top of the magnetozone. In this formula-tion the quantity (LAge 2 UAge)/T is the av-erage rate of sediment accumulation in thesection during the geomagnetic interval. Somelocalities lie in truncated topmost or bottom-most magnetozones, and interpolation cannotbe used to estimate their ages. In these casesages are extrapolated, using the rate of sedi-ment accumulation of the sub- or superjacentmagnetic interval and the stratigraphic dis-tance between the locality and the under- oroverlying magnetic transition.

Input data used in the calculations include(1) the stratigraphic position of the locality ina local lithological section, (2) the stratigraph-ic positions of the magnetic transitions in thesame section, and (3) the ages of the magnetictransitions, as derived from correlation of thesection to the Geomagnetic Polarity TimeScale. From these input data, the quantitiesused in the above formula can be calculated.We have two further refinements in our cur-rent implementation. The first is to derive sep-arate age estimates for the possible upper andlower limits of the stratigraphic positions forlocalities traced into sections. This allows ourestimate to incorporate the uncertainty in thecorrelation of the locality to a stratigraphicsection. When the stratigraphic position isprecisely determined, these age estimates willbe the same. The second refinement involvesusing the highest and lowest possible posi-tions of the magnetic transitions in the litho-logical sections. Magnetic transitions arebracketed by sites of different polarity, and byconvention the transition is placed at the mid-point between those sites. Noting the actualsite horizons incorporates uncertainties in thestratigraphic positions of the magnetic tran-sitions by generating two age estimates, onefor the highest position and one for the lowest.When the magnetic transitions are strati-


FIGURE 5. Differences between estimated maximumand minimum ages of 555 localities between 5.7 and 10.7Ma (inclusive).

graphically constrained, these age estimateswill converge. Taken together, these two re-finements produce conservative estimates fora pair of maximum and minimum ages foreach locality (i.e., the largest reasonable rangeof possible ages).

Five hundred fifty-five localities in our da-tabase are between 10.7 and 5.7 Ma and haveboth a minimum and a maximum age esti-mate (Appendix 2). The difference betweenthe maximum and minimum estimates variesconsiderably among localities, with the larg-est being 1.025 Myr (Fig. 5). However, the agesof most localities are better delimited, asshown by the frequency distribution of thedifferences in Figure 5. Nearly 60% (n 5 332)of the 555 dated localities have a temporal res-olution equal to 100 Kyr or less, whereas 82%(n 5 457) have a resolution of 200 Kyr or less.

For the purposes of our analyses, it is nec-essary to assign each locality to a single 100-Kyr interval and to identify localities toopoorly dated to be assigned. If interval bound-aries are drawn at 60.05 (e.g., the 7.5-Ma in-terval has boundaries at 7.45 and 7.55 Ma),then 149 (27%) of the 555 localities included inAppendix 2 have both maximum and mini-mum age estimates in the same 100-Kyr inter-val. These localities can unambiguously be as-signed to that interval. However, 183 (33%) ofthe localities straddle two intervals, eventhough the difference between their maximumand minimum ages is less than 100 Kyr du-ration. These localities cannot be so unambig-uously assigned. In addition, a third group of125 (23%) localities have differences that aregreater than 100 Kyr and less than 200 Kyr,but they straddle two or three intervals. These

latter two classes of localities do, however,overlap one interval by at least 50% of their to-tal duration. If the percent overlap with anyinterval is assumed equivalent to the proba-bility that the true age lies within the interval,then these two groups of localities can be as-signed to the interval with the greatest over-lap. This will be the same interval containingthe midpoint between the two age estimates.If a locality does not overlap with any intervalby 50%, then it is left unassigned and excludedfrom our analyses. This is the case for the 98(18%) localities with age spans greater than200 Kyr.

The boundaries between intervals are arbi-trary and it is possible that this arbitrary di-vision might affect the patterns we wish to de-scribe. The foregoing discussion used inter-vals centered at multiples of 0.1, with bound-aries at 60.05. To test the robustness of someour conclusions, we have redone some of ouranalyses using 100-Kyr intervals with bound-aries at the multiples of 0.1 and centered at0.05. (For instance, the 7.5-Ma interval hasboundaries at 7.4 and 7.5 Ma, and is centeredat 7.45 Ma.) These alternative analyses did notshow significant differences.

Data Quality in the Siwalik Fossil Record

Comparisons of species richness and espe-cially turnover depend on assumptions aboutthe ages of the fossils and the preservationalbiases of the assemblages. If the patterns ofspecies richness and faunal turnover are to beinterpreted usefully, it is necessary first toconsider the reliability and resolution of thedates, as well as the reliability of the fossil oc-currence data.

Reliability of Age Estimates for Fossil Sites

The algorithm used to determine the ages ofSiwalik localities makes several assumptionsthat affect the potential for errors in our esti-mates and the reliability of the ages. The mostbasic assumption is that the rate of sedimentaccumulation remains constant within eachmagnetic interval. Because sedimentation influvial systems is stochastically episodic, withmany short hiatuses that collectively may rep-resent more time than the accumulated sedi-ment (Sadler 1981), the assumption is known


FIGURE 6. Number of localities per interval. Data fromAppendix 2, n 5 457.

to be violated (Johnson et al. 1988). However,if no contained hiatus is large relative to aninterval’s duration and if the stratigraphic dis-tribution of the many small contained hiatus-es is random, then interpolation of ages canstill be done (Badgley et al. 1986). This as-sumption of approximate sedimentologicaluniformity is consistent with correlations be-tween our individual sections, which typicallyshow the same magnetic intervals having thesame relative stratigraphic thickness in sepa-rate sections.

Second, we assume that the probability of alocality lying somewhere between its upperand lower stratigraphic limits has a uniformdistribution, with zero probability above andbelow the marked range. This is a conservativeassumption, because it produces a widerspread between the maximum and minimumage estimates than if we assumed some otherprobability distribution, such as a normal dis-tribution around the center of the interval.

A third assumption is that the age estimatesof the magnetic transitions of the Geomagnet-ic Polarity Time Scale are correct. These agesare derived from interpolations between dat-ed tie points in the marine record, and theyassume constancy in oceanic ridge spreadingrates. The age interpolations are vulnerable toerrors in the ages of tie points, which in sev-eral cases are rather poorly constrained, andto error in identifying inflection points in thespreading-rate curves (Cande and Kent 1992,1995). (See Pilbeam et al. 1996 for discussionof a particular case.) Numerous publishedgeomagnetic polarity timescales are in use, re-flecting different choices of tie points, spread-ing rates, and inflection points. Each varianttimescale gives different estimates for the agesof Siwalik localities, with maximum differenc-es close to 500 Kyr for the interval at 10.7–5.7Ma. In previous publications we have used thetimescales of Mankinen and Dalrymple(1979), Berggren et al. (1985), and Cande andKent (1992). In this paper we use the nowwidely accepted timescale of Cande and Kent(1995).

Other potential sources of error in our ageestimates are not included in the expresseduncertainties. Mistakes in field correlationsbetween the localities and sections are impor-

tant in this respect. In placing localities strat-igraphically we have tried to express the un-certainties in our correlations and preservethem through the steps of our analysis. Nev-ertheless, the potential remains for undetect-ed errors. Errors in placement of magnetictransitions are also possible, although to someextent alleviated by the refinement of usingupper and lower positions of the transitions tocalculate a pair of ages. This is again a con-servative approach, and future paleomagneticwork refining the placement of transitions inour sections should improve the precision ofour estimates. Finally, there may be a sourceof error in our correlations between the localmagnetic sections and the geomagnetic time-scale. These correlations are fundamentally in-terpretations of data, and we have undoubt-edly misidentified some magnetic transitions.In this paper we present only a single favoredcorrelation, although there may be plausiblealternatives. In most cases additional field-work will resolve these problems.

Quality of Fossil Occurrence Data

Throughout the late Miocene nearly all Si-walik stratigraphic intervals have some fos-sils, and in many the record is really verygood. Nevertheless, there are considerable dif-ferences in the number of fossils between in-tervals, as well as in the length of gaps be-tween successive localities. This is illustratedin Figures 6 and 7A,B, which show, respec-tively, the number of localities and the numberof cataloged large- and small-mammal speci-mens in each 100-Kyr interval between 10.7


FIGURE 7. Number of catalogued fossils per interval. A, Large mammals (n 5 11,779). B, Small mammals (n 52730).

and 5.7 Ma. Figure 8 shows the distribution ofgaps over the same interval.

The bias in fossil recovery results from acombination of factors that include (1) the areaof outcrop and quality of exposure, (2) thetypes and mix of depositional environments,and (3) differences in the number of hoursspent prospecting and collecting at individuallevels. However, Figure 9, which shows the re-

lationship between the number of surveyinghours and the number of fossils encounteredduring biostratigraphic surveys (data fromTable 2), demonstrates that the greater abun-dance of fossils in some intervals is not duesolely to collecting effort or outcrop area.Rather, greater abundance is also due to theintrinsic productivity of the lithologies.

Figures 6 and 7 show that the late Miocene


FIGURE 8. Gap age against duration. Gap age is the midpoint of the gap. Gap duration is calculated as the differencein age between chronologically successive localities. Data from Appendix 2; n 5 456.

FIGURE 9. Number of mammalian fossils found in sur-vey blocks per hour of searching.

record on the Potwar Plateau is best over theapproximately 2-Myr period between about7.9 and 10 Ma, with all intervals having somefossils. In this same 2-Myr period, productiv-ity is usually higher than that of the adjacentperiods (Fig. 9), and the duration of gaps be-tween localities is very much lower (Fig. 8).The details differ depending on whether gaplengths, number of localities, or fossil abun-dance is considered, but the record is notice-ably poorer before 10 Ma and after 7.9 Ma;and it is especially so after 7.1 Ma.

This bias in the relative abundance of thefossils could in itself account for differences inthe patterns of first and last occurrences, andthus in our assessments of the number of spe-cies in each stratigraphic interval (Badgleyand Gingerich 1988; Badgley 1990). In previ-ous papers we have therefore used two mea-sures to determine whether data from differ-ent intervals could be compared and to esti-mate range extensions over intervals of poorrecord. These measures were interval-qualityscores (Barry et al. 1990, 1995) and an index ofcompleteness (Barry et al. 1995; Maas et al.1995). The approach taken in this study, how-ever, does not require either, because we nowuse a procedure (Shaw 1964; Koch 1987) to es-timate range extensions directly from the oc-currence data. This method uses counts of thenumber of fossils from each interval and thenumber of specimens in each taxon to assessthe likelihood of discovering a fossil of a spec-ified taxon at a specified level.

Other biases affecting the reliability of thefossil occurrence data result from diversetaphonomic processes and the methods used


to make the collections. The fragmentary na-ture of almost all Siwalik fossils also creates abias that makes more-easily identified taxa, orthose that can be identified from more skeletalelements, appear to be relatively more com-mon. All these biases limit paleoecological in-ferences, such as assessing relative abundanceof taxa or recognizing habitat specificity, butwe presume they are the same for all intervalsunder study and do not affect comparisonsbetween intervals.

Environmental Change

A fundamental question in paleobiology in-volves the relationship between environmen-tal change and faunal evolution. In the Siwaliksequence, we have the opportunity to examinediversity fluctuations and other faunal pat-terns in relation to shifts in depositional en-vironments and vegetation as well as to globaland regional climate change. In this next sec-tion we review the evidence and interpreta-tions for environmental change in the Potwarregion during the late Miocene. In doing so weuse several published sources, including stud-ies of the sediments and paleosols, as well asdata on stable isotopes from the paleosols andfossil mammals. Other than for very rare spec-imens, there is no plant record for the PotwarSiwaliks.

Because we focus foremost on the relation-ships between local environmental and faunalchange, our discussion is limited to the Po-twar Plateau (;30,000 km2). The small localgeographic scale makes it possible to testwhether changes in the fluvial system causedthe faunal changes.

The present climate of the Potwar is warmand humid, with very marked monsoon sea-sonality. Spring months are dry and very hot,summer months bring the monsoon rains(with half the annual precipitation), and au-tumn and winter are cool and drier. Precipi-tation is highly variable throughout the Pla-teau and from year to year. Rainfall averagesabout 500–700 mm/yr over much of the Pla-teau, but can vary from 250 to 1300 mm/yrfrom one year to the next (Anonymous 1985;Shukla 1987). Much of the annual variation isdue to differences in the intensity of monsoonrains, which are also highly irregular in time

of onset. As discussed below, climate appar-ently changed through the study interval, butit can be characterized during the middle andlate Miocene as warm and humid with astrong monsoon circulation and marked sea-sonality.

Lithological Evidence

Detailed environmental reconstructionshave been made for the Siwalik formations, re-constructions that, because of an inadequateplant record, draw heavily on sedimentologi-cal evidence and models based on the modernanalogs provided by the Indus and GangesRivers. Although some work has been done onthe Dhok Pathan Formation (Behrensmeyerand Tauxe 1982; Behrensmeyer 1987; Badgleyand Tauxe 1990), most of the relevant fieldstudies in the Potwar have focused on themiddle Miocene Chinji Formation and thelower part of the middle-to-late Miocene Na-gri Formation, with the transition betweenthese two formations at ca. 12–10.5 Ma beingof special interest (Retallack 1991; Willis1993a,b; Willis and Behrensmeyer 1994, 1995;Badgley and Behrensmeyer 1995b; Behrens-meyer et al. 1995; Zaleha 1997a,b). Here wealso use these studies to characterize the tran-sition between the late Miocene Nagri andDhok Pathan Formations. We believe this isjustified because comparisons of the Chinjiand Dhok Pathan Formations indicate similar-ities in their depositional environments and intheir transitional stratigraphic relationshipswith the intervening Nagri Formation (Beh-rensmeyer and Tauxe 1982; Behrensmeyer1987; Retallack 1991; Willis 1993b; Zaleha1997a,b). Particularly noteworthy are the sim-ilarities between the Chinji/Nagri transitionand the vertical and lateral changes docu-mented between the Nagri and Dhok PathanFormations. Ongoing research is documentinglateral and vertical change in lithofacies, pa-leosols, and isotopic composition within theDhok Pathan Formation in the Khaur and Has-not areas, with the continuing goal of inte-grating this information with the paleontolog-ical evidence we present below.

The Miocene Siwaliks were deposited in avery large scale fluvial system, one compara-ble in size to the modern Indus or Ganges sys-


tems. These modern rivers occupy divergentbasins paralleling the northern and westernmountains and bounded by the Indian cratonto the south and east. This arrangement ofmountains, basins, and major drainages is aresult of the tectonic movement of the subcon-tinent against Asia and existed throughout theMiocene. However, paleochannel directionsand paleocurrents indicate that the paleo-Ganges system may have extended fartherwest and drained the Potwar region (Beck andBurbank 1990), which now is connected to theIndus drainage. The reconstructed MioceneIndo-Gangetic system extended over 2000 kmto the east and 1000 km to the south, withfloodplain widths on the order of 100 to 500km. Thus, the Potwar Plateau encompassedonly a small part of this ancient foreland basinand can give us only limited information onthe whole system (Willis and Behrensmeyer1995). In the Miocene, the Potwar study areawas probably some 100–200 km to the south-east of the mountain front where its riversoriginated (Zaleha 1997b). This distance isabout the same as at present. At ca. 10 Ma, thePotwar region was at approximately 298N lat-itude, that is 48 south of its current position(Tauxe and Opdyke 1982).

Reconstructions of the Siwalik rivers depictdrainages coalescing into trunk rivers run-ning to the Arabian Sea and Bay of Bengalalong the axes of the two divergent basins.Large submarine sediment fans attest to theexistence of these Indus-size trunk rivers inthe Miocene (Rea 1992), but deposits of the ax-ial system draining the late Miocene Potwarare not preserved, either because it flowedeastward to the Bay of Bengal or, if flowingsouth to the Arabian Sea, was confined by theslope of the floodplain to the margin of the de-positional basin (Willis 1993a,b; Willis andBehrensmeyer 1995). Channel and floodplaindeposits of large and small tributaries to thetrunk river are preserved, however. These trib-utaries are interpreted to be of two subtypes.The first are deposits of large streams thatemerged from the mountains at widely spacedintervals (100–200 km) and flowed severalhundreds of kilometers southeast across thefloodplain to join the trunk river (Fig. 10).These second-order streams, which we refer to

as ‘‘emergent’’ or ‘‘upland-sourced’’ streams,carried relatively unweathered sediment fromthe mountains and deposited it to form dis-tally broadening, low-gradient megafans.Similar in size and morphology to modernPunjab rivers such as the Jhelum, the Mioceneupland-sourced streams were braided andtypically had channel belts more than 5 kmwide with individual channel widths on theorder of 200 to 400 m. They drained adjacentand distant mountain regions, flowedthroughout the year, and were prone to avul-sion during unusual flood events. Abandonedchannels then became sites of short-livedlakes or swamps with more gradual sedimen-tation, while the active channels moved tolower and perhaps quite distant areas on thefan. Because of the instability of the channels,alluvial ridges along channels did not form,nor did channels migrate laterally to form ex-tensive sheet bodies (Behrensmeyer et al.1995).

In contrast, the second subtype are depositsof generally smaller, braided rivers havingchannel belts ;1–2 km wide and channels 70–200 m wide. Some of these third-order streamsmay also have had mountain sources, whereasothers had sources in foothills or fromgroundwater on the floodplain (Fig. 10). Be-cause they were confined to areas between thefans of the larger emergent rivers and carriedfiner, reworked material eroded from thefloodplain, we refer to them as ‘‘interfan’’ or‘‘lowland-sourced’’ streams. Their sourceswere closer to or even on the second-orderfloodplains, so their flow was probably influ-enced more by floodplain precipitation thanby precipitation in distant, mountainous re-gions. Consequently, flow varied morethroughout the year than in the larger, emer-gent streams. Channels of the third-orderstreams also seem to have been less prone tovery large flood events but nevertheless had amore frequent rate of avulsion than the sec-ond-order, emergent streams (Willis 1993b;Zaleha 1997b).

As a result of the deposition on alluvial fansand interfan areas, the sloping floodplain wasa mosaic of low and high areas, with local re-lief on the order of 10 m and at most a fewhundred meters elevation difference across


FIGURE 10. Reconstruction of late Miocene Siwalik fluvial system in plan view. Numbers refer to orders of streamsas discussed in text. 1 5 axial trunk river, shown flowing eastward to the Bay of Bengal. 2 5 emergent or upland-sourced rivers, draining mountains and tributary to the axial trunk river. 3 5 interfan or lowland-sourced streams,arising on the floodplain or near the mountain front and generally tributary to the axial or second-order rivers. 45 small streams of the floodplain, often ephemeral. Note that the third- and fourth-order streams may also havedrained into low areas on the floodplain, creating seasonal swamps or ponds as indicated by gray stippled areas.The boxed inset approximates the size of the modern Potwar Plateau.


TA

BL

E3.

Ty

pes

ofSi

wal

ikp

aleo

sols

,th

eir

char

acte

rist

ics,

and

inte

rpre

tati

ons.

(Dat

aan

din

terp

reta

tion

sta

ken

from

Ret

alla

ck19

91.S

tud

yb

ased

ontw

osa

mp

lese

ctio

ns

at9.

4an

d8.

6M

a.)

Pal

eoso

lF

requ

ency

†To

pog

rap

hic

sett

ing

Par

ent

mat

eria

lD

rain

age

Mat

uri

tyD

istu

rban

ceC

arb

onat

eco

nte

nt

Infe

rred

veg

etat

ion

Sara

ng

24/

14(x

25

2.28

1,p

50.

131)

Poi

ntb

ar,l

evee

,an

dcr

evas

se-

spla

y

Stre

aman

dle

vee

dep

osit

sof

2nd

-an

d3r

d-o

rder

stre

ams

Wel

ld

rain

edV

ery

imm

a-tu

reV

ery

fre-

quen

tU

pp

eran

dlo

wer

hor

izon

sst

ron

gly

calc

ar-

eou

s;n

odu

les

not

pre

sen

t

Ear

lysu

cces

sion

;st

ream

sid

e

Lal

13/

0*( x

25

5.70

6,p

50.

017)

Old

,low

-ly

ing

floo

dp

lain

;dis

-ta

nt

from

stre

ams

Flo

odp

lain

allu

vi-

um

of2n

d-

and

3rd

-ord

erst

ream

s

Wel

ld

rain

ed,

wit

hoc

casi

onal

floo

din

g?

Mat

ure

tove

rym

a-tu

re

Infr

equ

ent

Mat

rix

ofu

pp

erh

oriz

ons

dec

al-

cifi

ed;l

ower

ho-

rizo

ns

wit

hsc

atte

red

fin

en

odu

les

Old

gro

wth

;de-

cid

uou

sw

ood

-la

nd

ord

ryfo

r-es

t

Nar

anji

0/2*

( x2

55.

256,

p5

0.02

2)L

inea

rd

epre

s-si

ons

onab

an-

don

edch

ann

els

Top

sof

coar

sesa

nd

sof

3rd

-an

d4t

h-o

rder

stre

amch

ann

els

Wel

ld

rain

ed,s

ea-

son

ally

brie

fly

wat

erlo

gg

ed?

Mat

ure

tove

rym

a-tu

re

Infr

equ

ent?

Mat

rix

oflo

wer

hor

izon

sm

od-

erat

ely

calc

are-

ous;

nod

ule

sn

otp

rese

nt

Seas

onal

lyd

ry;

tall

rive

rin

efo

r-es

t,w

ith

lush

un

der

gro

wth

Bh

ura

19/

2*( x

25

4.35

8,p

50.

037)

Ab

and

oned

chan

-n

els

and

leve

es,

flan

kin

gSa

ran

gso

ils

Ch

ann

elan

dle

-ve

ed

epos

its

of2n

d-

and

pos

si-

bly

3rd

-ord

erst

ream

s

Wel

ld

rain

edIm

mat

ure

tom

atu

reF

requ

ent?

Up

per

and

low

erh

oriz

ons

mod

-er

atel

yca

lcar

e-ou

s;w

ith

smal

lsc

atte

red

nod

-u

les

Stre

amsi

de

gal

-le

ryfo

rest

Son

ita

12/

13*

( x2

59.

289,

p5

0.00

2)Yo

un

gfl

ood

-p

lain

sn

ear

stre

ams

Lev

eean

dfl

ood

-p

lain

dep

osit

sof

2nd

-an

dp

ossi

bly

3rd

-or-

der

stre

ams

Wel

ld

rain

edan

dd

ryw

ith

only

occa

sion

alfl

ood

ing

Imm

atu

reto

mat

ure

Infr

equ

ent?

maj

orfl

ood

s

Up

per

hor

izon

sm

oder

atel

yd

e-ca

lcifi

ed;l

ower

wit

hsc

atte

red

,sm

all

nod

ule

s

Mid

-su

cces

sion

;d

ryw

ood

lan

dor

fore

st

Pil

a3/

0( x

25

1.19

5,p

50.

274)

Low

-ly

ing

aban

-d

oned

chan

nel

san

dd

epre

s-si

ons;

pro

xim

alcr

evas

se-s

pla

ys

San

dy

allu

viu

mof

3rd

-ord

erst

ream

s

Mil

dly

wat

er-

log

ged

onse

a-so

nal

bas

is

Mat

ure

?U

pp

eran

dlo

wer

hor

izon

sm

od-

erat

ely

calc

are-

ous,

scat

tere

d,

smal

ln

odu

les

Moi

st,l

owla

nd

;g

rass

yw

ood

-la

nd

orw

ood

edg

rass

lan

d

Kh

akis

tar

5/0

( x2

52.

020,

p5

0.15

4)A

ban

don

edch

an-

nel

s,sw

ales

,an

dfl

ood

pla

inla

kes;

away

from

acti

vech

ann

els

Cal

care

ous

and

clay

eyal

luv

ium

of2n

d-o

rder

stre

ams

Seas

onal

lyw

ater

-lo

gg

edan

dp

er-

hap

sse

ason

ally

dry

Imm

atu

reF

requ

ent,

floo

d-

pro

ne

Up

per

hor

izon

sve

ryw

eak

lyd

e-ca

lcifi

ed,v

ery

smal

lor

no

nod

ule

s

Seas

onal

lyw

et;

low

lan

dfo

rest


TA

BL

E3.

Con

tin

ued

.

Pal

eoso

lF

requ

ency

†To

pog

rap

hic

sett

ing

Par

ent

mat

eria

lD

rain

age

Mat

uri

tyD

istu

rban

ceC

arb

onat

eco

nte

nt

Infe

rred

veg

etat

ion

Pan

du

3/0

(x2

51.

195,

p5

0.27

4)E

leva

ted

,dis

tal

par

tsof

leve

eSa

nd

yan

dsi

lty

dep

osit

sof

2nd

-or

der

stre

ams

Wel

ld

rain

edm

ost

year

;wat

er-

log

ged

atti

mes

?

Ver

ym

atu

reIn

freq

uen

t?U

pp

erh

oriz

ons

calc

areo

us;

low

-er

wit

hab

un

-d

ant,

larg

en

od-

ule

s

Seas

onal

lyw

et;

fore

stor

woo

d-

lan

d

Kal

a1/

0( x

25

0.39

1,p

50.

532)

Smal

lab

and

oned

chan

nel

,floo

d-

pla

inla

ke,o

rm

oist

floo

d-

pla

in

Silt

yd

epos

its

of3r

dor

der

stre

ams

Seas

onal

lyw

ater

-lo

gg

ed,a

lso

seas

onal

lyd

ry?

Mat

ure

tove

rym

a-tu

re

Fre

quen

t,an

nu

ally

inu

nd

at-

ed?

Up

per

hor

izon

sd

ecal

cifi

ed;

low

erw

ith

abu

nd

ant

nod

-u

les

Gra

sses

orse

dg

esw

ith

scat

tere

dtr

ees;

wet

woo

ded

gra

ss-

lan

d

*St

atis

tica

lly

sig

nifi

can

td

iffe

ren

cein

rela

tive

freq

uen

cyb

etw

een

low

eran

du

pp

erse

ctio

ns.

†N

um

ber

ofo

ccu

rren

ces

inlo

wer

sect

ion

/n

um

ber

ofo

ccu

rren

ces

inu

pp

erse

ctio

n.

the floodplain (Willis 1993b). Low areas weresites of short-lived, perennial or seasonalponds and swamps. Such low areas wouldhave been rapidly filled by crevasse-splaysand levees associated with the larger second-and third-order channels, and with sedimentfrom even smaller fourth-order floodplainstreams (Fig. 10) (Willis 1993a; Zaleha 1997a).Preserved floodplain deposits therefore con-sist predominantly of low-angle crevasse-splay lobes that built out into low areas, someof which contained wetlands or lakes, alongwith minor contributions from levees andsmall floodplain channels (Behrensmeyer etal. 1995; Willis and Behrensmeyer 1995). Thepreserved features of the fourth-order flood-plain streams indicate that they typically cutinto preexisting floodplain deposits, had sin-gle rather than braided channels, and weremuch smaller (10 to less than 100 m wide)than the individual channels of the third-or-der streams. The preserved deposits are most-ly ribbon shaped, indicating that the channelsdid not migrate laterally, perhaps becausethey were stabilized by vegetation (Willis1993a). Flow in these smaller channels was ap-parently slower and more episodic through-out the year than in the major channels andmay have ceased at times as the smaller chan-nels became sites of ponded water (Willis1993a). In all three formations the smallerfloodplain channels are exceptionally impor-tant as sites of vertebrate bone accumulation.Most of the larger fossil localities are fromsuch depositional environments or from fine-grained fills associated with the tops of sec-ond- or third-order channels (Badgley andBehrensmeyer 1980, 1995; Behrensmeyer 1987;Badgley et al. 1995; Behrensmeyer et al. 1995;Willis and Behrensmeyer 1995).

Floodplain paleosols are well developed,common, and varied throughout the whole ofthe Siwalik sequence, although they have beenstudied in detail only very locally (Retallack1991). The Chinji and Dhok Pathan Forma-tions both have a larger number of well-dif-ferentiated paleosols than the Nagri, in keep-ing with overall differences in proportions ofchannel and overbank deposits. None of theformations have paleosols that are readily as-signed to modern soil types, although they are


broadly similar to modern soils of the Indo-Gangetic plain (Retallack 1991; Behrensmeyeret al. 1995; Quade and Cerling 1995), espe-cially those forming under 258C mean annualtemperature and 1400 mm/yr precipitation(Behrensmeyer et al. 1995). The nine paleosolseries described by Retallack (1991) (Table 3)differed primarily in topographic setting, fre-quency of disturbance, degree of maturity,drainage, and parent material, as well as in-ferred vegetation and precipitation. At allstratigraphic levels, however, the paleosolsgive evidence of seasonal differences in thewater table that produced waterlogging withformation of Fe-Mn nodules, followed by des-iccation with leaching and precipitation of cal-cium carbonates. Leaching of matrix carbon-ate was essentially complete on coarse depos-its and well-drained sites. There is also evi-dence for intense oxidation at the time of soilformation, which depleted organics and pre-sumably helped destroy bone in mature pa-leosols as well (Willis 1993a; Behrensmeyer etal. 1995; Zaleha 1997b).

Temporal changes in the Siwalik fluvial sys-tem have been documented, but the degree towhich they are related to climate or subsi-dence, or are due simply to autocyclic dynam-ics of the fluvial system, is not clear (Willis1993b; Zaleha 1997b). The transition betweenthe Chinji and Nagri Formations has been in-terpreted as due to the progradation of a sec-ond-order emergent system over a smaller,third-order interfan system (Willis 1993b; Za-leha 1997b). The Nagri and Dhok Pathan tran-sition, on the other hand, seems a well docu-mented case of the coexistence of two contem-poraneous systems (Behrensmeyer and Tauxe1982), the larger of which diverted from thestudy area between 10.1 and 9 Ma. Chinjifloodplains were apparently more poorlydrained than those of the Nagri Formation,with the abundance of lacustrine deposits andiron-concretion-rich paleosols indicating great-er seasonal waterlogging (Willis 1993b; Beh-rensmeyer et al. 1995; Willis and Behrensmey-er 1995). Floodplain deposition in the Chinji,and perhaps Dhok Pathan, Formation was alsomore episodic than in the Nagri, with longer-lived small floodplain channels. Initially, flowin these small channels was continuous year-

round, but in later stages it became more ep-isodic with periods of subaerial exposure asthe channels were abandoned and later infil-led. By contrast, in the Nagri Formation eventhe initial stages of channel flow may havebeen more seasonal, with deposition of sedi-ment occurring mostly during floods (Willis1993a; Willis and Behrensmeyer 1995). Specif-ic, important differences between the rivers ofthe Nagri and Dhok Pathan Formation includedecreased channel size and discharge in thelatter, as well as decreased avulsion period(Zaleha 1997b).

Our lateral facies studies in the Kaulial Kassection and nearby sections (Behrensmeyer1987) provide a general reconstruction of thelandscape in the Khaur region between 10.1and about 7.0 Ma. Before ca. 9 Ma, in the up-per part of the Nagri and lower part of theDhok Pathan Formations, channels hundredsof meters wide were separated by fine-grainedinterfluves with modest topographic relief.The highest areas in these interfluve flood-plains were associated with low-angle cre-vasse-splay fans, and the lowest areas withabandoned channels. Levees were generallypoorly developed and channels show little ev-idence of lateral migration, indicating fre-quent avulsion. Floodplain soils are carbon-ate- and iron-rich and oxidized, suggestingseasonal alternation between waterloggedand moist-dry with intense bioturbation. Theinterfluves were crossed by small channelstens of meters wide that were tributaries ordistributaries associated with the larger chan-nels and probably active mainly during floodperiods. The physical habitat between 10.1and ca. 9 Ma varied on a scale of tens to hun-dreds of meters from floodplain silt-clay soilsto sandy splay substrates to channel corridorswith wet to dry beds depending on the sea-son.

After ca. 9 Ma, in the upper part of the DhokPathan Formation, large scale, braided chan-nels draining foothill source areas were morecommon in the Khaur region, expanding thearea covered by sandy substrates (e.g., aban-doned or seasonally dry channels) at the ex-pense of silt/clay soils. Small-scale, erosionalfloodplain channels became much less fre-quent, reflecting a reduction in the role of such


channels in distributing floodwaters. This dif-ference could reflect a shift in the rainfall re-gime and/or an increased role of the majorrivers in controlling the topography anddrainage of the alluvial plain. Low-angle cre-vasse-splay fans continued to create minorpositive relief on this plain, but overall the in-terfluves were broader, flatter, and more ho-mogenous than in the upper Nagri and lowerDhok Pathan Formations. Lateral variation inthe physical habitat was on the order of hun-dreds to thousands of meters. Pedogenic car-bonate is more poorly developed in the upperDhok Pathan floodplain soils, which areleached of matrix CaCO3, and this may reflectmore effective soil drainage with less oppor-tunity for vertical transpiration and evapora-tion of water-logged soils during the dry sea-son.

This reconstruction of environmentalchange through time applies only to theKhaur region. It is likely that other parts of theSiwalik alluvial plain were changing at differ-ent rates and in different ways, providingvarying ecological opportunities for the floraand fauna. Mountain-proximal regions, suchas Khaur, were subject to the effects of en-croaching lowland-sourced fluvial systemsearlier than mountain-distal regions to thesouth. Nevertheless, the environmental devel-opments seen in the north are indicative ofwhat would be expected throughout the largerSiwalik system in the late Miocene, as low-land-sourced rivers became more dominantand the climate changed to greater seasonalaridity with more intense monsoons.

Retallack (1991) described and named ninepaleosol series from two short sections in thelower Dhok Pathan Formation at ca. 9.4 and8.6 Ma. In these two sections, which are with-in and just above the Nagri–Dhok Pathan tran-sition, he documented differences in the fre-quencies of soil types (Table 3). He interpretedthe dissimilarities as resulting primarily fromdifferences in parent material, local topogra-phy, and drainage. Seasonality of rainfall andlevels of disturbance may also have affectedsoil type (Retallack 1991). The Bhura Seriesand Sonita Series in particular are a couplet offrequently occurring, moderately developedpaleosols that formed in similar settings on

similar parent material but differed in drain-age and intensity of seasonal drying. The LalSeries and Naranji Series are another pair ofmature paleosols that differed primarily in to-pographic setting, parent material, and dis-tance from channels, and secondarily in drain-age. Retallack’s overall interpretation is thathis sample of younger paleosols (at 8.6 Ma)formed closer to the channels and were betterdrained, with perhaps more marked season-ality. These observations on the paleosols canbe related to the replacement of the emergentNagri system by an interfan system, and theymay also indicate a different climatic regime.However, although there are statistically sig-nificant contrasts between the older and youn-ger sections in the frequencies of Lal, Bhura,Sonita, and Naranji soils (Table 3), subsequentwork in the Kaulial area indicates that his con-clusions should be viewed with caution. Wefind Retallack’s sections were too short andtoo laterally restricted to include a represen-tative sample of the soils of the two time in-tervals. Furthermore, both intervals havemuch more variation in soil features associat-ed with lateral and vertical differences in thefluvial systems than can be accommodated byhis system of soil classification.

Retallack (1991) found no evidence for ex-tensive grasslands (e.g., ped structure, mas-sive fine roots), although he suggested thatthe Pila Series and Kala Series paleosolsformed under occasionally waterlogged,grassy woodlands. Paleosols belonging to ei-ther series are uncommon, occurring only fourtimes among 111 recognized paleosols, andare found only in the older section. Neverthe-less, yellow and light-colored soils becomemore common after 8 Ma. These could be ex-amples of his Pila or Pandu soils, or theymight be unrecognized soil types. Fewer pa-leosol types were reported by Retallack (1991)in the upper section (four versus eight in thelower), which he interpreted as evidence for aless varied landscape in the younger time in-terval. However, considering the much smallernumber of recognized soil horizons in the up-per section (31 versus 80), these proportionsare not different. Nevertheless, our own ob-servations suggest that the upper part of DhokPathan Formation (ca. 9.0–5.5 Ma) is less var-


FIGURE 11. Isotopic values of paleosol carbonates by age. A, d13Carbon. B, d18Oxygen. Ages from Appendix 3. Allisotopic data are from Quade and Cerling 1995.

ied than the lower part (ca. 10.1–9.0 Ma) intypes of paleosols; in the sizes, geometry, andfrequency of channels; and in local topogra-phy. The difference between the older andyounger parts of the Dhok Pathan Formationis also expressed in the decreased frequencyof fossil-preserving environments in the up-per Dhok Pathan (Fig. 6) (Behrensmeyer 1987).

It is noteworthy that at any stratigraphiclevel there is always considerable diversity insoil types, and by inference considerable di-versity in vegetation. Low areas containedseasonal swamps or ephemeral ponds, where-as the paleosols suggest that other areas hadextensive wet upland and lowland forests, dryold-growth deciduous woodlands, tall river-ine or gallery forest with lush undergrowth,and wet grassy wooded meadows, as well astracts of secondary growth and early succes-sion (Retallack 1991). These vegetation typeswould have been common, coexisting ele-ments of the landscape. It seems likely that or-ganisms concentrated in some areas, andmaybe especially in the lush vegetated corri-dors along the streams during the dry season.Animals may also have retreated to the highercrevasse-splay fans during flood periods ormigrated off the floodplain altogether. Afterca. 9 Ma, the decreased environmental vari-ability, increased soil drainage, and particu-larly the diminished network of vegetatedchannel corridors on the alluvial plain wouldhave reduced the diversity of habitats avail-able to the Siwalik flora and fauna.

Isotopic Evidence

Temporal changes have also been found inthe stable isotopic content of Siwalik paleosolsand fossil material (Quade et al. 1989, 1992;Morgan et al. 1994; Stern et al. 1994; Quadeand Cerling 1995). Quade and others (Quadeet al. 1989, 1992; Quade and Cerling 1995) firstreported differences in stable carbon and ox-ygen isotope ratios in soil carbonates in threePotwar sections, noting rapid shifts in bothd13C and d18O values in the late Miocene. Qua-de and others initially reported the ages oftheir samples using the Berggren et al. (1985)timescale. We have now recalculated the agesof the Kaulial and Jalalpur samples using theCande and Kent (1995) timescale (Appendix 3,Fig. 11A). (The upper, post-9-Ma part of theGabhir section is not included because of un-certainties in correlation to the magnetic po-larity timescale.) The new ages show that thebeginning of the d13C shift is slightly youngerthan 8.1 Ma, when a value greater than29.0‰ first appears in the Kaulial section.The shift is more or less completed by 5.9 Ma,after which values less than 22.0‰ no longeroccur. (These beginning and end dates are re-spectively, approximately 800 Kyr and 400Kyr older than previously reported.) Howev-er, as Quade and Cerling (1995) note, the shiftis complex and certainly not monotonic. Sam-ples that are strongly 13C depleted occur evenafter 7.4 Ma, which is when the oldest samplewith a value greater than 24.0‰ occurs. Val-ues greater than 22.0‰ are present as early


as 6.8 Ma, suggesting that the more restrictedinterval between 8.1 and 6.8 Ma was the timeof greatest change. There is no apparent dif-ference in the timing of the carbon isotopicshift from the northwest at Kaulial to thesoutheast at Jalalpur (Fig. 1).

The d18O shift noted by Quade and others(Quade et al. 1989; Quade and Cerling 1995)is also conspicuous and appears to be as rapidas the carbon shift, although beginning earlier(Appendix 3, Fig. 11B). In the well-dated Kau-lial section, the oldest available sample isgreatly enriched relative to older samplesfrom the Jalalpur section. It indicates that theshift in d18O values occurred as early as 9.15Ma. Unfortunately, the ages of the Jalalpursamples older than 8 Ma are imprecise and donot establish the age of the transition’s lowerboundary. With the exception of one samplewith an age between 10.7 and 10.5 Ma, all areless than 28.0‰, and the 9.15 Ma Kaulialsample might therefore be taken as marking apoint of significant change. This date is at least600 Kyr older than reported by Quade andCerling (1995) and 1.1 Myr older than the be-ginning of the shift in carbon values. The d18Oshift is also complex (Quade and Cerling1995). The transition takes place in phases andlasts longer than the carbon transition. Withthe exception of a strongly depleted sample at7.27 Ma, Ma d18O values after 8 are greatly en-riched compared with older ones, whereaspost-6-Ma values are even more enriched. Theinterval between 9.15 and 7.27 Ma may delin-eate a transitional period of unusually widevariation, with the most intense change com-ing before 8.0 Ma.

Paleosol d13C ratios reflect the isotopic com-position of vegetation growing at the site ofsoil formation (Cerling et al. 1989). Quade andothers consequently interpreted the reporteddifferences in carbon isotope ratios as due tolong-term changes in vegetation (Quade andCerling 1995). They considered d13C valuesmore positive than 29‰ to indicate the pres-ence of at least a small number of plants usingthe C4 photosynthetic pathway, while morenegative values indicated the presence of onlyC3 plants. Paleosol nodules with d13C valuesgreater than 22‰ formed under pure C4

grasslands, while those with values near

25‰ formed under mixed vegetation aver-aging about half C3 and half C4 during the pe-riod of soil formation (Quade et al. 1989; Qua-de and Cerling 1995). Accordingly, Quade andothers (Quade et al. 1989; Quade and Cerling1995) interpreted the carbon shift to be achange from C3-dominated plant communi-ties to C4-dominated ones; that is, fromclosed-canopy forest and woodland commu-nities to more open types of vegetation.

The transition between C3- and C4-domi-nated communities was rapid. Plant commu-nities composed predominantly of C4 speciesappear as early as 7.4 Ma, whereas pure C4

grasslands were present at 6.9 Ma. Commu-nities composed exclusively or predominantlyof C3 species apparently disappeared from therecord after 7.0 Ma, when the last isotopic val-ues less than 29.0‰ occur. Thus, althoughinitially the transitional plant communitieswere spatial and temporal mosaics with for-est, woodland, brush, and grassland, after 6.9Ma apparently there were only grasslandsand mixed C3 and C4 communities, with littleor no pure C3 vegetation. The available evi-dence also indicates that after 5.9 Ma C4 grass-lands were the dominant and maybe only veg-etation on the Siwalik alluvial plain (Quade etal. 1989; Quade and Cerling 1995). It shouldbe noted, however, that the inferred absence ofC3 communities is based on negative evidence,and there are numerous paleosols throughoutthis time interval that were not sampled be-cause they lack carbonate and/or free carbon(see also Table 3). These soils may have con-tinued to support C3 vegetation.

Because environmental and climatic factors,such as CO2 partial pressure, light intensity,rainfall, and temperature, govern growth andcompetition between C4 and C3 plants, the car-bon shift implies climatic and environmentalchange. Most C4 plants are grasses adapted forgrowth under higher temperature and lowerrainfall regimes than C3 grasses, trees, andshrubs. Increasing seasonality and reducedprecipitation have therefore been cited to ac-count for inferred temporal alteration in veg-etation (Quade et al. 1989; Quade and Cerling1995).

Carbon isotope ratios in mammalian toothenamel and avian eggshells (Quade et al. 1992;


Morgan et al. 1994; Stern et al. 1994) have alsobeen used to reconstruct ancient Siwalik veg-etation. In this case the d13C values of enameland eggshell reflect the isotopic compositionof the animal’s diet, but because animals canbe highly selective feeders, such isotopic datacannot directly document relative composi-tion of the vegetation. Rather, they only indi-cate plant availability. Enamel samples alsorepresent very short intervals of time withinthe animal’s life span (1–10 yr), whereas car-bonate nodules form over longer periods(100–1000 yr). The time-averaging of an iso-topic signal over longer periods in soil nod-ules could obscure any record of C4 vegetationpatches in a dominantly C3 environment. Bothmammal and avian isotopic data sets show atransition from C3-dominated diets to mixedC3/C4- and C4-dominated diets, with the tran-sition being most striking after 8 Ma, as in thepaleosol carbonate data. Nevertheless, bothMorgan et al. (1994) and Stern et al. (1994)found some evidence for a C4 dietary compo-nent before 8 Ma. Although the values arenear the threshold for pure C3 vegetation (Cer-ling et al. 1997), the animal data do suggestthat C4 vegetation was present earlier than thepaleosol evidence shows. This conclusiongains support from occlusal microwear dataon teeth, which also document herbivoreswith substantial amounts of grass in theirdiet, even in the middle Miocene (Morgan etal. 1994). Although seemingly in conflict withthe soil carbonate data, it is likely that certainkinds of habitats, and especially disturbedhabitats with vegetation in an early stage ofsuccession, are not represented in the paleosolisotopic record. Many Siwalik paleosols lackpedogenic carbonate or have dispersed nod-ules too small to sample, because either theyare immature or they formed on well-drainedor permanently waterlogged sites where car-bonate was not precipitated during pedogen-esis (Quade and Cerling 1995). Because car-bonate deficient soils are unrepresented in theisotopic samples, any vegetation restricted tothem is consequently also unrepresented (orunderrepresented) in the isotopically sampledrecord. Underrepresentation is especially like-ly for disturbed habitats with extensive openareas and grasses, because they are most likely

to have formed on immature or well-drained,nutrient-poor soils. Nevertheless, C4 vegeta-tion must have become more widespread onthe floodplain after 8.1 Ma, because 13C-en-riched nodules begin to occur at that time inmany different paleosols.

Paleosol d18O ratios are determined largelyby temperature, evaporation, and the d18O val-ues of precipitation at the site of soil formation(Quade and Cerling 1995), but interpretationof the shift in oxygen ratios reported in the Si-waliks is not straightforward. Quade and Cer-ling (1995) outlined several climatic factorsthat alone or together might have been re-sponsible for the change, including an in-crease in mean temperature by about 58C, adecrease in total amount or an increase in theseasonality of rainfall, a change in the sourcesof precipitation, or increased direct evapora-tion from the soil. Some of these same factorsare also implicated in the carbon isotopicshift, but the connection between the carbonand oxygen isotopic shifts is not understood.Significantly, although the beginnings of thetwo events are separated by over one millionyears, the interval of most intense d18O changeends just when the carbon shift starts at 8.1Ma. In addition, the most negative d13C valuesreported in the entire sequence are two sam-ples from the Kaulial section at 9.15 and 8.97Ma, which is when the d18O shift first starts.However, no simple relationship between car-bon and oxygen values exists, as many of the13C-depleted samples from the upper part ofthe Kaulial section are 18O-enriched, whereasothers are 18O-depleted (Quade and Cerling1995).

Correlations to Oceanic and HimalayanTectonic Events

Quade et al. (1995) have documented a sim-ilar carbon and oxygen isotopic shift begin-ning at ca. 7.0 Ma in terrestrial sediments inNepal, whereas terrestrially derived sedi-ments on the Bengal fan record a change ind13C values after ca. 7.8 Ma (France-Lanordand Derry 1994; date converted from Berg-gren et al. [1985] timescale). Mineral assem-blages and rates of deposition on the Bengaland Indus fans also show marked contrastsbetween ca. 8 and 7 Ma (Amano and Taira


1992; Rea 1992; dates converted from Berg-gren et al. [1985] timescale), whereas forami-niferal evidence from the Arabian Sea sug-gests the beginning of strong summer up-welling at 9.4 Ma that continues to increaseuntil about 8.2 Ma (Kroon et al. 1991, datesconverted from Berggren et al. [1985] time-scale). Although considerable uncertaintyover the chronology of Himalayan and Tibet-an Plateau uplift still exists, the consensus isthat a phase of uplift and rapid erosion oc-curred from as early as 11 Ma to as late as 6Ma (Amano and Taira 1992; Harrison et al.1993; Turner et al. 1993; Meigs et al. 1995; butsee also Copeland 1992; Burbank et al. 1993;Coleman and Hodges 1995). Although an el-evated Tibetan Plateau and Himalayas mayhave existed long before 11 Ma, late Mioceneuplift is seen as important in generating or in-tensifying the South Asian monsoon, as indi-cated by the onset of summer upwelling in theArabian Sea (Prell and Kutzbach 1992; Rea1992; but see also Ramstein et al. 1997). An in-tensified monsoon may then explain the ob-served d18O shift, through its effect on season-ality and precipitation in northern India andPakistan in the interval between 9.4 and 8.2Ma. Elevation of the Tibetan Plateau and Him-alayas also may have been a critical factor inthe evolution of late Miocene cooling on aglobal scale and consequent vegetationchange, through its effects on chemical weath-ering rates, organic carbon burial, and atmo-spheric CO2 content (Raymo and Ruddiman1992; Molnar et al. 1993; Filippelli 1997). At-mospheric CO2 content alone also has beenseen as a cause of the isotopic shifts, throughits direct effect on C3/C4 plant competitive re-lationships (Cerling et al. 1993, 1997). Paganiet al. (1999), however, found no evidence tosupport the claim.

Climate, Depositional System, andVegetation

Although climate and interspecific compe-tition are presumed to have been the primaryagents acting on the plant communities, thedepositional system also strongly affected thevegetation and landscapes of the 10.7–5.7-Mainterval. Preserved deposits include the sandychannels of large and small streams, their as-

sociated silty and sandy levees and crevassesplays, and the silt and clay fills of abandonedchannels, ponds, and other depressions (Beh-rensmeyer et al. 1995; Willis and Behrensmey-er 1995). Position on the floodplain relative tothe mountain front and axial trunk stream, aswell as distance from the channel belts, weretherefore important determinants of local to-pography, soil substrate, drainage, and fre-quency of disturbance. Because depositionwas on both alluvial fans and interfans, areasof low and high elevation should have been ir-regularly distributed and patchwork-like,with low areas being both near to and distantfrom the channels (Behrensmeyer et al. 1995;Willis and Behrensmeyer 1995). Similarly, con-ditions affecting soil formation varied withproximity to the channels. Some areas wouldhave been sites of continuing frequent depo-sition, whereas others would have had longperiods for soil development between epi-sodes of deposition. Presumably vegetationresponded to specific local conditions. Con-sequently, plant communities formed mosaicsthat changed in an irregular manner over dis-tances of hundreds of meters to a few kilo-meters, owing to edaphic factors such as to-pography, soil parent material, drainage, andheight of the water table. The amount and sea-sonality of precipitation also would have beencritical, although varying on a larger regionalscale. Finally, high and unpredictable levels ofdisturbance due to flooding, stream avulsion,and fire would have left tracts of vegetation indifferent stages of succession, adding to theoverall habitat mosaic. Forest glades, swamps,and meadows, as well as areas of secondarygrowth, should have been common features ofthe landscape, present within what were oth-erwise predominantly forests and woodlands.More speculatively, brushland and perhapssome grasslands also may have been presentprior to the widespread expansion of C4 veg-etation. Presumably, this variety of vegetationcreated a complex of habitats favorable to dif-ferent animals.

Climate change per se also may have beenthe primary agent affecting vegetation duringthe study interval. Increasing seasonality, re-duced precipitation, and an increase in the in-cidence of fire have commonly been cited to


account for inferred differences in vegetation,as well as differences in the Siwalik sedimentsand faunas (e.g., Krynine 1937; Quade et al.1989; Retallack 1991; Quade and Cerling1995). These effects, however, are difficult toseparate from those of the fluvial system’s dy-namics. The latter were especially likely tohave been instrumental in transforming land-forms and vegetation between 10.1 and 9 Ma,when a large emergent river system (NagriFormation deposits) was replaced across thestudy area by a smaller interfan system (DhokPathan Formation deposits). The lower DhokPathan interfan system had smaller streamsand narrower channel belts, with slower, moreseasonally variable flow and more frequentavulsions than the emergent Nagri system(Willis 1993b; Willis and Behrensmeyer 1995;Zaleha 1997b). The lower Dhok Pathan systemalso had a more widely fluctuating water ta-ble, more frequent flood disturbances, andperhaps a more variable spatial mosaic ofabandoned channels, soils, and topography.Small floodplain channels, which are impor-tant sites of vertebrate bone accumulation,were initially very common on the interfanDhok Pathan system but became progressive-ly less so after ca. 9 Ma (Behrensmeyer 1987),whereas the preserved soils seem to haveformed nearer to the channels, where theywere better drained and more susceptible todrying (Retallack 1991). After 9 Ma, the inter-fluves were broader with less topographic re-lief, and the physical environment appears tohave been more homogenous than in the lowerDhok Pathan Formation.

Because small floodplain channels were socommon, fossil productivity is much higherafter 10 Ma in the interfan lower Dhok Pathansystem then it is in the older emergent Nagrisystem. Differences in fossil productivity be-tween the older and younger parts of theDhok Pathan Formation are also expressed indecreased frequency of fossils in the upperDhok Pathan Formation, especially after 7.1Ma.

Biostratigraphic Record

In the following sections we examine as-pects of faunal change in the late Miocene ofthe Siwaliks. To reconstruct a picture of such

change, we ideally would like to know for eachtime interval the original relative speciesabundances within the community, as well asthe patterns of species turnover. However, thisis not possible with fossil taxa because of thetaphonomic filters that modify death assem-blages, fossil preservation, and recovery. Asdiscussed below, our approach assumes thateach species has a more or less constant abun-dance over its duration. At present we see noevidence for major fluctuations in abundancein more than a few species, and our recordcannot detect and evaluate more subtle differ-ences in abundance. At the family level, bio-stratigraphic surveys and skeletal-elementanalyses do show changing relative abun-dances. For example, the numbers of recov-ered astragali indicated that between 9.3 and6.3 Ma tragulids decrease and bovids increasein relative abundance (Barry et al. 1991; M. E.Morgan unpublished data). We include somediscussion of abundance patterns, but here weare more interested in the patterns of speciespresence or absence through time.

We begin by examining the patterns of ob-served first and last occurrences, that is, the dis-tributions of empirically observed fossil data,and using them to infer the distributions offirst and last appearances. We argue that theseinferred first and last appearances better ap-proximate the ‘‘true’’ appearances and dis-appearances of species in the Potwar. We thenuse several approaches—comparative anato-my, stable isotopes, occlusal microwear, bodysize—to infer diet and habitat preference in anumber of these species in order to assess theextent to which community structure changesor remains stable over time. Finally, we ex-amine the relationships between patterns offaunal change and indicators of environmen-tal change.

Because the Potwar sequence is a local se-quence, it is worth noting that first and last ap-pearances are likely to be due to geographicrange extensions and contractions, not to ac-tual speciation or extinction events. Thiswould be the case for any local sequence. Amoment’s reflection reveals the enormousamount and quality of data, in terms of geo-graphical breadth and chronological control,necessary to document a sequence of actual


species originations, extinctions, and species-wide dispersal events for these taxa. In whatfollows ‘‘extinction’’ refers both to local ex-tinctions resulting from range contractionsand to global extinctions. ‘‘Appearance’’ refersto immigration as well as speciation.

Fossil Material

The Siwalik fossil material consists primar-ily of isolated dental, cranial, and postcranialelements of mammals, reptiles, and fish. Mostof this material is fragmentary and difficult toassign to species; however, occasional more-complete specimens provide a comparativebasis for assigning fragmentary fossils to spe-cies. We include in our analysis 115 mamma-lian taxa from eight orders: Insectivora, Scan-dentia, Primates, Tubulidentata, Lagomorpha,Perissodactyla, Artiodactyla, and Rodentia(Appendix 4). The major taxa omitted are theCarnivora, Elephantoidea, and Rhinoceroti-dae. These latter taxa contributed importantlyto the Siwalik Neogene mammalian commu-nities, but we have eliminated them from anal-ysis primarily because comprehensive identi-fications of our material are still ongoing.

The stratigraphic occurrences reported hereare based on collections in the American Mu-seum of Natural History and the Yale PeabodyMuseum, as well as on collections made by theGSP-Harvard project. In some cases speci-mens that indicate significant range exten-sions are known from very poorly dated lo-calities but are not included because of the in-adequate dating. Such specimens include thetypes of Tragoportax salmontanus (ca. 7.9–8.1Ma) and Nycticeboides simpsoni (ca. 8.8 Ma),and a skull of Orycteropus browni (ca. 7.9 Ma).In addition, the first occurrence of Hexaproto-don sivalensis might be as old as 6.1 Ma or even7.2 Ma, the first occurrence of Elephas plani-frons as old as 5.9 Ma, and the first occurrenceof Hippohyus lydekkeri as old as 6.0 Ma.

With a few exceptions, the analyses of thispaper are done with species or presumed spe-cies lineages that are treated as one taxon. Themost important exception is the equids, whichare represented by two or three species at allbut the earliest horizons. Some recent pro-gress has been made on the species-level sys-tematics of the group, which indicates that

there were probably no more than six speciesin the Siwaliks during the late Miocene. Thebulk of the available material consists of iso-lated teeth and skeletal elements that cannoteasily be assigned to recognized species. Weinclude the equids, however, because they arean important ecological component of latestMiocene fossil assemblages. Other exceptionsinclude lorisids, tree shrews, and dormice, allof which are numerically insignificant. Whererelative abundances are cited, the figures arebased on counts of the number of catalogedspecimens, corrected for multiple specimensthought to belong to one individual. Manyspecies are new and not yet described, andothers have not yet been confidently identi-fied. In this paper we refer to such taxa by thenumber of a locality that has a typical speci-men (e.g., Dorcatherium ‘‘Y373 species’’ or’’?Tragelaphini/D013 species’’). Some taxathat we think may be new are referred to as‘‘unnamed.’’

Stratigraphic Ranges and Pattern of First andLast Occurrences

Appendix 4 gives the age of first and last oc-currence on the Potwar Plateau for the 115taxa occurring in the interval from 10.7 to 5.7Ma. For each species, we define as a first or lastoccurrence event (FO or LO) the locality onthe Potwar Plateau that we infer to be strati-graphically the lowest or highest in the Potwarcomposite stratigraphic column, and thereforethe oldest or youngest. Note that this is de-pendent both on the empirical observationthat a fossil at a locality represents a particularspecies and on the inference that a locality isthe lowest or highest in our composite section.As discussed earlier, such inferences are gen-erally firm. The ages used are the locality ageestimates discussed previously.

Figure 12 shows the observed stratigraphicranges ordered from oldest to youngest (datafrom Appendix 4). Seventy-two species haveobserved ranges (i.e., both the first and the lastoccurrences) lying entirely within the 10.7–5.7-Ma interval, and another 43 have either afirst or a last occurrence within the study in-terval, giving a total of 187 events. With threeexceptions, there is a strong positive relation-ship between the number of events and the


FIGURE 12. Observed and inferred stratigraphic ranges of 115 taxa (data from Appendix 4). The numbers identifytaxa in Appendix 4.

number of collected specimens within an in-terval, and this is especially so at 10.4 Ma, 10.0Ma, 9.3–9.2 Ma, and 8.7 Ma (Fig. 13A,B). Oneexception to the concordance of well-sampledintervals and occurrence events is at 8.1–8.0Ma, when there are many fossils of both largeand small mammals, but only a slightly aboveaverage number of events. The other two ex-ceptions are a broad interval of time between7.0 and 7.3 Ma and the 6.4-Ma interval, duringwhich unusually large numbers of events oc-cur despite only average or below average fos-sil abundance.

The correlation between interval samplesize and number of first/last occurrences (Fig.13C) introduces a bias, making interpretationof the pattern of occurrences challenging(Badgley and Gingerich 1988). That is, epi-sodes of low and high turnover may be ‘‘real,’’or they may be artifacts of taphonomic factorsthat affected fossil productivity. Although se-quences of first and last occurrences will rare-ly record the ‘‘true’’ pattern of first and lastappearance, it is the ‘‘true’’ pattern we need toknow to discuss faunal dynamics adequately.It is therefore necessary to take variation in thenumber of specimens (‘‘data quality’’) into ac-

count, and to eliminate or at least minimize itseffects. Establishing confidence limits on spe-cies ranges is thus an important step towardunderstanding the dynamics of faunal change.In the following we refer to these ‘‘true’’ firstand last appearances as ‘‘inferred first/lastappearances’’ (IFA/ILA), or when we are re-ferring to their confidence interval limits, ‘‘in-ferred first/last appearance limits’’ (IFAL/ILAL). Because we have a long, continuous,and relatively fossiliferous sequence in the Po-twar Plateau, we are able to estimate such lim-its for a significant fraction of first and last oc-currences.

Calculating Limits for ChronostratigraphicRanges

In earlier papers (Barry et al. 1990, 1991,1995) we analyzed the pattern of faunal turn-over in the Potwar sequence by half-million-year intervals, and we judged its reliability us-ing ‘‘interval-quality scores’’ and indices ofcompleteness. Our goal was to assess the ex-tent to which the actual fossil record reflectedthe original fauna in any particular interval.More recently (Pilbeam et al. 1996) we arguedthat, in addition to assessing data quality by


FIGURE 13. Relationships between number of specimens and number of events. A, Number of first and last occur-rence events (gray bars) and number of specimens (black line) per interval for large mammals. B, Number of firstand last occurrence events (gray bars) and number of specimens (black line) per interval for small mammals. C,Number of events against number of specimens for each interval (n 5 102).


time intervals for the fauna as a whole, it isalso necessary to do so for each individual tax-on. For each species we should ask, How con-fident are we that the empirically observedfirst or last occurrence estimates the actualfirst or last appearance? And, if we lack con-fidence in the first or last occurrence, what isa reasonable limit to the species true strati-graphic range?

There are several possible approaches to theproblem of inferring appearances from occur-rences. Strauss and Sadler (1989), among oth-ers, used a method based on the distributionof the size of gaps between occurrences of aprescribed species to (1) set expectationsabout gap lengths and (2) estimate a confi-dence interval for the position of the endpointof its stratigraphic range. We previously ap-plied this approach to two genera, Sivapithecusand Sayimys (Flynn et al. 1990). Marshall ex-panded the Strauss and Sadler model in vari-ous ways, most interestingly by first using anapproach that relaxed some assumptionsabout the distributions of gaps and allowedfor the calculation of upper and lower boundson the confidence interval (Marshall 1994),and subsequently using measures of fossilpreservation and recovery bias to calculateconfidence intervals (Marshall 1997). The dis-tribution-free model (Marshall 1994), howev-er, requires a rich fossil record, with at leastseven fossil localities per taxon to calculate a50% confidence interval with a confidenceprobability of 0.95. This is a condition not metfor many Siwalik species. In addition, thestratigraphic distribution of gaps in the Si-waliks (Fig. 8) plainly violates a remainingcritical assumption that there is no correlationbetween stratigraphic position and gap du-ration. To use gap-based approaches, there-fore, it is first necessary to develop an inde-pendent preservation and collecting bias func-tion (Marshall 1997). This we have not yetdone.

Here we use instead an approach modifiedfrom ideas developed by Shaw (1964) andKoch (1987). This approach uses the abun-dance of a prescribed taxon relative to theoverall abundance of the fossils within its ob-served range to determine an expectation forfinding the taxon (its probability of occur-

rence). We then use the number of specimensin adjacent 100-Kyr intervals outside the ob-served range to place confidence limits on thetaxon’s range endpoints by determining atwhat point the expectation is not reasonablymet. As with gap-based approaches, the re-sulting confidence limits identify only a strati-graphic interval over which the taxon appearsor disappears. They do not specify a shorterinterval in which the taxon is most likely tohave appeared or disappeared. However, wesubsequently cull the data, using in our anal-yses only species with short confidence inter-vals. We believe the confidence interval limitsof the taxa that are not culled closely approx-imate the true first or last appearances.

For each taxon we estimate the IFA and ILAlimits by first calculating the probability ofcollecting at least one specimen of the pre-scribed species as:

rP 5 1 2 (1 2 n/m)i

where Pi 5 probability of finding taxon i; n 5number of specimens attributed to taxon i overits total observed range; m 5 total number ofspecimens of either small or large mammalspresent during the observed range of the tax-on; and r 5 number of small or large mammalspecimens in the interval, or series of succes-sive intervals, adjacent to the first or last oc-currence of taxon i. (We evaluate small andlarge mammals separately because of the dif-ferent recovery methods—screenwashing ver-sus surface collection.)

Our formulation differs from that of Koch(1987) in two ways. First, he estimated theprobability of occurrence for a taxon (n/m) asthe number of times (n) it was present in a se-ries of m samples or ‘‘collections,’’ whereas weestimate the probability as the number ofspecimens of the taxon relative to the totalnumber of fossils in one collection spanningthe whole observed range of the taxon. Sec-ond, Koch (1987) calculated the probability ofthe taxon occurring in a second series of rsamples, whereas we calculate the probabilityof one occurrence in a single additional sam-ple containing r specimens. In this way ourapproach is closer to that of Shaw (1964). Weare, however, most interested in finding howlarge m has to be, that is how many adjacent


intervals have to be merged, before our expec-tation of finding one additional specimen be-comes unlikely (given some critical value ofPi). We assume that the relative abundance ofthe specified taxon does not change through-out its stratigraphic range. That is, n/m as cal-culated over the taxon’s whole range is a goodestimator of the probability of occurrence atthe ends of its range (Shaw 1964).

We have set the critical value for Pi to 0.8,for us a reasonable balance between having anadequate number of species for subsequentanalyses and having a high level of confi-dence. The number of intervals required to ob-tain a probability equal to or greater than 0.8below the FO and above the LO were thencounted—these endpoints represent the limitsto the IFA or ILA range extensions, respec-tively. As an example, the murid rodent cf.Parapelomys robertsi has a FO in 8.1 Ma and aLO in 7.2 Ma. Between 8.1 and 7.2 Ma, thereare 521 small-mammal specimens (m) in ourcollections, 71 (n) of which are cf. Parapelomysrobertsi. In the intervals at 8.2, 8.3, 8.4, 8.5, and8.6 Ma, there are 1, 0, 4, 1, and 77 small-mam-mal specimens respectively. The cumulativesums of the values for Pi are then: 0.136, 0.136,0.519, 0.585, and 1.00. Using a cutoff value of0.8, we would place the limit of the IFA con-fidence interval in the 8.5-Ma interval, sincethe high Pi value for the 8.6-Ma interval indi-cates that it is unlikely that the species waspresent in that interval. The true appearance,however, could lie anywhere between 8.1 and8.5 Ma. Similarly, there are 2, 17, and 2 small-mammal specimens in the intervals at 7.1, 7.0,and 6.9 Ma, giving successive probabilities of0.254, 0.938, and 0.954, and a limit for the ILAconfidence interval of 7.1 Ma.

Although the need to differentiate betweenobserved and inferred appearances/disap-pearances is clear, not all of the range exten-sions are equally useful. For example, the limitto the ILA range extension for cf. Parapelomysrobertsi falls within 100 Kyr of the observedLO, but the limit for the IFA range extensionis 400 Kyr older than the FO. We believe thatthe shorter ILA range extension is a more in-formative, and therefore more useful, estimatethan the longer IFA extension. This is becausethe closer the observed occurrence and in-

ferred limit, the smaller the confidence inter-val and the more likely that the inferred ap-pearances and disappearances, as equivalentsof statistical sample estimates, represent pop-ulation parameters (i.e., the ‘‘true’’ endpoints).Ideally we would prefer to use in our analysesonly those species for which the difference be-tween the observed occurrence and the in-ferred limit is less than 100 Kyr (the scale ofresolution we seek to achieve). Because doingso would reduce our data set too much, we usea 200-Kyr cutoff, which increases sample sizeand remains justifiable on the grounds thatconfidence intervals only twice the length ofthe interval scale still provide sufficiently use-ful information. Accordingly, we have culledour data by eliminating all taxa with range ex-tensions that differ from the correspondingFOs and LOs by more than 200 Kyr. The 78first or last appearances meeting these limi-tations are marked in Appendix 4 with an as-terisk, and the inferred limits are plotted instratigraphic order in Figure 14A, with thenumber of points per interval in Figure 14B.

The Problem of Rare Species

Before discussing the patterns of Figure 14,it is important to recognize that they are de-rived from a highly reduced data set, and toproceed with our analyses we must assumethat the culled data are representative of thewhole. This culling is particularly worrisomeif the reduced data set preferentially excludesclasses of species, such as those that are rareor have short ranges (McKinney et al. 1996).

The data-culling procedure we use tends toexclude rare species. This is because rare taxausually appear in only the most fossiliferousintervals, and the limits to their appearancesand terminations in less fossiliferous adjacentintervals will be known with less precision.The problem is demonstrated in Figure 15,which compares the frequency distributionsof abundances of species included to those ofspecies excluded from analysis. (Abundancewas determined as relative abundance overthe species’ observed stratigraphic range.) Al-though the figure has relatively more veryrare species in the excluded set (32% versus23%), and more common species in the in-cluded set (35% versus 23%), the effect is


FIGURE 14A. Seventy-eight culled first or last appearance events. Events in stratigraphic order. The numbers iden-tify taxa in Appendix 4, with first appearances being on the lower border and last appearances on the upper.

weak. Importantly, some very rare species areincluded in the analysis.

Figure 16, which compares the observeddurations of included and excluded taxa,shows that the two distributions are very sim-ilar, and that taxa with short durations, as wellas those with moderate and long durations,are in both data sets. (The few taxa with verylong durations are mostly undivided lineagesof successive species. These lineages are un-der study and will be progressively resolvedto the species level.) There are, however, inter-esting differences in the stratigraphic distri-butions of species when they are classified byduration. Species of short duration tend to ap-pear after 9 Ma, whereas species with long du-rations are more likely to have appeared be-fore 9 Ma. The pattern is apparent for both in-cluded (Fig. 17A) and excluded (Fig. 17B)taxa, whether observed or inferred ranges areconsidered. Flynn et al. (1995) previously re-ported the same changed pattern of specieslongevity.

Data Analysis

Patterns of Inferred First/Last Appearances.The two issues that we want to evaluate arethe following:

1. What is the pattern of appearances anddisappearances? That is, are the IFALs and IL-ALs distributed fairly evenly, suggesting a rel-atively steady rate of faunal change; or arethey clumped, suggesting a pulsed pattern offaunal change; or neither?

2. Is there detectable change in the ecologyof the species over this 5-Myr period and, if so,what is the nature of the change?

When we remove the taxa that do not meetour criteria, we are left with less than one-halfof our original data (Figs. 12, 14). As noted, webelieve that significant new biases have notbeen introduced by our selective culling andthat the remaining data should better repre-sent the actual pattern of species first and lastappearances in the Siwaliks than the completedata set. After culling, 78 faunal events re-main, including 38 IFALs and 40 ILALs. The


FIGURE 14B. Seventy-eight culled first or last appearance events. Number of events per interval. There are 38 firstappearances and 40 last appearances.

median number of events is 1 per 100-Kyr in-terval (mean: 1.5), with an observed rangefrom 0 to 11 events per interval. Figure 14Bshows that there are many intervals with noevents and many intervals with one, two, orthree events, and that such intervals are dis-tributed throughout the sequence. This sug-gests that faunal change occurred at all times,with a background level of three or fewerevents per 100 Kyr. However, a few intervalsseem to incorporate times of more elevatedfaunal change and, although there is no sharpdiscontinuity from background, they are ofspecial interest.

Among the inferred first appearances, onlytwo intervals (7.8 and 7.1 Ma) have more thanthree events. At 7.8 Ma, two of the five in-ferred appearances are small mammals, themurid Progonomys ‘‘new sp. at Y581’’ and therhizomyid Eicooryctes kaulialensis, and three

are large mammals, the tragulid Dorcatherium‘‘Y373 species,’’ the bovid ‘‘Bovidae/Y905 un-named species,’’ and the colobine cf. Presbytissivalensis. The first four of these also have theirfirst occurrences at 7.8, whereas the colobineis first known from the 7.6-Ma interval. In thiscase, the four coincident first appearances andfirst occurrences give us confidence in the tim-ing of this event. At 7.1 Ma, three of the in-ferred events are rodents, the murids Parape-lomys robertsi and Karnimata cf. huxleyi, and therhizomyid Rhizomyides sivalensis. The othertwo are as yet undescribed bovids, knownonly from a single interval. In this case, theprecise timing of the appearances is less cer-tain, as four of the five taxa have their actualfirst occurrences in the 7.0-Ma interval.

The inferred last appearances also show in-tervals of marked faunal change, with one at10.3 Ma and two others at 7.3 and 7.1 Ma. Five


FIGURE 15. Frequency distribution of relative abundances of included taxa compared with that of excluded taxa.(Relative abundance was determined over the species’ observed stratigraphic range as n/m, where n is the numberof specimens of the taxon and m is the total number of either small or large mammal specimens over the observedrange of the taxon.) Abundance classes at top; very rare: less than 0.1%; rare: 0.1%–1%; common: 1%–10%; abun-dant: greater than 10%.

of the six events at 10.3 Ma involve large mam-mals, including two suids, Conohyus sindienseand Listriodon pentapotamiae, two tragulids,Dorcatherium cf. majus and Dorcabune anthra-cotherioides, and the giraffe Giraffokeryx punja-biensis. The sole small mammal is an undes-cribed species of cricetid rodent. Three ofthese six taxa also have last occurrences at 10.3Ma and two others at 10.4 Ma, and the timingof the event seems reasonably placed at 10.3Ma. Disappearing taxa at 7.3 Ma include themurids Karnimata ‘‘Y024 unnamed species’’and Progonomys nr. debruijni, the rhizomyid

Kanisamys sivalensis, the large tragulid Dor-catherium ‘‘Y457 species,’’ the large bovid‘‘Large Boselaphini/Y927 species,’’ and thevery small bovid Elachistoceras khauristanensis.The first five taxa have their last occurrencesin the same 7.3-Ma interval. At 7.1 Ma thereare three large mammals, the suid Hippopota-modon sivalense, the tragulid Dorcatherium‘‘Y311 species,’’ and the giraffe Bramatheriummegacephalum, as well as the murid rodents werefer to as cf. Parapelomys robertsi, cf. Mus‘‘Y931 species,’’ and Karnimata cf. darwini.Only one of these six has a last occurrence in


FIGURE 16. Frequency distribution of observed durations of included taxa compared with that of excluded taxa.Median duration of 62 included taxa was approximately 0.9 Myr. Median duration of 53 excluded taxa was ap-proximately 1.4 Myr.

the 7.1-Ma interval; the other five are lastfound at 7.2 Ma.

For each taxon, we place the events on Fig-ure 14 at the limit of the potential range ex-tension, without attempting to consider wherewithin the extension the event might actuallyhave been. An alternative procedure is to as-sume that the probability of finding the ‘‘true’’first/last appearance is the same for each suc-cessive interval, and then to divide the eventequally among all the intervals in which itmight lie (Barry et al. 1990). For our purposesthis procedure is conservative, because it re-sults in the distribution of events over timethat most underestimates the magnitude ofturnover maxima and minima, and best vi-sually expresses the uncertainty of their ages.Figure 18 is the resulting plot of the inferredfirst and last appearances for the 78 events in

Figure 14, but with each event now parti-tioned as described above. Inspection of thisfigure leads us to the same conclusions. Thatis, many intervals have little or no faunalchange, and a very few have apparently ele-vated change that could be interpreted aspulses. Among the latter is a period centeredon 10.3 Ma, which is best characterized by dis-appearances, another at 7.8 Ma with the larg-est number of first appearances, and a muchbroader period from ca. 7.3 to 7.0 Ma, whichencompasses both appearances and disap-pearances. It is important to note that thethree pulses cannot be attributed to changinglevels of fossil abundance because they occurwithin intervals with consistently low num-bers of specimens (Fig. 7A,B).

With 40 last and 38 first appearances, if turn-over were constant we would expect 1.5 events


FIGURE 17A. Observed and inferred stratigraphic ranges of 115 taxa (data from Appendix 4). Taxa included inanalyses (n 5 62).

to occur in each 100-Kyr interval. That is, eachinterval would have slightly less than 2% of thetotal turnover. At 10.3 Ma there are 9 events, at7.8 Ma 5 events, and between 7.3 and 7.1 Ma 20events (Fig. 14A,B). These three periods consti-tute 11.5%, 6.4%, and 25.6% respectively of theobserved turnover (or summed 43.6%), againstexpectations of 1.9%, 1.9%, and 5.7% (total9.6%). A comparison to a uniform distributionwith 1.53 events per interval shows that this isa highly significant difference (Kolmagorov-Smirnov D0.5 5 0.22333, p K 0.01). We concludethat turnover is not constant between 10.7 and5.7 Ma.

We estimate that 43 species were present inthe 10.3-Ma interval because (1) they havebeen found in the interval, (2) their inferredrange extension overlaps with the interval, or(3) they are documented in both an older anda younger interval. Of these 43 species, 6 dis-appear in the interval and only 11 (30%) of the

remaining 37 survived to the 7.3-Ma interval.(The discrepancy between this number andFigure 14B is due to the inclusion of 14 speciesin the 10.3-Ma fauna that were excluded fromthe preceding analysis because their termi-nations are too poorly known, e.g., Oryctero-pus spp. Two of the 14 might range into the7.3-Ma interval.) Similarly, of 31 species in the7.3-Ma interval, only 11 (35%) persisted fromthe 10.3-Ma interval. Thus, when looked atfrom either perspective, about two-thirds ofthe species change between the two boundingevents, and turnover seems to be relativelyrapid. With the exception of the secondary-ap-pearance event at 7.8 Ma, turnover during theperiod from 10.3 to 7.3 Ma also seems rathersteady (Figs. 14A,B, 18). Of the 11 species per-sisting between 10.3 and 7.3 Ma, there is nosignificant difference in the number of largespecies (8) versus small species (3), comparedwith those that disappear (14 and 18) (x2 5


FIGURE 17B. Observed and inferred stratigraphic ranges of 115 taxa (data from Appendix 4). Taxa excluded fromanalyses (n 5 53). The numbers identify taxa in Appendix 4.

2.751, p 5 0.097). We conclude that Siwalikmammal assemblages do not exhibit stasis be-tween 10.7 and 5.7 Ma.

The peaks reported here are not the same asthe late Miocene turnover peaks reported byQuade and Cerling (1995) and Cerling et al.(1998). Those authors calculated faunalchange indices by weighing the number offirst/last occurrences by the estimated num-ber of species reported or inferred for each in-terval, and consequently they reported highlevels of turnover at 9.3 Ma and between 8.3and 7.8 Ma (dates converted to the Cande andKent [1995] timescale). Apart from theoreticaluncertainty about the expected relationshipbetween species richness and the number ofappearances on one hand and extinctions onthe other (Van Valen 1973; Webb 1984), thereare methodological problems with Quade andCerling’s indices. Quade and Cerling (1995)and Cerling et al. (1998) used observed occur-

rences instead of inferred appearances, as wellas a measure of species richness (‘‘standingrichness’’) that is, as discussed in the next sec-tion, at best poorly estimated and always un-reliable. As a result, the reported index valuesare very sensitive to the declining quality ofthe latest Miocene Siwalik record. Similarly,the turnover peaks reported by Kohler et al.(1998) are also largely artifacts of the chang-ing quality of the Siwalik record.

Number of Species per Interval. Species rich-ness, the number of species found or inferredto be present, is notoriously dependent on thenumber of specimens collected. It is, neverthe-less, of interest. In Table 4 and Figure 19A,Bwe present data on species richness for the Po-twar Siwaliks between 10.7 and 5.7 Ma. Sothey would be comparable to other publishedstudies, these data are based on the observedoccurrences, not the inferred extensions, andare calculated as ‘‘standing richness.’’ (Stand-


FIGURE 18. Seventy-eight culled first or last appearance events, partitioned equally among all the intervals includedin the species’ range expansion. Top, Number of inferred last appearances per interval (n 5 40). Bottom, Numberof inferred first appearances per interval (n 5 38).

ing richness is the number of taxa known be-fore and after the interval, plus the sum of firstoccurrences, last occurrences, and only occur-rences, divided by two. See Maas et al. [1995]for elaboration and justification of this meth-od.) The calculation assumes that an unre-corded taxon is present in an interval if it isknown from both older and younger intervals(the so-called range-through assumption).

As expected, Figure 19A,B shows strong ef-fects of specimen number, with intervals hav-ing few specimens registering few taxa. In aprevious study we concluded that intervalswith more than about 400 specimens of largemammals or more than 150 small mammalswere roughly comparable in data qualityacross intervals (Barry et al. 1995). If thesesame criteria are used, few of our much short-er intervals can be compared with each otherbecause few have that many specimens, andthere is consequently considerable uncertain-

ty about any trend in species richness overtime. Clearly the large decline in both datasets after 7.5 Ma is an effect of declining re-cord quality, as is the modest decline in smallmammals between 10.3 and 9.2 Ma. These re-sults are inconclusive as to whether the num-ber of species remains stable or changes in ei-ther direction between 9.2 and 7.4 Ma, al-though a separate analysis of small-mammalrichness recorded from single sites reinforcesthe perception of declining numbers of species(Flynn et al. 1998).

Other measures of species richness havebeen proposed that are less sensitive to sam-ple size (Magurran 1988). Among these isFisher’s a diversity index, which relates thenumber of species (S) to the number of spec-imens (N) using a log series model of speciesabundances. (The value of a approximates thenumber of species in the sample that shouldbe represented by a single individual. Note


TABLE 4. Standing species richness by interval.

Interval(Ma)

Largemammals

Smallmammals

5.75.85.96.06.16.26.36.46.56.66.76.86.97.07.17.27.3

5.04.53.02.02.04.57.06.55.05.05.06.07.5

10.011.013.516.0

1.01.01.01.01.01.01.04.55.05.05.05.05.57.07.08.5

10.07.47.57.67.77.87.98.08.18.28.38.48.58.68.78.88.99.0

17.018.017.517.017.017.019.018.518.017.517.519.020.021.522.521.020.0

12.012.012.012.010.510.512.011.010.511.011.010.012.013.013.013.013.0

9.19.29.39.49.59.69.79.89.9

10.010.110.210.310.410.510.610.7

20.021.021.520.020.020.020.020.020.020.518.517.016.516.516.517.017.5

13.014.511.5

9.09.0

10.011.011.011.014.517.018.520.019.519.019.019.0

also that here S is the number of species ac-tually recorded in the interval, not ‘‘standingrichness,’’ which includes ‘‘range through’’taxa.) Although the assumption that the abun-dance data should fit a log series is restrictive,the index is known to be robust, and inspec-tion of our raw data from each interval sug-

gests that the species abundance patterns ap-proximately fit log series. Figure 20A,B showsthe a values calculated for intervals with ad-equate samples for large and small mammalsplotted against time, with 95% confidence lim-its. (In the case of the small mammals, mini-mum sample size was 25 specimens, but in-tervals with a high proportion of unidentifiedmaterial also were excluded even if samplesize was large. Material from these excludedintervals had been selectively culled for iden-tification, and the subsample of identified fos-sils is therefore not representative of the wholeassemblage in relative proportions. In the caseof the large mammals, minimum sample sizewas 50, except for four intervals at the top andbottom of the section. These were the intervalsat 6.4, 6.8, 10.4, and 10.7 Ma.)

Although there is a wide range of values inboth data sets, there is a trend toward decreas-ing a in the younger intervals, most noticeablyin the small mammals. In both large and smallmammals, the magnitude of the decrease in-dicates a substantial reduction in the numberof species, although certainly not as large assuggested by Figure 19A,B. The trend of de-creasing species richness is most apparent ifonly the largest values are taken from thewhole series, but the wide total range in bothcases makes interpretation difficult—especial-ly because adjacent intervals may have signif-icantly different values. Inspection of theabundance data suggests that before 7.5 Ma,intervals with low a values either are domi-nated by a single site (particularly so for thesmall mammals) or have large numbers offragmentary equid teeth cataloged as separatespecimens. (The latter effect would include theintervals at 7.9, 8.3, and possibly 8.5 Ma.)Thus, we think that the very low values before7.5 Ma are artifacts of taphonomy, or in a fewinstances overzealous cataloging practicesthat led to overrepresentation of equids. Nev-ertheless, the low values after 7.5 Ma are likelyto reflect a real change in the number of spe-cies.

A second richness measure is the Margalefdiversity index (Magurran 1988), whichmakes use of the same information on thenumber of species and number of specimens,but as D 5 (S 2 1)/ln N. In both our large- and


FIGURE 19. Number of species (triangles) and number of specimens (bars) per interval. A, Large mammals. B, Smallmammals.

FIGURE 20. Species richness indices over time. Bars on ‘‘log series a’’ points are upper and lower 95% confidencelimits. A, Large mammals. B, Small mammals.

small-mammal data sets the Margalef indexclosely tracks Fisher’s a index (Fig. 20A,B), re-inforcing the impression of overall decliningspecies richness in both large and small mam-mals.

Relative Abundances within Fossil Assemblag-es. Other diversity indices emphasize differ-ences in the relative abundance of taxa, as op-posed to the number of species. Among theseare the Berger-Parker, Simpson, and ShannonEvenness indices, the first two of which mea-sure the dominance of common species rela-tive to the others, whereas the third measuresthe evenness of the species abundances (Ma-gurran 1988). The Berger-Parker index is usu-ally represented as the inverse of the propor-tion of the most common taxon relative to thetotal sample, but here (Fig. 21A,B) we show itas the basic proportion with 95% confidencelimits. (Confidence limits were calculated us-ing the formulas of Ehrenfeld and Littauer[1964].) The Simpson and Shannon Evennessindices (Fig. 22A,B) both use the proportionalabundances of all species. The Simpson indexdisproportionally weighs the most abundantspecies, whereas the logarithmic transforma-

tion in the Shannon Evenness index weighsspecies of intermediate abundance mostheavily. In using these indices we assume that,although the relative abundances of the fossilassemblages within a given interval do notclosely approximate those of the original liv-ing communities (Badgley 1986b), the under-lying taphonomic biases are constant betweenintervals (i.e., isotaphonomic). Thus, the com-parison of diversity indices between intervalsshould reveal something about contrasts be-tween the living communities, not just varia-tion in taphonomic conditions. Our assump-tion rests on the similarity through time of thefluvial environments that preserved the fos-sils.

Inspection of Figure 21A indicates that theBerger-Parker index values for the large-mam-mal assemblages delineate three distinct phas-es. One phase encompasses intervals 10.0 Maand older, which have faunas with low dom-inance; a second phase includes intervals from9.8 to 7.9 Ma, which have faunas with consis-tently higher (and sometimes markedly high-er) dominance; and a third consists of inter-vals 7.8 Ma and younger, which again have


FIGURE 21. Relative abundance (species diversity) over time. Triangles represent value of inverse of Berger-Parkerindex; bars indicate upper and lower 95% confidence limits. A, Large mammals. B, Small mammals.

FIGURE 22. Relative abundance (species diversity) over time. The Simpson index is a dominance measure; the Shan-non index is an evenness measure. A, Large mammals. B, Small mammals.

more evenly balanced faunas displaying lowdominance. Although there is considerableheterogeneity within each phase, x2 tests (witha sequential Bonferroni correction) comparingthe relative abundances of the taxa in the in-tervals at 10.0 and 7.8 Ma with those in the in-tervals between 9.8 and 7.9 Ma confirm thesignificance of the differences; the only non-significant pairs are the comparisons with the8.9-Ma interval (Table 5). In the same manner,the Simpson and Shannon Evenness indices(Fig. 22A) track the Berger-Parker index close-ly, also forming three more or less distinctphases. Notably, both indices show the samechanges in faunal dominance from evenly bal-anced to one-taxon-dominated and back tomore evenly balanced. As with the Berger-Parker index, the Simpson and Shannon Even-ness indices show the anomalous 8.9-Ma in-terval to have a more balanced fauna than ineither the immediately preceding or the fol-lowing intervals. The 9.6-Ma interval is alsoan outlier, registering on both indices as ex-tremely uneven. However, the exceptional in-dex values are probably artifacts of mistakenly

counting equid postcranials from one site asseparate individuals, with a resulting over-representation.

With very few exceptions, after 10.7 Maequids are the most common fossils. They of-ten attain very high relative frequency in ourcollections (Fig. 23), and their changing rela-tive abundance controls the various domi-nance-evenness indices. In our preceding dis-cussion of species richness, we noted that highequid relative abundance in the intervals at7.9, 8.3, and 8.5 Ma might be in part an artifactarising from overemphasis on collecting andcataloging equid teeth. However, such an ar-tifact cannot explain all of the high valuesfrom 9.8 to 7.9 Ma (Fig. 23), and there areplainly intervals in which equids stronglydominate the faunas and intervals in whichthey do not. The observed change in domi-nance is supported by our biostratigraphicsurvey data (Table 2, Fig. 24), which shows theproportion of equids relative to that of bovidsrising at 9.8 Ma and then falling at ca. 8.5–8.0Ma. We think it is likely that the shift after 10Ma from balanced to unbalanced equid-dom-


TABLE 5. Comparisons of relative abundances of largemammals at 7.8 and 10.0 Ma with relative abundancesin intervening intervals.

Inter-val

(Ma)

Mostcom-montaxon

Allothertaxa

x2

(7.8 Ma)p

(7.8 Ma)x2

(10.0 Ma)

p(10.0Ma)

7.87.98.08.18.38.58.78.99.29.39.59.69.79.8

10.0

20114251

389550

10339

336633

37150

4432

425

4781

1463838406051

348469

36452420

876

16.33626.282

5.99631.51810.27121.214

2.9769.090

19.4996.282

48.49616.44411.949

0.230

0.000*0.000*0.014*0.000*0.001*0.000*0.0840.003*0.000*0.012*0.000*0.000*0.001*0.631

0.23048.954

118.5309.666

78.43819.61058.526

4.30851.352

148.59110.054

140.36529.45318.633

0.6310.000*0.000*0.002*0.000*0.000*0.000*0.0380.000*0.000*0.002*0.000*0.000*0.000*

* Significant at 0.05 (with a sequential Bonferroni correction).

FIGURE 24. Relative abundance of equids and bovidsover time, calculated as percentage of total number oflarge mammals.

FIGURE 23. Relative abundance of equids over time contrasted with the Shannon Evenness index. Percent equidscalculated as the number of equid specimens over the total number of large mammal specimens in each interval.Bars on ‘‘percent equid’’ points are upper and lower 95% confidence limits. Intervals with very small samples ex-cluded.

inated faunas is the result of an increase in thenumber of equid species, which are treated asone taxon in our current analysis. We alsospeculate that the shift back to more balancedfaunas at 7.8 Ma may reflect a reduction in the

number of equid species. Finally, on Figures20, 21, and 22 we have included some ‘‘pre-hipparion’’ intervals to place trends in speciesrichness and abundance into a larger context.The large mammals of the 11.3-Ma intervalshow intermediate dominance, although thegiraffe Giraffokeryx punjabiensis is by far themost abundant taxon. This species persistsinto the 10.3-Ma interval, but at much reducedabundance.

None of the dominance/evenness diversityindices show a trend for the small mammals,


TABLE 6. Comparisons of relative abundances of smallmammals at 7.8 and 8.1 Ma with relative abundances inintervening intervals.

Inter-val

Mostcom-montaxon

Allothertaxa

x2

(7.8 Ma)p

(7.8 Ma)x2

(8.1 Ma)p

(8.1 Ma)

6.47.27.37.87.98.19.29.3

10.2

7017241613

155197

1521

11041441032

149276

2636

4.7927.8175.299

7.2701.0683.9853.9854.408

0.0290.005*0.021

0.007*0.3010.0460.0460.036

6.6529.1775.4821.0687.667

6.5122.9983.844

0.0100.002*0.0190.3010.006*

0.0110.0830.050

* Significant at 0.05 (with a sequential Bonferroni correction).

although the three indices closely track eachother. The assemblages at 7.8 and 8.1 Ma areseparated from the others on the Berger-Park-er and Simpson indices, but x2 comparisons(with a sequential Bonferroni correction) ofthe relative abundances of taxa between theseand the other intervals indicate few compari-sons are significant at a 0.05 level (Table 6). Asnoted, small-mammal assemblages tend to bedominated by single localities and could re-tain a strong taphonomic imprint of predationor hydraulic sorting.

Body Size. The overall pattern of size struc-ture within a group of related taxa can pro-vide additional information about guild re-sponses to change induced by abiotic and/orbiotic events. Within the two most abundantand well-studied groups, muroid rodents andartiodactyls, there is significant change inbody size structure between 10 and 9 Ma(Morgan et al. 1995 with revised age esti-mates; M. E. Morgan unpublished data).Among the rodents, notably larger species ofboth the Muridae and Rhizomyidae appear.Size estimates are derived from M1 area (seeMorgan et al. 1995 for data and methods). Mu-rids first appear at 13.7 Ma and do not changenotably in size until the appearance of Karni-mata ‘‘Y388 unnamed species,’’ estimated atabout 60 g and first occurring at 9.2 Ma (IFAL9.9). This species is more than twice the sizeof its presumed ancestor, Karnimata ‘‘Y450 un-named species,’’ estimated at about 25 g andlast occurring at 9.6 Ma (ILAL 9.4). The twolater species, cf. Parapelomys robertsi at 8.1 Ma

(IFAL 8.6) and Parapelomys robertsi at 7.0 Ma(IFAL 7.2), continue this increase. Among therhizomyids, there is size increase within thelong-lived Kanisamys lineage with the first oc-currence of K. sivalensis at 9.3 Ma (IFAL 9.5).In addition, several large species of the genusBrachyrhizomys (rhizomyids that were com-parable in size and habits to living, fossorialbamboo rats) also appear between 10 and 9Ma, and even larger species such as Brachyr-hizomys choristos appear at 8.4 Ma (IFAL 8.5).Other notable large rodent species appearinglater are the rhizomyid Protachyoryctes tatroti,thought to be a large derivative of Kanisamyssivalensis (Flynn 1982), at 7.8 Ma (IFAL 8.0),Rhizomyides sivalensis at 7.1 Ma (IFAL 7.1), andthe porcupine Hystrix sivalense at 8.0 Ma (IFAL8.0). Leporids, which typically weigh morethan 1 kg, have an IFAL of 7.8 Ma.

Body weight can be estimated for the fivedominant Siwalik artiodactyl families (an-thracotheres, suids, giraffes, tragulids, andbovids) from dimensions of the astragalus(Morgan et al. 1995.) Taken as a group, the fivetaxa occupy a nearly continuous range of sizesthat expands markedly just before 10 Ma. Inthe giraffes, the observed size change at 10.3Ma occurs at a species boundary, when Giraf-fokeryx punjabiensis, estimated at 233 6 8 kg (n5 51), is replaced by Bramatherium megacephal-um, estimated at 840 6 32 kg (n 5 27) (Fig. 25).Although most astragali cannot be attributedto species, a marked size increase also occursbetween 11 and 10.3 Ma with the first ap-pearances of the Selenoportax and Tragoceriduslineages (early populations estimated at 114 64.5 kg, n 5 26, and 53 6 1.3 kg, n 5 70 re-spectively) (Fig. 25). Both giraffes and bovidsexhibit continuing size increases after 10 Ma,but trends are difficult to quantify in the ab-sence of associated cranial and postcranial re-mains that could be used both to recognizespecies and to estimate size. Nevertheless,horn cores of cf. Protragelaphus skouzesi at 8.8Ma (IFAL 8.8) and ‘‘Large Boselaphini/Y927species’’ at 8.0 Ma (IFAL 8.1) record the ap-pearance of very large bovids and postcranialremains indicate the presence of speciesweighing nearly 250 kg. In addition, late oc-curring individuals of Bramatherium megace-phalum and the Tragoceridus lineage all tend to


FIGURE 25. Weight estimates for bovids and giraffes, based on distal width of astragalus (DWA). Bovid regression:2.894 log10(DWA) 1 0718. Giraffe regression: 2.725 log10(DWA) 1 0.776.* *

be noticeably larger than older conspecifics.Large suids of the genus Hippopotamodon alsoappear at 10.2 Ma (IFAL 10.2), and in the an-thracothere Hemimeryx lineage there is a sizeincrease at about 10 Ma.

Ecomorphological Analyses. Ecomorpholog-ical studies provide a means of assessing theecological requirements of species, which thenallow evaluation of differences in communitycomposition. Our objective is to determine towhat extent ecologically related groups oftaxa behave similarly in their patterns of firstand last appearances, or how their patternsmight be related to other documented differ-ences. That is, can we detect critical environ-mental thresholds for species occupying sim-ilar niches or sharing key limiting resourcessuch as certain types of vegetation and wateravailability? Ecomorphological approachesused here include a variety of dietary studies,such as qualitative morphological studies ofthe dentition, occlusal microwear studies, andisotopic analyses of enamel apatite, as well asbody size estimates from dental and postcra-nial remains (Gunnell et al. 1995; Morgan etal. 1994, 1995; M. E. Morgan unpublisheddata). We also use information from compar-ative and functional studies of closely related

extant taxa (Flynn 1982). We identified taxalikely to have arboreal and/or closed habitatrequirements and those with grazing and/oropen habitat requirements.

Thirty-eight taxa have IFALs between 10.7and 5.7 Ma. With the possible exception of thelarge hominoid Sivapithecus parvada, knownonly from one locality at 10.0 Ma, none of theremaining 37 IFALs can be characterized eco-logically either as restricted to forested envi-ronments or as completely arboreal. However,several newly appearing taxa might be indic-ative of more-open conditions, including aspecies of Giraffa (IFAL 9.1 Ma), the large an-tilopine cf. Protragelaphus skouzesi (IFAL 8.8Ma), the gerbil Abudhabia pakistanensis (IFAL8.6 Ma), a second very large bovid with high-crowned teeth that we refer to as ‘‘Large Bo-selaphini/Y927 species’’ (IFAL 8.1), and an in-determinate species of Hippotragini (IFAL 6.9Ma). Other appearing taxa not meeting thecriteria for inclusion in the turnover analysisthat have strong ecological signals include theleporids at 7.8 Ma—interpreted as able to uti-lize relatively open environments—and thehippopotamid Hexaprotodon, a likely grazerprobably present by 6.1 Ma and possibly asearly as 7.2 Ma. Rhizomyid rodents, which are


present throughout the middle and late Mio-cene in the Siwaliks, show an increase in hyp-sodonty at 9.5 Ma with the IFAL of Kanisamyssivalensis, and marked hypsodonty at 7.8 Mawith the IFAL of Eicooryctes kaulialensis (Flynn1982). Finally, although we cannot directly as-certain the individual diets of any of the bovidspecies because of difficulties in associatingteeth with horn cores (the basis of speciesidentifications), we can say that, overall, bo-vids show an increased reliance on C4 grasses,judging from stable carbon isotope analysis ofbovid teeth (Morgan et al. 1994).

Forty taxa have ILALs during this 5-Myr in-terval, including several with arboreal, closed,or moist habitat requirements. These includespecies of the Sivapithecus clade (ILAL 8.4 Ma),deinotheres and chalicotheres (both ILAL 8.0),and seven species of tragulids. The latter weredominant or equal in abundance to bovids un-til 10 Ma, but in the late Miocene they becamemuch rarer as well as less diverse (Barry et al.1991; M. E. Morgan unpublished data). Otherlocally disappearing taxa, which were culledfrom our turnover analysis but are still rele-vant, include the ctenodactylid rodent Sayimyschinjiensis (ILAL 9.4 Ma), the tree squirrel Ra-tufa (ILAL 8.8 Ma), lorisids (ILAL 8.8 Ma), cri-cetids (ILAL 8.2 Ma), and dormice (ILAL 7.3).

The concordant ILALs of three very largebodied mammals at 10.3 Ma (Giraffokeryx pun-jabiensis, Conohyus sindiense, and Listriodon pen-tapotamiae) is interesting considering theirhigh relative abundances and long species du-rations. These three taxa coexisted in the Si-waliks for at least 4 million years, including0.4 Myr after the first appearance of equids. Atthe same time, two new large herbivores ap-pear in the Potwar record, the giraffid Bra-matherium megacephalum and the suid Hippo-potamodon sivalense. Both may have evolved insitu. These two new species are much largerthan any earlier artiodactyl species. This ob-served size shift within the artiodactyls sug-gests restructuring of at least a segment of theherbivore guild. However, molar microwearpatterns of Bramatherium megacephalum and itspossible ancestor Giraffokeryx punjabiensis aresimilar (both are browsers), as are the mi-crowear patterns of H. sivalense and the dis-appearing suids C. sindiense and L. pentapota-

miae (M. E. Morgan unpublished data). Stablecarbon isotope analysis of Hippopotamodonsuggests that it included C4 grasses in its diet(Morgan et al. 1994).

The rodent faunas between 10.7 and 5.7 Mawere dominated by rhizomyids and murids.Cricetids are last recorded at 8.6 Ma with anILAL of 8.2 Ma. Among the murids, speciesturned over completely between 9.6 and 9.3Ma, and between 8.4 and 7.8 Ma. Most of thisturnover is attributed to congeneric replace-ments of species with short durations. In con-trast, the youngest cricetids are all long-rang-ing species that first appear in the middleMiocene. It is presently difficult to assess theecological significance of the observed Siwalikmurid and cricetid species differences.

Discussion and Conclusions

Previous analyses of Siwalik mammal as-semblages used coarser levels of time resolu-tion and concentrated on species composition,with only passing discussion of body size andspecies richness (Barry et al. 1991, 1995; Barry1995). In this paper we have used 100-Kyr in-tervals in place of 500-Kyr intervals and de-veloped additional measures of ecological di-versity and structure. One hundred thousandyears may be the finest level of resolution pos-sible for studies of terrestrial vertebrate faunasover periods of millions of years because ofsampling limitations inherent in the fossil re-cord. In other published analyses of terrestrialassemblages, observed first and last occur-rences, instead of inferred appearances, havebeen used to estimate species ranges andturnover. Thus the changing quality of the var-ious fossil records is not usually considered tobe important in shaping the patterns underanalysis. We believe this failure potentiallycompromises the results and conclusions ofany study, and a crucial part of our present ef-fort has been to allow for the incompletenessof the fossil record and the biases it introduc-es. To do this we have used observations offirst and last occurrences to infer local firstand last appearances, which are then used asthe estimators of the endpoints of speciesranges. We have also weighed the basic obser-vations by how well constrained the ages of lo-calities are, rejecting those that are too impre-


FIGURE 26. Summary diagram, showing temporal relationships of changes in the depositional system, isotopes,vegetation, and fauna. Horizontal bands delineate periods with greater faunal turnover. Hatched regions denoteapproximate time periods during which a feature of the depositional system changes. The arrow labeled ‘‘increasingseasonality’’ marks the interval between Retallack’s (1991) two paleosol sections.

cise. In the following we relate the biotic dif-ferences derived from our analyses to the in-dicators of environmental change.

Environmental and Faunal Change

Our previous analyses of Siwalik mammalassemblages documented brief periods ofhigh turnover between longer periods of low-level turnover, a pattern that could be inter-preted as demonstrating coordinated stasis,although the evidence for pulses in the lateMiocene was weak at best (Barry 1995; Barryet al. 1995). Our more resolved data set nowshows stronger evidence for three very briefperiods of high turnover in the late Miocene,and as we show in the following there is con-

cordance between faunal turnover and someSiwalik environmental events. In Figure 26 wegraphically summarize these relationships.

The depositional systems of the Potwar un-derwent substantial changes during the lateMiocene. The Nagri and Dhok Pathan For-mations are interpreted as having been de-posited by coexisting emergent and interfanriver systems, with the larger emergent Nagrisystem being displaced from the study areabeginning at ca. 10.1 Ma (Willis and Behrens-meyer 1995). After 9 Ma, the floodplains of theinterfan Dhok Pathan deposits were less welldrained, with smaller rivers having more sea-sonal flow than previously (Zaleha 1997b).Comparison of two paleosol sequences, at 9.4


and 8.6 Ma (Retallack 1991), also suggests re-organization of topography and drainage ac-companying a transition to a more seasonalclimate. A very few paleosols may haveformed under waterlogged, grassy wood-lands, but more extensive grasslands were notpresent in either of Retallack’s sample sec-tions. Similar sedimentological differenceshave been observed in younger parts of the se-quence. Most striking, after ca. 8 Ma smallfloodplain channels become less common,while lighter colored and carbonate-depletedpaleosols are more common (Behrensmeyer1987).

The stable carbon and oxygen isotopic evi-dence also indicates significant, acceleratingenvironmental change. The period of transi-tion begins at 9.2 Ma with a shift in d18O val-ues and continues to nearly the end of theMiocene for both the d18O and d13C records.The carbon isotope record demonstrates thatafter 8.1 Ma significant amounts of C4 grassesbegan to appear on the floodplains of the Po-twar Miocene rivers, and very open wood-lands with predominantly C4 plants were pre-sent by 7.4 Ma. Communities composed exclu-sively or predominantly of C3 plants disap-peared after 7.0 Ma, and by 6.8 Ma floodplainhabitats included extensive C4 grasslands. Thetwo most negative paleosol carbonate d13Cvalues reported for the Potwar Siwaliks occurat approximately 9.2 and 9.0 Ma (Quade andCerling 1995). Among other possibilities, thed18O record suggests changes in precipitationleading to a drier and more seasonal climate.After 7.3 Ma, d18O values typical of middleMiocene Siwalik rocks are absent (Quade andCerling 1995).

Inferred first and last appearances are themost sensitive indicators of faunal change.They show a constant, low level of faunal turn-over throughout the interval at 10.7–5.7 Ma,with three short periods of elevated turnoverat 10.3, 7.8, and 7.3–7.0 Ma (Fig. 26). The old-est period is characterized primarily by ex-tinction, the second by a large number of firstappearances, and the youngest by a combi-nation of extinction and first appearances. Thethree periods combined account for nearly44% of all turnover between 10.7 and 5.7 Ma,compared with the ;10% expected if turnover

were constant. Indices of species richness andecological diversity are much less sensitive in-dicators, but still of interest (Fig. 26). Thenumber of species, as gauged by Fisher’s a di-versity index, exhibits a decline for both largeand small species throughout the late Mio-cene. In the case of the large mammals, a di-versity declines by about one-third from theearlier values, whereas small-mammal a di-versity falls by more than one-half. In bothcases the decrease indicates a substantial re-duction in the number of species, althoughprobably not as drastic as suggested by thespecies richness values themselves. Relativeabundances of species also change abruptlybetween 10 and 9.8 Ma (Fig. 26), at which timethe large-mammal assemblages come to bedominated by equids, and then again between7.9 and 7.8 Ma, when equids cease to be sooverwhelmingly abundant. Body sizes of ro-dents and artiodactyls generally increase dur-ing the same period, with notable stepsamong the rodents between 10.3 and 9.2 Ma,and for artiodactyls at 10.3–10.2 Ma (Morganet al. 1995). The diversity and richness indicestogether with the data on body size stronglyimply that the ecological structure of themammal communities changed substantiallyshortly after 10 Ma, and then again at 7.8 Ma.The brief reappearance of more-balancedlarge-mammal assemblages in the 8.9 intervalis also noteworthy.

The taxa that undergo the changes notedabove are difficult to characterize ecologically,but the pattern of appearance and disappear-ance is highly selective with respect to pre-sumed habits. Some of the newly appearingrodents or other small species are hypsodont,or are likely to have fed on underground plantorgans or have been fossorial (Kanisamys siv-alensis, Eicooryctes kaulialensis, species of Bra-chyrhizomys, and Hystrix sivalensis), all char-acteristics that suggest open habitats. Otherspecies that appear belong to extant higher-level taxa usually found in open habitats (ger-bils, rabbits). Similarly, among the larger taxathe tragulid Dorcatherium Y373 species is highcrowned, and its successor ‘‘Tragulidae/L101unnamed species’’ is even more so, implyingdependence on grass. Extant species of Manis,Sus, and Giraffa are commonly found in open


woodlands, whereas Hexaprotodon sivalensis ispresumed to have been a grazer. Extant re-duncines are grazers associated with wetmeadows, floodplains, or permanent drain-ages, whereas hippotragines and antilopinesare grazers and mixed feeders of open terrain,or brush and scrub. Tragelaphines are brows-ers or mixed feeders and inhabitants of forestto lightly wooded areas near water. The ma-jority of these taxa have inferred appearancesyounger than 9.0 Ma.

Taxa that become locally extinct and aboutwhich we can make habitat inferences includespecies with extant relatives that are charac-teristic of closed vegetation (tree shrews, lor-isids, hominoids, dormice, some tree squir-rels, dendromurines, and several species oftragulids). Many other species disappear dur-ing the time interval, but with few exceptionsnone give indications that they were bound tomore-open habitats. The exceptions include cf.Presbytis sivalensis and one tragelaphine, onereduncine, and two antilopine bovids. The co-lobine monkey, the tragelaphine, and the re-duncine are taxa that first appeared after thecarbon isotopic transition started and subse-quently disappeared at a later stage. We as-sume they were inhabitants of the mixed C3/C4 communities or the wetter parts of thefloodplain that did not persist in the increas-ingly open conditions in the latest Miocene.Both of the antilopines, on the other hand, en-tered the Siwalik record before the beginningof the carbon transition and may have hadnarrow habitat tolerances. Almost three-fourths of the disappearances in Figure 14 areyounger than 9.4 Ma, with the pace of extinc-tion accelerating after the appearance of C4

vegetation on the floodplain.Turnover during the 10.3 Ma event differs

greatly from that of the two latest Mioceneevents with respect to the duration of the spe-cies involved. The event at 10.3 Ma comprisestaxa that were both common and of long du-ration, whereas the latest Miocene events in-clude more taxa that were shorter ranging andless common (Fig. 17A). This is true of boththe taxa that appear and those that disappear.That is, long enduring taxa both appear andgo extinct at 10.3 Ma, whereas in the latestMiocene taxa of brief duration appear and dis-

appear. This difference of mode develops be-tween approximately 9.0 and 8.5 Ma, at whichtime many short ranging and very rare speciesbegin to make appearances. It is therefore ageneral feature of latest Miocene faunalchange that contrasts to older periods, not justa peculiarity of the phase of more intenseturnover.

Correlations Between Environmental andFaunal Events

The 5-Myr period between 10.7 and 5.7 Mawitnessed a qualitative change in the mam-malian fauna of the Potwar Plateau. Some ofthe differences seem to reflect responses tochanging environmental conditions, but oth-ers are not easily correlated to recognized lo-cal environmental events.

The 10.3 Ma turnover event does not cor-relate to any obvious local environmental orclimate event. It is about 200 Kyr older thanthe onset of replacement of the emergent Na-gri system by the Dhok Pathan interfan sys-tem, and although approximately contempo-rary with the earliest isotopic dietary evidencefor C4 vegetation, it is more than 2 Myr olderthan paleosol evidence for extensive C4 vege-tation on the floodplain. It is also much olderthan the oxygen isotope shift marking a tran-sition to drier and more seasonal climates. The10.3 Ma event marks the extinction of the Gir-affokeryx-Listriodon fauna, which historicallyhas been linked to the first appearance ofequids in the Siwaliks. Nonetheless, the turn-over and extinction actually follow the firstappearance of equids by 400 Kyr and are atleast 300 Kyr older than the transition to hip-parionine-dominated faunas. The pattern ofspecies disappearance and appearance de-scribed above suggests that biotic interactionsmay have been more important than environ-mental change in creating a pulse of faunalchange around 10.3 Ma.

In contrast, the two latest Miocene turnoverevents show a close temporal correlation tochanges in floodplain deposition and vegeta-tion. The event at 7.8 Ma follows by 250 Kyrthe first paleosol evidence for widespread C4

vegetation on the floodplain and is approxi-mately contemporary with other changes infloodplain deposition (Behrensmeyer 1987).


The 7.8 Ma turnover is also coincident withthe shift from equid-dominated to more even-ly balanced large-mammal assemblages. Ap-pearing taxa include a hypsodont rodent anda high-crowned tragulid, implying a grassand possibly C4 diet. The carbon isotope eventindicates that habitats with C4 plants becamemore common on the floodplain. This isstrong evidence of changing habitat availabil-ity and is concordant with the appearance ofgrazing or mixed-feeding taxa. At the sametime, the paleosol carbon record indicates thatC3-dominated habitats still persisted and wereeven common, thus explaining the absence ofelevated extinction.

The younger turnover event between 7.3and 7.0 Ma is composed of both appearancesand disappearances. It begins very shortly af-ter the first occurrence of C4-dominated florasat 7.37 Ma and ends with the last occurrenceof C3-dominated vegetation at 7.04 Ma. Thesetwo isotopic thresholds are critical to under-standing the impact of this vegetation changeon the mammals, in that they record the ap-pearance and disappearance of end-membersfrom the vegetation spectrum. They thus in-dicate both the gain and the loss of types ofextreme open and closed habitats, and arelikely related directly to the appearance anddisappearance of individual species. Thissame broad interval also gives evidence of in-creasingly seasonal stream flow and perhapshigher levels of habitat disturbance.

The change in large-mammal diversity be-tween 10 and 9.8 Ma potentially could be re-lated to fluvial system dynamics, as it closelyfollows the initial appearance at 10.1 Ma of theinterfan Dhok Pathan system (Fig. 26). How-ever, although there are very few large local-ities associated with the emergent Nagri sys-tem on which to base a comparison, there doesnot seem to be a relationship between positionon the floodplain and composition of the fau-nal assemblages. Two large sites (Y258, Y251),both associated with the interfan Dhok Pathansystem and dated at 10.1 and 10.0 Ma respec-tively, have low numbers of equids (31% and28%) when compared with younger sites inthe same interfan setting. In contrast, two sites(Y309, Y317), closely associated with theemergent Nagri system and dated at 9.3 and

9.2 Ma, have numbers of equids (52% and 50%respectively) comparable to contemporarysites of the interfan system. This suggests thatbefore 9.8 Ma the habitats situated on theDhok Pathan interfan system were similar tothose of the Nagri emergent system. Further-more, the transitory reappearance of a morebalanced large-mammal assemblage in the8.9-Ma interval, which comes after the dis-placement of the emergent Nagri system, sug-gests there is no direct connection between de-positional system and faunal structure. Thistransitory reappearance of more balanced as-semblages closely follows the two most neg-ative carbon isotope values in the sequenceand the beginning of the oxygen isotope tran-sition.

General Patterns

Biotic change in the late Miocene Siwaliks isa combination of steady, low-intensity changein species composition and ecological struc-ture, punctured by brief, irregularly spacedintervals with accelerated species turnoverand ecological change. Episodes of acceleratedspecies turnover amidst longer quiescent pe-riods are characteristic of both the ‘‘coordi-nated stasis’’ (Brett et al. 1996) and the ‘‘turn-over pulse’’ (Vrba 1985, 1995) models of evo-lutionary change. The presence, however, of asteady background species turnover leadingto the loss of 65–70% of the members of theinitial fauna, coupled with steadily decliningdiversity and abrupt, unrelated changes inrelative abundance, does not fit the observedor predicted patterns of either model. Al-though there seem to be pulses of change,there is no attending stasis in this Siwalik re-cord. This difference in pattern comparedwith previous results is due in large part tothe inclusion of rare and low-abundance spe-cies in our analysis (McKinney et al. 1996).

Although interpretation of the events at 10.3Ma is uncertain, species turnover and ecolog-ical changes in the latest Miocene are closelytied to change in the vegetation, and, we be-lieve, through the vegetation change to cli-matic change. Evolution of the fluvial systemcould have played some minor role, but we fa-vor explanations based on climatic change forboth latest Miocene pulses. The strongest ev-


idence for this is the association of the mostsignificant faunal changes and expansion ofC4 plants on the floodplain with oxygen iso-topic and sedimentological evidence for in-creasingly drier and more seasonal climates.Displacement of the emergent Nagri systemby the interfan Dhok Pathan system seems notto be related to significant faunal differences,but climate change as well as tectonics mayhave affected evolution of the fluvial system.This makes the precise chain of causality dif-ficult to trace. Because high levels of back-ground turnover also occur within the periodof progressive climate change, the back-ground turnover may also be a result of cli-mate change. Our earlier analyses did not de-tect these coincidences of faunal, floral, andclimatic change because those analyses used aresolution of only 500 Kyr. The climatic andenvironmental changes considered here takeplace on timescales of a few hundreds of thou-sands of years or less, and testing the relation-ships needs faunal data with the same reso-lution. A further testing of these ideas willnow also require data gathered on a largergeographic scale—one large enough to in-clude coexisting megafan systems, each witha rich and long fossil record. Such tests areachievable in the Miocene of Pakistan and ad-jacent regions. It is, however, always neces-sary to recognize the limits of paleontologicaldata and its adequacy for resolving our ques-tions.

Acknowledgments

This paper is a result of a long collaborationwith the Geological Survey of Pakistan, andwe are grateful to the Directors General andmembers of the Survey who have over theyears supported our project. The contribu-tions of M. Anwar, W. Downs, E. Lindsay, K.A. Sheikh, and the late Noye Johnson are alsogratefully acknowledged, as is their compan-ionship. Financial support has come througha number of grants from the National ScienceFoundation (most recently SBR-9408664 andDBS-9196211 to J. Kelley), the SmithsonianForeign Currency Program, the Wenner-GrenFoundation for Anthropological Research,and the American School for Prehistoric Re-search. We also gratefully acknowledge the

abiding support of Ms F. Berkowitz of theSmithsonian Office of International Relations.We greatly appreciate careful reviews by D.Fox and J. Hunter, although we have not al-ways taken their advice.

Literature Cited

Alroy, J. 1996. Constant extinction, constrained diversification,and uncoordinated stasis in North American mammals. Pp.285–311 in Ivany and Schopf 1996.

Amano, K., and A. Taira. 1992. Two-phase uplift of Higher Him-alayas since 17 Ma. Geology 20:391–394.

Anonymous. 1985. Atlas of Pakistan. Survey of Pakistan, Ra-walpindi.

Badgley, C. 1986a. Counting individuals in mammalian fossilassemblages from fluvial environments. Palaios 1:328–338.

———. 1986b. Taphonomy of mammalian fossil remains fromSiwalik rocks of Pakistan. Paleobiology 12:119–142.

———. 1990. A statistical assessment of last appearances in theEocene record of mammals. In T. M. Bown and K. D. Rose,eds. Dawn of the age of mammals in the northern part of theRocky Mountain interior, North America. Geological Societyof America Special Paper 243:153–167.

Badgley, C. E., and A. K. Behrensmeyer. 1980. Paleoecology ofMiddle Siwalik sediments and faunas, northern Pakistan. Pa-laeogeography, Palaeoclimatology, Palaeoecology 30:133–155.

———. 1995a. Long records of continental ecosystems. Palaeo-geography, Palaeoclimatology, Palaeoecology Vol. 115.

———. 1995b. Preservational, paleoecological and evolutionarypatterns in the Paleogene of Wyoming-Montana and the Neo-gene of Pakistan. Pp. 319–340 in Badgley and Behrensmeyer1995a.

Badgley, C. E., and P. D. Gingerich. 1988. Sampling and faunalturnover in Early Eocene mammals. Palaeogeography, Pa-laeoclimatology, Palaeoecology 63:141–157.

Badgley, C. E., and L. Tauxe. 1990. Paleomagnetic stratigraphyand time in sediments: studies in alluvial Siwalik rocks ofPakistan. Journal of Geology 98:457–477.

Badgley, C., L. Tauxe, and F. L. Bookstein. 1986. Estimating theerror of age interpolation in sedimentary rocks. Nature 319:139–141.

Badgley, C., W. S. Bartels, M. E. Morgan, A. K. Behrensmeyer,and S. M. Raza. 1995. Taphonomy of vertebrate assemblagesfrom the Paleogene of northwestern Wyoming and the Neo-gene of northern Pakistan. Pp. 157–180 in Badgley and Beh-rensmeyer 1995a.

Barndt, J. 1977. The magnetic polarity stratigraphy of the typelocality of the Dhok Pathan Faunal Stage, Potwar Plateau, Pak-istan. Master’s thesis. Dartmouth College, Hanover, N.H.

Barndt, J., N. M. Johnson, G. D. Johnson, N. D. Opdyke, E. H.Lindsay, D. Pilbeam, and R. A. H. Tahirkheli. 1978. The mag-netic polarity stratigraphy and age of the Siwalik Group nearDhok Pathan Village, Potwar Plateau, Pakistan. Earth andPlanetary Science Letters 41:355–364.

Barry, J. C. 1984. A summary of data file structure and data stan-dards for the Yale-Geological Survey of Pakistan Project-Si-walik fossil collections. Pp. 71–74 in Shah and Pilbeam 1984.

———. 1995. Faunal turnover and diversity in the terrestrialNeogene of Pakistan. Pp. 115–134 in E. S. Vrba, G. H. Denton,T. C. Partridge, and L. H. Burckle, eds. Paleoclimate and evo-lution, with emphasis on human origins. Yale UniversityPress, New Haven, Conn.

Barry, J. C., A. K. Behrensmeyer, and M. Monaghan. 1980. A geo-


logic and biostratigraphic framework for Miocene sedimentsnear Khaur Village, northern Pakistan. Postilla 183:1–19.

Barry, J. C., N. M. Johnson, S. M. Raza, and L. L. Jacobs. 1985.Neogene mammalian faunal change in southern Asia: corre-lations with climatic, tectonic and eustatic events. Geology 13:637–640.

Barry, J. C., L. J. Flynn, and D. R. Pilbeam. 1990. Faunal diversityand turnover in a Miocene terrestrial sequence. Pp. 381–421in R. M. Ross and W. D. Allmon, eds. Causes of evolution: apaleontological perspective. University of Chicago Press, Chi-cago.

Barry, J. C., M. E. Morgan, A. J. Winkler, L. J. Flynn, E. H. Lind-say, L. L. Jacobs, and D. Pilbeam. 1991. Faunal interchange andMiocene terrestrial vertebrates of southern Asia. Paleobiology17:231–245.

Barry, J. C., M. E. Morgan, L. J. Flynn, D. Pilbeam, L. L. Jacobs,E. H. Lindsay, S. M. Raza, and N. Solounias. 1995. Patterns offaunal turnover and diversity in the Neogene Siwaliks ofnorthern Pakistan. Pp. 209–226 in Badgley and Behrensmeyer1995a.

Beck, R. A., and D. W. Burbank. 1990. Continental-scale diver-sion of rivers: a control of alluvial stratigraphy. Geological So-ciety of America Abstracts with Programs 22:238.

Behrensmeyer, A. K. 1982. Time resolution in fluvial vertebrateassemblages. Paleobiology 8:211–227.

———. 1987. Miocene fluvial facies and vertebrate taphonomyin northern Pakistan. In F. G. Ethridge, R. M. Flores, and M.D. Harvey, eds. Recent developments in fluvial sedimentolo-gy. Society for Economic Paleontology and Mineralogy Spe-cial Publication 39:169–176.

Behrensmeyer, A. K., and M. Raza. 1984. A procedure for doc-umenting fossil localities in Siwalik deposits of northern Pak-istan. Pp. 65–69 in Shah and Pilbeam 1984.

Behrensmeyer, A. K., and L. Tauxe. 1982. Isochronous fluvialsystems in Miocene deposits of northern Pakistan. Sedimen-tology 29:33l–352.

Behrensmeyer, A. K., B. J. Willis, and J. Quade. 1995. Flood-plains and paleosols of Pakistan Neogene and Wyoming Pa-leogene deposits: implications for the taphonomy and paleo-ecology of faunas. Pp. 37–60 in Badgley and Behrensmeyer1995a.

Behrensmeyer, A. K., N. E. Todd, R. Potts, and G. E. McBrinn.1997. Late Pliocene faunal turnover in the Turkana Basin,Kenya and Ethiopia. Science 278:1589–1594.

Berggren, W. A., D. V. Kent, J. J. Flynn, and J. A. Van Couvering.1985. Cenozoic geochronology. Geological Society of AmericaBulletin 96:1407–1418.

Brett, C. E., L. C. Ivany, and K. M. Schopf. 1996. Coordinatedstasis: an overview. Pp. 1–20 in Ivany and Schopf 1996.

Burbank, D. W., and R. A. Beck. 1991. Models of aggradationversus progradation in the Himalayan foreland. GeologischeRundschau 80:623–638.

Burbank, D. W., L. A. Derry, and C. France-Lanord. 1993. Re-duced Himalayan sediment production 8 Myr ago despite anintensified monsoon. Nature 364:48–50.

Cande, S. C., and D. V. Kent. 1992. A new geomagnetic polaritytimescale for the Late Cretaceous and Cenozoic. Journal ofGeophysical Research 97B:13917–13951.

———. 1995. Revised calibration of the geomagnetic polaritytimescale for the Late Cretaceous and Cenozoic. Journal ofGeophysical Research 100B:6093–6095.

Cerling, T. E., J. Quade, Y. Wang, and J. R. Bowman. 1989. Car-bon isotopes in soils and palaeosols as ecology and palaeoe-cology indicators. Nature 341:138–139.

Cerling, T. E., Y. Wang, and J. Quade. 1993. Expansion of C4 eco-systems as an indicator of global ecological change in the lateMiocene. Nature 361:344–345.

Cerling, T. E., J. M. Harris, B. J. MacFadden, M. G. Leakey, J.

Quade, V. Eisenmann, and J. R. Ehleringer. 1997. Global veg-etation change through the Miocene/Pliocene boundary. Na-ture 389:153–158.

———. 1998. Reply to: Miocene/Pliocene shift: one step or sev-eral? Nature 393:127.

Cheema, M. R., S. M. Raza, and H. Ahmed. 1977. Cainozoic. InS. M. I. Shah, ed. Stratigraphy of Pakistan. Memoirs of theGeological Survey of Pakistan 12:56–98.

Coleman, M., and K. Hodges. 1995. Evidence for Tibetan Pla-teau uplift before 14 Myr ago from a new minimum age foreast-west extension. Nature 374:49–52.

Copeland, P. 1992. Two-phase uplift of Higher Himalayas since17 Ma: Comment. Geology 21:188–189.

Ehrenfeld, S., and S. B. Littauer. 1964. Introduction to statisticalmethod. McGraw-Hill, New York.

Filippelli, G. M. 1997. Intensification of the Asian monsoon anda chemical weathering event in the late Miocene–early Plio-cene: implications for late Neogene climate change. Geology25:27–30.

Flynn, L. J. 1982. Systematic revision of Siwalik Rhizomyidae (Ro-dentia). Geobios 15:327–389.

Flynn, L. J., D. Pilbeam, L. L. Jacobs, J. C. Barry, A. K. Behrens-meyer, and J. W. Kappelman. 1990. The Siwaliks of Pakistan:time and faunas in a Miocene terrestrial setting. Journal of Ge-ology 98:589–604.

Flynn, L. J., J. C. Barry, M. E. Morgan, D. Pilbeam, L. L. Jacobs,and E. H. Lindsay. 1995. Neogene Siwalik mammalian line-ages: species longevities, rates of change, and modes of spe-ciation. Pp. 249–264 in Badgley and Behrensmeyer 1995a.

Flynn, L. J., W. Downs, M. E. Morgan, J. C. Barry, and D. Pil-beam. 1998. High Miocene species richness in the Siwaliks ofPakistan. In Y. Tomida, L. J. Flynn, and L. L. Jacobs, eds. Ad-vances in vertebrate paleontology and geochronology. Na-tional Science Museum Monographs 14:167–180. National Sci-ence Museum, Tokyo.

France-Lanord, C., and L. A. Derry. 1994. d13C of organic carbonin the Bengal Fan: source evolution and transport of C3 andC4 plant carbon to marine sediments. Geochimica et Cosmo-chimica Acta 58:4809–4814.

Gunnell, G. F., M. E. Morgan, M. C. Maas, and P. D. Gingerich.1995. Comparative paleoecology of Paleogene and Neogenemammalian faunas: trophic structure and composition. Pp.265–286 in Badgley and Behrensmeyer 1995a.

Harrison, T. M., P. Copeland, S. A. Hall, J. Quade, S. Burner, T.P. Ojha, and W. S. F. Kidd. 1993. Isotopic preservation of Hi-malayan/Tibetan uplift, denudation, and climatic histories oftwo molasse deposits. Journal of Geology 101:157–175.

Ivany, L. C., and K. M. Schopf, eds. 1996. New perspectives onfaunal stability in the fossil record. Palaeogeography, Palaeo-climatology, Palaeoecology Vol. 127.

Johnson, N. M., and V. E. McGee. 1983. Magnetic polarity stra-tigraphy: stochastic properties of data, sampling problems,and the evaluation of interpretations. Journal of GeophysicalResearch 88B:1213–1221.

Johnson, N. M., N. D. Opdyke, G. D. Johnson, E. H. Lindsay, andR. A. K. Tahirkheli. 1982. Magnetic polarity stratigraphy andages of Siwalik group rocks of the Potwar Plateau, Pakistan.Palaeogeography, Palaeoclimatology, Palaeoecology 37:17–42.

Johnson, N. M., J. Stix, L. Tauxe, P. F. Cerveny, and R. A. K. Ta-hirkheli. 1985. Paleomagnetic chronology, fluvial processesand tectonic implications of the Siwalik deposits near ChinjiVillage, Pakistan. Journal of Geology 93:27–40.

Johnson, N. M., K. A. Sheikh, E. Dawson-Saunders, and L. E.McRae. 1988. The use of magnetic-reversal time lines in strati-graphic analysis: a case study in measuring variability in sed-imentation rates. Pp. 189–200 in K. L. Kleinspehn and C. Pao-


la, eds. New perspectives in basin analysis. Springer, NewYork.

Kappelman, J., J. Kelley, D. Pilbeam, K. A. Sheikh, S. Ward, M.Anwar, J. C. Barry, B. Brown, P. Hake, N. M. Johnson, S. M.Raza, and S. M. I. Shah. 1991. The earliest occurrence of Si-vapithecus from the Middle Miocene Chinji Formation of Pak-istan. Journal of Human Evolution 21:61–73.

Koch, C. F. 1987. Prediction of sample size effects on the mea-sured temporal and geographic distribution patterns of spe-cies. Paleobiology 13:100–107.

Kohler, M., S. Moya-Sola, and J. Agustı. 1998. Miocene/Plioceneshift: one step or several? Nature 393:126.

Kroon, D., T. Steens, and S. R. Troelstra. 1991. Onset of mon-soonal related upwelling in the western Arabian Sea as re-vealed by planktonic foraminifers. Proceedings of the OceanDrilling Program, Scientific Results 117:257–263.

Krynine, P. D. 1937. Petrography and genesis of the Siwalik Se-ries. American Journal of Science 34:422–446.

Maas, M. C., M. R. L. Anthony, P. D. Gingerich, G. F. Gunnell,and D. W. Krause. 1995. Mammalian generic diversity andturnover in the Late Paleocene and Early Eocene of the Big-horn and Crazy Mountains Basins, Wyoming and Montana(USA). Pp. 181–207 in Badgley and Behrensmeyer 1995a.

Magurran, A. E. 1988. Ecological diversity and its measure-ment. Princeton University Press, Princeton, N.J.

Mankinen, E. A., and G. B. Dalrymple. 1979. Revised geomag-netic polarity time scale for the interval 0–5 m.y. B.P. Journalof Geophysical Research 84B:615–626.

Marshall, C. R. 1994. Confidence intervals on stratigraphic rang-es: partial relaxation of the assumption of randomly distrib-uted fossil horizons. Paleobiology 20:459–469.

———. 1997. Confidence intervals on stratigraphic ranges withnonrandom distributions of fossil horizons. Paleobiology 23:165–173.

McKinney, M. L., J. L. Lockwood, and D. R. Frederick. 1996. Rarespecies and scale-dependence in ecosystem studies. Pp. 191–207 in Ivany and Schopf 1996.

McMurtry, M. G. 1980. Facies changes and time relationshipsalong a sandstone stratum, Middle Siwalik Group, PotwarPlateau, Pakistan. B.S. thesis. Dartmouth College, Hanover,N.H.

McRae, L. E. 1990. Paleomagnetic isochrons, unsteadiness, andnon-uniformity of sedimentation in Miocene fluvial strata ofthe Siwalik Group, northern Pakistan. Journal of Geology 98:433–456.

Meigs, J. J., D. W. Burbank, and R. A. Beck. 1995. Middle-LateMiocene (.10 Ma) formation of the Main Boundary thrust inthe western Himalayan. Geology 23:423–426.

Molnar, P., P. England, and J. Martinod. 1993. Mantle dynamics,uplift of the Tibetan Plateau, and the Indian monsoon. Re-views of Geophysics 31:357–396.

Morgan, M. E., J. D. Kingston, and B. D. Marino. 1994. Carbonisotopic evidence for the emergence of C4 plants in the Neo-gene from Pakistan and Kenya. Nature 367:162–165.

Morgan, M. E., C. Badgley, G. F. Gunnell, P. D. Gingerich, J. W.Kappelman, and M. C. Maas. 1995. Comparative paleoecol-ogy of Paleogene and Neogene mammalian faunas: body-sizestructure. Pp. 265–286 in Badgley and Behrensmeyer 1995a.

Olson, E. C. 1980. Taphonomy: its history and role in commu-nity evolution. Pp. 5–19 in A. K. Behrensmeyer and A. P. Hill,eds. Fossils in the making: vertebrate taphonomy and paleo-ecology. University of Chicago Press, Chicago.

Opdyke, N. D., E. H. Lindsay, G. D. Johnson, N. M. Johnson, R.A. K. Tahirkheli, and M. A. Mirza. 1979. Magnetic polaritystratigraphy and vertebrate paleontology of the Upper Si-walik Subgroup of northern Pakistan. Palaeogeography, Pa-laeoclimatology, Palaeoecology 27:1–34.

Pagani, M., K. H. Freeman, and M. A. Arthur. 1999. Late Mio-

cene atmospheric CO2 concentrations and the expansion of C4

grasses. Science 285:876–879.Pilbeam, D., M. Morgan, J. C. Barry, and L. Flynn. 1996. Euro-

pean MN units and the Siwalik faunal sequence of Pakistan.Pp. 96–105 in R. L. Bernor, V. Fahlbusch, and H.-W. Mittmann,eds. The evolution of western Eurasian Neogene mammal fau-nas. Columbia University Press, New York.

Prell, W. L., and J. E. Kutzbach. 1992. Sensitivity of the Indianmonsoon to forcing parameters and implications for its evo-lution. Nature 360:647–652.

Prothero, D. R. 1998. Does climate control mammalian evolu-tion? A test of the turnover pulse hypothesis. Journal of Ver-tebrate Paleontology 18:70A–71A.

———. 1999. Does climate change drive mammalian evolution?GSA Today 9(9):1–7.

Quade, J., and T. E. Cerling. 1995. Expansion of C4 grasses in theLate Miocene of northern Pakistan: evidence from stable iso-topes in paleosols. Pp. 91–116 in Badgley and Behrensmeyer1995a.

Quade, J., T. E. Cerling, and J. R. Bowman. 1989. Developmentof Asian monsoon revealed by marked ecological shift duringthe latest Miocene in northern Pakistan. Nature 342:163–166.

Quade, J., T. E. Cerling, J. Barry, M. E. Morgan, D. R. Pilbeam,A. R. Chivas, J. A. Lee-Thorp, and N. J. Van der Merwe. 1992.A 16-Ma record of paleodiet using carbon and oxygen iso-topes in fossil teeth from Pakistan. Chemical Geology (Iso-tope Geoscience Section) 94:182–192.

Quade, J., J. M. L. Cater, T. P. Ojha, J. Adam, and T. M. Harrison.1995. Late Miocene environmental change in Nepal and thenorthern Indian subcontinent: stable isotopic evidence frompaleosols. Geological Society of America Bulletin 107:1381–1397.

Ramstein, G., F. Fluteau, J. Besse, and S. Joussaume. 1997. Effectof orogeny, plate motion and land-sea distribution on Eur-asian climate change over the past 30 million years. Nature386:788–795.

Raymo, M. E., and W. F. Ruddiman. 1992. Tectonic forcing of lateCenozoic climate. Nature 359:117–122.

Rea, D. K. 1992. Delivery of Himalayan sediment to the northernIndian Ocean and its relation to global climate, sea level, up-lift, and seawater strontium. American Geophysical UnionGeophysical Monograph 70:387–402.

Retallack, G. J. 1991. Miocene paleosols and ape habitats of Pak-istan and Kenya. Oxford University Press, Oxford.

Sadler, P. M. 1981. Sediment accumulation rates and the com-pleteness of stratigraphic sections. Journal of Geology 89:569–584.

Shah, S. M. I., and D. Pilbeam, eds. 1984. Contribution to theGeology of Siwaliks of Pakistan. Memoirs of the GeologicalSurvey of Pakistan No. XI. Geological Survey of Pakistan,Quetta.

Shaw, A. B. 1964. Time in stratigraphy. McGraw-Hill, New York.Shukla, J. 1987. Interannual variability of monsoons. Pp. 399–

463 in J. S. Fein and P. L. Stephens, eds. Monsoons. Wiley, NewYork.

Stern, L. A., G. D. Johnson, and C. P. Chamberlain. 1994. Carbonisotope signature of environmental change found in fossil rat-ite eggshells from a South Asian Neogene sequence. Geology22:419–422.

Stix, J. 1982. Stratigraphy of the Kamlial Formation near ChinjiVillage, northern Pakistan. A.B. thesis. Dartmouth College,Hanover, N.H.

Strauss, D., and P. M. Sadler. 1989. Classical confidence intervalsand Bayesian probability estimates for ends of local taxonranges. Mathematical Geology 21:411–427.

Tauxe, L., and C. Badgley. 1988. Stratigraphy and remanence ac-quisition of a paleomagnetic reversal in alluvial Siwalik rocksof Pakistan. Sedimentology 35:697–715.


Tauxe, L., and N. D. Opdyke. 1982. A time framework based onmagnetostratigraphy for the Siwalik sediments of the Khaurarea, northern Pakistan. Palaeogeography, Palaeoclimatology,Palaeoecology 37:43–61.

Turner, S., C. Hawkesworth, J. Liu, N. Rogers, S. Kelley, and P.van Calsteren. 1993. Timing of Tibetan uplift constrained byanalysis of volcanic rocks. Nature 364:50–54.

Van Valen, L. 1973. A new evolutionary law. Evolutionary The-ory 1:1–30.

Vrba, E. S. 1980. Evolution, species, and fossils: how does lifeevolve? South African Journal of Science 76:61–84.

———. 1985. Environment and evolution: alternative causes ofthe temporal distribution of evolutionary events. South Afri-can Journal of Science 81:229–236.

———. 1995. On the connections between paleoclimate andevolution. Pp. 24–45 in E. S. Vrba, G. H. Denton, T. C. Par-tridge, and L. H. Burckle, eds. Paleoclimate and evolution,with emphasis on human origins. Yale University Press, NewHaven, Conn.

Webb, S. D. 1984. On two kinds of rapid faunal turnover. Pp.

417–436 in W. A. Berggren and J. A. van Couvering, eds. Ca-tastrophes and earth history: the new uniformitarianism.Princeton University Press, Princeton, N.J.

Willis, B. J. 1993a. Ancient river systems in the Himalayanforedeep, Chinji Village area, northern Pakistan. SedimentaryGeology 88:1–76.

———. 1993b. Evolution of Miocene fluvial systems in the Hi-malayan foredeep through a two-kilometer-thick successionin northern Pakistan. Sedimentary Geology 88:77–121.

Willis, B. J., and A. K. Behrensmeyer. 1994. Architecture of Mio-cene overbank deposits in northern Pakistan. Journal of Sed-imentary Research B 64:60–67.

———. 1995. Fluvial systems in the Siwalik Miocene and Wy-oming Paleogene. Pp. 13–35 in Badgley and Behrensmeyer1995a.

Zaleha, M. J. 1997a. Fluvial and lacustrine palaeoenvironmentsof the Miocene Siwalik Group, Khaur area, northern Pakistan.Sedimentology 44:349–368.

———. 1997b. Intra- and extrabasinal controls on fluvial de-position in the Miocene Indo-Gangetic foreland basin, north-ern Pakistan. Sedimentology 44:369–390.


App

endi

x1

Stra

tig

rap

hic

sect

ion

s.

Reg

ion

/Se

ctio

nT

hic

kn

ess

(m)

Pm

agR

efer

ence

s*C

omm

ents

Ro

hta

sSa

ngh

oi/

Bas

awa

1850

Yes

1Ja

lalp

ur

Jarm

agh

alK

as21

00Ye

s1,

2Sa

me

asC

ham

bal

Kas

inR

ef.1

Has

no

tA

rea

An

dar

Kas

Kot

alK

un

d13

0018

00Ye

sYe

s2 2

Low

erD

hal

aN

ala

862

Yes

2,th

isp

aper

(Fig

.3)

Incl

ud

estw

olo

wer

exte

nsi

ons

tota

lin

g31

3m

that

are

not

incl

ud

edin

Ref

.2.

Low

er10

0m

wit

hou

tp

aleo

mag

Dh

okSa

ira

250

Yes

Th

isp

aper

(Fig

.2)

Rem

easu

red

sect

ion

for

up

per

par

tof

Ref

.2P

adh

riY

910

Sect

ion

Has

not

Eas

tof

Bh

and

ar

180 68 180

100

No

No

No

No

Th

isp

aper

Th

isp

aper

Th

isp

aper

Th

isp

aper

Pil

lar

IP

illa

rII

Pil

lar

III

Pil

lar

IVP

illa

rV

Pil

lar

VI

64 71 66 55 80 90

Yes

Yes

Yes

Yes

Yes

Yes

3 3 3 3 3 3Ta

trot

-Bh

and

arTa

trot

52 60Ye

sYe

s2,

31

‘‘Ta

trot

’’in

Ref

.3

Kh

aur

Are

aK

un

dva

lK

as50

0Ye

s4

Dh

okP

ath

anR

est

Hou

se49

7Ye

s5,

this

pap

er(F

ig.4

)B

hia

nw

ala

Kas

1692

No

5Sh

own

aslo

wer

par

tof

Dh

okP

ath

anSe

ctio

nin

Ref

.5U

tran

Utr

an(U

N)

Kot

Mal

iara

nK

otM

alia

ran

(KM

)C

hou

triw

ali

Kas

Ch

outr

iwal

iK

as(C

H)

Dh

okM

ila

Kh

aur

Kas

(KK

)

1300 70 60

0 6017

08 100

1000 27

0

Yes

No

Yes

No

No

Yes

Yes

Yes

4,6

7 4,5,

67 5 7 4,

65,

7

Rem

easu

red

seg

men

t

Rem

easu

red

seg

men

t

Rem

easu

red

and

resa

mp

led

up

per

seg

men

t

Pal

eom

agsa

mp

lin

gin

low

er75

mH

asal

Kas

(HL

)24

25Ye

s4,

6,7,

8M

ain

sect

ion

mea

sure

dby

G.

D.J

ohn

son

.‘‘B

ora

Kas

’’in

Ref

.8H

asal

Kas

2850

No

5Se

ctio

nof

Ref

.4re

mea

sure

dby

Raz

aan

dM

eyer

Y22

7Se

ctio

nD

ing

aK

as(D

K)

Mal

hu

wal

aK

as

75 120

1200

No

Yes

Yes

7 5,7,

84,

6


App

endi

x1.

Con

tin

ued

.

Reg

ion

/Se

ctio

nT

hic

kn

ess

(m)

Pm

agR

efer

ence

s*C

omm

ents

Mal

hu

wal

aK

as(M

KM

)22

30Ye

s5,

7,8

Up

per

838

mof

pal

eom

agsa

mp

lin

gof

Ref

.8tr

ansf

erre

dto

mea

sure

dse

ctio

nof

Ref

.5Ja

bbi

203

Yes

5,8

Up

per

par

tof

sect

ion

inR

efs.

5an

d8

has

ad

up

lica

ted

seg

men

tG

and

akas

Roa

d(G

KR

)G

and

akas

(GK

)B

hag

wa

Kas

Sou

th(B

WS)

Bh

agw

aK

as(B

W)

70 160 75 110

No

Yes

Yes

Yes

7 5,6,

7,8

7,9

5,7,

8,9

Is‘‘

Sect

ion

5’’

inR

ef.9

Is‘‘

Sect

ion

6’’

inR

ef.9

Y26

9Se

ctio

nSi

ltB

rid

ge

(SB

)R

ath

aK

asSo

uth

(RK

S)R

ath

aK

as(R

K)

Th

agen

wal

iD

hok

(TW

D)

Y26

0Se

ctio

n(K

AK

)

75 30 3510

06 170 75

No

No

No

Yes

Yes

Yes

7 7 7 5,7,

87,

87

Kau

lial

Kas

(KU

L)

Kau

lial

Kas

(KU

L)

3231 70

Yes

Yes

5,8

7Is

‘‘K

UL

’’an

d‘‘

LK

’’of

Ref

.8co

mbi

ned

Ad

dit

ion

alp

aleo

mag

add

edto

Ref

.8C

hin

jiA

rea

Gab

hir

2100

Yes

2,10

,11

Pal

eom

agsi

tes

ofR

efs.

2an

d10

pla

ced

onto

sect

ion

ofR

ef.1

1C

hit

apar

wal

a74

5Ye

s10

,12

*R

efer

ence

s1.

Op

dy

keet

al.

1979

.2.

Joh

nso

net

al.

1982

.3.

McM

urt

ry19

80.

4.B

arn

dt

1977

.5.

Bar

ryet

al.

1980

.6.

Bar

nd

tet

al.

1978

.7.

Beh

ren

smey

eran

dTa

uxe

1982

.8.

Tau

xean

dO

pd

yke

1982

.9.

Tau

xean

dB

adg

ley

1988

.10

.Joh

nso

net

al.

1985

.11

.Wil

lis

1993

a.12

.Sti

x19

82.


Appendix 2Locality data.


Appendix 2. Continued.


Appendix 3Recalculated ages for isotopic samples.

Samplenumber Section Fraction d13C* d18O* Age (Ma)

Minimumage (Ma)

Maximumage (Ma)

97339730972997019702968496989694969396899683966996769672967196659663965796569655

JalalpurJalalpurJalalpurJalalpurJalalpurJalalpurJalalpurJalalpurJalalpurJalalpurJalalpurJalalpurJalalpurJalalpurJalalpurJalalpurJalalpurJalalpurJalalpurJalalpur

CarbonateCarbonateOrganicCarbonateOrganicOrganicCarbonateCarbonateCarbonateCarbonateCarbonateCarbonateCarbonateOrganicCarbonateCarbonateCarbonateCarbonateOrganicCarbonate

2.521.9

214.51.3

214.3216.5

2.01.80.41.8

22.021.9

0.7214.6

0.621.020.7

0.5214.722.3

26.827.8

—26.6

——

26.326.226.126.525.727.226.5

—27.027.528.026.4

—24.9

2.2702.5752.5753.3033.3033.3943.3943.4253.4413.5023.6903.7434.3604.6804.6805.1275.7675.8945.9065.912

2.2032.5172.5173.1883.1883.3703.3703.3963.4103.4633.4633.5694.2214.5704.5705.0585.7555.8945.8945.906

2.3372.6332.6333.4183.4183.4183.4183.4533.4713.5413.9173.9174.4994.7894.7895.1965.7785.8945.9175.917

9654965296489646964396409639963791879186a9186b918591849183918296309181918091789179917991779625

JalalpurJalalpurJalalpurJalalpurJalalpurJalalpurJalalpurJalalpurKaulial KasKaulial KasKaulial KasKaulial KasKaulial KasKaulial KasKaulial KasJalalpurKaulial KasKaulial KasKaulial KasKaulial KasKaulial KasKaulial KasJalalpur

CarbonateCarbonateCarbonateCarbonateCarbonateCarbonateCarbonateCarbonateCarbonateCarbonateCarbonateCarbonateCarbonateCarbonateCarbonateCarbonateCarbonateCarbonateCarbonateOrganicHumateCarbonateCarbonate

0.121.320.3

0.023.521.423.020.727.029.528.827.9

210.123.629.525.0

210.227.828.6

224.2224.328.627.9

27.224.925.426.926.927.427.726.326.827.326.527.629.127.127.927.427.428.127.5

——

27.127.9

5.9385.9826.0696.1256.4296.5946.6126.7617.0427.0427.0427.1027.2677.3667.4757.4777.5237.5867.6277.6277.6277.7367.922

5.9245.9536.0136.0726.2556.5676.5866.7407.0407.0407.0407.0967.2327.3357.4567.1907.5057.5707.6077.6077.6077.7077.859

5.9526.0106.1256.1786.6036.6216.6386.7817.0457.0457.0457.1077.3027.3977.4937.7637.5407.6017.6467.6467.6467.7647.984

9624962291749175917396199168916291609161916191589613960796119764976297589757

JalalpurJalalpurKaulial KasKaulial KasKaulial KasJalalpurKaulial KasKaulial KasKaulial KasKaulial KasKaulial KasKaulial KasJalalpurJalalpurJalalpurJalalpurJalalpurJalalpurJalalpur

CarbonateCarbonateCarbonateCarbonateCarbonateCarbonateCarbonateCarbonateCarbonateOrganicHumateCarbonateCarbonateCarbonateCarbonateCarbonateCarbonateCarbonateCarbonate

211.228.828.9

211.6211.5210.8211.8212.9210.6224.0224.0212.6211.0210.0210.129.829.9

210.7210.8

26.228.027.028.927.429.025.428.727.3

——

25.429.929.929.729.428.029.528.8

7.9527.9848.0588.1098.3418.4018.6258.9699.0389.0389.0389.1549.2709.4119.642

10.13910.60111.67312.229

7.8897.9208.0408.0898.3178.0728.5878.9559.0229.0229.0229.1419.0089.1539.4729.940

10.50511.59312.146

8.0158.0478.0768.1298.3648.7298.6638.9829.0539.0539.0539.1679.5329.6689.812

10.33810.69611.75212.311

* Isotopic values from Quade and Cerling 1995.


Appendix 4Stratigraphic ranges.

Species

Inferredlast

appear-ance(Ma)

Lastoccur-rence(Ma)

Firstoccur-rence(Ma)

Inferredfirst

appear-ance*(Ma)

No. ofintervalsin whichobserved

Large mammals1 Colobinae2 Hominoidea3 Hominoidea4 Deinotheriidae5 Manidae6 Orycteropodidae7 Listriodontinae8 Tetraconodontinae9 Tetraconodontinae

10 Suinae11 Suinae

cf. Presbytis sivalensisSivapithecus spp.Sivapithecus parvadaDeinotherium spp.Manis ‘‘Y260 species’’Orycteropus spp.Listriodon pentapotamiaeConohyus sindiensisTetraconodon magnusPropotomochoerus hysudricus?Hippopotamodon ‘‘Y450 unnamed

species’’

6.58.4†

9.8†

7.9†

2.28.1

10.3†

10.3†

9.3†

6.510.2†

7.18.5

10.08.06.99.2

10.310.3

9.36.8

10.2

7.612.310.012.9

9.213.613.713.110.010.211.3

7.8†

14.010.0†

14.09.9

14.014.014.011.210.2†

11.4

311

116

28

1310

329

5

12 Suinae13 Suinae14 Suinae15 Suinae16 Suinae17 Doliochoerinae18 Anthracotheriidae19 Anthracotheriidae20 Anthracotheriidae21 Hippopotamidae22 Tragulidae23 Tragulidae24 Tragulidae25 Tragulidae26 Tragulidae

Hippopotamodon sivalensePotamochoerus ‘‘Y553 species’’?Sus ‘‘D013 species’’‘‘Small Suinae/Y311 species’’‘‘Small Suinae/Y406 species’’Schizochoerus gandakasensisMicrobunodon punjabienseHemimeryx spp.?Merycopotamus dissimilusHexaprotodon sivalensisDorcabune nagriiDorcabune anthracotherioidesDorcatherium ‘‘Y270 species’’Dorcatherium nagriiDorcatherium ‘‘Y259 species’’

7.1†

2.22.29.42.28.18.63.62.22.28.4†

10.3†

8.62.2

10.3†

7.25.43.3

10.07.48.79.26.23.33.58.5

10.59.26.8

10.4

10.25.86.4

10.08.9

10.112.313.6

5.85.9

10.412.811.1

9.312.3

10.2†

5.8†

6.811.49.2

11.214.014.06.16.1†

11.214.011.29.5†

14.0

21231368

4024

1096

123

27 Tragulidae28 Tragulidae29 Tragulidae30 Tragulidae31 Tragulidae32 Tragulidae33 Giraffinae34 Sivatheriinae35 Sivatheriinae36 Bovidae37 Bovidae38 Bovidae39 Bovidae40 Bovidae41 Tragelaphini

Dorcatherium ‘‘Y311 species’’Dorcatherium ‘‘Y373 species’’Dorcatherium ‘‘Y457 species’’Dorcatherium cf. majusDorcatherium majus‘‘Tragulidae/L101 unnamed species’’Giraffa punjabiensisBramatherium megacephalumGiraffokeryx punjabiensisSivoreas eremita‘‘Bovidae/Y166 unnamed species’’‘‘Bovidae/Y905 unnamed species’’‘‘Small Bovidae/Y581 unnamed species’’‘‘Bovidae/Y545 unnamed species’’‘‘?Tragelaphini/D013 species’’

7.1†

6.57.3†

10.5†

6.9†

2.22.27.1†

10.3†

10.38.47.46.37.45.8

7.26.87.3

10.77.03.37.27.1

10.310.7

8.77.87.08.06.2

10.37.87.3

11.310.4

5.98.9

10.313.611.3

8.77.87.08.06.4

10.3†

7.8†

7.711.410.6†

6.1†

9.1†

10.3†

14.014.09.17.8†

7.1†

8.46.7

21513

1943

2620

211113

42 Boselaphini43 Boselaphini44 Boselaphini45 Boselaphini46 Boselaphini47 Boselaphini

48 Boselaphini

49 Boselaphini50 Reduncini51 Reduncini52 Hippotragini53 Antilopini54 Antilopini

Elachistoceras khauristanensisSelenoportax spp.Tragoceridus spp.cf. Eotragus ‘‘Y166 unnamed species’’Tragoportax salmontanus‘‘Medium Boselaphini/Y581 unnamed

species’’‘‘Medium Boselaphini/Y195 unnamed

species’’‘‘Large Boselaphini/Y927 species’’‘‘?Reduncini/D013 species’’Dorcadoxa porrecticornis‘‘?Hippotragini/Y453 species’’Gazella spp.cf. Prostrepsiceros vinayaki

7.3†

7.9†

2.23.67.36.3

8.1

7.3†

2.26.52.22.27.4

7.47.96.28.78.47.0

9.3

7.36.47.06.26.27.9

11.510.211.2

9.39.37.0

9.3

8.06.48.16.8

11.38.3

14.010.3†

11.29.99.97.1†

9.9

8.1†

6.98.1†

6.9†

14.08.6

161222

231

1

2183

234



Species

Inferredlast

appear-ance(Ma)

Lastoccur-rence(Ma)


Inferredfirst

appear-ance*(Ma)


55 Antilopini56 Hipparionini57 Chalicotherini

cf. Protragelaphus skouzesi‘‘Hipparion s.l.’’ spp.Chalicotherium salinum

8.8†

3.78.0†

8.85.88.0

8.810.712.9

8.8†

10.914.0

14313

Small mammals58 Tupaiidae59 Galericinae60 Galericinae61 Crocidurinae62 Crocidurinae63 Crocidurinae64 Lorisidae65 Leporidae66 Ratufini67 Tamiini68 Petauristinae69 Petauristinae70 Petauristinae

Tupaiidae, spp.Galerix rutlandaeSchizogalerix ‘‘Y259 species’’cf. Crocidura ‘‘Y367 species’’cf. Crocidura ‘‘Y311 species’’cf. Crocidura ‘‘D013 species’’Lorisidae, spp.‘‘Leporidae/D013 species’’Ratufa ‘‘Y259 species’’Eutamias ‘‘259 species’’Hylopetes ‘‘Y388 species’’cf. Hylopetes ‘‘Y259 species’’‘‘Petauristinae/Y311 species’’

6.58.5†

9.35.76.55.78.83.68.86.55.79.39.3

8.18.7

10.26.47.36.49.23.69.27.07.3

10.010.0

13.013.610.412.711.1

6.411.1

7.310.410.4

8.711.510.0

13.613.611.013.611.27.1

12.27.8

11.011.09.1

13.610.3

1072

1211

14647351

71 Gerbillinae72 Copemyinae73 Copemyinae74 Copemyinae75 Copemyinae76 Copemyinae77 Copemyinae78 Copemyinae79 Copemyinae80 Dendromurinae81 Dendromurinae82 Tachyoryctinae83 Tachyoryctinae84 Tachyoryctinae85 Tachyoryctinae

Abudhabia pakistanensisDemocricetodon ‘‘species D’’Democricetodon ‘‘species B’’Democricetodon ‘‘species E’’Democricetodon ‘‘species H’’Democricetodon ‘‘species F’’ (young)Democricetodon ‘‘species F’’ (old)Democricetodon ‘‘species G’’ (young)Democricetodon ‘‘species G’’ (old)Dakkamys asiaticusParadakkamys chinjiensisKanisamys sivalensisKanisamys nagriiProtachyoryctes tatrotiEicooryctes kaulialensis

8.210.1†

9.310.3†

10.18.29.38.29.3

10.110.17.3†

9.4†

6.56.5

8.610.210.010.410.4

8.610.0

8.610.010.410.4

7.39.66.97.1

8.613.013.613.013.0

8.712.3

8.713.011.513.0

9.311.5

7.87.8

8.6†

13.613.613.613.68.8†

13.68.8†

13.613.613.69.5†

13.68.0†

7.8†

11010

99262846

10733

86 Tachyoryctinae87 Tachyoryctinae88 Rhizomyinae89 Rhizomyinae90 Rhizomyinae91 Rhizomyinae92 Rhizomyinae93 Rhizomyinae94 Rhizomyinae95 Murinae96 Murinae97 Murinae98 Murinae99 Murinae

100 Murinae

Rhizomyides sivalensisRhizomyides punjabiensisBrachyrhizomys ‘‘Y535 species’’Brachyrhizomys nagriiBrachyrhizomys pilgrimiBrachyrhizomys micrusBrachyrhizomys blackiBrachyrhizomys tetracharaxBrachyrhizomys choristosProgonomys ‘‘Y581 unnamed species’’Progonomys ‘‘Y259 unnamed species’’Progonomys debruijniProgonomys nr. debruijniKarnimata ‘‘Y450 unnamed species’’Karnimata ‘‘Y388 unnamed species’’

5.710.1†

8.2†

9.0†

5.78.28.25.78.2†

6.59.4†

8.27.3†

9.4†

8.2

6.410.2

8.49.28.09.29.27.98.27.09.68.67.39.68.7

7.110.4

8.49.69.39.29.29.38.47.8

11.59.38.1

11.39.2

7.1†

11.08.5†

9.910.310.310.310.38.5†

7.8†

13.69.98.5

11.49.9

321421142475363

101 Murinae102 Murinae103 Murinae104 Murinae105 Murinae106 Murinae107 Murinae108 Murinae109 Murinae110 Murinae111 Murinae

Karnimata darwiniKarnimata cf. darwiniKarnimata huxleyiKarnimata cf. huxleyiKarnimata ‘‘Y024 unnamed species’’Parapodemus ‘‘Y182 species’’Mus auctorcf. Mus ‘‘Y931 species’’cf. Parapelomys robertsiParapelomys robertsi‘‘Murinae/D013 species’’

8.7†

7.1†

5.76.57.3†

7.35.77.1†

7.1†

5.75.7

8.77.26.47.07.37.96.47.27.26.46.4

9.37.26.47.08.19.27.37.28.17.06.4

9.5†

7.2†

6.77.1†

8.3†

9.97.77.2†

8.57.1†

7.2

41113231531



Species

Inferredlast

appear-ance(Ma)

Lastoccur-rence(Ma)


Inferredfirst

appear-ance*(Ma)


112 Murinae113 Gliridae114 Hystricinae115 Ctenodactylidae

‘‘Murinae/Y311 unnamed species’’Gliridae, spp.Hystrix sivalensisSayimys chinjiensis

9.37.37.2†

9.4

10.07.97.3

10.0

10.413.6

8.012.3

11.013.68.0†

13.6

213

48

* Inferred ages of appearance for those taxa appearing before 10.7 Ma are only approximately determined.† Events included in analysis of first and last appearances. See text for details.

faunal and environmental change in the late miocene siwaliks of...

Documents