In the Sub-Himalaya, river incision is rapid, with values up to 10-15 mm/yr. It does not exceed ...... A major landslide, or alternatively a glacial tongue coming from ...
JOURNAL
OF GEOPHYSICAL
RESEARCH,
VOL. 106, NO. Bll, PAGES 26,561-26,591, NOVEMBER
10, 2001
Fluvial incision and tectonic uplift across the Himalayas of central Nepal J. Lavb• and J.P. Avouac Laboratoirede G•ophysique,Commisariath l'EnergieAtomique,Bruy•res-Le-ChJtel, France
Abstract. The patternof fluvial incisionacrossthe Himalayasof centralNepal is estimated fromthe distributionof HoloceneandPleistoceneterracesandfrom the geometryof modem channelsalongmajorriversdrainingacrossthe range.The terracesprovidegoodconstraints on incisionratesacrossthe Himalayanfrontalfolds(Sub-Himalayaor SiwaliksHills) where riversare forcedto cut downinto risinganticlinesandhaveabandonednumerousstrath terraces.Farthernorthandupstream,in the LesserHimalaya,prominentfill terraceswere deposited,probablyduringthe late Pleistocene,andwere subsequently incised.The amount of bedrockincisionbeneaththefill depositsis generallysmall,suggesting a slowrate of fluvial incisionin the LesserHimalaya.The terracerecordis lostin thehigh rangewherethe riversare cuttingsteepgorges.To complementthe terracestudy,fluvial incisionwas also estimatedfrom the modemchannelgeometries usingan estimateof the shearstressexerted by the flowing water at the bottomof the channelas a proxy for river incisionrate. This approachallowsquantificationof the effectof variationsin channelslope,width, and dischargeon the incisionrateof a river;the determination of incisionratesrequiresan additionallithologicalcalibration.The two approaches are shownto yield consiste• results when appliedto the samereachor if incisionprofilesalongnearbyparallelreachesare compared.In the Sub-Himalaya,river incisionis rapid,with valuesup to 10-15 mm/yr. It doesnot exceeda few millimetersper yearin the LesserHimalaya,andrisesabruptlyat the frontof thehigh rangeto reachvaluesof-4-8 mm/yr within a 50-km-widezonethat coincideswith thepositionof thehighestHimalayanpeaks.Sedimentyield derivedfrom the measurement of suspended loadin Himalayanriverssuggests thatfluvial incisiondrives hillslopedenudationof the landscape at the scaleof the wholerange.The observedpatternof erosionis foundto closelymimicuplift aspredictedby a mechanicalmodeltakinginto accounterosionandslip alongthe flat-ramp-flatgeometryof the Main HimalayanThrust fault.The morphologyof the rangereflectsa dynamicequilibriumbetweenpresent-day tectonicsandsurfaceprocesses. The sharprelief togetherwith thehigh uplift ratesin the HigherHimalayareflectsthrustingoverthe midcrustalrampratherthanthe isostaticresponse to reincisionof the TibetanPlateaudrivenby late Cenozoicclimatechange,or late Miocene reactivation
of the Main Central Thrust.
1. Introduction
Souriau,1988;Summerfieldand Hulton, 1994;Hovius,2000]. Mountainbuilding thus leadsto a complexloop with various The kinematicsof mountain building results from the feedbackmechanismslinking crustal deformation,denudacombinationof crustaldeformationand erosion,with the two tion, and climate. The Himalayas are often cited as the processes being possiblycoupled[e.g., Koons, 1989; Molnar modem archetypewhere this kind of coupling may be at and England, 1990; Willet et al., 1993; Avouac and Burov, work. For example,headwarderosionalongthe riverscutting 1996]. This couplingarisesbecausedenudationdependson the edgeof the TibetanPlateauwould have induceduplift of topography,while erosioninfluencestectonicprocessesby the Himalayan peaks through isostaticrebound,enhancing controllingboundaryconditionsat the Earth'ssurfacethrough orographicprecipitationand hence denudation[Molnar and deposition anddenudation. Anotherreasonfor thiscouplingis England, 1990; Burbank, 1992; Masek et al., 1994; that erosiondependson climate,which is itself submittedto Montgomery, 1994]. Such a process would explain the orographicforcing [e.g., Masek et al., 1994]. As a result, positionof the front of the high rangeparadoxicallywell to tectonicuplift and denudationgenerallytend to balanceeach the north of the main active thrust faults, i.e., the frontal other,sothathigh denudationratesare foundto correlatewith thrusts.It would also drive accelerateduplift, independentof zonesof active mountainbuilding [Anhert, 1970; Pinet and the kinematicsof activethrustfaulting,in responseto climate •Nowat Laboratoire de G•odynamique desChaines Alpines, changeduring the Cenozoic [Molnar and England, 1990; Grenoble, France. Burbank, 1992]. The building of the Himalayas would, moreover, have affected global climate by modifying the Copyright2001 by the AmericanGeophysicalUnion. chemistryof the oceanand atmosphere[e.g., Raymoet al., 1988;RaymoandRuddiman,1992].This lattereffectdepends Papernumber2001JB000359. 0148-0227/01/2001JB000359509.00 on the nature of the rocks that are uplifted, eroded, and 26,561
26,562
LAV•, AND AVOUAC:FLUVIAL INCISION ACROSSTHE HIMALAYAS
weathered [e.g.,Derry and FranceLanord,1997].A better understanding of the kinematics of crustaldeformation and the patternof erosionin the Himalayasmight therefore providebasicinsightinto mountainbuildingprocesses. It wouldalsohelpassess the eventualeffectof the Himalayan orogenyon weathering fluxesand the effectof Cenozoic
2. Overview of Himalayan Tectonics and Topography
climatechangeon uplift.
form the most frontal relief just north of the Indo-Gangetic
2.1. Geomorphological and GeologicalSetting
Differentgeographic domainscanbe distinguished across theHimalayas of Nepal(Figures1 and2). TheSiwaliksHills
In thispaper,we document thepatternof denudation in the Plain.They are composedof easilyerodibleNeogenemolasse Himalayasof centralNepal by analyzingfluvial incision accumulatedin the foreland and deformed by thin-skinned (Figure2). Theyformrowsof hills with elevations alongsomeof the major riversdrainingacrossthe range tectonics by narrowelongated (Figure1). We first reviewthe geological andgeomorpho- below 1000 m (Figure lb), separated logicalsettingof the studyareathatencompasses mostthe piggybackbasins(calledDun in Nepali).Justnorthof this foldbelt,thehigherrelief of Mahabarat range Himalayas of Nepal(Figures1 and2). We thenpresent the Sub-Himalayan is much more impressive reaching elevations up to 2500-3000 result of our survey of abandonedterraces.This study mainlyof schistandgneiss complements earlier investigations of abandoned fluvial m (Figurelb). Therangeconsists terracesin the SiwalikHills [Nakata,1972;Delcaillau,1992; intrudedby Late Cambrianto Ordoviciangranitesand LavdandAvouac,2000] andin theLesserHimalaya[Iwataet overlainby Cambrianto Eocene"Thetysian"sediments. al., 1982;Yamanaka andIwata, 1982;Fort, 1993].In thehigh Theseunitsbelongto a crystallinesheetoverlyingthe Lesser rangetheriversgenerallyflow alongsteepN-S gorges where Himalaya(LH). The rocksin the LH consistof low-grade (phyllite,quartzite,andlimestone of Devonian terracescould not be preservedor may even never have metasediments formed,with streamgradientssystematically steeperby a or older ages)forminga large antiformalduplexstructure factor 10 comparedto thosealongtheir upperand lower (Figure2) [e.g., Schelling,1992].North of a line trending reaches.Seeberand Gornitz [1983] previouslynoticedthis about N108øE (white dashed line on Figure lb), the risesabruptlyfromelevations around500-1000m systematic steepening whichtheyinterpreted as an indicator topography of rapidriverincision.In the absence of a terracerecordwe to morethan 6000 m. This breakin slopemarksthe front of followtheirline of thoughtandattemptto determine fluvial the HigherHimalaya(HH). The highestHimalayanpeakslie incisionfrom the geometryof the modemchannels.Using only-•30-60 km northof this line (Figureslb and 2). The unitsof theHH consist mainlyof medium-to highthe tie pointsprovidedby the terracesurveyin the Sub- crystalline plutonsof Mioceneage HimalayaandLesserHimalaya,we determine andcalibrate a gradegneisswith largeleucogranitic simpleempiricalrelationship basedon the estimated fluvial [e.g.,Le Fort, 1986;Searle,1999].TheTibetanPlateau,with shearstressexertedby the flowingwateron the streambed, its Thetysiansedimentarycover, extendsto the north at andwe useit as a proxyfor fluvial incisionalongmountain elevationsaround5000 m (Figure2). The boundariesbetween the different domains roughly rivers.We applythe methodto all the majorriversdraining thecentralHimalayas.We comparefluvialincisionwith mean coincidewith major thrustfaults. The Main CentralThrust the LH fromthe denudationratesas estimatedfrom measuredsuspended load (MCT) is a ductileshearzonethatseparates at gaugingstations alongthe studiedriversin orderto test HH [e.g.,Le Fort, 1986].The Main BoundaryThrust(MBT) whetherthe landscapeis erodedat the samerate as river marksthe limit betweenthe Sub-Himalayaand the LH, and downcutting. In section6 we expandontectonic implications the Main FrontalThrust(MFT) boundsthe southernlimit of in order to discriminate between different models of Cenozoic the Sub-Himalaya, at the frontof the SiwalikHills. All these at depth,that upliftandlandscape evolution anddiscuss thesignificance of thrustfaultsmayconnectto a singledetachment may be called Main Himalayan Thrust (MHT), as suggested thepresentmorphology of therange.
Figure1. (opposite) (a)Geological and(b)geomorphological setting of study area.Geology fromNepalese 1:50,000geological map(courtesy of theDepartment of MinesandGeology of Nepal),$chelling [1992], Brunel[1986],andStdicklin [1980]and,fortheTibetm• part,fromGansser [1964].Faults werereported from Yeats andLillie[1991]. MBT,MainBoundary Thrust; MCT,Maincentral Thrust; MFT,MainFrontal Thrust. Thickblacklinesfollowthe six trans-Himalayan riversfor whichfluvialshearstressis calculated in this study. Downvalleyextent(whitearcuate segment) of areashaped prominently by glacialerosion [Duncan et
al., 1998]definethebeginning of dominant valleyshaping byfluvialincision anddomains of validityof our fluvial incisionmodel. The white oval dotsshowlocationswhere datedHoloceneterracescould be usedto calibrateratesof fluvial incisionfor thismodel.TheN108øEwhitedashedline in Figurelb followsthebreak
in slope in frontof thehighrangewhereintense microseismicity tends to cluster (microseismicity recorded
between 1994and1998,courtesy of theSeismological laboratory, Department of MinesandGeology). This breakin slopelies30 to 60 km southof thehighest Himalayan peaks(whitetriangles). Fromwestto east >8000m highsummits arelabeled byletters: D, Dhaulagiri; A, Annapuma; M,Manaslu; X, Shisha Pangrna; C, ChoOyu;E, Everest;L, Lhotse;M, Makalu;K, Kanchengjunga.
LAVt• AND AVOUAC: FLUVIAL INCISION ACROSS THE HIMALAYAS
Quaternary Siwaliks
26,563
ii::11•:ii'!'i Thetysian series
.... ---•._rivers •
............ •-:•iTibetan slabgneisses
Palaeozoic
metasediments -• Tedlaw Himalayan 0 granites Precambrian
•
metasediments•'•-•' thrust faults
shearstress
profiles
calibration sites
for theshear stressmodel cross sections Figures 2and 4
28ON
'• MFT
A ,,
5o
I
84OE
86OE ',.
•/c'pt4,/) 7•
MFT
Down valley extent of area
..-•"•sh,aped prominently
*•'•by glacial erosion R•vem •'•,• Summits> 800Ore A
SummPts > 7500m
200
800
• 500
3000
i.......................................... LZ.:........ J.I•"-"'-•"•
hypsometw(m)
5000
7000
'\
NEPAL
iNDIA "½
"88oE
26,564
LAVI• AND AVOUAC' FLUVIAL 1NCISION ACROSS THE HIMALAYAS
Indo-GangeticSub-
Plain
Lesser Himalaya
Higher Himalaya I
Himalaya
'•!•' ...•: ......•'?•'"?
Tibet
I
"*½ •'¾•' '•'•.•..:, .......
?8000 ,•• .•.•
;
Mahabarat Kathmandu ..•, '•*"•: :J
hills
•*
......
.......... '•"
ß ....
• •o
'........
"•¾"':7""•¾'•:: - •?•:'-•:•.--•:F:'2 •1.5mmlyr•:.•::: .........
+ß
•-6•0
3ooo
ß ,'•:' '.• .......... :........ :•, ... basin .:•.•::•. ,•:/::.J
D/
•-9ooo
• •o•
?.:•:?•;;;::;
+
+
+
+
Upper siwaliks '• Middle siwaliksSubLowersiwaliks
metasediments
Indianuppercrust
Lesser Himalaya
Himalaya
•'•::q series ....... Thetysian
L_•
Higher Himalaya
Leucogranite
F-•-•I crust LPJ__,Lower
100 km
(a) Figure 2. N18øE (a) geologicalsectionand (b) topographyacrossthe Himalayasat the longitudeof Kathmandumodified from Brunel [1986]. FollowingSchelling[1992], the duplex structurein the Lesser Himalayawas derivedfrom the antiformdefinedby foliationplanes.The geometryof the Main Himalayan Thrust(MHT) was derivedfrom this assumption and from balancingvarioussectionsin the Sub-Himalaya [SchellingandArita, 1991;LavdandAvouac,2000]. The rampbeneaththe high rangeconnectswith a flat thatroughlycoincideswith the midcrustalreflectorobservedby INDEPTH1 experiment,300 km eastof the section[Zhaoet al., 1993]. The inferredgeometryof the Moho approximately followsthe elasticdeflection thatmatchesgravitydataat the longitudeof MountEverest[Lyon-CaenandMolnar, 1983].The linesin the Figure2a showaveragetopography (solidline) andlowestandhighestelevations (graydashedlines)withina 50-km-wideswathalongprofileAA' in Figure1a (verticalexaggeration 8:1). In its frontalpart,the MHT has beenslippingat 21 + 1.5 mrn/yrduringtheHoloceneaccording to terraces warpingacrossthe Sub-Himalaya [Lav• andAvouac,2000].
from structural observations[Schelling and Arita, 1991;
Schelling, 1992] and seismologicaldata [Seebet and Armbuster, 1981; Pandey et al., 1995]. The midcrustal reflectorimagedin a deepseismiccrosssectionbeneathsouth Tibet would correspondto the northwardextensionof the MHT [e.g.,Zhaoet al., 1993](Figure2). 2.2. Active Tectonics in the Himalayas of Central Nepal and lts Relation to Present Topography Evidence
for recent deformation
was found at various
localitiesalongthe MFT, within the Sub-Himalaya,alongthe MBT or in the LH [Nakata, 1989; Mugnier et al., 1994]. Recentinvestigations of activetectonicsacrossthe Lesserand Sub-Himalayasouthof Kathmandu [LavdandAvouac,2000] show that the MFT is clearly the major active fault there. Indeed, slip along the MFT at 21 + 1.5 mm/yr [Lav• and Avouac,2000] absorbsnearly all the presentshorteningrate acrossthe whole Himalayan range as estimatedfrom GPS measurements[Bilham et al., 1997; Larson et al., 1999; Jouanneet al., 1999]. Paradoxically,the highestHimalayan
peakslie parallelto the MFT but 100 to 150 km fartherto the north and 20 to 50 km north of the trace of the MCT (Figures 1 and 2). The reasonfor this paradoxremainsunclear,and severalexplanations havebeenproposed. Active thrustingin the upper crust along or closeto the MCT mightexplainthe presentmorphologyof the frontof the HH. This was argued,for example,by Seebetand Gornitz [1983] on the basisof the observationthat seismicactivity andriver knickpointsalsotendto follow the front of the high range.More recently,Bilhamet al. [1997] haveobservedthat the front of the high range also coincideswith a zone of interseismic uplift. They proposed that part of this interseismicuplift might be unrecoverableand contributeto buildingthe high rangein the long term by pervasivethrust faultingin the uppercrustaroundthe MCT. Anothersolution is that the MHT steepensand makesa ramp beneaththe high range, as suggestedby the antiformal structureof the LH [Schellingand Arita, 1991] (Figure 2). The front of the high range would then be maintained by overthrustingover a midcrustalramp. Accordingly,the total shorteningacrossthe rangewould equalthat accommodated by slip alongthe MFT
LAVI2AND AVOUAC' FLUVIAL INCISIONACROSSTHE HIMALAYAS andthe patternof activeuplift wouldbe primarilycontrolled by the geometryof the MHT at depth. If we now considerthatthepatternof erosionmightnot be in equilibriumwith activerock uplift, otherexplanations are possible.Harrison e! al. [1997] found evidence for late Miocenereactivationof the MCT and suggestthat the front the highrangewould,in fact,reflectthisreactivation.Finally, as already mentioned,it has also been proposedthat the presenttopographycould reflect the isostaticresponseto enhancederosionof the edgeof the TibetanPlateauby a Late Cenozoicmonsoonstrengtheningor by Quaternaryglaciations [Burbank, 1992; Masek et al., 1994; Montgomery, 1994]. In that case,the patternof uplift acrossthe Himalayas would not be relatedcloselyto the kinematicsof activethrust faulting. The determination of fluvial incisionalongthe majorrivers drainingacrossthe Himalayasand its relationto landscape denudationand to active tectonicsmight thus be used to discriminate
between
these different
models
of mountain
buildingandlateCenozoicupliftof thehighrange. 1.3. Characteristicsof the Studied Fluvial Systems
The Narayaniand SaptKosi watershedsdrain most of the Himalayas of central Nepal. They include several rivers cuttingacrossthe HH with headwaterson the Tibetan Plateau
(Figure lb). The base level of erosionis controlledby sedimentation and subsidence in the foreland.Upstreamof Tribenighat(Figure lb), whereit entersthe Gangabasin,the
Narayani drains an areaof 38,000km2. TheSapt_ Kosi watershed covers a muchlargerareaof 60,000kmz. The differenceis essentiallydue to the fact that the Arun River, which belongsto the Sapt Kosi watershed,drainsan area of
morethan20,000 km2ontheTibetan Plateau. All the rivers in the study area are cutting down into bedrockexceptlocally in the Sub-Himalaya,acrossthe Dun. We have studiedsix trans-Himalayanrivers within thesetwo watershedsthat are, from west to east, the Kali Gandaki, the MarsyandiRiver, the Buri Gandaki,the Trisuli River, the Sun
Kosi, and the Arun River (Figure 1). In the LH the rivers are characterizedby relatively even stream gradient and convolutedcourseswith large east-westtrendingdeflections (Figure 1). Acrossthe HH, all the rivers cut straightalong narrow and steep N-S gorges. Three of them, the Kali
26,565
terracespermit a precise quantificationof fluvial incision ratesandprovidethe basisfor a relativedatingof the terraces in the LH and HH.
3.1. Methodologyfor Estimating Long-Term Fluvial
Incision
Profiles
The simplestandmostdirectmethodfor estimating river incision is to date and measure the elevation of former river
bedswith respectto the presentriver bed.Fortunately, fluvial terracesare ubiquitousin the Sub-HimalayaandLH of central Nepal. FollowingBull [1991], we distinguishstrathand fill terraces.Strath terraces are characterizedby thin fluvial gravel ( red
weathering profile> 3 m P12
PI•
30-50
> 45
3 m > red
weathering profile> 5 m red weathering profile> 5 m
M5 M4 M3 M2 M•
Ft3
G2 Ft2
H
H
H
Ft2,
HH
Fh
(60>age> 35 kyr)
aChronology of the Sub-Himalayan Holoceneterraceshasbeenestablished from charcoaldating(parentheses); chronology of the SubHimalayan Pleistocene terraces wasderivedby comparing theirwarpingwith thatof well-datedHolocene terraces assuming constant uplift rate;chronology of theterraces alongriversdrainingtheLH wasinferredby comparing weathering profileswiththoseof theSub-Himalayan riversandsomesparsedating(parentheses, seetext for references).
26,568
LAVI• AND AVOUAC'FLUVIAL1NCISIONACROSSTHEHIMALAYAS 200
terrace top
Bagmati valley
strath terrace
charcoal 150
--
--
100
0
2
4
6
8
10
12
14
16
18
2O
22
N15E projecteddistancefrom the MFT (km)
Figure 4. (a) Elevationabovepresentriver bed, alonga N15øEprofile (profileBB' in Figure l a), of the strathsurfacesrecognizedalongthe BagrnatiRiver (Holoceneterracesin bold lines, undatedPleistocene
terraces in dashed line).Thesiteswheregoodchronological control couldbeobtained from'4Cdating of charcoalfragments arealsoreported[LavJandAvouac,2000].(b) Theoreticaluplift profile(dash-dotted line in Figure4a) derivedfrom the structuralsectionassuming fault bendfoldingat the MFT [LavdandAvouac, 2000]. !
basisfor an attemptat correlation(Plate 3a). River incision into the bedrockbeneaththe fill terracesappearsto increase upstream, suggesting a largeramountof fluvialincisionat the front of the HH where the river gradientrisesabruptly(Plate 2a, lower dashedorangeline in Plate 3a). The relatively abruptbreakin slopeof the terracetreadat km 65 (Plate3a) mightbe takento reflect somewarpingas well, but sucha conclusionis highly speculativeand dependent on the initial geometryof the terracetread.More generally,because of the difficultyin assessing agesand amountsof fluvial incision andin correlating the variousterracetreads,theterracerecord in theLH thusonlyprovidesqualitativeinformation. 3.3.2. Fluvial terraces along the Trisuli. In its lower reach the Narayani River cuts acrossthe Sub-Himalayan rangealongthe MFT just north of Tribenighat(Figure lb). There, somestrathterracessimilarto thosealongthe Bagrnati and Bakeya Rivers were identified and investigatedin the field. Two terraceswere dated from charcoalsamples:the highestone is early Holoceneat 9.2 cal. kyr B.P. and the secondcorresponds to a meanderabandoned at 2.7 cal. kyr
B.P. [Lavd, 1997]. Both samplesindicateconsistent incision rates of 6-7 mrn/yr. As along the BagrnatiRiver, fluvial incisionhasbeenforcedby activethrustingat the MFT. MDT activityhasalsobeenobserved from a terracestudyby Iwata and Nakata [1986] north of Naryanghatat the confluence betweenKali andTrisuli (Plates2b, 3b, and4). Accordingto our observationon the weatheringprofile of theseterraces, Iwata and Nakata's[1986] terracelevel V may haveformed duringthe earlyHolocene.Strathelevation(-25 m) of this terrace level would indicate an incision rate around 2-3
mrn/yr. In the LH thepatternis similarto thatalongtheArun River
with majorPleistocene fill terraceat confluence basins(Plate 4). Two prominentPleistocene treadscan be distinguished both with >120 m fill material of dominantlyfluvial origin. Numerous minor Pleistocene and Holocene fill-cut terraces
and lateral fans form generallyunpairedand disconnected remnantsthat cannot be correlatedeasily. The prominent terraceFh, well exposedat the confluencewith the Tadi Khola near Trisuli Bazaar (Plate 5a), displaysa 10-m-thick
LAVfi AND AVOUAC: FLUVIAL INCISION ACROSSTHE HIMALAYAS
i'
Burnlingh•t
W
E
ß
.
ß
o.
.
ß
ß
26,569
weatheringprofile, indicativeof an age probablyolder than 45 kyr (Table 1). At this localitythe Tadi Kholahasincised by
