JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 104,NO. B1, PAGES 1077-1093,JANUARY 10, 1999
Kinematics
from
GPS
of the
India-Eurasia
collision
zone
measurements
KristineM. Larson,
•'2 RolandBfirgmann,
• RogerBilham,4
and JeffreyT. Freymueller
•
Abstract. We use geodetic techniquesto study the India-Eurasia collision zone.
Six years of GPS data constrain maximum surface contraction rates across the
Nepal Himalaya to 18 4-2 mm/yr at 12øN 4-13ø (1or). Thesesurfacerates across
the 150-km-wide deforming zone are well fitted with a dislocationmodel of a buried
north dipping detachment fault striking 105ø, which aseismicallyslips at a rate of
20 :k i mm/yr, our preferredestimatefor the India-to-southern-Tibetconvergence
rate. This is in good agreementwith variousgeologicpredictionsof 18 4- 7 mm/yr
for the Himalaya. A better fit can be achievedwith a two-fault model, where
the western and eastern faults strike 112ø and 101ø, respectively,in approximate
parallelism with the Himalayan arc and a seismicity lineament. We find eastward
directedextensionof 11 4-3 mm/yr betweennorthwesternNepal Lhasa, alsoin
good agreement with geologicand seismicstudies acrossthe southern Tibetan
plateau. ContinuousGPS sites are used to further constrainthe style and rates of
deformation throughout the collisionzone. Sites in India, Uzbekistan, and Russia
agree within error with plate model prediction.
1.
Introduction
collisionis accommodatedby strike-slipfaulting in Asia
[Molnarand Tapponnier,
1975],with corresponding
east-
While the far-field plate motions across the Indiaward displacement or extrusion of Tibet and southern
Eurasia collisionzone are relatively well understood,the
China. Convergencealso clearly leads to crustal thickplate boundary encompassesa large area, the details
ening evidencedin the •5 km average elevation of the
of the deformation field are complicated, and geologic
Tibetan plateau and a crustal thicknessof •70 km. The
rates are poorly known (see Figure 1). Global plate
relative importance of these two mechanisms, crustal
motionmodels[DeMetset al., 1990;1994]predictthat
thickening and lateral extrusion, is still a question of
approximately50 mm/yr of northwarddirectedconveropen debate. Estimates of convergenceaccommodated
genceis taken up betweenIndia and Eurasia. Geologic
by extrusionaltectonicsrangefrom 15% [Englandand
Molnar,1997a],to 10-25% [Molnaret al., 1987],andup
convergenceis expressedas shortening across the Hito
50% [Avouacand Tapponnier,1993].Unfortunately,
malaya [Molnar and Deng, 1984; Armijo et al., 1986;
geologicmeasurementsacrossthe major strike-slip and
Molnar and Lyon-Caen,1989]. This meansthat nearly
and seismicevidencesuggests
that 18+7 mm/yr of the
two thirds of the convergenceis accommodatedelsewhere.
Two competingmechanismscan describethis accommodation.
It is clear that
some of the India-Eurasia
normal faults in the region, which would bound lateral
extrusion,are fairly uncertain[e.g.,Molnar et al., 1987;
A vouac and Tapponnier, 1993; England and Molnar,
1997a]. Resolvingthe crustalthickeningversuslateral
extrusiondebate has important consequences
for geody-
namic modelsof the region[e.g., Englandand House•Departmentof Aerospace
EngineeringSciences,
Univer- man, 1986; Avouac and Tapponnier, 1993; Houseman
sity of Colorado, Boulder.
and England, 1993; England and Molnar, 1997b; Roy2Nowat Departmentof Geophysics,
StanfordUniversity, den et al., 1997].
Stanford, California.
Many of these debates would be resolved if precise
aDepartmentof Geologyand Geophysics,
Universityof
measurements
of surface deformation throughout the
California, Berkeley.
plate
boundary
region were available. Currently avail4Departmentof GeologicalSciences
and CooperativeInstitute for Researchin EnvironmentalSciences,University able kinematic descriptions of the deformation field
of Colorado, Boulder.
have been based on either seismic or geologic data
5Geophysical
Institute, Universityof Alaska,Fairbanks. [Peltzer and Saucier, 1996; Avouac and Tapponnier,
1993; England and Molnar, 1997a; Holt et al., 1995].
Copyright 1999 by the American GeophysicalUnion.
While seismicand geologicdata directly relate to deformation at the Earth's surface,both are hampered by
large uncertainties. Furthermore, seismicobservations
Paper number 1998JB900043.
0148-0227
/ 99/ 1998JB
900043509.00
1077
1078
LARSON
70ø
/
ET AL.: INDIA-EURASIA
80ø
1
'
90ø
Talas-Ferghana
-- ..• •
13+_7
KINEMATICS
100ø
.•,•
--,.,..
•fau_lt-• __,--,.---,--,.....•••_'"'
•..•..-19.'---5
.,.__,..
•
•3 -.,. ---,--- TienShah_Q___ . •
FROM
GPS
110ø
-'"-
--,---....• .-,-40 ø
," Pamir
32
+8,,• .• 5+_3
x,
,
%
%?:%7' TM
30 ø
INDIA
80 ø
90 ø
100 ø
Figure 1. Tectonic setting and fault map of the India-Eurasiacollisionzone. Also shownare
major collisional
beltsand fault zoneswith estimates
of strike-slipor convergence
rates(all in
mm/yr). Theseratesandtheirstandard
deviations
aredetermined
froma varietyofgeologic
data
suchasoffsetlithologic,structural,or morphologic
unitsandcross-section
restorations
that span
variabletime periods.Geologic
evidence
suggests
that 18-+-7
mm/yr of arc-perpendicular
shorteningis accommodated
acrossthe Himalayaand 10-+-5mm/yr of east-westextensionis accommodatedacrosssouthernTibet. Sourcereferences
are as follows:Tien Shan,Molnar andDeng
[1984];Talas-Ferghana,
Burrmanet al. [1996];Himalaya,Molnar[1987]andLav• andAvouac
(submitted
manuscript,
1998);NorthernTibetgrabens,
England
andMolnar[1997b];
Southern
Tibet grabens,Armijo et al. [1986];Karakorum,
Avouacand Tapponnier
[1993];KarakorumJiali,Arrnijoet al. [1989];AltynTagh,AvouacandTapponnier
[1993]andPeltzeret al. [1989];
North Pamir, Haiyuan,Nan Shan,and Kunlun,EnglandandMolnar[1997b];LongmenShan,
Burchfielet al. [1996].
span a time too short to be representative
of the regional kinematics, whereasgeologicdata may include
featuresthat are not representative
of the currentlyactive system. Geodynamicmodelsthat rely on these
less preciseconstraintsare limited in their ability to
correctly interpret the kinematics of crustal deformation in the region. Geodesy provides the most accu-
rate meansof measuringsurfacedeformation,but the
observedrates are valid only duringthe shortmeasurement period. That is, nearplate boundaryfaults(such
as the southernHimalayathrust system)we commonly
capture the interseismicdeformationstagein the local
earthquakecycle,whereasaway from plate boundaries,
geodeticdata often agreewell with long-termgeologic
displacement rates.
modation of interseismicstrain where the Indian plate
underthrusts the southern Tibetan plateau. Data from
the Nepal Himalaya network together with seven regional continuousGPS stationsquantify the magnitude
of shorteningthat occursnorth of the Himalaya and reveal eastward extrusion of southernTibet. Finally, consideration of all currently available geodeticconstraints
on the regional kinematics allows us to compare and
contrastactivesurfacedeformationwith previouslyproposedmodels of the geodynamicsof the collisionzone.
The Nepal Himalaya embracethe central third of the
Himalayan arc, including the rupture zonesof the 1934
Bihar earthquake(Mw =8.1) andthe 1833Nepalearthquake(M--7.7) [Khattri, 1987; Bilham, 1995]. The
current slip potential of western Nepal and the Kumaon
We use Global PositioningSystem(GPS) measure- Himalaya(to the westof Nepal)is believedto be 6-15 m
ments to addressthe geodynamicsof the India-Eurasia
based on measuredconvergencerates and the absence
of great earthquakesfor at least the past 300 years of
along the Nepal Himalaya revealsdetailsof the accom- colonial history, which would be sufficient to drive a
collision at two scales.
A dense network across and
LARSON
ET AL.: INDIA-EURASIA
KINEMATICS
FROM
GPS
1079
magnitude-/•w 8.1-8.7 earthquake[Bilhamet al., 1995; surements lasting from 3 to 5 days. Thirty sites in
Khattri, 1987]. Geodynamically,
theseobservations
are the plate boundary region are included in our analyimportant becausethey offer a view of convergentplate sis. Nearly all the Nepalesesites were first observedin
processesboth in the hanging wall and in the footwall
of a megathrust, a view that is obscuredby oceansfor
most convergentplate boundaries.
Our current work buildson previousanalysesof geodetic data from the Nepal Himalaya. Studies of leveling data from eastern Nepal reveal a peak uplift rate
1991. Measurements were first made at Bangalore in
1991 (in a noncontinuous
mode);all the other far-field
stations began continuousobservationsabout 1995, as
did Lhasa. Locations of the sites in the Nepal Himalaya
and Tibet
network
are shown in Plate
I on a shaded
relief map of the topography. Our sites span a considof 5-8 mm/yr which can be interpretedas interseismic erable range in ellipsoidalheights,from lessthan 50 m
strain accumulation associatedwith slip deceleration at
a crustalramp [Jacksonand Bilham, 1994; Pandeyet
al., 1995]. Freymuelleret al. [1996]took a largerperspective and analyzed GPS data from one site in India
and one site in Nepal to confirmglobal predictionsof Indian plate motion independent of the Australian plate
[DeMetset al., 1990]. Bilham et al. [1997]presented
an analysis and interpretation of GPS data collectedin
the Nepal Himalaya over a 4.5 year period. They found
in the foothills of the Himalaya to more than 7800 m on
the South Col of Mount Everest. To provide additional
tectonic backgroundfor the network, Plate 1 also includesmaximumprincipalstressorientations[Zoback,
1992]andselected
focalmechanisms
[MolnarandLyonCaen, 1989].
The locations of all the sites used in this study are
listed in Table 1, along with the number of observations at each site and their temporal span. The first
a 17+2 mm/yr northerly directedsurfaceconvergence
GPS observationsof the Nepal Himalaya network were
rate between India and southernmost Tibet. Eighty
made
in April 1991. At this time the GPS constellapercent of this convergenceis concentrated in a zone
tion
consisted
of only 15 satellites,ambiguity resolu:
less than 120 km wide.
tion
was
difficult,
and the global GPS tracking network
Bilham et al. [1997]alsouseda 20 year time series
was
significantly
smaller
than the oneavailablein 1997.
of leveling data acrossthe peak uplift zone. The center
and maximum gradient of the horizontal velocity field The impact on the quality of geodeticresultsin the Hiis associatedwith the region of rapid uplift. The nar- malaya is straightforward:the precisionof each day's
row width of the horizontal and vertical velocity fields measurements in the east and vertical components is
is consistent with a two- dimensional
dislocation
model
markedly poorer than what we observefor data colin which the Indian plate is locked to the base of the Hilected in later years. For example, the standard deviamalaya down to depths of 15+5 km, north of which the tions for 5 daysof measurementsmade in 1991 were 2, 8,
Indian plate slides freely beneath the Tibetan plateau and 20 mm in the north, east, and vertical components
at a rate of 20+3 mm/yr.
for the 70 km baseline between JIRI and NAGA. In 1995
In this paper we use more recently collectedGPS data the valueswere 2, 2, and 10 mm, respectively.The 1991
to improve relative velocity estimates in the Nepal Hiresultswere negativelyaffectedby the receiversavailmalaya and have extended this local network with sites able at the time, primarily C/A codeL1/squaring L2
that were not previously available. One new site in receivers(Trimble SST). The WM-102 receiversused
particular, Lhasa, allows us to measure east-west ex- at TING and RONG observedfor shorter time periods
tension rates across southern Tibet.
We use the GPS
with more frequentphasebreaksthan thoseof the Trimand leveling observationsin the Himalaya to invert for
ble SSTs, resulting in east componentsfor TING and
the geometry and slip rates of three-dimensionaldislo- RONG much weaker than those for the other sites. We
cation models representinginterseismicstrain accumu- could not resolvecarrier phaseambiguitiesusing either
lation. In particular, we focus on implications of sig- the Trimble SST or WM-102 data. In 1992 a subset of
nificant along-arc variations in the present strain field.
six sites was observedfor 2 days. A secondmajor exWe have simultaneouslyanalyzed data for the local netperiment wasconductedin November1995. Thesemeawork with data from continuouslyoperating GPS sites
in Asia and have defined
all velocities
in a consistent
reference frame. This approach allows us to describe
surements were conducted in collaboration with Project
Idylhim [Bilhamet al., 1997].Carrierphaseambiguities
were easily resolvedfor the 1995 Nepalesedata.
Severalof the original Nepal Himalaya siteswere not
a much broader scale. Finally, we include published
observed
in 1995, including PHER, NAMC, KHAN,
present-daygeodeticresultsfrom Asia in order to proLUKL,
SIMA,
and MAHE. We were able to reoccupy
vide the reader with a more comprehensive description
most
of
these
sites
over the next year. In fall 1996
of the plate boundary deformation.
and spring 1997 we made additional measurementsat
2. GPS Network
and Observation
NAGA, SIMI, BIRA, and NEPA. We have also inthe kinematics
of the India-Eurasia
collision
zone over
History
The GPS data used in this study were collectedbetween 1991 and 1997, mostly in campaign style mea-
cluded measurements
collected in 1995 and 1997 from
the South Col (SCOL) of Mount Everest by another
group of scientists. The analysis thus includes GPS
data from January 1991 through September1997.
1080
LARSON
ET AL.- INDIA-EURASIA
KINEMATICS
FROM
..f
o
oß
I"-
03
(3o
ß
LO
o•
(3o
ß
•'
(.o
o•.
03
o
(3o
ß
Od
(.o
(3o
,.
•-'
•o
03,.
0
o
0
GPS
LARSON ET AL.: INDIA-EURASIA
KINEMATICS
FROM GPS
1081
Table 1. GPS Sites Used in This Study
Observing Sessions
No.
Name
Identifier
Latitude
deg
Longitude
deg
1
Airport
AIRP
27.69715
85.35791
2
Bhairhawa
BHAI
27.50734
83.41802
3
4
5
Bharatpur
Biratnagar
Bangalore
BHAR
BIRA
IISC
27.67783
26.48396
13.02116
84.42958
87.26440
77.57036
6
Irktusk
IRKT
52.21902
7
Janakpur
JANK
26.71064
85.92452
8
9
10
11
12
13
14
Jiri
Jomoson
Khadbari
Kitab
Lhasa
Lukla
Mahendre.
JIRI
JOMO
KHAN
KIT3
LHAS
LUKL
MAHE
27.63541
28.78069
27.37987
39.13476
29.65733
27.68333
28.96319
86.23035
83.71787
87.20560
66.88544
91.10398
86.72500
80.14795
15
Nagarkot
NAGA
27.69271
16
Namche
NAMC
27.80262
17
Nepalganj
NEPA
28.13409
81.57467
18
19
20
Pheriche
Pokhara
Bishkek
PHER
POKH
POL2
27.89373
28.19895
42.67976
86.82282
83.97768
74.69426
21
22
Ranj
Rongbok
RANJ
RONG
28.06254
28.19370
23
South
SCOL
27.97294
24
Shanghai
SHAO
31.09964
25
26
27
Simara
Simikot
Surkhet
SIMA
SIMI
SURK
27.16245
29.96698
28.58576
Col
104.31623
Height
m
First
Last
1283.848
1991.23
1995.87
1991.25
1995.87
1991.25
1991.23
1991.08
41.430
149.762
13.624
842.198
502.449
Epochs
10
2
8
3
1995.87
1996.97
1997.73
11
15
95
2
5
37
33
12
1995.75
1997.73
1991.23
1995.89
1878.205
2806.276
1088.499
622.599
3624.677
2768.700
161.000
1991.23
1991.23
1991.25
1994.80
1995.38
1991.24
1991.27
85.52121
2105.150
86.71514
3522.882
8
2
1995.89
1995.90
1996.29
1997.73
1997.73
1996.95
1996.79
15
28
11
82
89
4
7
4
5
2
41
30
2
3
1991.10
1997.28
95
30
1991.25
1996.31
12
4
1991.23
1996.86
16
6
4361.658
767.763
1714.199
1991.27
1991.27
1995.40
1997.28
1995.87
1997.73
7
10
43
2
3
19
82.57303
86.82739
1654.130
4953.110
1991.27
1991.25
1995.88
1995.86
8
5
2
2
86.92943
7861.972
1995.36
1997.39
3
2
1995.06
1997.39
86
38
1991.25
1991.25
1991.27
1996.97
1996.86
1995.89
10
13
14
4
4
4
215
97
10
4
9
2
121.20044
84.98172
81.82646
81.63519
121.53654
7.249
Total
89.214
22.142
68.433
2908.145
626.803
28
Taipei
TAIW
25.02133
1992.37
1997.73
29
Tansen
TANS
27.87384
83.55380
1391.444
43.945
1991.25
1995.90
30
Tingri
TING
28.62950
87.15521
4256.033
1991.25
1995.86
Latitudes, longitudes,and heights are referencedto the WGS84 ellipsoid. Each observingsessionconsistsof one 24-hour
period. Epochs are defined as the number of distinct 2-week periods during which at least one observingsessionoccurred.
3. GPS Data Analysis
The GPS data are analyzed with the use of the
GIPSY software developedat the Jet Propulsion Labo-
agreewell with the resultsof Larsonet al. [1997].The
scale, translation, and rotation of the epoch positions
and velocitiesare constrainedto agree with ITRF94 in
a weightedleast squaressense[Boucheret al., 1996].
ratory [Lichtenand Border,1987]. We usean analysis
The resultinghorizontaland vertical velocitiesof each
strategy discussedfully by Larson et al. [1997]. In
site in the region are listed in Table 2.
short, regional observationsfrom each 24-hour period
are analyzed simultaneouslywith data from the GPS
of the Collision
Zone
global tracking network. In addition to site coordinates, 4. Kinematics
we explicitly estimate initial conditionsfor each satel4.1. Regional Plate Motions and Plate
lite, satellite and receiver clocks, solar radiation presBoundary Deformation Zone
sure parameters, a tropospheric zenith delay parameter
To address the far-field plate motions in the region,
for each site, and carrier phaseambiguities. Station coordinates and covariances from each 24-hour solution
we comparethe GPS-derived velocitieswith global plate
are then used in a weightedleast squaresfit, where the model predictionsof the No-Net-Rotation NUVEL1-A
motion of each site is allowed to vary linearly. We use model, hereafter NNR-A [DeMets et al., 1990; 1994;
51 sites in our global velocity determination but dis- Argusand Gordon,1991]. NNR-A is a compilationof
cussonly those in the Eurasia-India collision zone. The angular velocity vectorsfor 14 tectonic plates, includvelocities of sites in western Europe, South America, ing Eurasia and India. As these plate motion rates are
the Pacific, Antarctica, Australia, and North America constrained only by seafloor spreading data, there are
1082
Table
Site
LARSON ET AL.: INDIA-EURASIA KINEMATICS FROM GPS
2. Station
Velocities
GPS
India NNR-A
Eurasia NNR-A
Observed
Prediction
Prediction
North
East
AIRP
BHAI
BHAR
BIRA
IISC
IRKT
JANK
JIRI
JOMO
KHAN
KIT3
LHAS
LUKL
MAHE
NAMC
NAGA
NEPA
PHER
POKH
POL2
RANJ
RONG
34.5 + 1.5
37.6 4- 1.6
37.5 4- 1.6
38.8 4- 1.3
39.8 4- 1.2
-10.2 4- 3.1
38.2 4- 1.6
29.9 4. 1.5
27.0 4- 1.5
32.8 + 1.5
2.6 + 2.0
16.6 4. 2.5
28.3 4- 1.4
36.6 4- 1.4
28.6 4- 1.5
35.2 4- 1.2
36.8 4- 1.3
27.1 4- 1.3
36.4 4- 1.6
3.7 4- 2.6
35.5 4- 1.6
22.1 4- 2.6
38.1 + 2.2
38.8 4- 2.3
36.4 4- 2.3
40.1 4- 1.9
42.2 4- 1.8
26.3 4- 3.7
39.1 + 2.6
35.9 + 2.2
35.8 4- 1.9
40.6 + 2.2
28.5 4. 2.3
44.2 4- 2.8
37.8 + 2.3
33.2 4- 2.1
39.7 4- 2.1
37.7 4- 1.6
35.1 4- 2.0
37.3 4- 2.1
36.9 4- 2.3
28.9 4. 3.0
32.3 4- 2.3
36.4 4- 5.4
SCOL
SHAO
SIMA
SIMI
SURK
TAIW
TANS
TING
28.84- 3.6
-15.0 4- 2.2
38.34. 1.4
24.2 4- 1.3
35.8 4. 1.6
-13.6 4- 1.3
36.0 4- 1.5
39.7 4- 3.8
31.8 q-2.5
35.74. 2.1
32.2 4- 2.0
34.1 4- 2.3
33.7 4- 1.4
35.4 4- 2.2
21.4 4- 2.2
36.5 + 4.2
Up
-5.2 + 5.9
-3.3 4- 5.9
2.0 4- 5.9
-4.4 4- 4.6
7.2 4- 3.0
4.6 4- 5.7
2.2 4. 6.6
2.4 4. 6.3
5.3 4. 4.1
6.6 4- 5.8
0.7 4. 3.1
3.2 4- 3.9
1.9 4. 6.9
-2.3 4- 5.3
2.4 4- 5.5
1.1 4- 2.4
-2.6 4- 4.6
7.6 4- 5.1
-7.0 4- 5.7
9.3 4- 4.4
3.4 4- 6.1
-18.6 4- 13.7
2.5 4- 17.9
-2.5 4. 3.2
0.1 4- 5.7
9.0 4. 4.2
1.5 4- 5.7
-6.9 4. 1.7
11.2 4- 5.6
-16.9 4- 8.7
North
East
North
42.3
42.2
42.3
42.4
41.5
41.2
42.4
42.4
42.2
42.4
38.9
42.5
42.4
41.8
42.4
42.3
42.0
42.4
42.2
40.9
42.1
42.4
42.5
36.5
42.3
42.0
42.0
36.4
42.2
42.4
36.6
36.1
36.3
37.7
40.1
34.6
37.2
37.0
35.6
37.4
23.0
37.9
37.1
34.3
37.1
36.7
35.1
37.1
36.0
24.1
35.5
37.0
37.1
48.3
36.7
34.4
35.0
48.5
35.9
36.9
-5.0
-4.5
-4.8
-5.5
-2.9
-9.8
-5.2
-5.3
-4.6
-5.5
0.2
-6.6
-5.4
-3.6
-5.4
-5.1
-4.0
-5.4
-4.6
-2.0
-4.3
-5.4
-5.5
-13.3
-4.9
-4.0
-4.0
-13.3
-4.5
-5.5
East
25.1
25.1
25.1
24.9
23.2
22.7
25.0
25.0
25.2
25.0
26.0
24.9
25.0
25.3
25.0
25.1
25.2
25.0
25.2
25.8
25.2
25.1
25.0
22.2
25.0
25.4
25.3
22.3
25.2
25.1
All velocities
arein mm/yr. Uncertainties
are1 standard
deviation.
NNR-A,No-Net-Rotation
NUVEL1-Amodel.
no direct data measuringthe convergence
rate between
Eurasia
and India.
Shanghai'svelocity with respectto Eurasia is 104-3
mm/yr towardN104øE.Thisisin goodagreement
with
In Table 2 we list the predicted NNR-A Eurasian and two independentanalysesof VLBI data from Shanghai:
Indian plate velocity for each site adjacent to the ob- Heki [1996]reportedll4-1 mm/yr towardNl12øE in
servedvelocities. Note that NNR-A predictsno vertical a Eurasiafixed frame; Molnar and Gipson[1996]estidisplacements.Dependingon one's preferredvantage, mated 84-1 mm/yr towardN116øE,alsoin a Eurasia
station velocities can be depicted in an "India fixed" fixed frame. Taipei's velocity is rotated slightly with
frame or a "Eurasia fixed" frame by subtractingthe pre- respectto Shanghai,114-2mm/yr eastwith respectto
diction for that plate. In Figure 2 we plot selectedresid- Eurasia. POL2, in the Tien Shan belt, has a velocity
ual vectorsto the Eurasian and Indian frames. Banga- of 74-3mm/yr towardN28øEwith respectto Eurasia.
lore (IISC) and four sitesin southernNepalagreewithin
3 mm/yr with Indian plate motion predictions.Velocities of these sites relative to Eurasia range from 404-2
mm/yr toward NlløE at MAHE to 464-2mm/yr to-
If we take into accountthe large uncertaintiesin di-
ward N20 øE at BIRA.
between
rection(4-30ø),this findingis in goodagreement
with
an analysisby Abdrakmatov
et al. [1996](Figurelb),
whichreported54-3mm/yr northwarddirectedmotion
POL2
and Eurasia.
We canalsocompareour analysisof data from LHAS
The remaining sites shown,including those in Kitab
(KIT3), Taipei (TAIW), Shanghai (SHAO), Lhasa with the recentwork of Karo et al. [1998]. They re(LHAS), and Bishkek(POL2), are shownwith respect port a Eurasiafixed velocityof 314-3mm/yr toward
to Eurasia. KIT3, IRKT, and POL2 do not showsignif- N53øE4-4ø, which agreeswell with our rate of 304-3
icant deformationwith respectto Eurasia at the 95% mm/yr toward N40øE4-5ø. The estimateddirections
confidencelevel, but their uncertainties are still quite disagreeby 13ø, but this variation is well within the
large becausethe data span a short period of time. 95% confidenceregion for the two estimates. Figure
LARSON ET AL.: INDIA-EURASIA
KINEMATICS
FROM GPS
1083
./
50 ø
Eurasia-fi
40 ø
kit3
•)
30 ø
NEP
Inda
fixed
20 ø
10 ø
60 ø
70 ø
80 ø
90 ø
1 O0 ø
110 ø
120 ø
Figure 2. Regionalvelocitieswith 95% confidence
ellipsesshownin an India fixed frame (capitalized station names)or Eurasiafixed (lowercasestation names)frame. The thick black line is
the NUVEL1-A plate boundary between India and Eurasia.
2 showsthat Lhasa's velocity is slightly rotated with
respect to velocities from sites in northern Nepal. In
Figure 3 we display selectedvelocity vectors with re-
over 200 km, indicating a wider contraction zone than
in east Nepal.
A change in convergencedirection between eastern
spectto Simikot(SIMI). This perspectivedemonstrates and western Nepal cannot be preciselydetermined from
east-westextensionof 11:k3mm/yr betweenSIMI and these data becauseof the large uncertaintiesin the esLHAS.
timated east componentsfor many of the GPS stations.
A rotation of the convergencedirection would be con4.2. Himalayan Collision Zone
To look at the details of the Nepal Himalaya deformation,we plot velocitieswith respectto Nagarkot
(NAGA) in Figure4. This projectionclearlyshowsthat
sitesin southernTibet (RONG and TING) are moving
south with respect to NAGA and that sites in southern Nepal (e.g., BIRA and JANK) are movingnorth
toward
NAGA.
Horizontal
contraction
across eastern
sistentwith earthquakefocal mechanisms
[Molnar and
Lyon-Caen,1989; Ni and Barazangi,1984]and maximum principalstressorientations[Zoback,1992],which
alsorotate counterclockwisefrom west to east along the
arc to remain in an arc-perpendicular orientation, as
shownin Plate 1. Arc-normal shorteningis also consistent with
east-west
extension
across southern
Tibet
[Armijo et al., 1986].
Nepal(betweenBIRA and TING) is 18:k2mm/yr ori-
Vertical velocity estimates derived from GPS are 2 to
3 times lesscertain than correspondinghorizontal veloccomplicated in western Nepal. While SIMI is mov- ity estimates. Vertical velocitiesderivedfrom campaign
ing southwest,as expected,sitesin southwesternNepal style GPS studies such as this one must be examined
(MAHE and NEPA) aremovingnegligiblywith respect skeptically. Vertical accuracy is reduced by blunders
to NAGA instead of northward as we see in eastern
in the measurementof true antenna height. Data anaNepal. This finding suggeststhat the zoneof active con- lysts may misinterpret field notationsof antenna height.
traction extends farther to the south in western Nepal. Even so, we have not found a large number of outliers
The velocity pattern in western Nepal is smooth, so in the vertical estimates. Excluding TING and RONG,
our interpretation of a changein deformation pattern only six of the 22 Nepal Himalaya sitesindicate signifidoesnot rely on a singlestation. The rate of shorten- cant velocitiesat the I standarddeviationlevel (Table
ing within the networkin westernNepal (betweenSIMI, 2). The 1991 observationsat TING and RONG are
ented N12øE:k12 ø over 240 km.
The situation
is more
NEPA, andMAHE) is 13:k2mm/yr orientedN10øE:k9ø
problematicbecausewe usedan antenna(WM-102) for
1084
LARSON
ET AL.' INDIA-EURASIA
KINEMATICS
FROM
GPS
34 ø
32 ø
30 ø
'Mah
.:.
28 ø
Sim
26 ø
.10.mm/yr-.•- INDIA
78 ø
80 ø
82 ø
84 ø
......INDi/•
. ....:...
....
86 ø
88 ø
90 ø
92 ø
Figure 3. Selectedregionalvelocities
shownin relationto Simikot(diamond)with 95% confidenceellipses. Historic seismicityfrom 1950 to 1997 suggestsa changein strike along the
Himalaya at 83ø longitude.India-Eurasiaplate boundaryand faults are shownin black.
which we have no calibration; we have therefore ex-
vations of surfacedeformation using three-dimensional,
cluded these sites from further
dislocation models.
discussion.
At one of
The INDEPTH
seismic reflection
profile is located near 90ø longitude and showsthat
a discretefault plane dipping "09ø extendsfrom 27.7ø
mandu Valley (betweenNAGA and AIRP) measured to about 29ø latitude, where it reaches"040 km depth
et al., 1996;Zhao et al., 1993]. Deformasubsidence
of ? mm/yr. This agreeswith GPS vertical [Makovsky
tion to the north of -029ø latitude below Tibet is likely
velocityestimatesat AIRP of-6+6 mm/yr.
to be accommodatedby more widely distributed shear
5. Discussion
and asthenosphericflow in the lower crust and upper
mantle. However, our data near the tip of the slip5.1.
Interseismic
Strain
Accumulation
Across
ping Himalayan thrust systemare quite insensitiveto
the Himalaya
the rate, geometry,or mechanisms
of deformationbelow
To better understand the processof strain accumu- Tibet. Therefore we postulate that the observedinterlation in the Nepal Himalaya, we interpret GPS obser- seismicdeformation is primarily causedby the aseismithe sites, AIRP, we have an independent measurement
of the vertical velocity. Leveling data from the Kath-
32 ø
0 mm/yr,••
Lhas
30 ø
Ting
28 ø
Nep
km
ß
•
100
ß
BhaiS
I
200
26 ø
80 ø
82 ø
84 ø
86 ø
88 ø
90 ø
92 ø
Figure 4. Regionalvelocitiesshownin relation to Nagarkot (diamond)with 95% confidence
ellipses.
LARSON
ET AL.: INDIA-EURASIA
cally slipping portions of the thrust system below the
Himalaya. The modelfaultsreachfar belowTibet to
avoid edge effectsfrom the northern tip of the finite
dislocations.
Our
first-order
models
assume that
the
deformation can be characterized by slip on rectangular dislocation planes buried in a homogenous,isotropic
KINEMATICS
FROM
GPS
1085
tion with uniform slip located at the downdip tip of the
lockedfault. We solvefor the best fitting singlerectangular dislocationplane to representaseismicfault slip
north of the tip line of the locked system. The width
and length of the fault plane are constrainedto be 1000
km and 1500 km, respectively. The nonlinear inver-
and elastichalf-space[Okada,
1985].Thesedislocations sionfor ninefault parameters(geometry+ dip slip and
strikeslip) of a singlefault, with relativelywide bounds
on the geometry,computesa dislocationstriking 106ø,
represent aseismicshear north of the locked portion of
the Himalayan detachment system.
Alternatively, interseismic deformation in underthrusting regimes can be modeled by the combination
of steady block motion acrossthe complete thrust surface with a dislocation that is slipping opposite to the
thrust motion representing the locked portion of the
dipping 4ø to the north, following the southern edge
of the Greater Himalaya at a depth of 18 km (Figure
5). We computea dip-sliprate of 19+1 mm/yr and a
strike-slipcomponentof 4+1 mm/yr. The misfit of this
model is 1.07; that is, the model provides an adequate
fit to the data within the given uncertainties.Allowing
plies that the actively slippingfault segmentsdowndip only for dip slip, the fault strike rotates to 105ø, and
of the back-slippingdislocationare parallel to the locked all other parameters changeinsignificantly,including a
segment. For multiple sourcesof deformation that are dip-sliprate of 20+1 mm/yr. If we invert only the GPS
not aligned we prefer modelingthose active fault planes data, the model fault depth changesvery little, to 19
as finite width dislocationsin an elastic half-space. We km, and the fault dip increases
to 8ø with 20+1 mm/yr
do not attempt here to model viscous relaxation pro- and 4+1 mm/yr of dip-slipand strike-slip,respectively.
cessesexpected to play a role in the crustal earthquake However, if we solve for dip slip only, the tip line deepcycle, and we do not accountfor buoyancy effects. Data
ensto 22 km, with 23+1 mm/yr of slip. Finally, if we
spanning many decadesmay eventually reveal unique fix the fault strike to 106ø and invert only the leveling
evidence to differentiate
such models from the elastic
data, we compute a tip line depth of 14 km and a diprepresentations
we utilize [Thatcher,1986].
slip rate of 15+2 mm/yr. The relative sensitivityto
We use a constrained nonlinear optimization algo- small variations in our model parameters suggeststhat
rithm, which allowsus to estimatethe geometry(pa- the uncertaintyin the detachmentdepth is ,-•3km. Figrameterized by length, depth, width, dip, strike, and
ure 5a showsthe surfaceprojectionsof the fault planes
location) and the rates of strike slip and dip slip of for the three caseswe considered:only GPS data, both
one or more faults that best fit the GPS and/or lev- GPS and leveling data, and only leveling data, along
eling data [Arnadottir and $egall,1994; Biirgmannet with model and observed velocity vectors. In Figures
al., 1997]. Specifically,we seek modelsthat mini- 5b-5d we representthe velocity vectorsin their fault
mizethe weig•hted
residualsumof squares,
WRSS parallel, fault perpendicular, and vertical components,
(dabs
-- dmad)
JP-1 (dabs
-- dmad),wheredabsand dmad along with the model predictionsin these directions.
are the observedand modeled velocity components,reAll model parameters are listed in Table 3.
spectively, and P is the data covariance matrix. In
Overall, the levelingdata appearto prefera shallower
some caseswe apply bounds (such as constraintson
and slowersourceof deformation. This preferencecould
the depth of faulting, the range of permissible fault
be causedby systematic errors in the data, may indistrikes, or allowingfor dip-slip faulting only) to find
cate differencesin the deformation during the somewhat
best fitting sourcesthat are additionally constrained
different time intervals coveredby the GPS and levelby geologicor other information. The model misfit is
ing data, or may be related to insufficientconsideration
fault [Savage,1983;1996]. Sucha back-slipmodelim-
v/WRSS/(n-p), wheren isthenumber
of dataandp of the correlationsin the levelingdata [Arnadottir et
is the number of free model parameters. We compare
our geodeticobservationswith relatively simple dislocation models to evaluate
first-order
characteristics
of the
collisionzone revealedin the geodeticdata. These models include a best fitting singlefault plane and alongstrike heterogeneityrepresentedby two adjacent fault
planes. We evaluateda wide range of additional models
of varying complexity that led to the choiceof models
we show here. For example, we found that the data do
not yet resolvesignificantdown-dip variation in dip or
slipmagnitudeof the activefault surface[Biirgmannet
al., 1997].
5.2.
Model 1: Single Fault Plane
al., 1994]. The lack of slopedependentbiasesin the
data [JacksonandBilham,1994],the consistency
of the
uplift pattern measured from 1977 to 1990 and from
1990 to 1995 along the northern 15 km of the level line,
and inversion
of the data with full consideration
of the
data covariancesappear to rule out these causes. Sav-
age [1998]showsthat higher rigiditiesat depth lead
to a systematic differencebetween inverted dislocation
source depths of vertical and horizontal displacement
data. For a horizontal detachment source placed in a
half-space underlying a less rigid layer, leveling data
would invert for a shallower and less rapidly slipping
dislocation
than
would
GPS
data
if we assume a ho-
The simplest model of interseismic strain buildup mogeneous
elastichalf-space[Savage,1998]. Thus the
across a dipping thrust fault consistsof one disloca- apparent shallowingof the inferred detachment source
1085
LARSON
ET AL.'
INDIA-EURASIA
KINEMATICS
FROM
GPS
o
.
-5
32 ø
observed
M3.
"-
model•
-20
-25
•> 10 mm/yr
,
I
,
,
,
I
,
,
,
I
,
,
30 ø
,
I
,
,
,
I
ß
i
,
..........
o
E -10
,
28 ø
'
i
,
i
ß .
26 ø
82 ø
84 ø
86 ø
88 ø
90 ø
,
,
i
,
,
,
i
ß
,
,
,
i
,
,
,
ß, ,ß •, ,ß ß, ßßß
,,' L,
ockede)
•'
80 ø
,
-2o
o •)-40
92
c•
,,,
-100
0
100
200
300
Distance from fault tip (kin)
Figure 5. (a) Observedhorizontalvelocitiesrelativeto Nagarkot(solidvectors)and predicted
horizontalvelocities(open vectors)from best fit singlefault plane modelusingboth GPS and
levelingdata. Nagarkotis shownas a soliddiamond.Uncertaintiesare 95% confidence
ellipses.
The surfaceprojectionsof the fault planeusingboth GPS and levelingdata (M1), only GPS data
(M2), and only levelingdata (M3) are shownin black. Alsoshownare the velocitiesof the GPS
sitesin (b) fault parallel(striking105.3ø),(c) fault perpendicular,
and (d) verticalcomponents,
with 1 standard deviation uncertainties. M1 model predictions are shown as a dotted line in
Figure 5b-5d. (e) Locationof the lockedzonein profile.Profilesin Figures5b-5e are shownwith
respectto the distancefrom the fault tip. For further descriptionof the fault model, seeTable 3.
Table
Fault
3. Model
Parameters
Data
Dept h,
Dip,
St rike,
Dip-slip,
St rike-slip,
Model
km
deg
deg
mm/yr
mm/yr
Misfit
Best Fitting One Fault Models
Single
Single
Single
Single
Single
G+L
G+L
G
G
L
18.3
17.6
18.7
22.0
14.0
-4.4
-4.7
-7.7
-9.7
-14.8
106.4
105.3
106.1
104.4
106.1
-19.1
-19.6
4- 10.9
3.6 4- 1.1
4- 1.0
1.07
1.12
-20.1
4- 1.1
-22.7
4- 1.2
1.32
-15.4
4- 1.6
0.93
3.5 4- 1.2
1.23
Best Fitting Two Fault Model
West
East
West
East
G+L
G+L
G
G
25.0
14.6
25.0
16.2
-8.3
-3.4
-8.4
-2.3
112.3
100.9
112.3
99.4
-23.4
-20.8
-23.6
-21.8
4444-
1.5
1.1
1.6
1.3
1.00
1.06
Two-Fault Model, Western Fault Aligned to Seismicity
West
East
West
East
G+L
G+L
G
G
25.0
14.9
25.0
16.2
-4.5
-3.4
-4.5
-2.3
120.0
100.9
120.0
99.4
-21.3
-19.6
-21.5
-19.4
4- 1.6
-F 1.1
4- 1.6
4- 1.2
Data usedare GPS (G) and/or leveling(L). Model misfit is definedin the text.
4.34-
1.6
1.05
4.2 4- 1.6
1.24
LARSON
ET AL.: INDIA-EURASIA
inverted from the levelingdata may suggestlayering
of the Earth's crust with increasingrigiditiesat depth.
Both horizontal
and vertical deformation
underestimate
the sourcedepth; that is, our inversionmay be somewhat biased toward shallow depths. As the precision
of geodeticdata improves,we are increasinglysensitive
to second-orderdetails of our model parameterization
including varying constitutive constants.
Other geologicand geophysical
evidencesupportsour
model basedonly on the geodeticdata. The depth, dip,
and location of the model fault plane agree quite well
with the Main Himalayan Thrust reflector imaged by
the INDEPTH seismicreflectionexperiment[Makovsky
et al., 1996; Zhao et al., 1993]. The data confirmthe
existenceof a relatively simpleaseismicallyslippingdetachment plane that accommodatesmost of the active
India-Tibet convergencenorth of -•28 ø latitude. Instead of representing a simple dislocation slip plane,
however, slip is probably accommodatedin a narrow,
strain-weakenedshear zone, similar to mylonitic shears
observedalong thrust faults exhumed from midcrustal
depth [e.g.,Ramsay,1980;Kirby, 1985;Harrison et al.,
1•].
Intermediate size earthquakesalong the Himalayan
arc consistently have focal mechanisms with a shal-
lowly northwarddippingnodal plane (Plate 1) and are
KINEMATICS
FROM
GPS
1087
not predictthe shift in displacement
patternsalongthe
southwesternedge of the network. As we move from
eastern to western Nepal, the southernmost site velocities changefrom a small northward motion relative to
NAGA to a morewesterlytrend with an insignificant
northcomponent.We thereforecontinueour investigation by allowingfor a variationin fault geometryalong
the Himalaya.
5.3.
Model
2: Two Adjacent
Fault Planes
The western Nepal segment of the Himalayan thrust
systemhas not experienceda significantearthquake for
at least 300 years[Bilharnet al., 1997]. It is therefore
important to evaluate whether this area behaves differ-
ently from the easternregion near Kathmandu, which is
located in the hanging wall of the 1934 Bihar earthquake
rupture. Variations could be due to the two regionsbeing at different stagesof the regional earthquake cycle
or may stem from structural segmentationof the fault
system. The seismicityband along the Himalaya experiences a sharp bend of about 20ø to the northwest of
83ø longitude, indicating a possiblesegmentboundary
near this longitude (Figure 3). We evaluate whether
significant lateral variations are indicated by the data
by modeling the deformation with two contiguousfault
planes. In particular, we are interested in differencesin
the width of the locked fault plane, the depth to the
fault, and variations in slip rate along the Himalayan
range front. For this purpose we restrict the searchto
dip-slip on two faults that lie approximately adjacent
southern tip of the Main Frontal Thrust in southern to each other. Their absolute position, strike, dip, and
central Nepal. The agreementof these long-term con- depths are unconstrained.
We find significantvariations in the depth of the two
traction rates across the southern tip of the detachment system with the interseismic strain accumulation model faults. The tip of the eastern fault is at a depth
rates we measure at the base of the Greater Himalaya of 15 km, but the western fault clearly prefers depths
suggeststhat detachment slip efficiently accommodates of 25 km or more. There is an apparent correlation bemostof the India-southernTibet convergence.This lack tween estimated slip magnitude and depth of faulting
of evidence for significant plastic strain north of the for the western fault. That is, within a certain range,
Main Frontal Thrust, the locked detachment south of deeper models with higher slip rates fit the data as well
thought to be eventson the detachment systemand as-
sociatedsecondaryfaults [Baranowskiet al., 1984;Molnat, 1990; Ni and Barazangi,1984]. Lav• and Avouac
(submitted manuscript, 1998) measured21.5 + 1.5
mm/yr of Holoceneshorteningacrossfoldsmarkingthe
the Greater Himalaya constrainedby the geodeticdata,
and the insignificant contribution of moderate earthquakes to detachment slip indicate that most of this
slip occursin great Himalayan detachment earthquakes
reachingfrom the southern Siwalik hills to the Greater
Himalaya (Lav• and Avouac, submitted manuscript,
1998).
The single-dislocationmodelfits most of the data well
within the uncertainties,the exceptionsbeing sites in
southwestern Nepal and LHAS. The model does not
fit the data from LHAS within 95% confidencelimits,
but this fit is not to be expected, given that LHAS is
affected by both extensional and contractional deformation. LHAS movesabout 5 mm/yr fastersouthward
than SIMI, showingthat a small amount of north-south
contraction may occur acrossthe south Tibetan graben
system. Removal of the LHAS data has little impact
on model estimates. The singlefault plane model does
as the
model
at 25 km.
We would
be able to more
confidently constrain the other fault parameters if additional constraints could be put on either the slip rate
or the depth based on complementary data. The outline
of the eastern and western faults, along with observed
and predicted horizontal velocities,are shownin Figure
6a. Dip slip on the western and eastern fault segments
are 23+2 mm/yr and 214-1 mm/yr, respectively.The
two model fault planes rotate into approximate paral-
lelismwith the Himalayanarc (striking ll2 ø and 101ø,
respectively). The western and eastern faults dip 8ø
and 3ø to the north, respectively. The model misfit is
reduced to 1.00. We evaluate the GPS data separately
to examine the importance of the leveling data in this
conclusion. We find that the resulting model is very
comparable to the previous one, which used both GPS
and levelingdata (Table 3). The GPS data fit a two
fault plane model significant.
ly better than a one fault
1088
LARSON ET AL.' INDIA-EURASIA
30 ø
KINEMATICS FROM GPS
/':.""""'"':'."._'••'
•' .../. o.b•,•ved:.
,,..1.0.
m.,m/.yr.
'/
28 ø
26 ø
80 ø
82 ø
84 ø
86 ø
88 ø
ß
5
I
,
I
,
I
90 ø
,
I
,
I
,
I
,
92 ø
I
,
I
,
I
EAST
0
b•
1"*"'
....
,>,
'
E -10
E -15
-20
-25
'--
'
i
'
'
'
i
'
'
'
i
'
'
i
lO
....
tE -10
/
i
I
i
i
'
i
,
,
.
i
.
,
,
i
,
,
,
I
I
,
I
'
•.,10
l i'•'•ti.......
•..• 'l•i'
I
'
'
ß.......................
I
,
,
,
I
,
,
,
I
,
,
,
I
,
E -10
.......
i
'
'
i
'
'
'
Locked
i
....
e)
e)
Locked
e -40
-1 O0
0
1O0
200
300
-1 O0
Distancefromfaulttip (km)
0
1O0
200
300
Distancefromfaulttip (km)
Figure 6. (a) Observed
horizontalvelocities
relativeto Nagarkot(solidvectors)and predicted
horizontalvelocities(open vectors)from best fit two fault plane model usingboth GPS and
levelingdata. Nagarkotis shownas a soliddiamond.Uncertainties
are 95% confidence
ellipses.
The surfaceprojectionsof the fault planesusingboth GPS and levelingdata are shownin black.
Alsoshownarethe velocities
ofthe GPSsitesin (b) faultparallel(striking112.3øand100.9ø),(c)
fault perpendicular,and (d) verticalcomponents,
with I standarddeviationuncertainties.Model
predictionsfor the westernand easternfault planesare shownas a dotted line in Figures 6b-6d.
(e) Locationof the lockedzonein profile. Profilesin Figures6b-6e are shownwith respectto
the distancefrom the fault tip. For further descriptionof the fault model, seeTable 3. Historic
seismicity from 1950 to 1997 is also shown.
plane model, with a reduction in the model misfit from
1.32 to 1.06.
How
well does this
dislocation
model
fit the verti-
cal GPS observationsshownin Figure 6d? The WRSS
for 22 site velocities improvesfrom 18.6 to 11.8 when
the model predictions are used instead of a null pre-
diction. The improved fit suggeststhat some of the
measured vertical GPS rates are in fact causedby the
India-Eurasia collision, although becauseof the large
uncertaintiesthe vertical velocityestimatesdo not constrain
the dislocation
models.
The preferred "segmentboundary" is near the rather
LARSON ET AL.: INDIA-EURASIA
abrupt changein the strike of the seismicitynear 83ø
KINEMATICS
FROM GPS
1089
longer time periods will be neededto uniquely resolve
longitude[Tan•ukar ½i aL, 1997]. It also coincides the contribution of viscousand gravitational processes
with the southern
extension of the north-south
strik-
ing Thakkholagrabensoutheastof Jomoson(JOMO),
and a basementhigh (the BundelkhandMassifand FaizabadRidge)on the underthrusting
Indianplate [/(hailri, 1987]. This findingsuggests
that there may be a
to the observed
5.4.
surface deformation.
Kinematics
of the
India-Eurasia
Collision
Zone
Our analysis has focused on the Himalayan frontal
significantdiscontinuitybetweenthe westernand eastern Nepal deformationzone,alongwhat is otherwisea thrust and southern Tibet but also includes more dissmoothsmall-circlearc of high elevations,topographic tributed, continuouslyoperating sites in the India-Eurroughness,
and seismicity[Bilham et al., 1997; Seebet asia collision zone. Additional published results are
and Gornitz,1983]. Interactionof the underthrusting
available
basementhigh with the overriding plate may be the
causefor this observedlateral heterogeneity.
Notwithstandingthis apparentrelation betweenwestern Nepal seismicity and inferred deformation kinematics, we note that the geometryof the western112ø striking model fault plane doesnot exactly follow the observedseismicityband striking 120ø. We have there-
without
fore tested another two-fault
Roydenet al., 1997]and our own resultstogetherwith
model in which we force
for some areas in central
a formal
combination
and east Asia.
of the networks
Even
we can
continue our discussionby evaluating existing geodetic
constraints
in the broader
context
of the kinematics
of
the India-Eurasia collision. In Figure 7 we show shortening, extension, and strike-slip rate estimates from
publishedgeodeticstudies[Abdrakraatov
et al., 1996;
Karo et al., 1998; King et al., 1997; Michel et al., 1997;
the westernmodelthrust to morecloselyfollowthe 120ø
striking seismicityband (Table 3). Allowingonly dipslip deformation, we cannot fit the data for SIMI and
JOMO, with a correspondinglylarge model misfit of
1.30. If we allow strike slip as well as dip slip, we
have a good fit to the observations,with a misfit of
1.05, comparedto 1.00 for the best fitting two fault
plane model with dip slip only. This model implies
that about 4 mm/yr of right-lateralstrikeslipis accommodated acrosswestern Nepal. Mapped right-lateral
strike-slipfaults striking ESE through westernNepal
selected station velocities referenced to Eurasian plate
motion. While many questionspertaining to the kinematics of the collision zone remain unanswered,we can
As we noted above, the differencesbetween motions
in western and eastern Nepal could also be due to these
locities at the 95% confidence limit in relation to the
ation model includes a viscous lower crust from 25 to
ments from western Nepal fully account for the conver-
addressthree regionalissuesincluding(1) the northern
limit of the broader India-Eurasia collisionzone, (2)
east-westextensionin southernTibet, and (3) the extent of eastward
motion
of south China.
A number
of
geodeticexperimentsare in progressthat will eventually
resolvemany of the remaining unansweredquestionsof
how plate boundary strain is distributed in this region.
Continuous
sitesKIT3 (3+2 mm/yr) andIRKT (4+2
[Upreti,1996]presumably
respondto this inferreddex- mm/yr) and the station AZOK (2+3 mm/yr) reported
tral shear.
by Abdrakhmatovet al. [1996]have insignificantveEurasian plate. This finding showsthat only a minimal
regionsbeing in differentstagesof the earthquakecy- amount of plate boundary strain may exist northwest
cle. Preliminary layered viscoelasticmodels that con- of the Pamirs and north of the Tien Shan and Mongosiderthe possibleeffectsof continuedpostseismicrelax- lia, consistentwith seismologicand geologicestimates
ation from the 1934 Bihar earthquake(F. Pollitz and [Englandand Molnar, 1997b]. The broad distribuR. B/irgmann, unpublishedmodels, 1998) predict up tion of north-south shortening acrossthe India-Eurasia
to 5 mm/yr of transientnorthwardmotion and local- collision zone is best resolved in the west. Results
ized contraction and uplift in eastern Nepal, somewhat from the Pamirs [Michel et al., 1997; G. Michel, perconsistent with the observed difference between western
sonalcommununication,
1998],the Tien Shan [Michel
and easternNepal motions. This gravitational relax- et al., 1997;Abdrakmatov
et al., 1996],andour measure-
40km depth(withviscosity
y = 10TMPa s), hasincreas- gencebetween India and Eurasia, with observationsof
ing shear and bulk moduli with depth, and accountsfor
gravitationaleffects[Pollitz, 1997]. If we removethe
modeled contribution
of viscous relaxation
50+6 mm/yr comparedwith model predictionsof 44
mm/yr. This findingsuggests
that Himalayanshorten-
of the 1934
ing acrossnorthwest India and Kashmir, located south
of the Pamirs, is comparable to that observed across
invert once more for the best fitting two-fault disloca- the Nepal Himalaya to the east. Even so, the uncertion model, we find that the inverted depth of the east- tainties are quite high, and determination of how strain
ern fault plane deepensto greater than 20 km. That is partitioned in detail and what role rotating blocks
is, instead of being related to a structural heterogene- and strike-slip faults may play will require additional
ity along the Himalayan arc, the lateral variations of measurements,denserstation spacing,and formal intethe displacementfield may reflect residual effectsof a gration of the individual networks.
large earthquake60 yearsago that is superimposedon
In addition to contraction across the Himalaya, we
Bihar earthquake from the observedGPS velocities and
the secularaseismicfault slip. However,data spanning
measure
east-west
extension
in southern
Tibet
of 11+3
1090
LARSON ET AL.- INDIA-EURASIA
:i••,
,:.velocity
rel.
KINEMATICS
FROM GPS
Eurasia
to Eurasia
shortening/
<:•• extension
"•=strii<e
slip
4O N
amir
India
20 N
?.:
:.0: 500:
60 E
80 E
100 E
120 E
Figure 7. Summary of geodetically determined deformation rates acrossthe India-Eurasia
collisionzoneintegratingour resultswith publishedstudies. Large solidarrowsindicate velocities
relative to Eurasia from this study. Arrows with dashedoutline are sites from other publications
suchas Taejeonon the Korean Peninsula[Karo et al., 1998] and Azok in the northernmost
Tien Shan [Abdrakhmatov
et al., 1996]or are inferredassuminga rigid south China block (east
and west of Xianshuihefault) and Indian subcontinent(northwestIndia). Arrowsin circlesare
used to show velocitiesthat show no significantmotion with respect to Eurasia at the 95%
confidencelimit. Small gray arrowsshowactive deformationrates determinedfrom severalyears
of GPS measurements
acrossthe Tien Shan[Abdrakhmatov
et al., 1996],the northernPamir and
westernTien Shan [Michelet al., 1997;G. Michel,personalcommunication,1998];the Himalaya
[Bilham ctal., 1997;this study], the Xianshuihefault [King et al., 1997; Roydenet al., 1997],
and the LongmenShah [King et al., 1997]. Shorteningacrossthe Indian subcontinentsouthof
the Himalaya (not shown)is constrainedto be lessthan 2 mm/yr [Paul et al., 1995; Bilham et
al., 1995].
mm/yr betweenSIMI and LHAS, which is consistent SIMI and JOMO (3+2 mm/yr) is not significant,sugwith geologicand seismicreports for the region. The gestingeither that east-westextensionis not uniformly
rate of extension across southern Tibet has been estidistributed along the range or that the extensionoccurs
only
north of JOMO and TING.
mated as 5-10 mm/yr from seismicmomentreleasein
earthquakes
[Baranowski
et al., 1984]and 10+5 mm/yr
The
motion
of LHAS
is eastward
in a Eurasian
or
from extensivegeologicmappingand Landsatimagein- southwestTibetan referenceframe (Figures2 and 3),
terpretation[Armijo et al., 1986]. On the basisof geo- suggestingthat east-westextensionof southern Tibet
accommodateseastward displacementsrather than westmetric considerationsof arc-normal thrusting of Tibet
ward
motion as suggestedby Molnar and Lyon-Caen
overthe Indian plate, Molnar andLyon-Caen[1989]estimate extensionratesof 18+9 mm/yr and •10 mm/yr [1989].LHAS is locatedsouthof the BengCo and Jiali
in southernand central/northernTibet, respectively. echelonsegmentsof the right-lateral Karakorum-Jiali
As they included the full arc width from 75ø to 95ø fault zone,whichis thoughtto slip at about 15 mm/yr
longitude, this higher value is consistentwith our esti- [Armijo et al., 1989]. If so, centraland northernTibet
mate. Using similar geometricarguments,Bilham et al. would have to be moving eastward at an even higher
[1998]estimatean along-arctensilestrainrate in south- rate between the left-lateral Altyn Tagh and Kunlun
fault zone.
ern Tibet of 0.012/•strain per year, corresponding
to 12 faults and the Karakorum-Jiali
mm/yr betweenSIMI and LHAS. Althoughthis close
We can also examine the motion of SIMI, JOMO,
agreement is an encouragingresult, the current rate of
extensionbetweenSIMI and TING (2+4 mm/yr) and
TING, and LHAS with respect to the local plate convergencedirection. The local direction of convergence
LARSON ET AL.: INDIA-EURASIA KINEMATICS FROM GPS
1091
can be determinedby subtractingthe Eurasianand In- shallower depth. A changein strike of a lineament of
dian predictionsfor eachsite (Table 2). Residualvec- small to moderate earthquakesalong the Himalaya and
tors can then be resolvedinto plate convergentand of[- the inferred western edge of the 1934 Bihar earthquake
convergentdirections.The off-convergent
componentof
rupture coincide with this possiblesegmentboundary.
However, models that attempt to accountfor the effects
mm/yr),JOMO (4+2 mm/yr),TING (54-4mm/yr)to of viscousrelaxation followingthe 1934 earthquakesugLHAS (134-3mm/yr), with LHAS showing
significant gest that the differencescould also be due to remnant
eastward divergence. We will need sites to the west of postseismicdeformation from that event.
motionslowlyincreases
aswemoveeastfromSIMI (1+2
SIMI to test whether these sites show westward motion
with respectto the localplate convergence
direction.
King et al. [1997]findinsignificant
contractionacross
the LongmenShanand 12-15+4 mm/yr strikeslipon
the left-lateral Xianshuihe-Xiaojiang
fault. This result
suggeststhat eastward displacementof southern and
central Tibet is partially accommodated
by southeast
directed corner flow or clockwise rotation around the
easternsyntaxisof the collisionzone(Figure7), assug-
gestedby Roydenet al. [1997]andKing et al. [1997].
Eurasia-fixed velocities of sites on either side of the Xi-
anshuihefault shownin Figure 7 are dependenton the
premisethat there is insignificantstrain betweeneither
We find east-west
extension
across southern
Tibet
at
a rate of 11+3 mm/yr betweennorthwesternNepal and
Lhasa. The spatial distribution of extension acrossTibet is not well resolved by our data. However, extension is apparently occurring north of the strain accumulation zone of the Himalayan detachment system.
The observed motion of Lhasa suggeststhat portions
of southern Tibet displace eastward ahead of the Indian indenter and that somenorth-south shorteningoccurs across southern Tibet. If the east-west striking,
right-lateral Karakorum-Jiali fault zone north of Lhasa
slipsat rates suggestedby geologicestimates,then eastern central Tibet south of the Kunlun fault zone may
SHAO or TAIW and King et al.'s [1997]GPS refer- move at 20 mm/yr to the east. Resultsfrom King et
ence station at Chengdu, which is located east of the
al. [1997]suggestthat about half the eastwardmotion
Xianshuihefault and the LongmenShan. Neithergeologicnor seismicstudiesfind evidencefor significant
strainacrossthe SouthChinablock[Englandand Molnat, 1997b;Holt et al., 1995]
The lack of shorteningacrossthe Longmen Shan
[Kinget al., 1997]andthe --•10mm/yr eastwardmotion
of Tibet
of SHAO, TAIW, and a site in Korea relative to Eurasia
is transferred
south of the south China
block
toward Indochina by the southeaststriking, left-lateral
Xianshuihe-Xiaojiang fault zone. Eastward motions of
about 10 mm/yr of sitesin Shanghaiand westernTaiwan and the lack of discernibleshorteningacrossthe
LongmenShan [King et al., 1997]and the southChina
[Kato et al., 1998]suggestthat coherenteastwarddis-
block suggestregional eastward displacementof Tibet
and south China away from the collisionzone. No sig-
placement of Tibet and south China amounts to about
nificant
strain
is observed north
of the Tien
Shan and
10mm/yr (Figure7). It hasto be pointedout, however, Mongolia.
that the easternsitescouldpossiblybe affectedby subWe have successfullydetermined the pattern and
duction and continentalcollisionprocessesto the east. mechanism of strain accumulation acrossthe Nepal HiContinuouslyoperatingGPS sitesat Wuhan and X'ian malaya and have provided some additional constraints
locatedwell westof SHAO [Kato et al., 1998]will soon
on motions
provide the constraints that are needed to test these
the
patterns.
versy still focuseson the contributions of crustal thickening and extension, strike-slip faulting and eastward
extrusion, as well as crustal block rotations in the ac-
6.
Conclusions
total
in southern
width
commodation
of the
Tibet
collision
of continental
and South
zone.
China
Much
collision between
and
contro-
India
and
North-south contractionacrossthe Nepal Himalaya Eurasia[Molnar and Deng,1984;Englandand Molnar,
occursat a rate of 18+2 mm/yr, suggesting
that about 1997a; 1997b; Avouac and Tapponnier, 1993; Peltzer
20 mm/yr of strain is accumulatingacrossthe locked and Saucier, 1996; Davy and Cobbold,1988; England
portion of the Himalayan thrust system south of the
and Houseman,1986; Peltzer et al., 1989]. GPS data
Greater Himalaya. Our three-dimensionalmodel inare particularly suitable for resolvingthe active threeversions show that the detachment fault is locked over
dimensionalkinematicsin a global referenceframe. On--•140km width alonga 500 km reachof the Nepal Hi- going and future GPS studies will provide answersto
malaya. Future M8 earthquakeswill eventuallyrelieve many of the outstanding questionspertaining to the
the accumulating slip deficit. Of particular concernis geodynamicsof this collision zone.
the region between the 1905 Kangra and 1934 Bihar
earthquakeruptures that may have built up as much
Acknowledgments.
We would like to thank the IGS
as 6-15 m of slip potential. Our modelssuggestthat
significantdifferencesin the active displacementpat- and their many support personnelfor providingoutstanding
GPS data and data products. We also thank P. Athans, J.
terns between west and east Nepal can be explained Avouac, R. Bendick,W. Berg, M. Bhatterrai, F. Blume, P.
by a major structural segmentboundary near 83ø lon- Bodin, V. Gaur, M. Jackson,S. Jade, D. Mencin, P. Molnar,
gitude, east of which the detachment is located at a P. Oli, B. O'Niell, M.R. Pandey, J. Paul, G. Rosborough,B.
1092
LARSON
ET AL.:
INDIA-EURASIA
Shrestha,F. Sigmundsson,B. Stephens,W. Wang, B. Washburn, UNAVCO, and all the members of Project Idylhim
for help in various parts of this project. Figures were pro-
ducedusingGMT [Wesseland Smith, 1995]. JPL provided
the GIPSY software. This research was supported by NSF
grants EAR 9506677 and EAR 9727652 and NASA grants
NAG-1908 and NAG5-6147. Earthquake locations are from
KINEMATICS
FROM
GPS
zoicslip on the Talas-Ferghana
fault, the Tien Shan,central Asia, Geol. Soc. Am. Bull., 108, 1004-1021, 1996.
Davy, P. and P. Cobbold,Indentationtectonicsin nature
and experiment,1, Experimentsscaledfor gravity,Bull.
Geol. Inst. Uppsala, 1J, 129-141, 1988.
DeMets, C., R. Gordon,D. Argus, and S. Stein, Current
plate motions,Geophys.J. Int., 101,425-478, 1990.
the Councilof the National SeismicSystem(CNSS) world- DeMets, C., R. Gordon, D. Argus, and S. Stein, Effect of
wide earthquake catalog provided by the Northern Califorrecent revisions to the geomagneticreversal time scale on
nia EarthquakeData Center (NCEDC). We are gratefulto
estimates of current plate motions, Geophys. Res. Lett.,
Jim Savage,Gilles Peltzer, and Bob King for carefulreviews
21, 2191-2194, 1994.
of the manuscript.
England, P.C., and G.A. Houseman, Finite strain calculation of continental deformation, 2, Comparison with the
India-Asia collision zone, J. Geophys. Res., 91, 36643676, 1986.
References
Abdrakhmatov, K., et al., Relatively recent construction
of the
Tien
Shan
inferred
from
GPS
measurements
of
present-day crustal deformation rates, Nature, $8•, 450453, 1996.
Armijo, R., P. Tapponnier, J.L. Mercier, and H. Tonglin,
Quaternary extension in southern Tibet: Field observations and tectonic implications, J. Geophys. Res., 91,
13,803-13,872, 1986.
Armijo, R., P. Tapponnier,and T. Han, Late Cenozoicrightlateral strike-slip faulting in southern Tibet, J. Geophys.
England, P.C., and P. Molnar, Active deformation of Asia:
From kinematics to dynamics, Science, 278, 647-650,
1997a.
England, P.C., and P. Molnar, The field of crustal velocity
in Asia calculated from Quaternary rates of slip on faults,
Geophys. J. Int., 130, 551-582, 1997b.
Freymueller, J., R. Bilham, R. Bfirgmann, K.M. Larson,
J. Paul, S. Jade, and V. Gaur, Global Positioning System measurements of Indian plate motion and convergence across the lesser Himalaya, Geophys. Res. Lett.,
23, 3107-3110, 1996.
Res., 9•, 2787-2838, 1989.
Harrison, T.M., F.J. Ryerson, P. Le Fort, A. Yin, O.M.
Argus, D., and R. Gordon, No-net rotation model of curLovera, and E.J. Catlos, A late Miocene-Plioceneorigin
for the Central Himalayan inverted metamorphism, Earth
rent plate velocities incorporating plate motion model
Planet Sci. Lett., 1J6, El-E7, 1997.
NUVEL-1, Geophys. Res. Lett., 18, 2039-2042, 1991.
Arnadottir, T., and P. Segall,The 1989 Loma Prieta earth- Heki, K., Horizontal and vertical crustal movements from
three-dimensional very long baseline interferometry kinequake imaged from inversionof geodeticdata, J. Geophys.
Res., 99, 21,835-21,855, 1994.
Avouac, J.-P., and P. Tapponnier, Kinematic model of active
deformation in Asia, Geophys. Res. Lett., 20, 895-898,
rates
1993.
Baranowski, J., J. Armbruster, L. Seebet, and P. Molnar,
Focal depths and fault plane solutions of earthquakes and
active tectonics of the Himalaya, J. Geophys. Res., 89,
6918-6928, 1984.
Bilham, R., Location and magnitude of the 1833 Nepal
earthquake and its relation to the rupture zones of contiguous great Himalayan earthquakes, Curr. Sci., 69,
101-128, 1995.
Bilham, R., P. Bodin, and M. Jackson, Entertaining a
great earthquake in western Nepal: Historic inactivity and
geodetictests for the developmentof strain, J. Nepal Geol.
Soc., 11, 73-88, 1995.
Bilham, R., K. Larson, J. Freymueller, and Project Idylhim
Members, GPS measurementsof presentday convergence
rates in the Nepal Himalaya, Nature, 336, 61-64, 1997.
Bilham, R., F. Blume, R. Bendick, and V.K. Gaur, Geodetic constraints
matic reference frame: Implication for the reversal timescale revision, J. Geophys. Res., 101, 3187-3198, 1996.
Holt, W.E., M. Li, and A.J. Haines, Earthquake strain
on the translation
and deformation
of In-
dia: Implications for future great Himalayan earthquakes,
Curr. Sci., 7J, 213-229, 1998.
Boucher, C., Z. Altamimi, M. Feissel,and P. Sillard, Results
and analysis of the ITRF 94, IERS Tech. Note 20, Int.
Earth Rotation Serv. Cent. Bur., Obs. de Paris, Paris,
1996.
Burchfiel, C., Z. Chen, Y. Liu, and L. Royden, Tectonicsof
the Longmen Shan and adjacent regions, Central China,
Int. Geol. Rev., $7, 661-735, 1996.
Bfirgmann, R., K.M. Larson, and R. Bilham, Model inversion of GPS and leveling measurements across the
Himalaya: Implications for earthquake hazards and future geodetic networks, ConferencePaper for the Indo-US
Workshop on Paleoseismicity,Dehra Dun, India, Wadia
Inst. of Himalayan Geol., March 26-28, 1997.
Buttman, V.S., S. F. Skobelev, and P. Molnar, Late Ceno-
and
instantaneous
relative
motions
within
central
and eastern Asia, Geophys. J. Int., 122, 569-593, 1995.
Houseman,G.A., and P.C. England, Crustal thickeningversus lateral expulsion in the India-Asia continental collision, J. Geophys. Res., 98, 12,233-12,249, 1993.
Jackson, M., and R. Bilham, Constraints on Himalayan Deformation inferred from vertical velocity fields in Nepal
and Tibet, J. Geophys. Res., 99, 13,897-13912, 1994.
Kato, T., et al., Initial results from WING, the continuous
GPS network in the western Pacific area, Geophys. Res.
Lett., 25, 369-372, 1998.
Khattri, K.N., Great earthquakes, seismicity gaps, and potential for earthquake disaster along the Himalaya plate
boundary, Tectonophysics,138, 79-92, 1987.
King, R.W., F. Shen, B.C. Burchfiel, L. Royden, E. Wang,
Z. Chen, Y. Liu, X. Zhang, J. Zhao, and Y. Li, Geodetic measurement of crustal motion in southwest China,
Geology,25, 179-182, 1997.
Kirby, S.H., Rock mechanics observations pertinent to the
rheology of the continental lithosphere and the localization of strain along shear zones, Tectonophysics,119, 127, 1985.
Larson, K.M., J. Freymueller, and S. Philipsen, Global plate
velocities from the Global Positioning System, J. Geophys. Res., 102, 9961-9982, 1997.
Lichten, S., and J. Border, Strategiesfor high-precisionGPS
orbit determination, J. Geophys. Res., 92, 12,751-12,762,
1987.
Makovsky, Y., S. Kleinpeter, L. Huang, and D. Lu, Structural elements of the southern Tethyan Himalaya crust
from wide-angle seismic data, Tectonics, 15, 997-1005,
1996.
Michel, G., C. Reigber, D. Angermann, J. Klotz, CATS
Team, and GEODYSSEA Team, Ongoing and recent de-
LARSON ET AL.: INDIA-EURASIA
KINEMATICS
FROM GPS
1093
formation in central and SE Asia (abstract), Eos Trans.
AGU, 78(•6) Fall Meet. Suppl., F172, 1997.
Ramsay, J.G., Shear zone geometry: A review, J. Struct.
Molnar, P., Structure and tectonics of the Himalaya: Constraints and implicationsof geophysicaldata, Annu. Rev.
Earth Planet. Sci., 12, 489-518, 1984.
Molnar, P., Inversion of profilesof uplift rates for the geometry of dip-slip faults at depth, with examples i¾omthe
Alps and Himalaya, Ann. Geophys.,5, 663-670, 1987.
Molnar, P., A review of the seismicityand the rates of active
underthrusting and the deformation of the Himalaya, J.
Himalayan Geol., 1, 131-154, 1990.
Molnar, P., and Q. Deng, Faulting associated with large
earthquakesand the averagerate of deformationin central
and eastern Asia, J. Geophys. Res., 89, 6203-6228, 1984.
Molnar P., and J. Gipson, A bound on the rheologyof continental lithosphereusingvery long baselineinterferometry:
The velocity of south China with respect to Eurasia, J.
Geophys. Res., 101, 545-553, 1996.
Molnar, P., and H. Lyon-Caen, Fault plane solutions of
earthquakes and active tectonics of the Tibetan plateau
and its margins, Geophys. J. Int., 99, 123-153, 1989.
Molnar, P., and P. Tapponnier, Cenozoictectonics of Asia:
Effects of a continental collision, Science, 189, 419-426,
Royden, L.H., B.C. Burchfiel, R.W. King, E. Wang, Z.
Chen, F. Shen, and Y. Liu, Surfacedeformation and lower
1975.
crustal flow in eastern Tibet, Science, 27'6, 788-790, 1997.
Savage,J.C., A dislocationmodel of strain accumulation
and release at a subduction zone, J. Geophys. Res., 88,
4984-4996, 1983.
Savage, J.C., Comment on "The stressstate implied by
dislocation models of subduction deformation," Geophys.
Res. Left., 23, 2709-2710, 1996.
Savage,J.C., Displacementfield for an edgedislocationin
a layered half-space,J. Geophys.Res., 103, 2439-2446,
1998.
Seeber, L., and V. Gornitz, River profiles along the Himalayan arc as indicators of active tectonics, Tectonophysics, 92, 335-367, 1983.
Tandukar, R.P., et al., Microseismicepicentremap of Nepal
Himalaya and adjoining region, scale 1:2,000,000, Nat.
Seismol. Cent., Dep. of Mines and Geol., Kathmandu,
Nepal, 1997.
Thatcher, W., Cyclic deformationrelated to great earthquakesat plate boundaries,Bull. R. Soc. N.Z., œ,i,245272, 1986.
Molnar, P., B.C. Burchfiel,L. K'unangyi, and Z. Ziyun, Geomorphic evidence for active faulting in the Altyn Tagh
and northern Tibet and qualitative estimatesof its contribution to the convergenceof India and Eurasia, Geology,
15, 249-253, 1987.
Ni, J., and M. Barazangi, Seimotectonicsof the Himalayan
collision zone: Geometry of the underthrusting Indian
plate beneath the Himalaya, J. Geophys. Res., 89, 11471163, 1984.
Okada, Y., Surface deformation due to shear and tensile
faults in a half-space,Bull. Seismol. Soc. Am., 75, 11351154, 1985.
Pandey, M.R., R. P. Tandukar, J.P. Avouac, J. Lav•, and
J.P. Masot, Interseismic strain accumulation on the Hi-
malayancrustal ramp (Nepal), Geophys.Res. Left., 22,
751-754, 1995.
Paul, J., et al., Microstrain stability of Peninsular India,
1864-1994, Proc. Indian Acad. Sci. Earth Planet. Sci.,
10,•, 131-146, 1995.
Peltzer, G., and F. Saucier, Present day kinematics of Asia
derived from geologicfault rates, J. Geophys.Res., 101,
27,943-27,956, 1996.
Peltzer, G., P. Tapponnier, and R. Armijo, Magnitude of
late Quaternary left lateral displacementsalong the north
edge of Tibet, Science, 2•6, 1285-1289, 1989.
Pollitz, F.F., Gravitational viscoelastic postseismic relaxation on a layered sphericalEarth, J. Geophys.Res., 102,
17,921-17,941,
Geol., 2, 83-89, 1980.
1997.
Upreti, B.N., Stratigraphyof the westernNepal, LesserHimalaya: A synthesis,J. Nepal Geol. Soc., 13, 11-28, 1996.
Wessel,P. and W. Smith, New versionof the GenericMapping Tools released,Eos Trans. A G U, ?'6,329, 1995.
Zhao, W., et al., Deep seismicreflectionevidencefor continental underthrustingbeneath southernTibet, Nature,
366, 557-559, 1993.
Zoback, M.L., First and second-orderpatterns of stressin
the lithosphere:The World StressMap Project, J. Geophys. Res. 97, 11,703-11,728, 1992.
R. Bilham, Department of Geological Sciences and
CIRES, University of Colorado, Boulder, CO 80309. (bilham@stripe.colorado.edu)
R. B/irgmann, Department of Geology and Geophysics, Univeristy of California, Berkeley, CA 94720.
(burgmann@seismo.berkeley.
edu)
J. T. Freymueller, Geophysical Institute, University of
Alaska,Fairbanks,AK 99775. (jeff@giseis.alaska.edu)
K. M. Larson, Department of Geophysics,Stanford University, Stanford, CA 94305.
(ReceivedApril 24, 1998; revisedAugust 17, 1998;
acceptedSeptember23, 1998.)