Geophys. 1. Znt. (1989) 99, 123-153
Fault plane solutions of earthquakes and active tectonics of the
Tibetan Plateau and its margins
Peter Molnar* and H&ne Lyon-Caent
* Department of Earth, Atmospheric and Planetary Sciences, Massachusetts Institute of Technology, Cambridge, MA 02139, USA (Permanent
address); Laboratoire de Gkophysique Znterne et Tectonophysique (CNRS U.A. 733), Institut de Recherches Interdisciplinaires de Ge'ologie et
Me'canique, Universitk Joseph Fourier, BP 53X, 38041 Grenoble Ce'dex, France and Department of Earth Sciences, Oxford University, Parks
Road, Oxford OX1 3PR, U K
t Laboratoire de Sismologie, Institut de Physique du Globe, 4, Place Jussieu, 75230 Paris Ce'dex 05, France
Accepted 1989 April 10. Received 1989 April 10; in original form 1988 November 26
SUMMARY
Fault plane solutions of earthquakes within and on the margins of the Tibetan Plateau show
diverse styles of faulting and deformation, with thrust faulting and crustal shortening normal
to the margins of the plateau and with normal and strike-slip faulting resulting in roughly
east-west crustal extension within the plateau. The direction of overthrusting of the Himalaya
onto the Indian Shield is radially outward, varying from southwest in the western Himalaya to
south-southeast in the east. Assuming that the Indian Shield behaves rigidly, this requires a
west-northwest divergence of western Tibet from southeastern Tibet at a rate of 1 8 f
9 mm yr-', comparable with the rate of convergence at the Himalaya. Fault plane solutions of
earthquakes in the southern portion of the Tibet Plateau consistently show large components
of normal faulting on roughly north-striking planes and corroborate such extension. Within
the high plateau, where elevations exceed 5000m, normal and strike-slip faulting occur so
that an overall east-southeast-west-northwest extension of the region (at about 10 mm yr-')
is partitioned into roughly equal parts of crustal thinning and north-northeast-southsouthwest crustal shortening (about 5 mm yr-I). In general, strike-slip faulting characterizes
solutions for earthquakes within eastern Tibet, where mean elevations drop below 45005000 m, but the orientations of the strike-slip faults vary across the region. In central Tibet,
left-lateral slip occurs on planes trending roughly northeast, but for earthquakes farther east,
the orientations of that plane become progressively east-west and then southeast. This
variation in orientation implies a rotation of material along curved left-lateral shear zones.
Thus, the eastward extrusion of Tibet appears to be facilitated not only by rapid left-lateral
shear, but also by large clockwise rotations of the material in eastern Tibet. The rate of
eastward extrusion of material in eastern Tibet, relative to the Tarim Basin to its north, is
roughly 30-40 mm yr-'. Fault plane solutions of earthquakes in the northern and eastern
margins of Tibet show large components of thrust faulting, with the P-axes, oriented radially
outward from the plateau and approximately perpendicular to the regional topographic
contours of the plateau. The orientation of this crustal shortening is northeast-southwest on
the northeastern margin, east-west on the eastern margin, and northwest-southeast in the
Longmenshan on the southeastern margin. Thus, at least some of the extrusion of eastern
Tibet out of India's northward path into Asia is absorbed by crustal shortening on the margins
of the plateau. The variation from normal faulting in the high Tibetan Plateau, where
elevations exceed 5000 m, to dominantly strike-slip faulting farther east where elevations are
lower, and then to thrust faulting on the margins of the plateau, where elevations drop below
3000 m, surely results, at least in part, from a decrease in the value of the vertical stress: the
magnitude of the east-west compressive stress need not vary across the plateau.
Key words: active tectonics, earthquakes, fault plane solutions, Tibetan Plateau
1
INTRODUCTION
Crustal shortening within the ancient southern margin of
Asia, coupled with some underthrusting of India beneath
southern Tibet and of the Tarim Basin beneath northern
Tibet have led to crustal thickening and the creation of a
deep crustal root that buoys
- up
- the Tibetan Plateau. This
plateau is the largest and most spectacular topographic
manifestation of the collision and subsequent penetration of
India into the rest of Asia (Fig. 1) and constitutes a focal
123
124
P. Molnar and H . Lyon-Caen
X
%
4
..
...;....
/:
J.
-**
f
Figure 1. Simplified topographic map of Asia, showing areas higher than 5000 m (in black) and than 2500 m (dotted line) and the region that
we studied outlined in black (from Cauet er al. 1957).
point for understanding how India's penetration affects the
tectonics of the rest of Asia. Because of the limited amount
of geologic work in Tibet, however, the nature, distribution,
amount and timing of the crustal shortening are poorly
constrained. Nevertheless, crustal thickening is clearly at
present a very minor process within the high plateau, if it
has not stopped altogether. Field investigations of recent
and Quaternary structures, analyses of Landsat imagery,
and fault plane solutions of earthquakes concur to show that
the active tectonics is characterized by roughly east-west
crustal extension, manifested by normal and strike-slip
faulting (e.g. Armijo et al. 1986; Molnar & Chen, 1983;
Molnar & Tapponnier 1975, 1978; Ni & York 1978; Rothery
& Drury, 1984; Tapponnier et al. 1981).
India's penetration has not stopped, and the convergence
in Asia is presently absorbed in part by crustal shortening on
the northern and southern margins of Tibet and farther
north in the Tien Shan and in part by eastward extrusion of
crust out of India's northward path (e.g. Armijo,
Tapponnier & Han Tonglin 1989; Kidd & Molnar, 1988;
Molnar & Deng Qidong 1984; Molnar & Tapponnier 1975;
Molnar, et al. 1987; Tapponnier et al. 1982; Tapponnier,
Peltzer & Armijo 1986). Estimates of rates of crustal
shortening in the different ranges of Asia and of rates of slip
on major faults corroborate the inference that significant
fractions of India's convergence are absorbed by each of
these processes (e.g. Armijo et al. 1986, 1989; Molnar &
Deng Quidong 1984; Molnar et al. 1987; Tapponnier et al.
1986), but both the kinematics and the total amount of the
extrusion remain poorly defined.
A quantitative description of the kinematics of this
extrusion is a prerequisite for a quantitative understanding
of the physical properties and the dynamic aspects
controlling this process. In a first approximation, Armijo et
al. (1989; Tapponnier et al. 1986) treated Tibet as a rigid
block in order to estimate an average rate for its extrusion,
but obviously internal deformation within Tibet could
modify that estimate. Our purposes here are, first, to offer a
refined image of the kinematics of this extrusion, which
takes into account the pervasive deformation of eastern
Tibet, and, second, to estimate rates at which the various
margins of Tibet move with respect to one another.
Active deformation of eastern Tibet is not neglibible.
Left-lateral slip at rates of at least l ~ m m y r - ' seem to
characterize the displacement associated with three major
fault systems, the Altyn Tagh fault along the northern
margin of Tibet, and the Kunlun and Xianshuihe faults that
cross eastern Tibet (Fig. 2). Peltzer (1987) recognized
consistent offsets of 200-350m of stream valleys along the
Altyn Tagh fault, which almost surely reflect slip since the
last glaciation and therefore at a rate of 20mmyr-' or
more. Kidd & Molnar (1988) estimated an average rate of
slip during Quaternary time on the major strand of the
Kunlun fault system of 13mmyrY' (between 10 and
F i 2. Map of Tibetan plateau showing the 1000, 3000 and 5000 m contours, major fault systems, epicenters of earthquakes for which fault
plane solutions have been determined, and the style of faulting associated with each. Lower hemisphere diagrams of focal spheres for each
show fault plane solutions; darkened quadrants include compressional P-wave first motions, and open quadrants include dilatational first
motions. Dates above each allow them to be identified in Tables 1 and 2, where source parameters are listed (for instance, 62 521 indicates
1962 May 21). Lines through or surrounding epicenters summarize the tectonic inferences for the various earthquakes: parallel lines give the
sense of slip for strike-slip faulting; diverging arrows give the orientation of the T-axis, and therefore of crustal extension, for an earthquake
showing largely normal faulting; a line through the epicenter without arrows gives the orientation of the P-axis, and therefore of crustal
shortening, for an earthquake showing thrust faulting, but for which it is difficult to know which nodal plane is the fault plane; and a single
arrow emanating from the epicenter gives the direction of overthrusting for earthquakes for which we think we can identify the fault plane.
The Tibetan Plateau and Microfiche GJI 99/1,2
125
126
P. Molnar and H . Lyon-Caen
20 mm yr-') and an average rate for Holocene time on the
minor strand also of about 10mmyr-' (between 5 and
20 mm yr-l). Abundant geologic evidence (Allen et al.
1989; Tang Rongchang et al. 1984), fault plane solutions of
two major earthquakes (Tapponnier & Molnar 1977; Zhou
Huilan, Liu & Kanamori 1983; Zhou Huilan, Allan &
Kanamori 1983) and surface faulting associated with other
major earthquakes (e.g. Molnar & Deng Qidong 1984)
indicate rapid left-lateral strike-slip movement on the
Xianshuihe fault, also at a rate of 10 to perhaps 20 mm yr-'
(Allen et al. 1989; Molnar & Deng Qidong 19841.
Deformation of eastern Tibet cannot be ascribed solely to
slip on these major fault systems. Left-lateral slip on two
east-northeast trending faults in the Nan Shan indicate rates
of slip of about 5 mm yr-' (Peltzer 1987; Peltzer et al. 1987),
and there may be other important faults. Left-lateral slip at
about 8 f 2 mm yr-l on the east-southeast trending Haiyuan
fault (Burchfiel et al. 1989a; Zhang Peizhen et al. 1988), on
the northeast margin of the Tibetan Plateau (Fig. 2), seems
to describe the most rapid, but certainly not the only, style
of deformation in this area. Moreover, left-lateral
strike-slip faulting was measured for 75 km along a fault
south of and parallel to the Xianshuihe fault following the
Litang earthquake of 1948 (Molnar & Deng Qidong 1984).
Thus the entire region probably is pervaded by left-lateral
shear.
While left-lateral shear occurs within eastern Tibet,
right-lateral shear on northwesterly planes occurs in western
Tibet. Peive et al. (1964) inferred 250 km of right-lateral slip
on the Karakorum fault (Fig. 2), a fault clearly visible on the
Landsat imagery (Molnar & Tapponnier 1978). Fault plane
solutions of earthquakes northeast of it are consistent with
right-lateral slip on parallel northwest-striking planes (see
below).
While normal and strike-slip faulting occur within Tibet,
the margins of the plateau are undergoing crustal
shortening. Thrust faulting on planes dipping northward
beneath the Himalaya has built this range on the southern
margin of Tibet. Along at least part of the northern margin
of Tibet, thrust faulting occurs on roughly easterly trending
planes (Molnar et al. 1987). Folding with northwesterly
trending axes and thrust faulting on northwesterly trending
planes is prevalent along the northeastern margin of Tibet
(Deng Qidong et al. 1984; Tapponnier & Molnar 1977;
Zhang Peizhen et al. 1989a,b). Finally, northwest-southeast
crustal shortening in the Longmenshan, along the
southeastern margin of the Tibetan Plateau seems to be the
primary style of deformation there (Tapponnier & Molnar
1977).
Faulting associated with earthquakes provides another
direct measure of the style of active deformation. If one can
identify the nodal plane of a fault plane solution that was
the plane of rupture, then the sense of slip on that plane can
be determined from the fault plane solution. Regardless of
whether that non-uniqueness can be resolved, however, the
orientations of the P-, B- and T-axes define the orientations
of the three principal strains accumulated by deformation
associated with the earthquake (Kostrov 1974). Fault plane
solutions corroborate the image of pervasive normal and
conjugate strike-slip faulting in the high parts of Tibet,
left-lateral shear within eastern Tibet, and crustal shortening
on the margins of the plateau.
Fault plane solutions alone are insufficient to allow an
evaluation of the strain field, or more importantly, of the
deformation gradient tensor (e.g. Malverne 1969; McKenzie
& Jackson 1983), but, in principle, this can be done with a
summation of seismic moment tensors
Mi, = Mo(nibj + nibj),
where Mo = M u = scalar seismic moment, in which p is the
shear modulus, A is the area of the rupture and u is the
average slip on the fault; n, is the unit vector normal to
the fault plane and b, is the unit vector in the direction of
the slip vector. Kostrov (1974) showed that the average
strain e,, due to slip on faults of different orientations is
given by
eii = C Mij12pV,
where the summation is over the number of earthquakes,
and V is the volume of the region. If one considers
earthquakes that occurred in a well-defined period, then the
average strain rate during that period can be estimated by
dividing the strain by the duration. Although we estimated
seismic moments for each earthquake, and we discuss briefly
the average strain rates implied by them, the small number
of sufficiently large earthquakes with well-determined
seismic moments does not allow their summation to provide
a very useful constraint on the magnitudes of the strain rates
(see also Ekstrom & England 1989). These analyses merely
offer additional corroboration of the orientation of strain
inferred from the fault plane solutions.
We use the fault plane solutions of earthquakes to refine
the qualitative image of the strain field of Tibet given above.
With that description and with measured Quaternary and
Holocene rates of slip on major strike-slip faults, we
analyze crudely the budget of crustal extension in Tibet and
crustal shortening on its margins. Toward these ends we
studied 38 earthquakes (Table l), which occurred between
1962 and 1986 within or on the margins of Tibet, using first
motions of P-waves and comparisons of synthetic and
recorded P- and SH-waveforms recorded by the WorldWide Standardized Seismograph Network (WWSSN). This
list includes eight earthquakes studied by Tapponnier &
Molnar (1977) using only P-wave first motions and another
26 events also analyzed by Ekstrom (1987), Jackson &
Yielding (1983), Ni & Barazangi (1984), Petterson &
Doornbos (1987) or Zhou Huilan et al. (1983a). The list is
augmented by published parameters from 36 other
earthquakes (Table 2), nearly all of which are constrained
by the synthesis of body wave forms.
The bulk of our effort was expended in studying the 38
events listed in Table 1, but the lengthy and tedious
presentation of this analysis is best relegated to an Appendix
on microfiche. Thus, we begin with a short description of
the procedure used to analyze these earthquakes, giving a
few examples, and then we consider their interpretation.
2
PROCEDURE
To constrain the fault plane solutions we used two
approaches. First, we relied on first motions of P-waves.
Second, we synthesized shapes and amplitudes of P- and
SH-waveforms and compared them with those recorded by
stations at distances between 30" and 80°, for which signals
The Tibetan Plateau and Microfiche GJI 99/1,2
127
Table 1. Source parameters of earthquakes studied.
Date
hr
Lat
Long
Depth
(ON)
(OE)
(W
95.73
96.44
95.50
103.03
98.67
95.06
96.47
96.42
104.02
72.91
106.32
90.97
95.73
88.39
82.28
81.75
97.26
80.3 1
88.65
95.60
88.55
81.21
81.09
75.75
75.80
82.14
101.15
91.42
73.60
82.28
99.92
62 May 21
63 Apr 19
64 Mar 16
70 Feb 24
71 Mar 24
71 Apr03
72 Aug 30
72 Aug 30
73 Aug 11
74 Dec 28
76 Sep 22
77 Jan 01
77 Jan 19
77 Nov 18
78 Apr04
78 Jul 31
79 Mar 29
79 May 20
80 Feb 22
80 Jun 01
80 Jun 24
80 Jul 29
80 Jul 29
80 Aug 23
80 Aug 23
80Oct 07
81 Jan23
81 Jun09
81 Sep 12
82 Jan 23
82 Jun 15
12
07
01
02
13
04
15
18
07
12
20
21
00
05
00
11
07
22
03
06
07
12
14
21
21
09
21
22
07
17
23
37.13
35.53
36.95
30.58
35.46
32.26
36.72
36.60
33.00
35.06
40.02
38.19
37.02
32.69
33.03
35.38
32.44
30.03
30.55
38.91
33.00
29.34
29.63
32.96
32.90
35.62
30.89
34.51
35.68
31.68
31.85
85 May 20
86 Apr 26
86 Jun 20
86 Jul 06
86 Jul 16
86 Aug 20
86 Aug 26
15
07
17
19
22
21
09
35.56 87.20
32.13 76.37
3 1.24 86.85
34.42 80.16
30.48 78.19
34.57 91.63
37.72 101.50
* Fault Plane Solution is not unique.
11 f 6
10f4
10 f 3
7f3
7f3
9f3
15 f 5
19 f 6
4 + 41- 2
12 f 3
8f3
8f3
14f4
11 f 4
11 f 4
6f4
12 f 3
16f3
6f3
12f4
11 f 3
14 f 3
18 + 31- 5
14 f 3
13 + 51- 2
4 + 61- 2
7f3
9f3
7f2
9f3
9f3
10f4
8f3
13 f 3
9f3
5 +5/- 3
13 f 3
11 f 5
7f3
Strike
(“1
285 f 20
277 f 15
70 f 15
41 + 161- 10
283 f 12
260 f 10
90 f 30
91 f 30
326 f 10
156 f 20
230 + 15
288 f 25
305 + 201- 30
236 f 15
327 f 15
236 f 10
270 + 101- 5
109 f 10
188 + 201- 10
128 f 10
71 f 10
95 + 251- 10
113 + 101- 15
130 + 151- 20
140 f 20
186 f 20
322 f 12
86 f 10
138 f 25
210 + 71- 12
55 f 20
1 0 0 f 10
234 + 121- 8
143 + 151- 10
138 f 10
248 f 15
125 + 101- 15
261 f 10
331 f 10
Dip
Rake
(“1
(“>
39f 5
8 0 f 10
77 f 10
58f 8
74 + 101- 12
79 f 10
62 f 15
58 f 15
85 f 15
44f 4
75 f 10
36f 4
38 f 5
68 + 51- 10
78 f 10
77 + 51- 10
84 + 61- 4
77 + 71- 5
48 + 61- 3
53 + 71- 5
75 + 151- 25
61 + 101- 5
65 f 4
80 + 201- 15
85 f 10
40 + 101- 8
8 0 f 10
83 + 51- 13
42 f 7
68 + 71- 10
7 0 f 10
81- 10
75 -I77 + 81- 10
8 4 f 10
78 f 7
51 f 2 0
82 f 8
88 + 12/- 10
59 + 111- 8
74f 25
350 f 10
50 + 701- 20
53 + 301- 20
5f12
355 + 201- 10
60 + 201- 45
38 + 501- 25
10 + 101- 15
77 + 81- 12
249 f 30
82 f 25
75 f 15
331 f 10
196 + 101- 8
352 + 151- 20
355 + 51- 10
100 + 201- 30
276 + 15/- 25
48 f 12
345 f 10
88 + 201- 35
92 + 251- 15
100 f 30
90 f 30
283 + 201- 30
5 f 12
354 + 61- 12
104f 15
281 + 301- 15
270 f 30
5 f 10
3 f 10
105 + 151- 20
178 f 7
333 f 20
90 f 30
349 + 81- 20
107 + 151- 20
MO
1 0 1 7 ~ ~
98.
141.
1.
2.9
13.
6.0
1 .o
1.1
6.2
14.
1.3*
7.1
5 .O
33.
6.8
2.0
5.5
2.4
22.
1.5
3.4
1.1
43.
1 .o
1.2
5.0
62.
5.9
10.4
36.
1.4**
3.6**
4.0
1.3
11.
6.1
0.5
25.
7.6
Strike-slip solution cannot be eliminated.
** Multiple event with different solutions s e e m to be required.
were sufficiently strong that they could be digitized
accurately. In all cases, the ranges of solutions permitted by
each approach overlapped one another, so that parameters
could be found that agreed with virtually all reliable P-wave
first motions and that also yielded matches of synthetic and
recorded seismograms.
2.1 Analysis of waveforms
To use wave forms to determine source parameters, we used
McCaffrey & Abers’s (1988) version of Nabelek’s (1984)
inversion procedure, which seeks a minimum misfit in a
weighted least squares sense of the synthetic and recorded
wave forms (see also McCaffrey & Nabelek 1987). Synthetic
wave forms were generated for a point source, buried in a
half space with a P-wave velocity of 6.2 km s-l, an S-wave
velocity of 3.6 km s-l and a density of 2800 kg m-’. These
values were used to calculate reflection coefficients for
evaluating the relative amplitudes of P, pP and sP, and of
SH and sSH, and to calculate take-off angles that match the
ray parameters given by the Je&eys-Bullen P- and S-wave
travel-time tables. Amplitudes were adjusted for geometric
spreading, a simple function of epicentral distance
(Langston & Helmberger 1975), and for attenuation using
Futterman’s (1962) operator with t* = 1s for P and I* = 4
for SH, for all but a few earthquakes in the western
Himalaya. For these we used 0.7s and 2.0s in order to
improve the fits of especially short pulse widths, but the
inferred source parameters were not affected by using
different values of t*.
Seismograms were weighted according to the differences
in azimuths to other stations, so that stations clustered near
one another were given smaller weights than those of
isolated stations (McCaffrey & Abers 1988, p. A-13). Where
amplitudes of SH phases were notably larger than those of
128
P . Molnar and H . Lyon-Caen
Table 2. Source parameter* of earthquakes from eastern Tibet and its margins.
str
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
62May21
63Apr19
64Mar16
64Sep26
64Oct21
65 Jan 12
66M.r 6
66Jun27
66Aug 15
66Oct 14
6 6 h 16
67Fcb20
67Mar 14
67AugM
67AugM
70Feb 19
70Feb24
71Mar24
71 Apr 3
71 Mav3
71 Ma; 22
72 Jul 22
72Aug30
72Aug30
72 Seo 3
73F& 6
73 Feb 7
73Jul 14
73 Jul 14
73Auc 11
73Aui 16
73 Sep 8
74Mar24
74h28
75 Jan 19
75 A p 2 8
75May 5
75 May 19
75Jun 4
75 Jul 19
75 Jul 29
76Aug 16
76Aug21
76Augn
76 S m 2 2
77 I$ 1
77 Jan 19
77Nov 18
78Apr 4
78 Ju131
79Mu29
79May20
791un 19
80Feb22
80Jun 1
8OJun24
8OJul 29
8Olul 29
8OAug23
80Auc23
80Oc;7
8ONav 19
81 Jan 23
81Jun 9
81 S a 1 2
82JL23
82Jun15
85May 20
86Ap26
86Jun20
86 Jul 6
861d 16
86Augu)
86Aug26
37.13
35.53
36.95
29.96
28.04
27.40
31.49
29.62
28.67
36.45
29.62
33.63
28.41
31.61
31.57
27.40
30.58
35.46
32.26
30.79
32.39
31.43
36.72
36.60
35.94
31.33
31.50
35.18
35.26
33.00
33.24
33.29
27.73
35.06
32.39
35.82
33.09
35.16
35.87
31.92
32.56
32.72
32.61
32.48
40.02
38.14
37.02
32.69
32.98
35.47
32.44
30.03
26.74
30.55
38.91
33.00
29.34
29.63
32.96
32.90
35.62
27.40
30.89
34.51
35.68
31.68
31.85
35.56
32.13
31.24
34.42
30.48
34.57
37.72
95.73
96.44
95.50
80.46
93.75
87.84
80.50
80.83
78.93
87.43
80.79
75.33
94.29
100.26
100.33
93.96
103.03
98.67
95.06
84.33
92.12
91.49
96.47
96.42
73.33
100.49
100.33
86.48
86.60
104.02
86.84
86.82
86.11
72.91
78.50
79.92
92.92
80.80
79.85
78.61
78.46
104.09
104.15
104.10
106.32
91.00
95.69
88.39
82,26
82.00
97.26
80.31
87.48
88.65
95.60
88.55
81.21
81.09
75.75
75.80
82.14
88.80
101.15
91.42
73.60
82.28
99.92
11
10
10
18
15
15
8
15
20
8
12
10
15
8
10
10
7
7
9
8
8
8
15
19
12
10
10
6
7
4
8
9
16
12
9
7
7
8
9
6
8
12
5
8
8
8
14
11
11
6
12
16
20
6
12
I1
14
18
14
13
4
44
7
9
7
9
9
10
87.20 8
76.37 13
86.85 9
80.16 5
78.19 1 3
91.63 11
101.50 7
285
277
70
310
265
270
0
277
132
25
290
341
273
245
233
257
276
283
260
190
58
212
90
91
341
305
210
81
37
326
160
118
275
354
0
169
250
248
180
180
210
165
215
165
230
288
305
236
327
236
270
251
75
188
128
71
279
288
265
320
186
214
322
86
138
210
55
100
234
254
138
248
305
261
331
PlSms
dip
P-axis
pl
dip rake
str
rake
IU
39
125
9
325
130
85
90
180
119
312
205
103
190
160
90
90
90
270
206
233
189
220
175
180
100
202
222
295
200
60
183
238
143
167
156
238
216
74
350
50
90
90
90
270
70
270
270
90
105
90
290
270
90
134
53
80
42
67
87
75
45
65
59
24
110 66
136 38
10
93 80
174 48
45
53 40
50
77 8 5
5
41 58
47
74 5
192 85
7 9 355 3 5 1 85
58 270 10 32
90 3
328 87
65 343 309 75
321 40
62 60
58 38
339 59
55 105 136 38
87 181 215 89
60 270 30 30
60 325 190 60
68 304 156 40
235 80
85 10
55 205 55 70
60 199 18 74
90 95 88
2
47 102 157 44
50 270 180 40
62 211 63 63
78 346 343 76
46
66 310 4
62 239 52 41
50 235 47 51
55 270 30 35
63 40
54 55
35 30
60 90
65 40
55 54
75 249 107 26
36 82
118 54
38 75
144 54
68 331 338 63
78 196 234 74
77 352 327 82
84 355 0 8 5
109 77
16 53
45 277 245 46
48 276 359 42
53 48
4
54
75 345 165 76
29 94
95 61
113 65
25 8 6
130 80
14 45
90
140 85
5
40 283 349 51
71 12
120 78
80 5
231 85
8 3 354 176 84
42 104 299 50
68 281 3 24
70 270 145 20
75 5
9
85
77 3
143 87
16 22
143 84
78 178 228 88
5 1 333 356 69
8
90 125 82
8 8 349 79 79
59 107 120 35
80
77
23
3
15
45
27
31
66
24
55
Sourcc parmeters of all enthquakes are conslrained. .L least in part. by
by an asterisk.
a BOIMOWS~~
et d. (19841
b Molnar and Chcn [I9831
c Molnar et d. (19771
d Zhou et d. [1983a]
e Jones et al. (19841
f Ni and Bar.ungi [I9841
b l r ( l m (19871
P, we halved the pre-assigned weights of SH phases, before
normalizing all weights to average to 1.0. In general, the
strikes, dips, and rakes inferred using different weights
differed by only a few degrees, and the depths by only
1-2 km.
Among the source parameters are the depth of focus and
three angles that describe the fault plane solution: the dip
-
270
270
90
70
90
251
270
90
53
164
191
270
100
180
206
133
142
70
357
270
215
216
L 75
322
329
90
77
270
328
192
215
314
304
270
146
90
149
3 23
96
193
173
201
35
60
170
120
45
349
100
205
348
193
186
12
335
185
75
270
26
206
205
43
23
120
28 8
305
288
114
204
226
196
191
192
225
100
151
263
263
132
196
88
92
72
101
100
90
259
160
170
187
78
245
270
I65
167
105
12
222
90
182
64
the
151
66
28
186
20 1
21 1
230
21 1
168
277
41
38
139
325
55
189
220
3
219
215
216
49
7
14
22
22
42
30
90
19
76
69
21
9
35
76
85
40
6
8
11
77
2
29
12
0
9
3
75
45
54
3
41
34
43
2
85
41
18
51
60
64
80
5
I5
6
55
9
8
36
20
I5
8
31
85
85
0
21
16
20
34
40
80
5
4
9
4
65
65
7
7
37
7
44
37
9
12
T-axis
B-axis
pl
az
pl
LZ
82
143
301
40
355
0
90
49
42
115
20
295
3
141
323
347
256
146
125
280
283
79
314
305
295
260
300
315
103
1Y1
78
0
43
68
48
60
0
69
14
21
69
75
55
2
5
50
59
15
4
13
2
6
61
48
75
298
5
1
15
0
16
11
110 9
71
5
334
90
116
297
310
292
1 I4
300
23
125
25
337
51
96
288
9
47
81
5
1
1
12
12
I
10
47
75
45
27
80
78
3
100 3
101 4
135
32
340
274
336
298
0
27
52
50
87
76
186
311
150
292
145
323
98
69
94
1
57
0
3
58
0
74
70
54
50
6
22
11
1
80
22
25
14
11
49
10
118 11
35 53
307 6
280 71
81
10
76
39
130
87
90
0
295
132
25
0
0
0
0
9
0
0
C*
110
0
B
152
93
51
53
77
63
355
12 a
0
a
a
a
b
a
b
a
14 d
0 d
0
IS
31
73
78
10
0
58
338
105
125
152
17
30
225
203
352
21 1
173
95
165
0
207
30
49
196
204
30
193
35
192
236
294
317
22
3
358
40
287
250
4
156
207
96
292
308
320
356
271
25
217
308
26
55
171
311
321
238
17
305
71
142
87
60
26
42
12
87
0
40
31
79
48
55
c*
b
b
b
a
d
d
b
b
b
b
O
a
9
O
49
72
36
27
26
O
43
b
b
b
b
b
b
b
e
O
e
44 e
20
5
9
54
70
75
82
10
5 f'
4
32
69
2
2
10
0
8
68 g
79
81
9
10
0
74
77
15
78
44
0
79
15
synthesis of body-wave form. except those denoted
and strike of one nodal plane and the rake of the slip vector
lying in that plane. The source time function is described by
the amplitudes of a series of overlapping symmetrical,
triangularly shaped pulses, the number and duration of
which we selected. The inversion routine yields amplitudes
for each triangular pulse. Both the shapes and the
amplitudes of the wave forms were used as constraints on
The Tibetan Plateau and Microfiche GJI 99/1,2
the solution, so that the scalar seismic moment was also
determined. Thus, a minimum of five parameters can be
examined by McCaffrey & Abers’s (1988) inversion routine;
focal depth, strike, dip, rake and seismic moment. For the
larger earthquakes, the relative amplitudes of separate
elements of the time function can also be estimated.
The inversion routine proved to be remarkably stable, in
the sense that estimates of the source parameters often
proved to be independent of one another and that the
removal of wave forms with small signal-to-noise ratios did
not lead to very different estimates of the source
parameters. Specifically, estimates of focal depths for a wide
range of both allowable and unacceptable orientations of
nodal planes generally lay within a few km of one another.
Similarly, if the focal depth was fixed to a value within
3-5 km of the value yielding the minimum misfit of recorded
and synthesized wave forms, the inversion routine invariably
returned values for the strike, dip and rake that were within
a few degrees of those associated with the minimum misfit.
The orientations of nodal planes were found to be similarly
independent of the duration of the source time function, but
in some cases we found the focal depth and the source time
function to be coupled slightly. To fit some sharp pulses with
depths greater than 10 km, short source time functions were
needed, but for shallower sources, such that the intervals
among P , p P , and sP, or between S and sS, are short,
source time functions longer than these intervals were
allowed. Finally, the values of the seismic moments clearly
depended on the durations of the source time functions,
with larger moments associated with longer assumed source
time functions for the same earthquakes. Because our
primary focus was on constraining the orientations of nodal
planes and, to a lesser extent, the focal depths, nearly all of
which are between 5 and 15km, we did not concern
ourselves much with the values of the seismic moments and
their uncertainties, except for the largest earthquake, that of
April 19 1963.
We used the inversion routine first as an expedient to
finding an acceptable (alias ‘best’) set of source parameters
and then to carry out numerical experiments in order to
place bounds on the allowable values of the various
parameters. Finding a ‘best’ solution is merely a prelude to
scientific inquiry, and in the absence of useful constraints on
the uncertainties in the ‘best’ fitting values of parameters,
‘best fits’ are rarely adequate, no matter how good the fits
are. Moreover, despite the frequent success of the inversion
routine in yielding source parameters that both agree with
P-wave first motions and are geologically sensible, in a
couple of cases, the inversion routine returned source
parameters that do not agree with P-wave first motions that
we consider reliable. Thus, the parameters in Table 1 are
not in all cases those that minimize in a least-squares sense
the misfit between recorded and synthetic seismograms.
Given that different, arbitrarily selected weightings of
different wave forms can affect somewhat the inferred
parameters, some arbitrariness must be introduced in the
procedure, and we chose to do this after estimating the
uncertainties of the parameters.
2.2 Assignment of uncertainties
Having found a set of acceptable source parameters for an
earthquake, we then used the inversion routine to carry out
129
experiments with the source parameters, varying them one
at a time and visually examining the quality of fits between
recorded and synthesized wave forms. We first examined the
source time function, inverting data for source time
functions of various durations. For most events, the sharp
pulses could be fit only with short time functions; the
inversion routine usually returned source time functions
with the first triangular pulse five times (or more) larger
than the others. We then fixed the depths, and inverted for
the other parameters; in general, the durations and shapes
of the pulses were visibly poorly matched for depths 3-5 km
different from the value found when the depth was left free.
Finally, we fixed one of the strike, the dip or the rake to a
series of values and inverted for all of the others. The wave
forms at most stations turned out to be mildly sensitive to
the particular source parameters used, and the degradation
in the root-mean-square misfit of recorded and synthesized
wave forms was, in general, due mostly to poor fits at a
small number of the stations whose records we could
digitize. From a visual examination of the fit of recorded
and synthesized wave forms, we assigned bounds to each of
these parameters. Thus, the uncertainties given in Table 1
were not derived statistically.
Figures 3, 4 and 5 illustrate the results of this procedure
for three representative earthquakes, one each with thrust,
strike-slip and normal faulting. Figures 3(a), 4(a) and 5(a)
show the fit of synthesized and recorded wave forms for the
inferred ‘best’ parameters and the P-wave first motions on
lower hemisphere projections of the focal sphere. Figures
3(b), 4(b) and 5(b) show the variations in such fits of a
subset of the wave forms for a series of poorly fitting source
parameters. The lengthy captions discuss the results in
detail, and Appendix B on microfiche presents similar
illustrations and discussions for all of the events that we
studied.
Although the wave forms from each earthquake presented
special conditions that forbid precise generalizations as to
what is necessary to ensure well-defined fault plane solutions
and focal depths, some simple generalities are possible.
First, for most events, focal depths are constrained largely
by P-waveforms, whose sharper pulses allow better
resolution of reflected phases than SH phases. For
strike-slip faulting on a steep plane, the wave forms do not,
in general, constrain the orientations of nodal planes better
than P-wave first motions, but there are two notable
exceptions. First, for some events, P-wave amplitudes are
too small to be digitized at more than one or two stations,
and reliable P-wave first motions are few, but numerous,
large and simple SH wave forms require dominant
strike-slip components. Second, a poor azimuthal coverage
does not permit P-wave first motions to resolve the dip of
one of the two steeply dipping nodal planes for the
earthquake of 1971 March 24, but SH-waveforms confirm
that both nodal planes dip steeply. For events with thrust or
normal faulting on planes dipping 25” to 65”, P-wave first
motions and P-waveforms rarely provide strong, direct
constraints on the orientations of either of the planes. Both
the initial motions of the SH phases and their shapes and
amplitudes turn out to be vital for constraining the
orientations of the nodal planes for such solutions.
Moreover, although large ranges in the values of the strike
and rake are permitted for some earthquakes characterized
130
P. Molnar and H . Lyon-Caen
21 May 1962
285/39/74/11/978
Fignre 3. Results of the analysis of the earthquake of 1962 May 21, showing thrust faulting. In (a), P-wave first motions and wave forms are
plotted with directions to stations plotted on lower hemisphere projections of the focal sphere. Compressional and dilatational first motions are
plotted as darkened and open circles. Surrounding the focal sphere, recorded wave forms are shown as continuous traces, and synthetic wave
forms are dashed. The parameters used to synthesize these wave forms are given below the date: strike, dip, rake (in degrees), focal depth (in
kilometers) and scalar seismic moment (in units of 10l6Nm). Waveforms are scaled as if recorded by an instrument with a magnification of
3OOO at a distance of 40".The corresponding amplitude scale (in pm) is shown by the vertical bar. The duration of the wave forms used in the
inversion procedure are delimited by vertical lines on the traces, and two time scales are shown: one for the source time function and the other
for the seismograms. Lines on the focal spheres define the nodal surfaces for direct P and SH phases. Only a fraction of the WWSSN had been
installed in 1962, when this event occurred, and only a small number of phases could be analysed. P-wave first motions (Tapponnier & Molnar
1977) require a large thrust component but leave the strikes of the planes unconstrained. The breadth of the signals requires a longer source
time function than was necessary for most events that we studied, and the double peak of the initial half cycle of most P phases requires that
the source time function consist of two pulses, but neither the duration nor the spacing and relative sizes of the two pulses are well constrained.
The inferred depth depends strongly on the duration of the source time function; for short source time functions of 2, 3 and 4 s, the inferred
depths would be 18, 16 and 14 km, respectively, but for long source time functions depths of as little as 7.5 km were found to yield fits
indistinguishably different from our best fitting values. Thus its uncertainty of 6 km is greater than usual. In (b), comparisons of synthetic and
recorded waveforms at a few stations are shown for several different fault plane solutions. The top traces are the same as those in (a), and the
others correspond to the parameters (strike, dip, rake, depth and scalar moment) written above the focal spheres. Again, solid and dashed
tines give recorded and synthetic waveforms, respectively. Notably poor fits are distinguished by the dark bars next to the wave forms. The
The Tibetan Plateau and Microfiche GJI 99/1,2
ATU Pw
RAB Pw
ATU SHw
COP SHw
ADE SHw
131
RAE SHw
(b)
Figure 3. (Continued.)
orientations of the nodal planes were found to be quite stable. The P-waveforms offer little constraint on the strike or dip, but SH-waveforms
are very sensitive to differences of 20" or more from the 'best' fitting strike of 285" and of 5" from the 'best' fitting dip of 39" of the northeasterly
dipping plane. For a strike of 305",for a relatively small dip of 34", or for nearly pure thrust faulting with a rake of 99". synthetic SH phases at
ATU and ADE are too large and too small, respectively. For the more nearly east-west strike of 265", the synthetic P and S H phases at ATU
are both too small, while the SH at COP is too large. For a relatively steep northeasterly dip of 44",the calculated amplitudes of P at ATU and
RAB mismatch those observed, and synthetic SH phases at ATU and RAB are too small. Finally, for a rake of 49", corresponding to a large
strike-slip component, the calculated amplitudes of P phases at ATU and COP (not shown), as well as those of SH at COP and RAB fit
poorly. Note that with these limits, the P-axis is constrained to lie within 6" of 206".
by large dip-slip components, their values turn out to be
coupled so that the orientatiop of the P- or of the T-axis is
more tightly constrained than the orientation of either
plane.
Having determined ranges of possible source parameters
using both P-wave first motions and wave forms of P and
SH phases, we used the overlap in the ranges to limit the
aIIowable range of each (TabIe 1).
2.3 Comparison with results of others
Many of the earthquakes listed in Table 1 have been studied
by other investigators (e.g. Ekstrom 1987; Jackson &
Yielding 1983; Ni & Barazangi 1984; Petterson & Doornbos
1987; Zhou Huilan et al. 1983a). Comparisons of specific
earthquakes are given in Appendix B. Here we give a more
general comparison not only to corroborate our results, but
also to confirm the reliability of Centroid Moment Tensor
(CMT) solutions (Dziewonski, Chou & Woodhouse 1981),
which rely on longer periods and more phases, but generally
from fewer stations, than we have used. Specifically, we
consider Ekstrom's (1987) improvements to the CMT
solutions. He added short-period signals to construct
broad-band records of P phases, which resolved focai depths
much more precisely than was possible with long-period
recordings alone. This is particularly important for the
Tibetan region, because focal depths of the CMT solutions
tend to be systematically overestimated. Accordingly, when
Ekstrom could not construct useful broad-band signals, he
fixed the depths at reasonable values (10-15 km), and
recalculated CMT solutions.
Appendix A shows plots of our parameters versus
132
P. Molnar and H . Lyon-Caen
Ekstrom's (1987) for the 27 earthquakes that we both
studied, and here we merely summarize the important
features. To compare fault plane solutions, we plotted
differences in strikes, in dips and in rakes between our
estimates and those of the similar planes of his best fitting
double-couple solutions. Of 81 differences, 55 (68 per cent)
lie within the uncertainties that we assigned, and 76 (94 per
cent) lie within twice those uncertainties. Although our
uncertainties were not determined in a statistically rigorous
fashion, this comparison suggests that on the average our
estimated uncertainties are between one and two standard
deviations. Ekstrom's and our fault plane solutions agree
best for those events since 1978, with seismic moments in
excess of 2 x 10'' Nm, and for which he could use two or
more broad-band P phases. In only three cases do
Ekstrom's depths lie outside the ranges that we determined.
All but two of our seismic moments are within a factor of 2
of Ekstrom's, and in nearly all cases, they differ by less than
a factor of 1.3. His use of longer periods, which are a more
stable indicator of the moment, probably make his scalar
moments more reliable than ours.
As one last comparison, we show the results of our
analysis of one particularly troublesome earthquake (1982
June 15), which seems to consist of two subevents with
23 January 1982
2 10/68/28i /9/355
Rw
:
M--
Figore 4. Results of the analysis of the earthquake of 1982 January 23, showing normal faulting. Layout as in Fig. 3. This earthquake
constitutes a good example of a well recorded event. (a) Both the P-wave first motions and their relative amplitudes require a large component
of normal faulting with one nodal plane dipping steeply northwest. The orientation of the other plane is constrained largely by SH wave forms.
The poor fit of P at RIV may reflect an error in the magnification written on the seismogram, and those for SH at MUN and ADE are
probably due to their proximity to a nodal surface. The very small initial motions of P-waves at stations to the northwest, e.g. KEV in (b), are
comparable with the noise at most stations and could not be fit in detail without a more complex source. (b) A strike of the northwest dipping
plane 7" more northeasterly than its 'best' value introduces a marked degradation in the fit of SH phases recorded in Europe (ATU, IST,
AQU), but a larger difference in the other sense, 12" more northerly, which is permitted by the P-wave first motions, yields a smaller
degradation in the fit at most of these stations. The relatively steep dip of 75". 7" greater than the 'best' value of 68", is ruled out by the
polarities and amplitudes of P-waves in Africa (BUL) and Europe (KEV), but a relatively gentle dip of 58", is demonstrably poor only by its
misfitting of the backswing of the P phases to the west (e.g. BUL and KEV) and by the poor match of SH at ATU. This solution, however,
yields an improved fit of SH at SEO, denoted by the star. The rake, and hence the strike and dip of the gently dipping plane, are constrained
largely by the SH phases with calculated amplitudes of SH in Europe (IST and AQU) and at SEO being too large for a rake of 311", but not
301". and with those for sSH in Europe being too large for a rake of 266".
The Tibetan Plateau and Microfiche GJI 99/1,2
BUL Pw
210 68/281/9
355
211 75/2a6/a
335
ATU SHw
IST SHw
AQU SHw
133
SEO SHw
L-
Figure 4. (Continued.)
different solutions. As discussed in the caption of Fig. 6,
Ekstrom's (1987) moment tensor can be decomposed into
two subevents similar to those that we obtained.
The punchline of these comparisons is that the agreement
is sufficiently good that the tectonic implications discussed
below would be unchanged if we had not bothered to study
these 27 earthquakes and instead had relied on Ekstrom's
(1987) parameters. Thus, if one is interested in the general
properties of moderate-sized earthquakes in a particular
region, the CMT solutions will probably serve the
investigator well, but to know how well constrained the
solution of a particular earthquake is, it probably is
necessary to analyze it oneself.
3
RESULTS
The pattern of faulting in and around Tibet varies markedly
from one region to another (Fig. 2), and a discussion of the
results is facilitated by considering individual parts of the
Tibetan Plateau and its margins separately. Using Fig. 2 as a
reference, we begin with a brief discussion of the Himalaya,
focussing on the western part of the range, where several
earthquakes have occurred since the completion of most
published studies of earthquakes from the Himalaya. Next
we consider the high part of the plateau, between roughly
78"E and 90°E, where normal and strike-slip faulting are
common. We then discuss eastern Tibet, between 90"E and
100"E and between 30"N and 36"N, where strike-slip
faulting dominates. Finally, we consider the northern and
eastern margins of the plateau, where thrust faulting
prevails.
3.1 Thrust faulting in the Himalaya
Numerous studies of fault plane solutions of earthquakes in
the Himalaya show that in general one nodal plane strikes
parallel to the chain and dips steeply south or southwest and
the other dips gently beneath the range (Baranowski et al.
1984; Chandra 1978; Ekstrom 1987; Fitch 1970; Molnar,
Fitch & Wu 1973; Molnar et al. 1977; Ni & Barazangi 1984;
Rastogi 1974). Given the preponderance of thrust faulting
on gently northward-dipping planes in the Himalaya, the
nodal plane dipping beneath the range is commonly
assumed to be the fault plane. Most of the studies cited
above confined their attention to the area east of roughly
80"E, because the few events farther west showed a
somewhat different pattern: thrust faulting on planes
dipping 35" to 55", commonly with the steeper plane dipping
northeast (e.g. Baranowski et al. 1984; Jackson & Yielding
1983). Thus, the earthquakes east of 80"E seem to imply a
134
P. Molnar and H.Lyon-Caen
continued gentle underthrusting of the Indian Shield
beneath the Himalaya (see also Molnar & Chen 1982;
Seeber, Armbuster & Quittmeyer 1981), but those farther
west seem to reflect crustal shortening within the Himalaya,
probably by more distributed deformation in the upper
crust.
The strikes of the steeply dipping nodal planes for
earthquakes east of 80"E are oriented nearly parallel to the
local trend of the range and consequently vary along the
range from about N150"E near 80"E to N85"E near 95"E
(Figs 2 and 7). Thus, the orientation of underthrusting
seems to vary along the chain. Assuming India to behave as
a rigid plate, this divergence requires an east-west extension
of southern Tibet, at about half the rate of underthrusting of
India beneath the Himalaya (Armijo et al. 1986; Baranowski
er al. 1984; Molnar & Chen 1982). The lack of earthquakes
showing underthrusting of western India beneath the
Himalaya and the poor constraints on the strikes of the
planes dipping 35" to 55", however, made it risky, if not
impossible, to examine the direction of underthrusting or
convergence west of roughly 80"E.
Since the publication of most of the studies cited above,
four moderate-sized earthquakes occurred between 75.5"E
and 78.5"E, and Ekstrom (1987) obtained solutions
consistent with northeastward underthrusting of India
beneath the Himalaya. Motivated by his work, but unaware
of its reliability, we also studied these four earthquakes, as
well as three others near 80"E (1979 May 20 and 1980 July
29a,b) and the two farther west for which SH phases had not
been used as constraints (1974 December 28 and 1981
September 12).
The results of this analysis, like Ekstrom's (1987), suggest
that the rough perpendicularity of slip vectors to the local
trend of the range, observed east of 80"E, continues to the
24 June 1980
71 /75/345/11/34
6
45 STF
405
(a)
m
e 5. Results of the analysis of the earthquake of 1980 June 24, showing strike-slip faulting. Layout as in Fig. 3. P-wave first motions
imply a large strike-slip component, but enough of them are dubious that they alone are not very convincing. Only at one station (UME) was
the P phase large enough to digitize, and its first motion is not clear. Large SH phases at stations to the west and northwest, however,
demonstrate a large strike-slip component. The SH at BAG* was not used because the residual is very large (15 s), and we are uncertain that
this is the SH phase. HLW and JER lie very close to the nodal surface of SH, and the first large signal at these stations is sSH. The direct SH
phase becomes increasingly stronger at successively larger azimuths from ATU to COP and NUR, and the gradient in these amplitudes also
places a strong constraint on the solution. Thus a more east-west strike (81") of the easterly trending nodal plane or a steep dip (90")of it
cause a more rapid gradient in the amplitudes of SH at ATU, COP and NUR than for the 'best' fitting value. A more northeasterly strike
(61"), or a gentle dip (50") yield SH nodal surfaces that pass too far from HLW and JER. The amplitude of P at UME is sensitive to the strike
and dip of the northerly trending plane, and hence to the rake on the easterly plane. Although a large normal component of slip is tolerated by
the wave forms, the azimuth of the T-axis is tightly constrained to be 298" f 11".
The Tibetan Plateau and Microfiche GJI 99/1,2
UME Pw
HLW S H w
JER SHw
ATU S H w
COP S H w
135
NUR S H w
Figure 5. (Continued.)
west. The earthquakes of 1980 August 23 at 32.96"N,
75.75"E and 32.90°N, 75.8OoE, of 1986 April 26 at 32.13"N,
76.37"E, and of 1986 July 16 at 30.48"N, 78.19"E are small,
and the fault plane solutions are not as well constrained as
those for other earthquakes in the Himalaya. In fact, for the
first three of these, we cannot eliminate the possibility that
the steeply dipping nodal plane dips northeast, so that
normal faulting occurred. Nevertheless, if these planes do
dip steeply southwest, as we and Ekstrom (1987) suppose,
then the direction of underthrusting is roughly N45"E
(*lo"), roughly perpendicular to the local trend of the
range.
The mean orientation of the P-axes of the four
earthquakes west of 75.5"E is N60"E (f20"), also
perpendicular to the trend of the range. The orientations of
the P-axes, or the nodal planes, of these earthquakes,
however, are less well constrained than those east of 80"E.
Moreover, our solutions differ from those determined by
others, and consequently, there is some reason to doubt the
inference that the radially oriented crustal shortening
applies to the region west of 75.5"E. Specifically, although
wave forms for the earthquake of 1981 September 12 cannot
be matched well at all stations (Fig. B29), the strikes of our
nodal planes differ by 30" from Ekstrom's (1987). Similarly,
the strikes of nodal planes obtained by Jackson & Yielding
(1983) using P-waveforms alone also differ by about 30"
from those of two events that we studied (1974 December 28
and 1981 September 12) and of another (1972 September 3),
analyzed by Baranowski et al. (1984), using both P- and
SH-waveforms.
A plot of azimuths of slip vectors of earthquakes east of
75"E and P-axes to the west versus the longitudes of the
epicenters (Fig. 7) shows a clear monotonic variation,
consistent with the variation in the strike of the range. The
difference in azimuths between 74"E and 94"E is 65" f 20°,
but note that the exclusion of the four earthquakes with
P-axes instead of slip vectors does not change this inference,
provided the three events near 76"E reflect thrust, and not
normal, faulting.
If India behaves as a rigid plate, this variation in the
direction of overthrusting, or crustal shortening, in the
Himalaya not only requires roughly east-west extension in
southern Tibet, but it yields a quantitative relationship
between the rate of convergence at the Himalaya and the
rate of extension in southern Tibet (Armijo et al. 1986;
Baranowski et al. 1984; Molnar & Chen 1982). If the rate of
convergence of India toward the Himalaya vH were constant
along the range, then, ignoring curvature of the Earth and
136
P. Molnar and H . Lyon-Caen
15 June 1982
15 June 1982
98/85/355/10/30
55/70/270/9/14
qq
0 0
N
O
I,
I
9"-7,
STF
s2-
SP
6"'Qs
LP
(a)
r - T 2 S
SP
(b)
Figure 6. The analysis of the relatively complex, and small, earthquake of 1982 June 15. Layout as in Fig. 3(a). This earthquake frustrated us
for a long time, in part because of its small size and low signal-to-noise ratios, but also because of its complexity. P-wave first motions recorded
by both long- and short-period instruments suggest a large component of normal faulting with planes dipping southeast and northwest. Such a
solution yields a satisfactory match of long-period P phases at J E R and UME and short-period P phases at NAI and AQU, but it fails to give
the second compressional pulse for P at BAG or to match the amplitudes of SH at all stations (a). Beginning from this solution, the inversion
routine always returned a solution with a large strike-slip component. Both the shape of the P-waveform at BAG and the relative amplitudes
of P and SH are suggestive of strike-slip faulting. The best fitting solution, in a least squares sense, shows nearly pure strike-slip displacement
with left-lateral slip on the east-southeast trending nodal plane (b). This solution fits the SH-waveforms quite well and gives a larger second
than first pulse in the P phase at BAG, but it violates P-wave first motions at stations to the northwest and yields unacceptable misfits to P
phases at JER and UME. Both solutions include one SH nodal surface near all of the stations with digitized SH phases, but the amplitudes of
SH are much larger than those for P only for the strike-slip solution. Thus the P-wave forms matched well by one solution are matched poorly
by the other, and it ought not be surprising that a combination of these solutions yields a rather good fit at all stations. To obtain such a
solution (c), we fixed the strike, dip, rake and depth of the subevent with normal faulting, and we inverted simultaneously for its source time
function and seismic moment and for all of the source parameters for a second subevent, which turned out to be the larger and characterized
by dominantly strike-slip faulting. Two additional facts lend credence to the inferred solutions of both subevents. First, both are consistent
with the regional tectonics of this area. Normal faulting with such an orientation characterizes the solutions of three nearby earthquakes (1967
August 30a,b and 1973 February 7), and strike-slip faulting with a roughly similar orientation can be seen on the Landsat imagery (Tapponnier
& Molnar 1977). Second, Ekstrom (1987) studied this event using long-period waves of the Global Digital Network and two broad-band
recordings of P phases. His best fitting double-couple is very similar to our solution with strike-slip faulting, but the polarities of his two
broad-band P phases, recorded in Europe, are blatantly violated by this solution. His solution for the full moment tensor, which fits these two
wave forms well, however, shows a combination of normal and strike-slip faulting with orientations of both pairs of planes similar to ours. In
Appendix B, we discuss the logic of the uncertainties that we assigned.
The Tibetan Plateau and Microfiche GJI 99/1,2
15 June 1982
1 :55/70/270/9/14
2:100/75/5/10/36
Figure 6. (Continued.)
using the law of cosines, we obtain a rate of displacement
between western and southeastern Tibet uT of
U T = VH(2
- 2 cos @)"*.
(3)
For @=65"f20", uT=vH (1 4~0.3). The azimuth of
relative displacement would be roughly N110-115"E. In a
later section we consider the plausible range of uH in order
to discuss the range of uT, and we compare the rate and
direction of vT with those estimated by Armijo et al. (1986)
from geologic studies in southern Tibet.
3.2 Normal and strike-slip faulting in southern and
central Tibet
Fault plane solutions of 23 earthquakes north of the crest of
the Himalaya, between about 77"E and 90"E, and south of
the Tarim Basin show large components of normal and
strike-slip faulting (Fig. 2), with T-axes consistently
oriented between east-west and northwest-southeast, a
137
result that was well established by studies of fewer
earthquakes (Molnar & Chen 1983; Molnar & Tapponnier
1978; Ni & York 1978) and by analyses of Landsat imagery
(Armijo et al. 1986, 1989; Rothery & Drury 1984;
Tapponnier et al., 1981). Nodal planes of solutions showing
large components of normal faulting commonly strike within
30" of due north, and those with large components of
strike-slip faulting commonly strike roughly northeastsouthwest or northwest-southeast.
There is some geographical variation in the style of
faulting; the seven earthquakes that occurred closest to the
Himalaya (from west to east in Fig. 2: 1975 July 29, 1975
January 19, 1975 July 19, 1966 March 6, 1982 January 23,
1971 May 3 and 1980 February 22) show nearly pure normal
faulting, but farther north strike-slip and normal faulting
seem to be equally common. The apparent preponderance
of purely normal faulting in southern Tibet might be a
consequence of poorly constrained solutions of several
earthquakes, including the two largest, which consist of
multiple ruptures (Molnar & Chen 1983). Nevertheless, the
approximately east-west to east-southeast-west-northwest
azimuths of the T-axes (Fig. 8) accords with the inference
that the active tectonics of southern Tibet is dominated by
normal faulting along major northerly trending graben
systems (Armijo et al. 1986; Tapponnier et al. 1981).
Unfortunately, a sum of the moment tensors of these seven
earthquakes is not very revealing, because one event (1975
January 19) contributes 80 per cent of the corresponding
strain, and neither the nodal planes nor the scalar moment
of that event are well constrained (Molnar & Chen 1983).
The seemingly random distribution of normal and
strike-slip faulting farther north surely reflects a mosaic of
interlaced normal and strike-slip faults, which divide the
upper crust of Tibet into relatively small blocks. The clear
variations in the orientations of the nodal planes requires
that there be numerous faults of different orientations, and
the different orientations of the possible slip vector suggests
that there must be numerous blocks that move in somewhat
different directions with respect to one another. Moreover,
the close proximity of earthquakes showing large components of normal and of strike-slip faulting, such as those
of 1980 October 7 (35.62"N, 82.14"E) and 1978 July 31
(35.47"N, 82.0°E), require that many of faults be short
(tens of kilometers or shorter) and that edges of blocks be
irregular. The different slip vectors for these earthquakes
implies that they cannot both have occurred on the common
boundary between two blocks, despite the epicenters being
only 20 ( f 2 0 ) km apart. Although 25yr of seismicity is
obviously inadequate to map the boundaries of such blocks,
the distances of 10-100 km between events with different
fault plane solutions and different slip vectors suggests that
the dimensions of such blocks are of that same order or
smaller, and therefore similar to those in the Basin and
Range Province of the western United States.
In most cases we cannot choose which of the nodal planes
is the fault plane, but from the proximity of a few
earthquakes to strike-slip faults, mapped from the Landsat
imagery (Armijo et al. 1989; Molnar & Tapponnier 1978),
one nodal plane seems the more likely to be the fault plane.
Specifically, for the event of 1978 April 4 (32.98"N,
82.26"E), right-lateral slip on the northwesterly trending
plane, parallel to the Karakorum fault, is probably the fault
P. Molnar and H . Lyon-Caen
138
N90"E
Azimuths of Slip Vectors (0)
and P-axes
along the
(w)
Himalaya.
N60"E
I
-
I-
a
N30"E
4 I-+ -1
\
\
NO"E
\
.
\
N30"W
75"E
85"E
80"E
95"E
90"E
LONGITUDE
F i i r e 7. Azimuths of slip vectors or P-axes of earthquakes in the Himalaya plotted as a function of the longitudes of the epicenters. For most
earthquakes, for which one nodal plane dips gently beneath the Himalaya, we assume that plane is the fault plane and plot the azimuth of the
normal to the other, the auxiliary, plane. For four earthquakes in the western Himalaya with nodal planes dipping 30" to W, we cannot choose
the fault plane, and we show azimuths of the P-axes. Note the monotonic increase in azimuth to the west. The dashed line through the data
shows the azimuth approximately normal to the chain; note that the overthrusting is roughly perpendicular to the chain all along it.
plane. Left-lateral slip on the northeasterly trending planes
probably characterized the faulting associated with the event
of 1978 July 31 (35.47"N, 82.00"E) near the Altyn Tagh
fault, and with the events of 1985 May 20 (35.56"N, 87.20"E)
SOUTHERN TIBET
and 1980 June 24 (33.00"N, 88.55"E), which occurred west
of regions of left-lateral strike-slip faulting (Fig. 2).
Although it seems likely that right-lateral slip is associated
with the events of 1986 June 20 (31.24"N, 86.85"E) and 1972
July 22 (31.43"N, 91.49"E), which occurred near major
northwesterly trending right-lateral faults in southern Tibet
(Armijo et 01. 1989), we abstained from making that choice
on Fig. 2.
The azimuths of nearly all of the T-axes lie between
east-west and northwest-southeast (Fig. 9), suggesting
crustal extension with this orientation. A sum of the
moment tensors, equation (2), corroborates this impression.
If all 16 earthquakes are included,
C Mii=
Figure 8. Lower hemisphere projection of P-axes (open circles) and
T-axes (closed circles) from the seven earthquakes in southern
Tibet. Note the roughly east-west to east-southeast-west-northwest
orientations of the T-axes, implying crustal extension with that
orientation.
(
11.0 -10.9
2.8)
-10.9
-2.4
2.5
2.8
2.5 -8.6
X
1018Nm,
where the 1-, 2- and 3-axes correspond to east, north and
up, and extension is positive. This corresponds to an axis of
maximum extension oriented N129"E, with shortening at
about half the rate of extension in the perpendicular
direction. The complicated signals of the largest earthquake
(1973 July 14a) make the elements of its moment tensor
quite uncertain (Molnar & Chen 1983), and if this
earthquake is excluded then
8.8 -3.8
-0.7
-0.7
-0.7
-4.0
The Tibetan Plateau and Microfiche GJI 99/1,2
HIGH PLATEAU
I
0
0
00
0
Figure 9. Lower hemisphere projection of P-axes (open circles) and
T-axes (closed circles) from the 16 earthquakes in the high part of
the Tibetan Plateau, between roughly 78"E and 90"E and north of
the belt of grabens and normal faulting in southern Tibet. Note the
roughly east-west to northwest-southeast orientations of the
T-axes implying crustal extension with that general orientation. The
scattering of P-axes reflects the comparable importances of normal
and strike-slip faulting.
This corresponds to maximum extension oriented N109"E,
again partitioned roughly equally into crustal shortening in
the perpendicular direction and crustal thinning. The
magnitudes of the strains are smaller than what we would
expect if eastern and western Tibet moved apart at
10 mm yr-', probably because the 25-yr period is not
representative of a longer duration during which much
larger earthquakes occasionally occur. Thus, the most
robust results are the approximately equal partitioning of
crustal extension into shortening and thinning and the
roughly east-southeast-west-northwest orientation of the
extension.
3.3 Strike-slip faulting in the eastern Tibetan Plateau
Most fault plane solutions of earthquakes in eastern Tibet,
south of 37"N, east of roughly 9WE, but west of the margin
of the plateau defined roughly by the 3000 m contour, show
large components of strike-slip faulting. In general, the
sense of slip on the more easterly trending nodal plane is
left-lateral. A few earthquakes occurred close to major
strike-slip faults, and for two events surface faulting reveals
the fault plane unambiguously. For the others logical
arguments suggest, but do not prove, that left-lateral
faulting occurred. Since the orientations of the planes vary
spatially across the region, we consider three separate
subareas of eastern Tibet: the Kunlun fault system, the
Xianshuihe fault, and the area southwest of the Kunlun fault
and west of the Xianshuihe fault.
3.3.1 The Kunlun fault system
Two events (1963 April 19 and 1971 March 24) occurred
near the surface trace of the principal strand of the
139
left-lateral Kunlun strike-sIip fault system. The fault plane
solutions of these two earthquakes, based on P-wave first
motions alone, were among evidence used to infer a
left-lateral sense of slip (Tapponnier & Molnar 1977), and
more recent field observations in different segments of the
fault have corroborated that inference (Cui Zheng-yuan &
Yang Bin 1979; Kidd & Molnar 1988; Li Long-hai & Jia
Yun-hong 1981). The synthesis of wave forms yield bounds
on the focal depths and seismic moments and confirm :hat
the dips of both nodal planes for the event of 1971 March 24
are steep. A more southeasterly trend of the east-west
trending plane of the event 1971 March 24 (277" f 15" versus
283" f 15" for the event of 1963 April 19) is consistent with
the curvature of the fault mapped with the Landsat imagery
and in the field (Fig. 2) but is too uncertain to quantify this
curvature.
The earthquake of 1973 August 11 occurred near the
southeast projection of the main straind of the Kunlun fault
system, and left-lateral slip on the northwesterly trending
plane might indicate a continuation of this fault (Fig. 2). In
such a case, the fault must curve southward, for the strike of
the nodal plane trends more northwesterly than the mapped
trace of the Kunlun fault to the northwest. This plane is also
crudely parallel to the eastern end of the Haiyuan fault,
some 400 km to the north (Fig. 2), along which an average
of about 8 m of left-lateral slip occurred in 1920 (Deng
Qidong et al. 1984; Zhang Weiqi et al. 1987). Thus, the
northwest trending plane is likely to have been the fault
plane, and the orientation of left-lateral simple shear, which
characterizes the area to the west, may curve southeastward
and continue to the epicenter of this earthquake.
3.3.2 The Xianshuihe fault
Besides fault plane solutions of earthquakes, both
Quaternary and Holocene offsets (Allen et al. 1989; Tang
Rongchang et al. 1984) and surface faulting associated with
major earthquakes (Tapponnier & Molnar 1977; Zhou
Huilan et al. 1983a,b) demonstrate active left-lateral slip on
this fault. The Xianshuihe fault can be recognized, both on
the Lanhat imagery (Tapponnier & Molnar 1977) and on
the ground (Allen et al. 1989), as part of an arcuate fault
system, concave to the southwest (Fig. 2). It strikes
northwest along the Xianshuihe (Fresh Water River) valley,
and farther southeast it curves to strike nearly north-south.
The variation in strike along the Xianshuihe fault is also
indicated by the differing fault plane solutions of the two
major earthquakes that occurred on it in the last 30yrs.
Allen et al. (1989) gave an orientations of 305" ( f 3 " , in our
opinion) for the surface trace of the western, Luhuo,
earthquake (1973 February 6) and 319" for the eastern,
Daofu earthquake (1981 January 23). We cannot assess the
uncertainty of the latter strike, but it accords with 322" f 12"
determined from the wave forms. Thus the difference
between the strikes for the two earthquakes is resolvable,
despite the epicenters being only about 75 km apart.
The fault plane solutions of three earthquakes (1967
August 30a,b and 1973 February 7) near 31.5", 100.3"E, just
west of the surface rupture of the Luhuo earthquake, show
large components of normal faulting with northwesterly
trending T-axes (Tapponnier & Molnar 1977; Zhou Huilan
et al. 1983a). A fault strand, clear on the Landsat imagery,
140
P. Molnar a n d H . Lyon-Caen
continues west from the area south of these epicenters; thus
the normal faulting occurs in a left step in the left-lateral
fault system, and a pull apart basin seems to be developing
there (Tapponnier & Molnar 1977). The complicated fault
plane solution of the earthquake of 1982 June 15 (31.85"N,
99.92"E) concurs with that interpretation. As discussed in
the caption of Fig. 6, we, like Ekstrom (1987), infer a
mixture of normal and strike-slip faulting, with left-lateral
slip on a roughly east-west plane (strike = 100" f 10"). This
orientation suggests that the curvature of the Xianshuihe
fault system continues west of the left step in the main
strands near 30"N, 100"E.
3.3.3 The area west of the Xianshuihe fault and southwest
of the Kunlun fault
As noted above, the fault plane solutions for earthquakes in
this area (roughly between 30"N and 35"N and between 90"E
and 99"E) seem to be characterized by left-lateral strike-slip
faulting, but the orientations of the nodal planes vary across
the region. Consider first the earthquake of 1979 March 29
(32.44"N, 97.26"E), which occurred just south of the western
extension of the Xianshuihe fault. The strike of the
east-west nodal plane, 270" (+10"/-5"), is nearly parallel to
the strike of that fault (-285"), and slip on it would have
been left-lateral. The nodal plane with left-lateral slip for
the event of 1971 April 3 (32.26"N, 95.06"E), about 200 km
farther west, strikes slightly more northeasterly (260" f 10').
Although it would be risky to presume that these two events
ruptured the same fault, the apparent differences in the
strikes suggests that the orientation of simple left-lateral
shear curves in the same sense as it does along the
Xianshuihe and Kunlun faults.
This varying orientation of left-lateral shear, occurring on
planes striking northwest east of 100"E to roughly east-west
near 9TE, continues westward so that near 92"E, left-lateral
shear occurs on planes striking northeast (Fig.2). Such an
interpretation is sensible for the event of 1975 May 5
(33.09"N, 92.92"E) but is not corroborated by additional
evidence. The proximity of the earthquake of 1971 May 22
(32.39"N, 92.12"E) to a clear northeast trending, active fault
zone (Kidd & Molnar, 1988), however, renders left-lateral
slip on that plane, instead of right-lateral slip on the
conjugate plane, very likely.
Fault plane solutions of two adjacent earthquakes farther
north are consistent with left-lateral slip roughly east-west
planes: 86"f 10" for the event of 1981 June 9 (34.51"N,
91.42"E), and 81" f 10" for the event of 1986 August 20
(34.57"N, 91.63"E). The indistinguishable latitudes of the
epicenters and the longitudinal separation of 15-20 km
clearly favors slip on the east-west plane over the conjugate
planes.
Thus left-lateral simple shear on planes trending
northwest in easternmost Tibet and trending roughly
east-west near 95" to 100"E, seems to continue to the west,
with the orientation curving to become northeast near 90"E.
Moreover, taken together, all of these solutions suggest that
at least the northern two-thirds of the eastern part of the
Tibetan Plateau are undergoing left-lateral shear. Therefore, the right-lateral shear of southern Tibet (Armijo et al.
1989; Tapponnier et al. 1986) appears to be confined to the
southernmost part of this region.
3.4 Crustal shortening dong the northern and eastern
margins of Tibet
Evidence of active thrust faulting can be seen on the ground
and on the Landsat imagery along the northern,
northeastern, and southeastern margins of the Tibetan
Plateau (Tapponnier & Molnar 1977), but, to our
knowledge, it has been investigated thoroughly only in a
small area northeast of the Haiyuan fault (Zhang Peizhen et
al. 1989a,b). Thus, the large components of thrust faulting
associated with earthquakes occurring south of the Altyn
Tagh, throughout the Nan Shan, and within and north of the
Longmenshan (Fig. 2) are not surprising. The most
significant result is the confirmation that the orientation of
crustal shortening is roughly perpendicular to the regional
average contours of the topography (England & Houseman
1986).
3.4.1 The Altyn Tagh
Active thrust fault scarps are clear along the bases of several
roughly east-west trending ranges in the Altyn Tagh, at the
west end of the Qaidam Basin (Molnar et al. 1987) and the
P-axis of the earthquake of 1977 January 1, oriented N24"E
(204" + 15"/-23") implies north-northeast-south-southwest
shortening.
3.4.2 The Nan Shan and the Qaidam Basin
Broad, low amplitude folds can be seen trending northwest
on the Landsat imagery of the Qaidam Basin (Tapponnier &
Molnar 1977), and roughly linear basins and ranges in the
Nan Shan seem to be bounded by thrust (or reverse) faults
with a similar trend. Bohlin (1940,1960) noted folding and
thrust faulting of Cenozoic sedimentary rock in a few areas
of the Nan Shan, and he inferred that the deformation
began in Miocene time, but we are aware of no detailed
investigations of this crustal shortening.
Among the earthquakes that we studied, fault plane
solutions of the three largest show nearly pure thrust
faulting. Probably the northeast dipping nodal plane
ruptured in the two events on the southern edge of the Nan
Shan (1962 May 21 and 1977 January 19), such that the Nan
Shan was thrust onto the Qaidam Basin, and probably the
southwest dipping plane ruptured in the event near the
eastern edge of the range (1986 August 26), with the range
overthrust onto the plain to the northeast. In any case
the northwest-southeast orientations of the P-axes are
well resolved: 206" (f5"), 226" (+15"/-23"), and 229"
(+30"/- lSo), respectively.
Thrust faulting is by no means the only style of
deformation occurring in the Nan Shan. Peltzer et al. (1987)
inferred a Holocene slip rate of 5.5mmyrF' on a fault
striking roughly 70". This fault is one strand of the
strike-slip fault system associated with the Changma
earthquake of 1932 in the northern Nan Shan. From SPOT
imagery, Peltzer (1987) inferred a comparable rate on
another left-lateral strike-slip fault, which curves southward
from the Altyn Tagh fault. Moreover, given the existence of
major left-lateral strike-slip faults north and south of the
Nan Shan, such faulting with in the range ought not to be
surprising.
The Tibetan Plateau and Microfiche GJI 99/1,2
NAN SHAN
141
The topography northeast of the Longmenshan does not
exhibit the dramatic step from a high range to a low basin,
but if the 3000-m contour marks the rough limit of the
Tibetan Plateau, then the eastern margin of the plateau
trends roughly north-south. The solutions for the three
relatively large Songpan earthquakes (magnitudes of 7.2, 6.7
and 7.2) (Fig. 2) near 33"N, 104"E indicate east-southeastwest-northwest crustal shortening (Jones et al. 1984). They
are associated with oblique southeastward overthrusting of
the plateau onto the area to the east on an echelon series of
northerly trending faults, and with east-southeast trending
P-axes (Jones et a f . , 1984). These orientations are quite
different from those of earthquakes in the Nan Shan, but
they too are aligned roughly perpendicular to the average
contours of the topography, which trend roughly north in
this area.
3.4.4 Yinchuan graben
Figure 10. Lower hemisphere projection of P-axes (open circles)
and T-axes (closed circles) from the eight earthquakes in the Nan
Shan, including those for the event of 1977 January 1 southwest of
the Qaidam Basin. Note the roughly northeast-southwest
orientations of the P-axes, implying crustal shortening with that
orientation.
Substantial components of left-lateral slip on roughly
east-west trending planes are clearly permitted by the fault
plane solutions for the earthquakes of 1964 March 1916 and
1972 August 30a,b, but none of the solutions is well
constrained. Moreover, despite the abundant field evidence
for left-lateral slip, components of right-lateral slip on
conjugate northwest trending planes, associated with
counter-clockwise rotations of basins and ranges, seems
quite plausible. Such conjugate slip and rotations occur in
broad shear zones in Greece (Jackson & McKenzie 1984)
and in southern California (Nicholson et al. 1984). In any
case, the azimuths of the P-axes for the two earthquakes of
1972 August 30 and for the small earthquake of 1980 June 1
lie in the northeast quadrant (Fig. 10).
Thus an overall northeast-southwest crustal shortening in
the Nan Shan seems well established (Fig. 10). Because the
seismic moment of the earthquake of 1962 May 21 is more
than 10 times greater than any of the other events, a
summation of the moment tensors is not very revealing.
3.4.3 The Longmenshan and Songpan
The orientation of crustal shortening on the eastern and
southeastern margins of the plateau is between east-west
and northwest-southeast. The Longmenshan, bounding the
southeast margin of the plateau, constitutes a relatively
high, northeast-trending mountain range bounded abruptly
on its southeast by the low, flat Sichuan Basin. There is little
doubt that the Longmenshan is being actively thrust onto
the Sichuan Basin, as suggested also by the fault plane
solution for one earthquake (1970 February 24)
(Tapponnier & Molnar 1977). Slip on the northerly dipping
nodal plane would entail such a southeastward overthrusting, with a component of right-lateral slip parallel to the
range.
We studied one other earthquake in the area shown in Fig.
2, that of 1976 September 22 (40.02"N, 106.32"E). The fault
plane solution is not well constrained and indicates either a
large component of normal faulting with a northwestsoutheast trending T-axis, or nearly pure strike-slip
displacement also with the T-axis oriented southeast and
with the P-axis northeast. Because of the proximity of the
epicentre to the north-northeast-trending, tectonically active
Yinchuan Graben, we assume that the solution showing
normal faulting is appropriate. In either case, the style of
faulting is different from that observed to the southwest
along the margins of the Tibetan Plateau.
4
TECTONIC IMPLICATIONS
The fault plane solutions described here call attention to
several patterns. The style of faulting varies across Tibet,
from large components of normal faulting in the highest part
of the plateau, west of roughly 90"E (Figs 1 and 2), to an
increasing prevalence of strike-slip faults at lower elevations
and finally to primarily thrust faulting in the margins of
Tibet. This presumably results from the obvious facts that
thrust faulting results in crustal thickening and hence an
increase in elevation, and that normal faulting affects a
thinning of the crust and subsidence. The roughly
east-southeast orientation of crustal extension in the plateau
is approximately perpendicular to the direction of
underthrusting and convergence at the Himalaya. Thus
some north-south shortening occurs within Tibet, by slip on
conjugate strike-slip faults, but not by crustal thickening.
The radially outward orientations of slip vectors, and
therefore in the direction of overthrusting, vary along the
Himalaya, as they must if southern Tibet undergoes
extension. Similarly, the orientations of P-axes of
earthquakes vary along the northern and eastern margins of
Tibet, so as to be aligned roughly perpendicular to the
regional contours of the topography. Finally, left-lateral
shear on planes with a systematically varying orientation
appears to dominate the active strain field across eastern
Tibet.
Many of these patterns are well known, and our results
merely confirm them and broaden the areal extent over
which they are known to apply. Let us discuss them briefly,
142
P. Molnar and H . Lyon-Caen
and then examine in more detail the implications of the
sense and orientation of strain for the kinematics of
deformation within and surrounding the Tibetan Plateau.
4.1 Variations in style of deformation and mean
elevation across Tibet
The style of deformation varies from (i) crustal shortening in
the western Himalaya and overthrusting of the Himalaya
onto the Indian Shield along most of the range to (ii)
predominantly normal and conjugate strike-slip faulting in
the high Tibetan Plateau, between about 78"E and 90"E
where the elevation nowhere drops below 4500m and
exceeds 5000m in most of the region to (iii) primarily
strike-slip faulting in eastern Tibet, where elevations
decrease gradually from about 4500 to 3000 m, and then to
(iv) thrust faulting on the margins of the plateau where
elevations drop below about 3000 m (Fig. 2).
The variation from northeast-southwest crustal shortening and crustal thickening in the western Himalaya to
normal faulting and east-west extension in the higher
plateau and then to thrust faulting and then east-west
compression on the eastern margin of Tibet does not in any
way require a lateral variation in the mean horizontal stress.
The mean horizontal stress could be constant along an
east-west profile across Tibet, just as it could be across the
Andes where normal faulting is observed at high altitudes
and thrust faulting is observed on the margins (Dalmayrac &
Molnar 1981; Molnar & Lyon-Caen 1988; Sebrier et al.
1985, 1988; SuBrez, Molnar & Burchfiel 1983). Instead, this
variation in tectonic style across Tibet is almost surely, at
least in part, a consequence of the greater vertical stress (at
any depth below sea level) where elevations are higher, than
in the areas surrounding Tibet where they are lower. An
augmented vertical stress will enhance normal faulting, and
a diminution of it will allow thrust faulting. We do not mean
to imply that we have used sophisticated seismological
techniques merely to map differences in mean elevations in
Tibet, and therefore that the study reveals nothing that
could not be learned more simply and inexpensively from a
thoughtful look at an atlas. Instead, what is noteworthy is
that the mean east-west horizontal stress is sufficiently large
to cause crustal shortening on margins of Tibet but
insufficiently large to prevent east-west extension of the
high plateau.
4.2 The orientations of slip vectors and convergence at
the Himalaya
The direction of overthrusting of the Himalaya onto the
Indian Shield varies along the range and is roughly radially
outward (Fig. 7) (Armijo ef al. 1986; Baranowski ef al. 1984;
Molnar & Chen 1982). In the eastern Himalaya slip vectors
point south-southeast. Between 85"E and 90"E, they point
due south. At about 80"E, they point south-southwest, and
farther west, near 75"E, the northeast-southwest orientation
of the P-axes implies crustal shortening with this orientation.
Because India behaves as a rigid plate, this variation
requires that southern Tibet undergo crustal extension with
a roughly east-southeast-west-northwest orientation. If the
rate of underthrusting were constant along the Himalaya,
then, as discussed above, the rate of extension in southern
Tibet, vT, would be approximately the rate of underthrusting, vH.
4.3 The orientations of P-axes of earthquakes on the
margins of Tibet
These orientations vary from north-northeasterly in
northern Tibet (1977 January l), to northeasterly in the Nan
Shan on the northeastern margin of Tibet, to eastsoutheasterly in eastern Tibet (Songpan earthquakes of 1976
August: Jones et al. (1984)), to southeasterly in the
Longmenshan in southeastern Tibet (Fig. 2). They are
aligned roughly perpendicular to regional topographic
contours and therefore roughly parallel to the gradients in
topography, as England & Houseman 119861 inferred from
fewer solutions. The P-axes clearly are not parallel to the
direction of India's convergence with Eurasia, but the
earthquakes and deformation associated with them are
surely related to that convergence.
The orientations of the P-axes alone do not place a tight
constraint on the physical mechanisms by which India's
penetration of Asia causes an extrusion of eastern Tibet, but
this pattern of P-axes radiating from the Tibetan Plateau
was one of the reasons that Molnar & Tapponnier (1975)
rejected a description of continental tectonics in terms of
microplates and appealed to one in terms of continuous
deformation (Tapponnier & Molnar 1976). Moreover, the
alignment of the P-axes perpendicular to topographic
contours is predicted by numerical calculations of the finite
deformation of a thin viscous sheet deformed in a gravity
field (England & McKenzie 1982; England & Houseman
1986; Houseman & England 1986; Vilotte et al. 1986). In
this context, the penetration of India into Eurasia manifests
itself, at least in part, as a radial expulsion of Tibetan crust
and as crustal shortening on its margins.
4.4 East-southeast crustal extension in the high plateau
Normal and strike-slip faulting prevail where altitudes are
high. In southern Tibet, where Armijo et al. (1986)
recognized seven major, roughly north-south trending
grabens, normal faulting dominates the strain field. Fault
plane solutions of the seven earthquakes from this area
show primarily normal faulting with east-west to eastsoutheast-west-northwest T-axes. Farther north, both
normal and strike-slip faulting occur with the strain field
characterized by approximately equal fractions of each.
The simultaneous east-west extension in central Tibet
and east-west shortening in eastern Tibet allow a range of
possibilities for the relative displacement of blocks on either
side of Tibet. If the extension were the more rapid, then
western Tibet would move away from southeast China. If it
were less rapid than the crustal shortening in eastern Tibet,
then southeast China would move toward western Tibet. If
the rates were the same, then the eastward extrusion of
Tibet would be absorbed entirely by crustal shortening on its
eastern margin. Thus, what would be much more interesting
to know than lateral variations in the orientations of
horizontal principal strains, or stresses, are the strain rates,
or the deformation gradient tensors, across Tibet.
Unfortunately we cannot evaluate sufficiently accurately
either the rate of extension within Tibet or the rate of
The Tibetan Plateau and Microfiche GJI 99/1,2
crustal shortening on its eastern margin to know whether
southeast China and western Tibet move apart from one
another or toward one another. Thus we cannot evaluate
quantitatively and directly the eastward component of
southeast China's movement with respect to India. We
suspect that there is an eastward component; India moves
nearly due north with respect to Eurasia (e.g. Minster &
Jordan 1978), but slip during earthquakes in Asia since 1900
suggests that southeast China moves east-southeast with
respect to Eurasia (Molnar & Deng Qidong 1984).
Nevertheless, an evaluation of southeast China's velocity
with respect to India from estimates of rates of deformation
or of slip in eastern Tibet cannot yet be made, and the
description of where much of the extruded material goes
necessarily remains vague.
Extrusion
143
by Laterol Translation
4.5 Left-lateral strike-slip faulting and rotation in
eastern Tibet
The fault plane solutions of earthquakes within the Tibetan
Plateau and measurements of slip rates on the major
strike-slip faults give a hint as to how deformation in eastern
Tibet transfers the crustal extension in the high plateau to
crustal shortening of the eastern margin of the Plateau.
First, they suggest that this area is pervaded by left-lateral
shear (Fig. 2). If right-lateral shear is important in eastern
Tibet, as it is in southern Tibet west of 95"E (Armijo et al.,
1989), this shear must be confined to the southernmost part
of eastern Tibet.
More importantly, the curved trajectories of left-lateral
shear imply that the material extruded from Tibet does not
merely translate eastward relative to the rest of Asia, but
also undergoes a marked rotation. The left-lateral faults,
although roughly east-west, are not strictly parallel to one
another. The major faults are curved so as to be concave
toward the south, and slip on them must include a rotation
of the material on one side with respect to that on the other.
Moreover, there is a suggestion of systematic variations in
the orientations of left-lateral shear across Tibet. In
northern Tibet, the Altyn Tagh fault trends east-northeast,
but in southeastern Tibet, the Xianshuihe fault trends
southeast. The Kunlun fault system trends nearly due east
between 88"E and 92"E, but farther east it splays with both
splays curving to trend east-southeast. The main strand
might continue as far east as near the epicenter of the
earthquake of 1973 August 11 (Fig. 2), for which left-lateral
slip could have occurred on a plane striking northwest. In
east-central Tibet, near 32"N, 90"E, left-lateral slip seems to
occur on northeasterly trending planes, but farther east the
trend becomes more easterly and then southeasterly on the
Xianshuihe fault. This variation in the orientations of
left-lateral faults also implies a rotation of material in this
area. Thus we suspect that the extrusion of eastern Tibet
occurs both by left-lateral slip on roughly east-west trending
planes and simultaneously by rotation of the material
between the faults (Fig. 11).
This emphasis of left-lateral shear and rotation differs
from the point of view taken by Armijo et al. (1989) and
Tapponnier et al. (1986). They demonstrated the existence
of east-southeast trending right-lateral strike-slip faults in
southern Tibet and inferred that these fault segments are
part of a major east-trending right-lateral shear zone. As a
E x t r u s i o n w i t h L a r g e Amounts
01
Rotation
Figure 11. Simple patterns for eastward displacement of material
on strike-slip faults. (Top) Left-lateral slip on straight, easterly
striking faults and eastward translation of blocks requires a
compensating right-lateral component in the south. (Bottom)
Rotation on curved left-lateral strike-slip faults alows eastward
extrusion without a large right-lateral component near the southern
(lower) margin.
first approximation, they then described the extrusion of
Tibet as the expulsion of a rigid block bounded on the south
by the right-lateral shear zone and on the north by the
left-lateral Altyn Tagh fault.
Rapid left-lateral slip on a series of roughly east-west
trending faults does not require that southernmost Tibet
move rapidly eastward with respect to northern Tibet, or
with respect to India. This would be so if the faults were
straight and parallel to one another and if there were no
deformation between them (Fig. l l a ) , for in this case the
roughly 10-20 mm yr-' or more of lateral slip on the Altyn
Tagh, Kunlun, and Xianshuihe fault systems would require
30 mm yr-' or more of right-lateral slip on a fault zone in
southern Tibet, even faster than that inferred by Armijo et
al. (1986, 1989) and Tapponnier et a/. (1986). With curved
faults and fault zones and with rotation of material between
them, however, the eastward motion of southern Tibet with
respect to northern Tibet need not be large (Fig. llb).
Obviously, the simple case drawn in Fig. l l ( b ) does not
include all important aspects of deformation in Tibetdeformation between major faults, the apparent dying out
of these faults, north-south convergence (by strike-slip
faulting) across Tibet or right-lateral shear in southern
Tibet. We show it primarily to dispel the image illustrated
144
P. Molnar and H . Lyon-Caen
by Fig. ll(a). Note also that such rotation can allow a
substantial amount of north-south convergence between
northern and southern Tibet without thrust faulting and
crustal thickening.
Thus, the east-west extension of the high part of the
Tibetan Plateau grades into left-lateral shear on roughly
easterly trending planes and then into crustal shortening on
the margins of the plateau. The change in the style of
deformation is surely associated with a variation in
elevation, which manifests itself as a diminution of the
vertical compressive stress from the interior of the plateau
to its margins. The preponderance of left-lateral shear on
curved faults may allow large rotations and eastward
transport of material without major right-lateral shear in
southern Tibet. At the same time, at least part of the
eastward extrusion of Tibet manifests itself as crustal
shortening on the northern, eastern, and southeastern
margins of the plateau.
5 THE KINEMATICS OF DEFORMATION
OF TIBET
The inferences of senses of shear and types of faulting
combined with rates of slip on particular faults can be used
to place constraints on the relative velocities of different
parts of Tibet and its surroundings.
5.1
Rates of extension in southern Tibet
The radially outward overthrusting of the Himalaya onto
the Indian Shield (Fig. 7) requires extension in southern
Tibet (Armijo et al. 1986; Baranowski et al. 1984; Molnar &
Chen 1982). As noted above, a variation in the direction of
65" f20" along the Himalaya requires a rate of extension,
vT, comparable with the rate of convergence, uH:
grabens, near 78"E (1975 January 19, Fig. 2). Thus the
greater rate of 1 8 f 9 m m y r C ' results from considering the
Himalaya farther west, where convergence has a larger
east-west component than farther east (Fig. 7).
The estimated rate of 18mmyr-' of divergence between
western and southeastern Tibet should include a component
of right-lateral slip on the Karakorum fault (Fig. 2). This
fault approaches the Himalaya near 80"E to 82"E, and if
rapid slip did occur on it, the azimuths of the slip vectors of
earthquakes in the Himalaya east and west of this area
should be different. The maximum difference allowed by the
variation and scatter in azimuths of slip vectors is about 30"
(Fig.7). For a rate of underthrusting of 18*7mmyr-'
beneath the Himalaya both east and west of where the
Karakorum fault intersects the range, the rate of slip on the
Karakorum fault should be less than 9 f 4 mm yr-'.
The rates deduced here are subject not only to India
behaving as a rigid plate but also to the convergence rate
being roughly constant along the Himalaya. If that rate
varied by a factor of 2 from one end of the range to the
other, then the calculated rate of extension should be
roughly 15 per cent smaller than the 18 f 9 mm yr-' given
above.
The extension in southern Tibet does not contribute to
the eastward extrusion of material in front of the advancing
Indian subcontinent. Recall that India underthrusts the
Himalaya roughly due northward at the eastern part of the
range, where India's motion with respect to Siberia is also
approximately due north. Hence, southern Tibet has at most
a negligible component of eastward displacement with
respect to India or Siberia. Moreover, normal faulting is not
prevalent east of roughly 90"E. Thus the extension in
southern Tibet is, kinematically, a reflection of westward
extrusion of material over the advancing Indian Shield.
vT = l . O ( f 0 . 3 ) ~ ~ .
5.2 Rate of eastward extrusion of eastern Tibet
From the variation in ages of the oldest sediment in the
Ganga Basin south of the Himalaya, we inferred a rate of
underthrusting of the Indian Shield beneath the southern
edge of the Himalaya of 15*5mmyr-' (Lyon-Caen &
Molnar 1985). If there were shortening within the Himalaya
of a few mm yr-', the rate of convergence would be 18 (57)
mmyr-' (Molnar 1987). Thus the rate of divergence
between eastern Tibet (near 90"E to 95"E) and western
Tibet, or strictly the Karakorum (near 75"E) would be
18 f 9 mm yr-'. The calculated direction is roughly N115"E.
This direction accords with that of N105"E (*loo)
estimated by Armijo et uf. (1986) from strain in southern
Tibet, and although our rate of 18 f 9 mm yr-' appears to
exceed theirs of 10 f 5 mm yr-', the difference is due largely
to the consideration of strain over regions of different
dimensions. Armijo et al. (1986) estimated a rate of
extension of 1.5 f 0.6 mm yr-' along one graben system in
southern Tibet and multiplied this by 7, assuming this rate
to be an appropriate average for the seven major graben
systems in southern Tibet. As the westernmost graben is
near 80"E, their estimate of 10 f 5 mm yr-' applies only to
the region to the east and ignores extension farther west.
Normal faulting clearly occurs west of 80"E; the largest
earthquake in Tibet since 1962 and with a fault plane
solution showing normal faulting occurred west of the main
Given estimates of rates of slip on the major faults in
northern and eastern Tibet, and taking into account the
curvature of these faults, we can estimate the rate that
material is extruded eastward with respect to Siberia or
India. As noted above, however, from these data we cannot
estimate the rate at which Southeast China moves with
respect to other parts of Asia.
We can use the known angular velocity describing India's
motion with respect to Siberia as a bound on velocities, but
we must account for the displacement of the Tarim Basin
with respect to Siberia and the crustal shortening in the Tien
Shan. The direction of convergence between the Tarim
Basin and Siberia seems to be roughly north-northwest
(Nelson, McCaffrey & Molnar 1987), with a rate of about 13
( f 7 ) mmyr-' (Molnar & Deng 1984). This direction is
roughly parallel to the trend of small circles about the axis
for India and Siberia (Eurasia), so let us describe India's
convergence toward the Tarim Basin ,QTB as a rotation
about Minister & Jordan's (1978) axis at 19.71°N, 38.46"E,
but with a rate of 9.1 x 10-9yr-', 75 per cent of the rate of
India's convergence with Eurasia.
Because the direction of underthrusting of India beneath
the Himalaya at the eastern end of the range is nearly due
north, let us describe the motion of India with respect to the
Himalaya and southern Tibet ,QH as a rotation about O"N,
The Tibetan Plateau and Microfiche GJI 99/1,2
145
40
38
36
34'N
34
32'N
32
30' N
30
28" N
28
e
I
88"E
I
90'E
/
1
92"E
m
94'E
I
96'E
\
98"E
- 26'N
I
100"E
I
102'E
Figure 12. Map showing epicenters and major faults in eastern Tibet and lines drawn normal to slip vectors of earthquakes on the Kunlun and
Xianshuihe faults and west of these faults. Note that most of the lines normal to slip vectors pass near 27"N, W E (closed circle), including that
from the earthquake of 73/8/11 near 33.O"N, 104"E and the line normal to the Kunlun fault near its eastern end. Lines normal to slip vectors of
earthquakes west of the Xianshuihe fault do not intersect those from earthquakes near it, but those from three of the easternmost earthquakes
pass near 28S0N, 98"E (open circle). To estimate rates of deformation, we used these two locations (open and closed circles) as axes of
rotation for slip along the Kunlun and Xianshuihe faults.
WE, with a rate of 2.67 X lO-'yr-',
corresponding to a
convergence rate of 18 mm yr-'.
The regions between the Altyn Tagh, Kunlun and
Xianshuihe faults and the right-lateral shear zone in
southern Tibet do not behave as rigid plates, but let us
ignore deformation between these faults, by assuming that it
accounts for a small (130 per cent) of the overall
displacement field. Treating the first three of these faults
crudely as plate boundaries, we can estimate angular
velocities for the Qaidam Basin with respect to the Tarim
Basin, the area west of the Longmenshan with respect to the
Qaidam Basin, and the area of western Sichuan and Yunnan
south of the Xianshuihe fault with respect to the
Longmenshan. Then as a check, we can calculate the
velocity at which western Sichuan and Yunnan moves with
respect to the Himalaya and southern Tibet and compare it
with the strike of the right-lateral shear zone in southern
Tibet and the rate of 10-20 mm yr-' estimated by Armijo et
al. (1989).
For motion of the Qaidam Basin with respect to the
Tarim Basin oL2TB, we use the axis at 12"S, l W E , given by
Armijo et al. (1989), and rates of -4.10X low9 or
-6.15 x lO-'yr-',
corresponding to slip rates of 20 or
30 mm yr-' (Peltzer 1987).
To determine an axis for the Longmenshan with respect
to the Qaidam Basin L M S Q o , we use the approximate
intersection, at 27"N, 96"E, of great circles perpendicular to
the slip vectors along the Kunlun faults (Fig. 12). For the
Kunlun fault, we use the slip vectors of the two earthquakes
that clearly occurred on that fault, that for the earthquake
of 1973 August 11 southeast of where the fault is clear on
the Landsat imagery, and also the strike of the fault at its
eastern end. The angular speed of -13.8 x lO-'yr-'
corresponds to slip at 13mm yr-' (Kidd & Molnar 1988) on
the main strand of the fault, east of 95"E.
For the rotation of Yunnan with respect to the
Longmenshan YL2LMS, we use normals to slip vectors of all
earthquakes showing strike-slip faulting on the Xianshuihe
fault and west of it. Even with large uncertainties assigned
to each, they cannot intersect at one point. Thus we
consider two possibilities: the same axis as used for the
Kunlun fault, which is an adequate average for all of the slip
vectors, and an axis farther northeast at 28.5"N, 98"E, which
accords with the three least uncertain slip vectors among the
146
P. Molnar and H . Lyon-Caen
Table 3. Assumed Angular velocities between various fragments in Asia.
InEUR
IQTB
IRH
2%
LMSRQ
YRLMS
YQLMS
YRLMS
YRLMS
Fixed Fragment
(or plate)
Eurasia
Tarim Basin
Himalaya
Tarim Basin
Tarim Basin
Qaidam Basin
Longmenshan
Longmenshan
Longmenshan
Longmenshan
Moving Fragment
(aPlate)
India
India
India
Qaidarn Basin
Qaidam Basin
Longmenshan
YUnnan
YUnnan
YUnniUl
YUnnan
Angular speed
Position of axis
(x 10-9ia)
h ( O N ) Long(OE) ("/Ma)
19.71
38.46
0.699
12.0
a
19.71
38.46 0.516
9.0
b
0.
0.
0.153
2.67 c
-12.
106.
-0.235
-4.10 c
-12.
106.
-0.352
-6.15 e
-0.790 -13.8
f
27.
96.
-0.860 -15.0 g
27.
96.
-1.290 -22.5 h
21.
96.
-1.290 -22.5
i
28.5
98.
-1.937 -33.8 j
28.5
98.
a. From Minster and Jordan [1978].
b. The same axis as for India to Eurasia, but with a fraction deleted for convergence between the Tarim
Basin toward Siberia at the Tien Shan at 13 mm/a [Molnar and Deng, 19841.
c. Calculating assuming due north convergence of India toward the Himalaya with a rate of 18 mm/a
[Lyon-Caen and Molnar. 19851.
d. Axis from Armijo et al. [1989],with the angular speed based on slip on the Altyn Tagh fault at 20
mm/a [Peltzer, 19871.
e. Axis from Arrnijo et al. [1989].with the angular speed based on slip on the Altyn Tagh fault at 30
mm/a [Peltzer, 19871.
f. Axis deduced from slip vectors of earthquakes along the Kunlun fault (Figure 12), with the angular
speed based on slip on the Kunlun fault east of 95OE at 13 mm/a [ K i d and Molnar, 19881.
g. Axis based on slip vectors from the Xianshuihe fault and the area farther west (Figure 12), with the
angular speed based on slip on that fault at 10 mm/a [Allen et al., 1989;Molnar and Deng. 19841.
h. Axis based on slip vectors from the Xianshuihe fault and the area farther west (Figure 12), with the
angular speed based on slip on that fault at 15 mm/a [Allen et al., 1989;Molnar and Deng, 19841.
i. Axis based on slip vectors only horn the Xianshuihe fault (Figure 12),with the angular speed baked on
slip on that fault at 10 mm/a [Allen et al., 1989;Molnar and Deng, 19841.
j. Axis based on slip vectors only from the Xianshuihe fault (Figure 12). with the angular speed based on
slip on that fault at 10 mm/a [Allen et al., 1989;Molnar and Deng, 19841.
four easternmost earthquakes (Fig. 12). The slip vector for
the small, complicated, multiple earthquake of 1982 June 5
(Fig. 7) does not fit any sensible axis. For rates about these
two axes, we used 14.5 x
and 22.5 % 10-9yr-' for the
first, and 22.5 X
and 33.8X 10-'yr-'
for the
northeastern axis, corresponding to rates of slip on the
Xianshuihe fault of 10 and 15 mm yr-', respectively (Allen
et al. 1989; Molnar & Deng Qidong 1984).
Before discussing the results of this analysis, we
emphasize that with the obvious violations of the basic
assumption of rigid plates, with the large uncertainties in
rates of slip, and with the strong dependence of the angular
speeds on the orientations of the rotation axes, a rigorous
statistical analysis is premature. We simply consider the two
positions of the axes for the motion of the Longmenshan
with respect to Yunnan and the two rates of slip along the
Altyn Tagh and Xianshuihe faults to explore the plausible
rates of extrusion of southeastern Tibet.
We calculated the rate of angular velocities of Yunnan
with respect to the Himalaya yQH using
UQH
= YSZLMS + LMSQQ + QnTB
- IQTB - Ha,,
where all of the other angular velocities are known (Table
3). For the two different axes for Yunnan with respect to the
Longmenshan, we calculated pairs of angular velocities
yQH, corresponding to slow (20 and 10 mm yr-') and to fast
(30and 15 mm yr-') slip on the Altyn Tagh and Xianshuihe
faults, respectively. All calculated axes for Yunnan with
respect to the Himalaya lie just south of the eastern
Himalaya. The calculated angular speeds are all negative,
corresponding to southeastward right-lateral displacement of
Yunnan with respect to the Himalaya, and angular speeds
range from 38 to 56 X
yr-' about axes near 24"N,91"E.
These calculated angular velocities yQH yield rates of
right-lateral slip at the eastern end of the Himalaya of
30-40 mm yr-', where Armijo et al. (1989) estimated such
slip at 10-20 mm yr-'. This range of calculated rates surely
does not bound the range of possible rates. If the rate of
convergence at the Himalaya (18?~7mmyr-') or at the
Tien Shan (13f 7 mm yr-') were 7 mm yr-' lower or higher
than the values used, the calculated slip rates would be
20-30% different. The failure to match the range of rates
estimated by Armijo et al. (1989) is probably a result of
deformation in eastern Tibet and distributed shortening
across this region.
At the longitude of the east end of the Himalaya and
relative to the Tarim Basin, the Qaidam Basin is displaced
east-northeast at roughly 20-30 mm yr-', and the areas
northwest of the Longmenshan both north and south of the
Xianshuihe fault are displaced eastward at a comparable
rates of 30-40 mm yr-'. The eastward displacement of the
Qaidam Basin apparently is largely absorbed by crustal
shortening in the Nan Shan (e.g. Burchfiel et al. 1989b), but
it is not clear how much of the displacement of material
south of the Kunlun fault system is absorbed at the
Longmenshan.
The Tibetan Plateau and Microfiche GJI 99/1,2
At the longitude of 97"E, there appears to be an eastward
flux of material with respect to either India or Eurasia,
which converge with a north-south orientation at this
longitude. This flux is given by the product of rate times
north-south lateral extent times thickness. Let us consider
two parts, the area between about 30"N and 35.5"N where
material is displaced eastward at about 30-40 km M yr-'
(mm yr-') and the crust is thick, 70 km, and the area farther
north, between 35.5"N and 39"N, where the eastward
component of displacement is less, 15-25 km M yr-' and the
crust is probably thinner, 40 km. Together, using rates of
35 km M yr-l and 20 km M yr- ', these give an eastward flux
of about 1.8 x lo6 km3 M yr-'.
We can use this value to evaluate the role played by such
extrusion in absorbing the penetration of India into the rest
of Eurasia. India converges with Himalaya at a rate of about
18 ( f 7 ) km M yr-'. Thus the average rate of northward
displacement of southern Tibet is about (50-18)
32 km M yr-', and the northward flux of material into the
rest of Eurasia across a zone between the western edge of
Tibet and the eastern end of the Himalaya is given roughly
by 32 km M yr-' x 70 km x 1700 km, equal to 3.8 x
lo6km3 M yr-I. The ratio of these two fluxes suggests that
the extrusion of Tibet could account for roughly half of the
convergence of southern Tibet with respect to Eurasia, and
therefore roughly a third of the convergence of India with
respect to Eurasia, as Armijo et at. (1989) also concluded.
Note, however, that because of the prevalent normal
faulting within the plateau, not all of the eastward flux
accommodates active penetration of India into Eurasia;
some of the flux is due to crustal thinning within the high
plateau.
5.3 Rate of deformation of the Tibetan Plateau
One question to which we seek the answer is the extent to
which the convergence between India and Eurasia is
absorbed by contraction of Tibet's north-south dimension
and to what extent it is absorbed by deformation elsewhere
in Asia. Only part of the eastward flux described above is
due to north-south shortening within the Tibetan Plateau.
Much of the extrusion can be ascribed simply to
east-northeast, effectively rigid-body, translation of Tibet
along the Altyn Tagh fault. As Armijo et al. (1989) showed,
even if one ignores the deformation within Tibet and slip on
the Kunlun and Xanshuihe faults, the displacements on the
Altyn Tagh and right-lateral faults in southern Tibet require
a substantial eastward flux of material. Although one could,
in principle, take the difference between the estimate of flux
given above and that due to rigid body displacement
associated with slip on these fault systems to estimate the
amount of extrusion due to crustal thinning, this would
involve taking the difference between two quite uncertain,
but comparable, numbers and therefore is not enlightening.
The rate of eastward displacement, with respect to the
Tarim Basin, of eastern Tibet south of the Kunlun fault
system is roughly 10 mm yr-' faster than that farther north.
This difference arises directly from slip on that fault, but the
nature of rotation of material schematically illustrated by
Fig. 11 leads to comparable average rates of eastward
transport of material south of the Kunlun fault and south of
147
the Xianshuihe fault. Ignoring possible variations in the rate
of slip along the Altyn Tagh fault between 92"E and 80"E,
which are not obviously large (Peltzer 1987), and internal
rotations within Tibet, this crudely suggests that eastern
Tibet moves roughly lomrnyr-' eastward away from
western Tibet. For a north-south dimension of this area of
600km and a thickness of 70 km, the flux associated with
this contribution to the extrusion would be 0.4 x
106km3Myr-', only a small part of the total of
1.8 X lo6 km3 M yr-I.
The fault plane solutions of
earthquakes within Tibet show a mixture of normal and
strike-slip faulting such that roughly half of the
east-southeast extension of the plateau is accommodated by
crustal thinning and half by north-northeast crustal
shortening. Thus, half of this 0.4 X lo6 km3 M yr-' of flux is
likely to be due to shortening of the plateau, and half is due
to crustal thinning. Were these values precise, they would
suggest that the extrusion of material out of India's way
amounts to 1.6 X lo6 km3 M yr-', with 0.2 x lo6 km3 M yr-'
of additional extrusion due to the extension and crustal
thinning of Tibet. Obviously, the 0.2 x lo6 km3 M yr-'
subtracted from 1.8 X lo6 km3 M yr-' deduced above is
negligible compared with the uncertainties in the total flux.
Assuming that the rate of extension across Tibet of
10 mm yr-' is reasonable, then the distribution of faulting
suggests that north-south, or north-northeast-southsouthwest, contraction of the plateau occurs at roughly
5 mm yr-'. The uncertainty is clearly at least 2 mm yr-'.
and east-southeast
The rate (roughly 10 mm yr-')
orientation of extension in the high part of the plateau
appear to be comparable with those in southern Tibet,
where grabens and normal faulting dominate. Above we
argued that the extension in southern Tibet seems to involve
a westward displacement of material with respect to India or
Eurasia, not an eastward displacement. Between these two
areas, Armijo et at. (1989) and Tapponnier et at. (1986)
mapped a system of west-northwest trending strike-slip
faults that they visualize as parts of a major right-lateral
shear zone marking a major boundary between two tectonic
provinces and along which central Tibet is translated
eastward with respect to southern Tibet. The image that we
offer here is of a zone between two regions both of which
are undergoing roughly west-northwest-east-southeast extension. Hence these faults might not be parts of a
through-going shear zone but rather simply strike-slip
segments within the mosaic of deformation in Tibet. Thus
rates of displacements on the various right-lateral faults are
likely to be different from one another, as well as variable
along the individual faults, as is the case for instance with
the Garlock fault in California (Davis & Burchfiel 1973).
5.4 Summary of kinematics of the Tibetan region
Fault plane solutions of earthquakes show a radial direction
of overthrusting of the Himalaya onto the Indian Shield,
and given a convergence rate of 1 8 f 7 m m y r - ' at the
Himalaya (Lyon-Caen & Molnar 1985; Molnar 1987), this
implies west-northwest-east-southeast extension of southern
Tibet (at 18 f 9 mm yr-'). This direction is corroborated by
fault plane solutions of earthquakes in southern Tibet, and
both the rate and direction agree with those from field
148
P. Molnar and H . Lyon-Caen
t
+
-k
%
40"N
30' N
20"N
80'E
9O'E
100'E
Figure l3. Summary of rates of deformation in Asia. The rate of convergence at the Himalaya of 18 f 7 mm yr-' includes Lyon-Caen &
Molnar's (1985) estimate for the rate of underthrusting at the Himalaya, corrected for possible crustal shortening within Himalayan thrust
sheets (Molnar 1987). The rate for the Tien Shan is taken from Molnar & Deng (1984) and is based solely on the seismicity of this century. For
the strike-slip faults within and surrounding Tibet, rates are taken from work of Peltzer (1987) on the Altyn Tagh fault, from Kidd & Molnar
(1988) on the Kunlun faults, from Burchfiel ef a!. (1989a) and Zhang Peizhen et ul. (1988) for the Haiyuan fault, from Allen et 01. (1989) and
Molnar and Deng (1984) for the Xianshuihe fault, and from Armijo el al. (1989) for the right-lateral faults in southern Tibet. The rate of
extension of southern Tibet of 18 f 9 mm yr-' is deduced from the variation in orientations of slip vectors in the Himalaya (Fig. 7) and the
convergence rate of 18 f 7 mm yr-'. The rates of extension and contraction of 10 and 5 mm yr-l are based on the rates of slip in eastern Tibet,
the calculated eastward flux of material, and the seismic moments of earthquakes in Tibet showing roughly equal amounts of crustal shortening
and crustal thinning (Fig. 9). Note that the 50-60 mm yr-l of convergence between India and Siberia is partitioned into comparable fractions of
convergence at the Himalaya, at the Tien Shan, and across Tibet, with the latter manifested as eastward extrusion of material out of India's
northward path.
studies in southern Tibet (Armijo et al. 1986). Fault plane
solutions of earthquakes in the high plateau of western Tibet
show a mixture of normal and strike-slip faulting, and from
the deduction that eastern Tibet moves away from western
Tibet at roughly lomrnyr-', we infer the rate of crustal
shortening within Tibet to be roughly 5 mmyr-', with a
north-northeast orientation. Fault plane solutions show
left-lateral strike-slip faulting on planes with varying
orientations in eastern Tibet; thus, the eastward extrusion of
material out of India's northward path into the rest of
Eurasia occurs not only by eastward translation along these
strike-slip faults, but also by clockwise rotation. Using rates
of slip on the major strike-slip faults, we estimate that the
eastward component of displacement of eastern Tibet with
respect to both the Tarim Basin and India is roughly
30-40 mm yr-' at the longitude of 97"E. Approximately
10 mm yr-' seem to be due to extension within Tibet, and
the rest corresponds to slip on the Altyn Tagh fault. Fault
plane solutions of earthquakes on the northeastern and
eastern margin of Tibet show thrust faulting and crustal
shortening roughly perpendicular to the margins, but we
cannot evaluate the rate of shortening and therefore how
rapidly southeast China moves away from Tibet, if, in fact,
it does so at all. Because of clockwise rotation of material in
eastern Tibet, however, the large eastward flux of material
at 97"E, with respect to Eurasia or India, need not reflect a
large eastward translation of southeast China with respect to
these plates.
In the overall partition of India's convergence with
Eurasia, roughly one third occurs at the Himalaya, a second
The Tibetan Plateau and Microfiche GJI 99/1,2
third between southern Tibet and the Tarim Basin, and the
remainder in the Tien Shan (Armijo et al. 1989). Most of
the shortening between southern Tibet and the Tarim Basin
occurs by strike-slip displacement along the Altyn Tagh
fault. Strike-slip displacements on northwest- and
northeast-striking faults contribute the only north-south
crustal shortening within Tibet itself, apparently at a slow
overall rate of about 5 mm yr-’.
ACKNOWLEDGMENTS
This research would not have been done by us without the
help of numerous kinds from R. McCaffrey, whose program
and guidance were essential. J. B. Minster offered helpful
suggestion on an early draft, G. Ekstrom critically reviewed
the manuscript, and K. H. Jacob and I. Miele offered
hospitality during seismogram-gathering marathons at
Lamont-Doherty Geological Observatory. One of us (P.M.)
worked in three different institutions, while the work was
carried out. We analysed the first 11 earthquakes in 1987
while he was in France and was supported in part by the
Centre National de la Recherche Scientifique. We studied
the 18 earthquakes occurring in and near Tibet since early
1977, while he was at M.I.T. in 1988, and the nine
earthquakes from the Himalaya, while he worked at
Oxford, supported in part by the National Environmental
Research Council and by a fellowship from the Royal
Society. Throughout the duration of the work, he received
support from the National Aeronautical and Space
Administration through grant NAG5-795, and by the
National Science Foundation of the US through grant
8500810-EAR. H.L.-C. was supported by the CNRS.
REFERENCES
Allen, C. R., Gillespie, A. R., Nan Yuan, Sieh, K. E., Zhang
Buchang & Zhu Chengnan, 1984. Red River and associated
faults, Yunnan Province, China: Quaternary geology, slip
rates, and seismic hazard, Geol. SOC.A m . Bull., 95, 686-700.
Allen, C. R., Luo Zhuoli, Qian Hong, Wen Xueze, Zhou Huawei
& Huang Weishi, 1989. Field study of a highly active fault
zone: The Xianshuihe fault of southwestern China, Geol. SOC.
Am. Bull., submitted.
Armijo, R., Tapponnier, P., Mercier, J . L. & Han, T., 1986.
Quaternary extension in southern Tibet: Field observations and
tectonic implications, J. geophys. Res., 91, 13803-13872.
Armijo, R., Tapponnier, P. & Han Tonglin, 1989. Late Cenozoic
right-lateral strike-slip faulting across southern Tibet, 1.
geophys. Res., 94, 2787-2838.
Baranowski, J., Armbruster, J., Seeber, L. & Molnar, P., 1984.
Focal depths and fault plane solutions of earthquakes and
active tectonics of the Himalaya, J. geophys. Res., 89,
6918-6928,
Bohlin, B., 1940. Notes on the Hydrography of Western Kansu,
Reports from the Scientific Expedition to the Northwestern
Provinces of China under the Leadership of Dr. Sven Hedin,
111 Geology 3, Tryckeri Aktiebolaget Thule, Stockholm, 54 pp.
Bohlin, B., 1960. Geological Reconnaissance in Western Kansu and
Kokonor, Reports from the Scientific Expedition to the
Northwestern Provinces of China under the Leadership of Dr.
Sven Hedin, 111 Geology 8, Statens Etnografiska Museum,
Stockholm, 79 pp.
149
Burchfiel, B. C., Zhang Peizhen, Wang Yipeng, Zhang Weiqi, Jiao
Decheng, Song Fangmin, Deng Qidong, Molnar, P. & Royden,
L., 1989a. Geology of the Haiyuan Fault Zone, Ningxia-Hui
Autonomous Region, China and its Relation to the Evolution
of the Northeastern Margin of the Tibetan Plateau, Geol. Soc.
Am. Bull., in press.
Burchfiel, B. C., Deng Qidong, Molnar, P., Royden, L., Wang
Yipeng, Zhang Peizhen & Zhang Weiqi, 1989b. Geology, in
press.
Cauet, R., Serryn, P. & Vincent, M., 1957. Petit A t l a Bordas: la
France. le Monde, SCIP, Paris.
Chandra, U., 1978. Seismicity, earthquake mechanisms and
tectonics along the Himalayan mountain range and vicinity,
Phys. Earth planet. Znt., 16, 109-131.
Cui Zheng-yuan and Yang Bin, 1979. On the Tuosuohu-Maqu
active fault zone (in Chinese), Northwest. Seismol. J., 1(2),
57-61.
Dalmayrac, B. & Molnar, P., 1981. Parallel thrust and normal
faulting in Peru and constraints on the state of stress, Earth
planet. Sci. Lett., 55, 473-481.
Davis, G. A. & Burchfiel, B. C., 1973. Garlock fault, an
intracontinental transform structure, southern California, Geol.
SOC. A m . Bull., 84, 1407-1422.
Deng Qidong, Sung Fangmin, Zhu Shilong, Li Mengluan, Wang
Tielin, Zhang Weiqi, Burchfiel, B. C., Molnar, P. & Zhang
Peizhen, 1984. Active faulting and tectonics of the Ningxia-Hui
Autonomous Region, China. J. geophys. Res., 89, 4427-4445.
Dziewonski, A. M., Chou, T. A. & Woodhouse, J. H., 1981.
Determination of earthquake source parameters from waveform data for studies of global and regional seismicity, J.
geophys. Res., 86,2825-2852.
Ekstrom, G . A., 1987. A broad band method of Earthquake
Analysis, PhD Thesis, Harvard University.
Ekstrom, G. & England, P. C., 1989. Seismic strain rates in regions
of distributed deformation, J. geophys. Res., in press.
England, P. C. & Houseman, G. A., 1986. Finite strain calculations
of continential deformation 2. Comparison with the India-Asia
Collision, J. geophys. Res., 91, 3664-3676.
England, P. C. & McKenzie, D. P., 1982. A thin viscous sheet
model for Continental deformation, Geophys. J. R. astr. Soc.,
70, 295-321.
Fitch, T. J., 1970. Earthquake mechanisms in the Himalayan,
Burmese, and Andaman regions and continental tectonics in
central Asia, J. Reophys. Res. 75, 2699-2709.
Futterman, W. I., 1962 Dispersive body waves, J. geophys. Res.,
67, 5279-5291.
Houseman, G . A. & England, P. C., 1986. Finite strain calculations
of continental deformation, 1 . Methods and general results for
convergent zones, J. geophys. R e x , 86, 6115-6132.
Jackson, J. & McKenzie, D., 1984. Active tectonics of the
Alpine-Himalayan belt between western Turkey and Pakistan,
Geophys. J. R. astr. Soc., 77,185-264.
Jackson, J. & McKenzie, D., 1988. The relationship between plate
motions and seismic moment tensors, and the rates of active
deformation in the Mediterranean and the Middle East,
Geophys. J . , 93, 45-73.
Jackson, J. & Yielding, G., 1983. The seismicity of Kohistan,
Pakistan: Source Studies of the Hamran (1972.9.3), Dare1
(1981.9.12)
and
Patan
(1974.12.28)
earthquakes,
Tectonophysics, 91, 15-28.
Jones, L. M., Han, W., Hauksson, E., Jin, A., Zhang, Y. & Luo,
Z., 1984. Focal mechanisms and aftershock locations of the
Songpan earthquakes of August 1976 in Sichuan, China, J.
geophys. Res., 89, 7697-7707.
Kidd, W. S. F. & Molnar, P., 1988. Quaternary and active faulting
observed on the 1985 Academia Sinica-Royal Society
geotraverse of Tibet, Phil. Trans. R. SOC.Lond., Series A , 321,
337-363.
Kostrov, V. V., 1974. Seismic moment, and energy of earthquakes,
and the seismic flow of rock (in Russian), Zzu. Acad. Sci. USSR
Phys. Solid Earth, 1,23-44.
Langston, C. A. & Helmberger, D. V., 1975. A procedure for
modelling shallow dislocation sources, Ceophys. 1. R. astr.
soc., 42, 117-130.
Lee, W. H . K., Wu, F. T. & Jacobsen, C., 1976. A catalog of
150
P . Molnar and H . Lyon-Caen
historical earthquakes in China compiled from recent Chinese
publications, Bull. seism. Soc. A m . , 66, 2003-2016.
Li Long-hai & Jia Yun-hong, 1981. Characteristics of the
deformation band of the 1937 Tuosuohu earthquake (M = 7.5)
in Qinghai (in Chinese), Northwest. Seismol. J., 3(3), 61-65.
Lyon-Caen, H. & Molnar, P. 1985. Gravity anomalies, flexure of
the Indian plate, and the structure, support and evolution of
the Himalaya and Ganga Basin, Tectonics, 4, 513-538.
Malverne, L. E., 1969 Introduction to the Mechanics of a
Continuous Medium, Prentice-Hall, Englewood Cliffs, N. J.
McCaffrey, R. & Abers, G., 1988. SYN3: A program for inversion
of teleseismic body wave forms on microcomputers, Air Force
Geophysics Laboratory Technical Report, AFGL-TR-88-0099,
Hanscomb Air Force Base, MA, 50 pp. 1988.
McCaffrey, R. & Nabelek, J., 1987. Earthquakes, gravity, and the
origin of the Bali Basin: an example of a nascent continental
fold-and-thrust belt, J. geophys. Res., 92, 441-460.
McKenzie, D. &Jackson, J . , 1983. The relationship between strain
rates, crustal thickening, paleomagnetism, finite strain and fault
movements within a deforming zone, Earth planet. Sci. Letr,
65, 182-202.
Minster, J. B. & Jordan, T. H., 1978. Present-day plate motions, J.
geophys. Res., 83, 5331-5354.
Molnar, P., 1987. Inversion of profiles of uplift rates for the
geometry of dip-slip faults at depth, with examples from the
Alps and the Himalaya, Ann. Geophys., 5 , 663-670.
Molnar, P. & Chen, W.-P., 1982. Seismicity and mountain building,
in Mountain Building Processes, pp. 41-57 ed. Hsii, K . ,
Academic Press, London.
Molnar, P. & Chen, W.-P., 1983. Focal depths and fault plane
solutions of earthquakes under the Tibetan plateau, 1. geobhys.
Rex, 88, 1180-1196.
Molnar, P. & Deng Qidong, 1984. Faulting associated with large
earthquakes and the average rate of deformation in central and
eastern Asia, J. Geophys. Res., 89, 6203-6227.
Molnar, P. & Lyon-Caen, H., 1988. Some simple physical aspects
of the support, structure, and evolution of mountain belts, in
Processes in Continental Lithospheric Deformation, Geol. Soc.
Amer. Spec. Pap. 218, 179-207.
Molnar P., Tapponnier, P., 1975. Cenozoic tectonics of Asia:
Effects of a continental collision, Science, 189, 419-426.
Molnar P., Tapponnier, P., 1978. Active tectonics of Tibet, J.
geophys. Res. 83, 5361-5375.
Molnar, P., Fitch, T. J. & Wu, F. T., 1973. Fault plane solutions of
shallow earthquakes and contemporary tectonics of Asia, Earth
planet. Sci. Letr., 16, 101-112.
Molnar, P., Chen, W.-P., Fitch, T. J., Tapponnier, P., Warsi, W.
E . K. & Wu, F. T., 1977. Structure and tectonics of the
Himalaya: A brief summary of relevant geophysical observations, in Collogue lnternationaux du CNRS, No. 268,
Himalaya: Sciences de la Terre, pp. 269-294, Editions du
Centre National de la Recherche Scientifique, Paris.
Molnar, P., Burchfiel, B. C., Liang K'unangyi & Zhao Ziyun, 1987.
Geomorphic evidence for active faulting in the Altyn Tagh and
northern Tibet and qualitative estimates of its contribution to
the convergence of India and Eurasia, Geology, 15, 249-253.
Nabelek, J., 1984. Determination of earthquake source parameters
from inversion of body waves, PhD Thesis, M.I.T.
Nelson, M. R., McCaffrey, R. & Molnar, P., 1987. Source
parameters for 11 earthquakes in the Tien Shan, Central Asia,
determined by P and S H waveform inversion, J. geophys. Res.,
92, 12629-12648.
Ni, J. & Barazangi, M., 1984. Seismotectonics of the Himalayan
collision zone: Geometry of the underthrusting Indian plate
beneath the Himalaya, 1. geophys. Res., 89, 1147-1163.
Ni, J. & York, J. E., 1978. Late Cenozoic extensional tectonics of
the Tibetan Plateau, J . geophys. Res., 83, 5377-5387.
Nicholson, C., Seeber, L., Williams, P. & Sykes, L. R., 1984.
Seismic evidence for conjugate slip and block rotation within
the San Andreas fault system, southern California, Tectonics,
5, 629-648.
Peive, A. V., Burtman, V. S., Ruzhentsev, S. V. & Suvorov, A. I.,
1964. Tectonics of the Pamir-Himalayan sector of Asia, in
Report of the Twenty Second Session, India, International
Geological Congress, Part XI, Himalayan and Alpine Oro-
geny, pp. 441-464, International Geological Congress, New
Delhi.
Peltzer, G., 1987. Contribution ti I'ttude de la collision Inde-Asie,
2) Estimation des deplacements quaternaires sur le bord nord
du Tibet, Thbe d'Etat, UniversitC de Paris.
Peltzer, G., Tapponnier, P., Zhang Zhitao & Xu Zhiqin, 1984.
Neogene and Quaternary faulting in and along the Qinling
Shan, Nature, 317, 500-505.
Peltzer, G., Tapponnier, P., Gaudemer, Y., Meyer, B., Guo
Shunmin, Yin Kelun, Chen Zhitai & Dai Huagung, 1987.
Offsets of late Quaternary morphology, rate of slip and
recurrence of large earthquakes on the Chang Ma fault (Gansu,
China), J . geophys. Res., 93, 7793-7812.
Pettersen, 0. & Doornbos, D. J., 1987. A comparison of source
analysis methods as applied to earthquakes in Tibet, Phys.
Earth planet. Int., 47, 125-136.
Rastogi, B. K., 1974. Earthquake mechanisms and plate tectonics in
the Himalayan region, Tectonophysics, 21, 47-56.
Rothery, D. A. & Drury, S. A., 1984. The neotectonics of the
Tibetan Plateau, Tectonics, 3, 19-26.
Sebrier, M., Mercier, J. L., MCgard, F., Laubacher, G. &
Carey-Gailhardis, E., 1985. Quaternary normal and reverse
faulting and the state of stress in the central Andes of south
Peru, Tectonics, 4, 739-780.
Sebrier, M., Mercier, J. L., MacherC, J., Bonnot, D., Cabrera, J.
& Blanc, J . L., 1988. The state of stress in an overriding plate
situated above a flat slab: the Andes of central Peru, Tectonics,
7, 895-928.
Seeber, L., Armbruster, J. & Quittmeyer, R., 1987. Seismicity and
continental collision in the Himalayan arc, in Zagros,
Hindu-Kush, Himalaya, Geodynamic Euolution, Geodynamics
Series, Vol. 3, pp. 215-242, American Geophysical Union,
Washington, DC.
SuArez, G., Molnar, P. & Burchfiel, B. C., 1983. Seismicity, fault
plane solutions, depth of faulting and active tectonics of the
Andes of Peru, Ecuador and southern Colombia, J. geophys.
Res., 88, 10403-10428.
Tang Rongchan, Huang Zuzhi, Qian Hong, Deng Tiangang, Jiang
Lenqiang, Ge Peiji, Liu Shengli, Cao Yangguo & Zhang
Chenggui, 1984. On the recent tectonic activity and earthquake
of the Xianshuihe fault zone, in A Collection of Papers of the
International Symposium on Continental Seismicity and
Earthquake Prediction, pp. 347-" $3, Seismological Press,
Beijing.
Tapponnier, P. & Molnar, P., 1976. ;?-line field theory and
large-scale continental tectonics, Nai' . e, 264, 319-324.
Tapponnier P. & Molnar, P., 1977. Active faulting and tectonics in
China, J. geophys. Res., 82, 2905-2930.
Tapponnier, P., Mercier, J . L., Armijo, R., Han, T. & Zhou J.,
1981. Field evidence for active normal faulting in Tibet,
Nature, 294, 410-414.
Tapponnier, P., Peltzer, G., Le Dain, A. Y., Armijo, R. &
Cobbold, P., 1982. Propagating extrusion tectonics in Asia:
New insights from simple experiments with plasticine, Geology,
10, 611-616.
Tapponnier, P., Peltzer, G. & Armijo, R., 1986. On the mechanics
of the collision between India and Asia, in Collision Tectonics,
pp. 115-157, eds Coward, M. P. & Ries, A. C., Geological
Society, London.
Vilotte, J. P., Madariaga, R., Daignitres, M. & Zienkiewicz, O.,
1986. Numerical study of continental collision: influence of
boundary forces and an initial stiff inclusion, Geophys. J. R.
astr. Soc., 89, 279-310.
Zhang Peizhen, Molnar, P., Burchfiel, B. C., Royden, L., Wang
Yipeng, Deng Qidong, Song Fangmin, Zhang Weiqi & Jiao
Decheng, 1988. Bounds on the Holocene slip rate of the
Haiyuan fault, north-central China, Quaternary Res., 30,
151-164.
Zhang Peizhen, Burchfiel, B. C., Molnar, P., Zhang Weiqi, Jiao
Decheng, Deng Qidong, Wang Yipeng, Royden, L. & Song
Fangmin, 1989. Amount and style of late Cenozoic
deformation in the Liupan Shan area, Ningxia Autonomous
Region, China, Geol. SOC.A m . Bull., in press.
Zhang Peizhen, Burchfiel, B. C., Molnar, P., Zhang Weiqi, Jiao
Decheng, Deng Qidong, Wang Yipeng, Royden, L. & Song
The Tibetan Plateau and Microfiche GJI 99/1,2
which he did not use broad-band P phases. Summaries of
these figures are given in the captions.
The most notable difference in fault plane solutions is for
the earthquake of 1977 November 18 (Fig. B14), for which
we obtained a large strike-slip component and Ekstrom
obtained a large normal component, regardless of whether
or not he included the two broad-band P phases. We had
rejected a solution similar to his, because it does not fit the
numerous well recorded, but relatively small, P phases at
stations to the northwest in Europe. Ekstrom specifically
singled this earthquake out as one that he had difficulty
matching the two broad-band P phases (both of which are
from stations to the east) with the CMT solution, and we
suspect that there were too few digital stations operating at
the time of this earthquake to yield a reliable solution.
Among the other notable disagreements, we obtained a
large component of strike-slip faulting for the earthquake of
1978 July 31 (Fig. B16), and he reported a large normal
component, with a T-axis in the same azimuth, using only
the CMT solution. For a strike-slip solution, the polarities
of SH and sSH are the same, but for his solution with one
steep plane and one gentle one, the polarities are opposite
to one another and combine to form signals that are out of
phase with those that we digitized. Also, his solution yields
relatively large P-waves compared with those of SH. Thus
we are confident in rejecting his solution. For two
earthquakes with large strike-slip components, his solutions
and ours are similar, except that the dip directions differ.
Fangmin, 1989b. Late Cenozoic tectonic evolution of the
Ningxia-Hui Autonomous Region, China, Geol. SOC.Am.
Bull., in press.
Zhang Weiqi, Jiao Decheng, Zhang Peizhen, Molnar, P., Burchfiel,
B. C., Deng Qidong & Wang Yipeng, 1986. Displacement
along the Haiyuan fault associated with the great 1920
Haiyuan, China, earthquake, Bull. sebm. SOC. Am., 77,
171-181.
Zhou Huilan, Liu, H.-L. & Kanamori, H., 1983a. Source processes
of large earthquakes along the Xianshuihe fault in southwestern China, Bull. sebm. SOC.A m . , 73, 537-551.
Zhou Huilan, Allen, C. R. & Kanamori, H., 1983b. Rupture
complexity of the 1970 Tonghai and 1973 Luhuo earthquakes,
China, from P-wave inversion, and relationship to surface
faulting, Bull. sebm. SOC.Am., 73, 1585-1597.
APPENDIX A COMPARISON O F OUR
SOURCE PARAMETERS WITH THOSE OF
EKSTROM (1987)
We studied 27 earthquakes for which Ekstrom (1987)
reported source parameters. Among those 27 events, he
used broad-band P phases for 22 earthquakes and
redetermined CMT source parameters (Dziewonski et al.,
1981) using arbitrarily chosen, but reasonable, fixed depths
for the remaining five events. Figures Al-A3 show
a comparison of his source parameters with ours; in all cases
those points marked with an X indicate earthquakes for
t
/
I
-
151
=
i
7711119
I
/
I
/
b
0
5
-
1
1
i
7711 11 18
10
Depth (km)
I
15
(Molnar and Lyon-Caen)
Figure Al. Our depths (horizontal axis) and those of Ekstrom (1987). In only three cases do Ekstrom’s depths lie outside the ranges that we
determined. In one case, the earthquake of 1978 July 31, the fault plane solutions differ significantly, and for the others, he could construct
only one or two broad-band P phases for which one of p P or sP is not very clear.
152
P. Molnar and H . Lyon-Caen
Figure A2. Differences between our strikes (top), dips (middle) and rakes (bottom) and those of Ekstrom (1987), plotted as a function of our
scalar seismic moment. In all cases we chose the strike, dip and rake of the plane of his best fitting double-couple solution that most closely
agreed with ours. Of 81 differences, 55 (68 per cent) lie within the uncertainties that we assigned, and 76 (94 per cent) lie within twice those
uncertainties. Although our uncertainties were not determined in a statistically rigorous fashion, this comparison suggests that on the average
our estimated uncertainties are between one and two standard deviations.
The Tibetan Plateau and Microfiche GJI 99/1,2
I
10"
Seismic Moment (Nt-m)
1O W
(Molnar
and Lyon-Caen)
Because he could construct only one broad-band P phase for
one event (1985 May 20) (Fig. B32) and none for the other
(1986 August 20) (Fig. B37), we suspect that his dips are not
well constrained, but we may have underestimated our
uncertainties. The strikes of our planes differ by roughly 30"
for one of the earliest events studied by Ekstrom (1977
January 1) (Fig. B12) and for another large well recorded
earthquake (1981 September 12) (Fig. B29), both with large
components of thrust faulting on planes dipping roughly 30"
to 60".The shapes and amplitudes of SH phases constrain
153
F i r e A3. Our scalar seismic moments (horizontal axis) versus
those of Ekstrom (1987) (vertical axis). All but two of our seismic
moments differ by less than a factor of 2 of Ekstrom's, and most are
within a factor of 1.5 of each other. In nearly all cases, his moments
are the larger, on the average by about 1.3 times. Part of this might
be because he used a smaller r* (0.6s) for P-aves than did we
(l.Os), but the main reason derives from his use of longer periods,
which are a more stable indicator of the moment. Hence, on the
average, his scalar moments are slirely better than ours.
our solutions, and we think that they rule out his solutions,
but again perhaps we have been overzealous in assigning
uncertainties. Finally for two very small events (1980 June 1,
Fig. B20, and 1986 July 16, Fig. B36), the orientations of
our planes differ by large amounts.
Ekstrom's and our fault plane solutions agree best for
those events since 1978, with seismic moments in excess of
2 x lO"Nm, and for which he could use two or more
broad-band P phases.