Earthquakes Beneath the Himalayas and Tibet: Evidence for Strong Lithospheric Mantle

See allHide authors and affiliations

Science  25 Jun 2004:
Vol. 304, Issue 5679, pp. 1949-1952
DOI: 10.1126/science.1097324


Eleven intracontinental earthquakes, with magnitudes ranging from 4.9 to 6, occurred in the mantle beneath the western Himalayan syntaxis, the western Kunlun Mountains, and southern Tibet (near Xigaze) between 1963 and 1999. High-resolution seismic waveforms show that some focal depths exceeded 100 kilometers, indicating that these earthquakes occurred in the mantle portion of the lithosphere, even though the crust has been thickened there. The occurrence of earthquakes in the mantle beneath continental regions where the subduction of oceanic lithosphere ceased tens of millions years ago indicates that the mantle lithosphere is sufficiently strong to accumulate elastic strain.

How the mechanical strength of the continental lithosphere varies as a function of depth has far-reaching implications for geodynamics. For instance, a weaker lower crust, sandwiched between a stronger upper mantle and upper crust, would facilitate delamination of the mantle lithosphere, which in turn may cause rapid wholesale uplift of the crust or even global climate change (1). In addition, the flow of a weak lower crust is inferred to explain the lateral growth of Tibet to the east (2), emplacement of post-tectonic granites (3), and rapid exhumation and metamorphism due to denudation (4).

The occurrence of intracontinental earthquakes beneath the crust would indicate a mantle lithosphere strong enough to sustain the accumulation of elastic strain needed for generating earthquakes. Several regions of recent intracontinental convergence seem to show two distinct zones of seismicity, one in the upper crust and the other in the uppermost mantle (5). Such a pattern is particularly striking in southern Tibet, where seismicity in the upper crust is no deeper than 5 to 10 km (6). Meanwhile, extrapolations of data from experimental rock mechanics to geologic strain rates and field scales suggest that ultramafic mantle rocks are stronger than crustal rocks under similar pressure and temperature conditions (7). Thus, seismic and laboratory observations suggest a bimodal distribution in the strength of the continental lithosphere, with seismicity marking peaks of strength in the upper crust and the uppermost mantle (Fig. 1A) (8).

Fig. 1.

(A) Schematic diagram showing the variation of mechanical strength as a function of depth in the continental lithosphere (5, 21). Shaded areas mark seismogenic (earthquake-generating) regions. Strength in the brittle regime is controlled by a linear relation between shear and normal stresses (Byerlee's rule), whereas ductile strength is mainly controlled by distinct flow laws for crust and mantle materials. Maps (B and C) and cross sections (D and E) are for regions H and K highlighted in fig. S1. In (B) and (C), dashed lines mark locations where cross sections were taken (along N25°W from 34.20°N/70.95°E, and along N15°E from 35.09°N/78.54°E, for regions H and K, respectively). In (C), curves I to III mark the locations of seismic arrays whose data were used by Wittlinger et al. (17) to construct a composite image of the lithosphere from receiver functions [see (E)]. Three-digit numbers identify individual seismic stations, with triangles marking key stations. Focal mechanisms of earthquakes (large round symbols) are represented by equal-area projections of the lower and the western focal sphere in maps and cross sections, respectively. Shaded quadrants represent compression first motions of P waves, whereas solid and open circles show the P and T axes, respectively. Encircled crosses mark precise hypocenters without fault plane solutions. Other symbols and labels are explained in the legend and the caption of fig. S1. In (D), the vertical bar marks maximal variations of Moho depths predicted by Airy isostasy, if extreme ranges in density contrasts are used. (E) For direct comparison, the composite seismic image is normalized by using the same values of VP and VS from which focal depths are calculated. Moho (RF) marks the position of the Moho interpreted by Wittlinger et al. (17). The most pertinent depth of the Moho (about 70 km) is near station 114. The deepest Moho (about 80 km) is under stations 127 and 198, which is at least 200 km farther to the east of relevant earthquakes (K1, K2, K10, and K12).

Here we present a comprehensive study of focal depths, fault plane solutions, and crustal thickness in and around the Himalayan-Tibetan collision zone to provide up-to-date constraints on relative strengths of the different layers of the continental lithosphere. To ensure comprehensiveness, we examined seismograms in the public domain for all earthquakes that occurred in and around the Himalayan-Tibetan region between 1963 and 1999, focusing on large to moderate-sized events with body-wave magnitude (mb) of 5.0 and above, as relocated by Engdhal et al. (9). For events since 1978, digital seismograms are available from several global networks, including many high-resolution broadband data for events since 1990. For the other events, we used the venerable, albeit analog, data of the World-Wide Standardized Seismograph Network. For waveforms with high signal-to-noise ratios (at least 3), we determined focal depths and fault plane solutions by matching observed seismograms with synthetic seismograms generated by the WKBJ algorithm (10). In this well-established approach, key constraints come from comparing the timing and amplitude of the depth phases pP, sP, and sS with those of direct arrivals P and S (1113). The final data set contains a total of 43 events (tables S1 and S2).

Unusually deep earthquakes occurred in four broad regions beneath the Himalayan-Tibetan orogen (fig. S1). A particularly noteworthy event occurred in 1990 beneath the western Himalayan syntaxis (event H8; Fig. 1, B and D). This large (mb ∼ 6) intracontinental earthquake had a focal depth of 100 ± 8 km, which is evident from the long time intervals between distinct depth phases and direct arrivals (Fig. 2). The high signal-to-noise ratio of broadband seismograms allowed us to match synthetic seismograms with unfiltered data that are proportional to ground velocity. This approach emphasizes high-frequency signals with sharp onsets to facilitate precise timing of relevant seismic phases, leading to precise determination of focal depths.

Fig. 2.

Examples of observed (top traces of each row) and corresponding synthetic (bottom traces) seismograms. Time intervals between depth phases (pP, sP, and sS) and direct arrivals (P and S) constrain focal depths. All traces are unfiltered broadband seismograms of ground velocity. Positions of observations on the focal sphere are plotted with the same convention as that in Fig. 1.

In calculating all focal depths, we made conservative, low estimates by allowing for the effect of a thick crust (70 km) with low average P- and S-wave speeds (VP and VS) of 6.0 and 3.46 km/s, respectively [Compare values of 6.2 to 6.56 km/s and 3.56 to 3.72 km/s for VP and VS, respectively, in other studies (14, 15)]. By assuming a thick crust and low values of VP and VS, the focal depth for event H8 is likely to be underestimated by as much as 10 km, because the earthquake occurred beneath the Lesser Himalayas, where the average elevation is about 2 km. Event H8 is likely to have occurred in the mantle, because the crust is unlikely to be 90 to 100 km thick in this region or elsewhere. This earthquake occurred in a region of low background seismicity and was 150 km southward from the dense zone of earthquakes associated with remnant subduction zone(s) near the Hindu Kush. A clear east-west trend of the Hindu Kush seismic zone (HKSZ) and its concentration of seismicity between depths of 150 to 250 km [centered near 36.3°N, 71°E (fig. S1) (16)] further negates any possibility of linking event H8 with the HKSZ.

Seven earthquakes occurred at depths greater than 90 to over 100 km to the south of the western Kunlun Mountains (Fig. 1, C and E). For instance, a moderate-sized (mb ∼ 5.4) event (K13) showed long delays of depth phases with respect to direct arrivals that are comparable to those of event H8, requiring the focal depths of both events to be close to 100 km (Fig. 2). Such a depth is about 30 km below the Moho imaged by Wittlinger et al. (17) using receiver functions (near station 114, Fig. 1C). In receiver functions, there is a strong tradeoff between depth and values of VP and VS (18, 19). For direct comparison, we normalized the seismic image by using the same values of VP and VS from which focal depths were calculated. The deepest Moho in this region (about 80 km under stations 127 and 198) occurs at least 200 km farther to the east of relevant earthquakes (K1, K2, K10, and K12).

The fault plane solutions of these deep events are not characteristic of seismicity associated with subducted lithosphere, so mantle earthquakes in region K are unlikely to be along a subduction zone. For instance, over 500 km farther to the west of the cross section shown in Fig. 1E, dense seismicity forms a subvertical zone between depths of about 80 and 250 km beneath the Hindu Kush region, where fault plane solutions show a tight classic pattern of down-dip extension for subduction zones (16, 20). In contrast, mantle earthquakes under the western Kunlun Mountains show a subhorizontal distribution with no indication of either down-dip extension or compression (Fig. 1E). Furthermore, background seismicity decreases appreciably to the east of the 74°E meridian (fig. S1), suggesting that mantle earthquakes under the western Kunlun Mountains are unrelated to remnants of subduction in the Hindu Kush or recent thrust faulting along the Pamir thrust belt.

In the central Himalayas and southern Tibet, seven deep earthquakes have been documented (6, 2123). For earthquakes throughout the foreland, crustal structures are too complex to unequivocally resolve whether most of the unusually deep earthquakes occurred in the mantle lithosphere. In the hinterland, joint inversion of crustal receiver functions and dispersion of surface waves help eliminate tradeoffs between crustal thickness (H) and the average speeds of seismic waves in the crust (V) (19). These recent results show that beneath Xigaze, where three deep earthquakes occurred close to each other, crustal thickness is constrained to be no more than about 66 to 70 km, whereas focal depths are about 80 to 90 km and are therefore in the mantle (fig. S2C).

Previous estimates of crustal thickness using receiver functions (15) suggested that these earthquakes may have occurred in the lower crust (24), because crustal thickness under Tibet could reach 80 km or more. In addition to the inability of receiver functions to resolve the tradeoff between H and V, some studies did not recognize strong lateral variations in crustal thickness under Tibet (15, 24). For instance, thick crust as much as 80 km deep exists near Lhasa, but this thickened crust is more than 200 km east of where mantle earthquakes occurred (fig. S1).

We have verified the difference in crustal thickness between Xigaze and Lhasa. Group velocities of dispersed Rayleigh waves are similar over the two regions, indicating equal average seismic wave speeds in the crust. In contrast, the arrival time of P-to-S conversion off the Moho (with respect to the direct P wave) is almost 1.5 s earlier at Xigaze than at Lhasa, requiring a correspondingly thinner crust beneath Xigaze (14) (fig. S3). Pronounced east-west variations in crustal thickness under Tibet have also been independently confirmed by surface-wave tomography (25) and other detailed work on receiver functions (17, 26, 27).

Among earthquakes that occurred near the Shillong Plateau and southeastern Tibet, event S8 has the greatest focal depth, near 70 ± 6 km (fig. S2D). The fact that this event was considerably shallower than events H8 and K13 is evident from the modest time intervals between depth and direct phases (fig. S4B). Precise estimates of crustal thickness are available only in regions about 100 km to the north of the epicenter and show a large drop of 20 km from Lhasa to Linzhi (station GANZ). As such, the relative position of the hypocenter with respect to the Moho is unknown. The same ambiguity applies to earthquakes that occurred to the south of the Himalayan front. In such cases, Gaur et al. (28) used receiver functions to estimate crustal thickness whose value trades off with average seismic wave speeds in the crust. Thus, we assigned a nominal uncertainty of ±5 km to the position of the Moho (fig. S2D). It seems that seismicity concentrates near the Moho (24, 29), but the current data set is unable to resolve whether most earthquakes occurred in the uppermost mantle or in the lowermost crust beneath the Shillong Plateau and southeastern Tibet.

Throughout southern Tibet [including the Tethyan Himalaya, events T12, T8, T11, T3, T5, and S8 (fig. S2, C and D)], fault plane solutions of unusually deep earthquakes do show a consistent pattern of east-west extension. This pattern is similar to active extension near the surface and faulting associated with earthquakes in the top 10 km of the crust, but is distinct from seismicity along subduction zones (6, 30). This observation works against any hypothesis that attributes deep earthquakes in southern Tibet to processes related to subduction.

Large to moderate-sized earthquakes occur at depths of 100 km or more beneath the western Himalayan syntaxis and the western Kunlun Mountains. Such focal depths are likely to be in the mantle, indicating that the uppermost mantle of the continental lithosphere is strong enough to sustain the accumulation of elastic strain required for causing earthquakes. The thickened crust of Tibet appears to vary in thickness by up to 20 km over distances of a few hundred kilometers, so whether every unusually deep earthquake is in the mantle remains uncertain. Nonetheless, a bimodal distribution of focal depths, peaking in the shallow crust and near the Moho, strongly suggests that the two seismogenic regions of the continental lithosphere correspond to maxima in mechanical strength.

Supporting Online Material

SOM Text

Figs. S1 to S4

Tables S1 and S2

References and Notes

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article