Himalayan Seismic Hazard

See allHide authors and affiliations

Science  24 Aug 2001:
Vol. 293, Issue 5534, pp. 1442-1444
DOI: 10.1126/science.1062584

F ive major earthquakes have visited India in the past decade (1), culminating in the devastating Bhuj earthquake of 26 January 2001. That earthquake in particular called attention to the hazards posed by buildings not designed to withstand major but obviously probable earthquakes. It also focused the eyes of the public away from a part of India where even worse damage and loss of life should be expected—the Himalayan arc (see the figure). Several lines of evidence show that one or more great earthquakes may be overdue in a large fraction of the Himalaya, threatening millions of people in that region.

Danger zone.

This view of the Indo-Asian collision zone shows the estimated slip potential along the Himalaya and urban populations south of the Himalaya (United Nations sources). Shaded areas with dates next to them surround epicenters and zones of rupture of major earthquakes in the Himalaya and the Kachchh region, where the 2001 Bhuj earthquake occurred. Red segments along the bars show the slip potential on a scale of 1 to 10 meters, that is, the potential slip that has accumulated since the last recorded great earthquake, or since 1800. The pink portions show possible additional slip permitted by ignorance of the preceding historic record. Great earthquakes may have occurred in the Kashmir region in the mid-16th century (21) and in Nepal in the 13th century (8). The bars are not intended to indicate the locus of specific future great earthquakes but are simply spaced at equal 220-km intervals, the approximate rupture length of the 1934 and 1950 earthquakes. Black circles show population centers in the region; in the Ganges Plain, the region extending ∼300 km south and southeast of the Himalaya, the urban population alone exceeds 40 million. (Inset) This simplified cross section through the Himalaya indicates the transition between the locked, shallow portions of the fault that rupture in great earthquakes and the deeper zone where India slides beneath southern Tibet without earthquakes. Between them, vertical movement, horizontal contraction, and microearthquake seismicity are currently concentrated (46).

A wealth of geophysical evidence demonstrates that south of the Himalaya, the top surface of India's basement rock flexes and slides beneath the Himalaya—not steadily but in lurches during great earthquakes (see the inset in the figure) (2, 3). This pattern resembles that found where lithospheric plates beneath oceanic regions converge rapidly: At deep-sea trenches, where the ocean floor flexes down seaward of the trench, the entire oceanic lithosphere plunges deep into Earth's mantle, and great earthquakes occur most commonly. Extreme examples are the great earthquakes in Chile in 1960 and in Alaska in 1964. Only during such earthquakes does the entire plate boundary rupture.

Second, Global Positioning System (GPS) measurements show that India and southern Tibet converge at 20 ± 3 mm/year (4). A 50-km-wide region centered on the southern edge of the Tibetan Plateau strains to absorb about 80% of this convergence. This region also shows localized vertical movement (5), and small earthquakes are most common here (6). The surrounding Himalaya accommodates the remaining 20%. Two meters of potential slip in earthquakes thus accumulate each century. In contrast, control points in southern India and southernmost Nepal approach each other no faster than a few millimeters per year (7). As the Bhuj earthquake shows, this deformation, although slow, is far from negligible.

Third, in the Himalaya, the potential slip accumulates almost entirely as elastic rather than inelastic strain, which would permanently deform the rock. Analyses of deformed river terraces in the foothills of the Himalaya demonstrate an advance of 21 ± 3 mm/year in southern Nepal (8) during the past 10,000 years. The minor difference between this rate, measured at the southern edge of the Himalaya and applicable to durations spanning many great earthquakes, and the 20 ± 3 mm/year measured with GPS implies that at most a small fraction (<10%) of the strain could be inelastic. Earthquakes must therefore release most, if not all, of India's 2 m per century convergence with southern Tibet.

Little is known about Himalayan earthquakes in the 18th century and before. Great earthquakes in the Himalayan region occurred in 1803, 1833, 1897, 1905, 1934, and 1950 (see the figure). The 1803 earthquake caused damage between Delhi and Lucknow. Recent reevaluations of the 1833 Nepal (9) and 1905 Kangra earthquakes (10, 11) indicate that rupture lengths were less than 120 km, smaller than previously believed (2, 12). An analysis of geodetic deformation during the 1897 earthquake (13) confirms that it occurred 100 km south of the Himalaya and therefore did not relieve strain in that belt. Thorough studies of the destruction and thus the intensity of shaking for the 1934 Bihar-Nepal earthquake were carried out in Nepal (14) and India (15). Together with geodetic constraints (16), they imply that a 200- to 300-km-long segment of eastern Nepal ruptured (17). Similarly, locations of aftershocks of the 1950 Assam earthquake imply a rupture zone ∼200 km long, with complexities at its eastern end (3, 18).

Although the major earthquakes that have occurred along the Himalaya since 1800 differed in dimensions, there is no doubt that they destroyed vast regions along the front of the Himalaya. More important today, however, is that less than half of the Himalaya (see the figure) has ruptured in that period.

Surface ruptures have not been found for any of these events. There are thus no geological constraints of recent ruptures, and geologists are concerned that paleoseismic investigations across Himalayan surface faults may yield misleadingly long recurrence intervals. Moreover, repeat surveys of trigonometrical points installed before the 1905, 1934, and 1950 earthquakes have yet to be made with modern techniques. The amplitudes of long-period seismic waves have provided quantitative measures of the seismic moments (a measure of earthquake size) of the 1934 and 1950 earthquakes (18). Knowledge of the lengths of the ruptures and sensible estimates of the width from various sources yield ∼4 m of slip in 1934 and ∼8 m of slip in 1950 (19). Uncertainties in these estimates permit slip as small as 2 m in 1934 and as high as 16 m for 1950, but such amounts would be unusual for earthquakes of their magnitude. These less direct measurements thus imply an average slip of ∼4 m during great earthquakes.

Despite the diverse quality of data in the past two centuries, we can be sure that we are not missing any great event since 1800. This permits us to estimate the minimum slip potential that has accumulated along the Himalaya since the last great earthquake (see the figure). We divide the central Himalaya into 10 regions, with lengths roughly corresponding to those of great Himalayan ruptures (∼220 km). With a convergence rate of 20 mm/year along the arc, six of these regions currently have a slip potential of at least 4 m—equivalent to the slip inferred for the 1934 earthquake. This implies that each of these regions now stores the strain necessary for such an earthquake. Moreover, the historic record (2022) has no great earthquake throughout most of the Himalaya since 1700, suggesting that the slip potential may exceed 6 m in some places.

Given that geological investigations of the 1905 and 1934 ruptures did not reveal surface ruptures but that river terraces have been warped and the foothills have grown during prehistoric great earthquakes, we cannot rule out the possibility that parts of the Himalaya have not ruptured in major earthquakes for 500 to 700 years and will be associated with slip exceeding 10 m. The mid-Himalayan 20th century earthquakes would then have been atypically small.

The weakest link in the arguments above is the uncertainty in the amount of slip during great earthquakes. Yet, because the longer the time since the previous earthquake, the larger the potential slip will be to drive the next one, the more severe those less frequent great earthquakes will be. Even if only one segment has stored potential slip comparable to that of the 1950 Assam earthquake, the largest intracontinental earthquake in recorded history (19), a replication of that earthquake along the more populous segments of the Himalaya would be devastating.

The population of India has doubled since the last great Himalayan earthquake in 1950. The urban population in the Ganges Plain has increased by a factor of 10 since the 1905 earthquake, when collapsing buildings killed 19,500 people (10). Today, about 50 million people are at risk from great Himalayan earthquakes, many of them in towns and villages in the Ganges plain. The capital cities of Bangladesh, Bhutan, India, Nepal, and Pakistan and several other cities with more than a million inhabitants are vulnerable to damage from some of these future earthquakes.

The enforcement of building codes in India and Pakistan mitigates the hazards to this large population, but a comparison between fatalities in the 1819 Kachchh and 2001 Bhuj earthquakes is not encouraging. The population of Kachchh has increased by a factor of 10. Two thousand fatalities occurred in 1819 (23), compared with the 19,000 confirmed fatalities this year. The implemented seismic code apparently did not lessen the percentage of the population killed. Like the Himalayan earthquakes, the Bhuj event occurred in an identified zone of heightened seismic hazard. Projecting these figures to just one of the possibly several overdue Himalayan earthquakes (for example, a repeat of the Kangra 1905 event) yields 200,000 predictable fatalities. Similar conclusions have been reached by Arya (24). Such an estimate may be too low by an order of magnitude should a great earthquake occur near one of the megacities in the Ganges Plain.

References and Notes

Navigate This Article