The October 8, 2005 Kashmir earthquake was the second major event with moment magnitude (Mw) ∼7.6 within 5 years to have rocked the west Indian continent and the third since 1935. Each of these events exacted a heavy toll on life and property with 35,000 dead in 1935, 20,000 in 2001, and 75,000 in 2005, leaving many more homeless and destitute. These earthquakes occurred in a specific tectonic context of the Indian plate, although powered by the same process—its approximately 2 m per century north-eastward penetration into the Tibetan plate whose elevation and bulk is sustained by the dynamics of the collision process. To the north, the collision results in massive thrust earthquakes along the Himalayas as India slides beneath Tibet. The indentation of the Indian plate into Asia results in earthquakes along its eastern and western flanks that separate it from the Asian plate by broad zones of horizontal and compressional tectonics similar to fault systems that prevail in California. To the west-southwest, the submarine plate boundary consists of a simple oceanic fracture zone between the Indian plate and the Arabian plate, and to the southeast an oceanic apron of heavier rocks dives obliquely beneath southeast Asia and the modest Andaman plate (Fig. 1). In December 2004, almost the entire southeast plate boundary ruptured in a 1600-km-long magnitude 9.2 earthquake, creating giant tsunamis that struck the Indian ocean shores and killed more than 300,000 people, raising the fatality count of the three twenty-first century Indian earthquakes far above the total of all historical events.
The 2005 Kashmir earthquake was caused by approximately 5 m of southwest-directed slip of the western Himalayan ranges along a 30° northeast dipping fault that had been mapped by geologists, but had not slipped in historical times. The resultant slip raised the mountains locally by 2.5 m. Long-term slip on the fault is clearly responsible for the presence of substantial step in the mean topography in this part of the range. The surface rupture was mapped for at least 90 km along the strike, but aftershocks suggest that the rupture in the subsurface extended 20 km farther to the northwest. The rupture zone partly filled a previously identified seismic gap at the western extremity of the Himalayan collision zone, leaving a much larger gap to its east remaining to be relieved (Fig. 2). The 2005 rupture was outside the region considered typical of earthquakes in the central Himalayas, being close to the transition from pure radially directed thrusting to strike-slip faulting along the Chaman fault system of Afghanistan.
Its epicenter at 34.402°N, 73.560°E, as determined by the U.S. Geological Survey (USGS), lay about 19 km northeast of Muzaffarabad, the capital town of the Pakistan-administered Kashmir, which also suffered the heaviest casualties. The time of its occurrence, just before 9 o'clock in the morning [03:50:38 UTC (Coordinated Universal Time)] when children were at school and older people in bed after the predawn Ramadan meal, resulted in very large numbers of people being buried under the debris of collapsed buildings. As a result, 7300 schools were destroyed. Massive landslides reamed many hill slopes, wiping out entire villages and roads and aggravating rescue missions in difficult terrain. The severity of the rupture process resulted in damage over a large area, in one case bringing down a 10-story residential building housing 60 families in Islamabad, Pakistan's capital, about 100 km south-southwest of the epicenter. This was the earthquake's farthest reach southward, targeting a structure that was very poorly built in this modern city of more than a million people. Aftershocks continued for several months, including one of magnitude 6.2. Early aftershocks were sufficiently severe to further damage already weakened structures.
Tectonics of earthquakes surrounding the Indian plate and within the Indian subcontinent
As is now well known, earthquakes are the result of a mechanical instability in the Earth's upper, colder brittle crust, where elastic strains steadily created by relative plate movements can accumulate up to about 1 part in 10,000 before fracturing rock along faults. Earthquakes are thus understood to constitute a periodic phenomenon with a cyclicity determined by the rate of strain accumulation in the region and failure strength of the potential rupture surface(s). In principle, knowledge of this strain rate and the precise strain required to promote failure would permit earthquakes to be forecast. Global Positioning System (GPS) geodesy, when compared with geological data on plate motions, indicates that the velocity of tectonic plates have changed little over the past 3 million years. We therefore have a good present-day measure of surface strain accumulation rates. The absence of definitive knowledge about rock failure conditions or the absolute level of strain, however, makes it currently impossible to identify the location and timing of future earthquakes.
Earthquakes along the northern compression boundary generally occur on a northward gently dipping interface between the underthrusting Indian plate and the overlying Himalayan stack of thrust sheets formed by slivers shaved off its relentlessly penetrating front (upper right inset in Fig. 2). The largest of these earthquakes occur on a ≈100-km-wide gently dipping surface (6–9°) lying within 20 km of the ground surface, but the Kashmir earthquake rupture was steeper and narrower and penetrated to greater depth.
Earthquakes along India's eastern boundary, such as the 2004 Andaman-Sumatra event, occur on an obliquely compressed interface separating the denser descending oceanic Indian plate from regions of lighter Asian continental plate or those of recently formed buoyant oceanic crust that form the Andaman plate. Steady descent of the old and therefore colder Indian plate beneath south Asia creates a stick-slip cycle of strain accumulation and release in the upper brittle crust. No similar megaquake is known to have occurred along this boundary in the past 200 years, but there are numerous examples of M > 7 earthquakes, which in 1881 and 1941 resulted in tsunamis along the Indian coast. Oblique motion was indeed observed on the main plate boundary during the 2004 earthquake, but moderate seismicity (Mw 7.5) in the previous century is characterized by thrust earthquakes on the subduction zone and strike-slip earthquakes on the Sumatra fault and its offshore continuation northwards. The process of separation of these components of slip is known as strain partitioning that, once established, carves out grooves on the fault zone that preferentially facilitate slip along them.
Earthquakes along India's western boundary permit northward slip of the plate relative to a promontory of Asia in Afghanistan and Baluchistan at a rate of 2–4 cm/yr. Though the plate boundary here is marked by a well-defined strike-slip fault (the Chaman fault) to the west, at places it is more than 150 km wide, and earthquakes in the Siestan ranges accommodate the compressional stress of indentation that result in a fold/thrust belt with a contraction rate of probably less than 1 cm/yr. The convergent and strike-slip components of this complex motion, as in the Nicobar/Andaman segment of India's eastern boundary, are separated into pure thrust on steep east- or west-dipping faults, or sinistral sliding motions on vertical faults. In 1931, a sequence of Mw > 7 earthquakes occurred that demonstrate the importance of strain partitioning. The 1931 Mach/Sharigh earthquakes were reverse slipping events, allowing roughly 1 m of extension of the Siestan range front. The transient decompression of the range resulted 3.5 years later in a reduction of fault-normal stress on a strike-slip fault near Quetta that apparently triggered slip in a Mw 7.7 earthquake. Although the reason for the delay is not yet understood, the association of these events is too unusual to be a coincidence and holds promise for forecasts of similar large strike-slip events to the north and to the south of the Quetta sequence.
Far from the plate boundaries, earthquakes such as the 1819 and 2001 Rann of Kachchh Mw > 7.6 earthquakes are also caused by the release of accumulated compressive strain. But since they are less frequent, their causal mechanisms are less clear. A proposed physical mechanism (Fig. 3) for earthquakes within the continent interior invokes the role of flexural stresses associated with the collision of India with the Tibetan plateau. A 450-m-high bulge has long been known to exist in central India, and was originally termed “the hidden range” by India's early geodesists because of the anomalous gravity field it produces where it crosses India's central plateau. The bulge is partly caused by the buckling forces of collision and partly by the weight of the plateau pushing down on India's northern edge, which has resulted in a 4-km depression beneath the frontal thrusts of the Himalayas, now filled by the sediments of the Punjab, Ganges, and Brahmaputra river systems. Tensile stresses occur near India's surface north of the bulge, resulting in normal faulting earthquakes with magnitudes less than Mw = 6.5. At depth below a neutral axis, stresses at around 20 km are compressive, giving rise to thrust or strike-slip earthquakes such as the 1988 Udaypur event on the Nepal-India border. A secondary depression occurs south of the bulge, and this depression results in reverse faulting near the surface and tensile normal faulting at depth. It is in this region that the devastating Latur earthquake occurred in 1993, causing 7500 deaths. Compressional stresses in a concave depression of the Indian plate are largest at the surface. It is therefore no accident that the Latur earthquake was the only earthquake on the Indian continent to produce a surface rupture.
New understanding and implications
First results from the epicenter of the Kashmir earthquake were in the form of studies of teleseismic waveforms that provided estimates of moment release, magnitude, and mechanism. These were followed shortly afterward by images of the epicentral region, using SAR (synthetic aperture radar) and ASTER (advanced spaceborne thermal emission and reflection radiometer), that confirmed the strike, the presence of near-surface rupture, and the extent of uplift. Some weeks later a sparse set of preseismic GPS observations were repeated, providing coseismic displacements south, west, and north of the epicenter. These GPS measurements confirmed the SAR data and suggested secondary faulting of unusual complexity.
In recent years, geophysicists have become accustomed to examining surface deformation through interferograms formed from pairs of temporally separated radar images, known as interferometric synthetic aperture radar (InSAR). The steep terrain and unstable slopes as well as vegetative changes of the Kashmir Himalayas meant that normal InSAR procedures were not amenable and special techniques had to be applied to recover the deformation field from ENVISAT (European Space Agency satellite) and ASTER imageries. However, the large vertical and horizontal displacements in the epicentral region facilitated these approaches. A useful guide to the methodology may be found at http://comet.nerc. ac.uk/news_kashmir.html.
While the 90-km-long northwest-southeast strip of coseismic deformation from Balakot to Muzaffrabad, gleaned from SAR, data coincide with the trace of a mapped active plate boundary fault, its rupture geometry differs markedly from those of Himalayan arc earthquakes to the east. In particular, its steeply dipping rupture plane at 37° is much larger than the ∼9° gently dipping rupture planes that are known to fit the observed data for Himalayan earthquakes elsewhere along the arc. Another significant feature of this earthquake is a prominent cluster of shallow aftershocks along a more westerly trend within the earlier-identified Indus-Kohistan seismic zone, which is offset from the main rupture by about 30 km to the southwest (Fig. 4). This implies that the Kashmir earthquake was caused by a complex rupture process involving subsurface accommodation of northeast/southwest convergence.
Analysis of GPS data around the region gathered from a half-dozen sites in June 2001 and repeated immediately after the earthquake in October-November 2005 using a composite model of the main rupture as well as that constrained by the distribution and focal mechanism of aftershocks (Fig. 4) suggest that the Kashmir rupture may have triggered a further slip of 1.8 m on a blind (subsurface) wedge thrust to its northwest. Although this result is controversial, the interpretation is consistent with the spatial distribution of aftershocks as well as the trend of the historically active Indus-Kohistan seismic zone and reveals a pattern of slip propagation across weak sedimentary strata unsuspected in Himalayan ruptures.
Significantly, all models of this rupture predict large consequent increases in the Coulomb stress to the southeast and therefore enhanced hazard potential of the entire region west of the Kangra rupture. This is best appreciated by a reference to Fig. 1 and Fig. 5 which show that the 2005 Kashmir rupture flanked by the smaller 1974 Pattan and 1885 Kashmir earthquakes filled only a small part of the large seismic gap west of the 1905 Kangra rupture that itself had been only partially relieved by the 1555 Kashmir Mw 7.6 or larger earthquake. The history of Himalayan earthquakes, which would allow one to set the date from which the strain budget along various segments of the Himalayas may be reckoned, is fairly well known since 1800 from writings in Persian, Arabic, and Tibetan texts, and travelers' accounts but less so before that time. The latter are rarely quantitative enough to estimate magnitudes and rupture areas. By piecing together the various threads of consistent accounts, a chronology of Indian plate-boundary earthquake ruptures has been attempted.
These estimates, plotted in Fig. 2, clearly draw attention to seismic hazard both in western and Central Himalayas, each bordering populous cities. A conservative estimate of the spatial extent and slip of earthquakes in the Kashmir region in 1501 and 1555 implies that the recent Kashmir earthquake released only a small part, <25%, of the accumulated slip along the segment flanked by the 1974 Pattan and the 1905 Kangra earthquake ruptures, and in so doing significantly augmented the prevailing stress at its unruptured eastern end. While we are unable to forecast when this eastern Kashmir segment will rupture and with what magnitude, there is little doubt that compressional stresses in the Panjal ranges are sufficiently mature to drive a great earthquake. Estimates of its magnitude range from a low of Mw 7.8 to a high greater than Mw 8.0. Studies of slip in trenches in the Himalayan foothills suggest that a megaquake had occurred near and west of Dehra Dun around 1400 that may have exceeded Mw 8.4 and may have extended west of the Kangra earthquake rupture into the present seismic gap.
Earthquakes in ancient times used to be considered “acts of God” through the trauma they exert on innocent populations. Today, we are no longer ignorant of the causes and probable locations of future earthquakes. And while successive earthquakes come as no surprise to scientists, each one delivers a punishing and unacceptable blow to local communities. There is thus an ironic gulf between what the scientific community knows and what society does with this knowledge.
Part of the reason for this regrettable gap is that scientists have yet to forecast earthquakes with any useful degree of credibility. For example, Fig. 2 contains no information about the timing of future earthquakes, but only about how big they could be were they to occur today. An extensive history of earthquakes extending back several thousands of years is a basic requirement for making a probabilistic assessment of hazard, so that effective measures are taken to minimize vulnerabilities of the communities exposed to earthquake risk. An authentic historical record of repeat earthquakes in India has yet to be created. This is an unfortunate situation that offers little hope of early remedy. The discovery of palaeo-earthquakes in the Himalayas in the past 10,000 years will no doubt correct some of this deficiency, albeit the uncertainties may still be large and poorly constrained.
Notwithstanding these shortcomings, much can be done to mitigate the possible adverse impact of future earthquakes on communities that are arguably at grave risk, notably those in the populous towns situated on the soft alluvial plains bordering the Himalayan plate boundary. Earthquake risk to local populations is primarily posed by buildings liable to collapse when subjected to ground shaking. Regrettably, although building codes exist in many parts of India and Pakistan, they are rarely applied stringently enough to prevent shoddy construction. Corruption can be overt in the form of contractors bribing officials not to report code infringements, or covert in the form of poor materials being incorporated into ongoing construction and hidden from inspection. More often, unsound building practices are incorporated from lack of knowledge about building assembly among residents anxious to save money. The scarcity of inexpensive wood for building houses and its replacement by concrete is much to blame for many recent disasters.
Thus, while natural hazards cannot be prevented, their disaster potential and much of the attendant risk to populations can be greatly minimized. It is essential that new construction incorporate earthquake resistance, usually incurring no more than a mere 10% increase in construction costs. If this policy were universally adopted, the replacement of building stock through old age would likely eliminate half of the world's most vulnerable buildings in the next 50 years. The retrofit of existing houses would be a more costly undertaking, especially since many structures would require rebuilding. But responsible government has an important role in retrofitting schools, hospitals, and critical support systems, whose collapse in earthquakes is inexcusable but currently widespread.
If the warning offered by the three disastrous earthquakes in India in recent years is unheeded by city planners, politicians, and architects, the disastrous impact of a future earthquake could be far more grievous than any we have yet seen. Three inexorably growing trends of our social dynamic conspire to force this conclusion: increased populations (10 to 100 times larger than that exposed to previous Indian earthquakes), increased vulnerability (the building stock is taller and far less well-constructed than ever before), and the demographic pattern of high population density in regions expected to experience abnormally high ground-shaking intensity (thick sediments near river valleys like the Ganges and Brahmaputra systems, and ancient lakes like those in the Katmandu and Kashmir valleys). Given that several Mw > 8 earthquakes are anticipated along the Himalaya, the population at risk exceeds 50 million people. An expedidited program to ensure and wherever desirable enforce the incorporation of earthquake-resistant design and construction practices, using all possible means and retrofitting of all community buildings and support systems, has the promise of safeguarding a large proportion of those potentially at risk and preventing trillions of dollars of economic damage that a future earthquake in the region may otherwise exact.
[Acknowledgements: This article has greatly benefited from numerous discussions with Rebecca Bendick, Peter Molnar, and Roger Bilham.]