The great earthquake belt which stretches from the Mediterranean through the Middle East into Central Asia results from the ongoing collision between the Eurasian plate and the African, Arabian and Indian plates to the south. Through much of this belt, the topography is created and controlled by fault movement in earthquakes. Many habitations are located at the foot of the fault-controlled mountain range-fronts that bound inhospitable deserts or elevated plateaus, in positions that are favourable for trade-routes, strategic control of access or for water supply. As a result, they are vulnerable to earthquakes, which often seem to have targeted population centres precisely. For many centuries, an uneasy accommodation was reached between human needs and the earthquake-controlled landscape, sometimes brilliantly exploited by local hydrological engineering, as in Iran. Occasional earthquakes would occur, killing a shocking proportion of the population, but the populations of the settlements themselves would be relatively small. Many once-small rural communities have now grown into towns, cities or megacities, while retaining their vulnerability through poor building standards. Earthquakes that occur in these places today now kill many more than they did in the past, as we have witnessed in the last few years. Extreme catastrophes have been rare only because the exposure of modern megacities to earthquake hazards has been relatively short (approx. 50 years); an increase in the number of such catastrophes now seems to be inevitable.
1. A tale from the desert
In February 1994, the small desert village of Sefidabeh in southeast Iran was destroyed by an earthquake of moderate size (magnitude, Mw, 6.1). Most of the approximately 300 buildings in Sefidabeh collapsed, having been built of adobe, or sun-dried mud-brick, the traditional indigenous building material from which both walls and heavy roof domes are constructed. Adobe is a notoriously dangerous material in earthquakes (Ambraseys & Jackson 1981), and the fact that only six people died in this case is attributed to the lucky chance of a foreshock 24 h beforehand, and to the local time of the mainshock (11.30), so that many people were outdoors anyway. But there is more to this story.
Sefidabeh is a desperately remote and inhospitable location (figure 1a), sandwiched between the two deserts of the Dasht-e-Margo (lit. ‘desert of death’) in Afghanistan and the Dasht-e-Lut (lit. ‘barren desert’) of southeast Iran; one of the very few stops on a long, lonely trans-desert trade route between northeast Iran and the Indian ocean. It is the only habitation of any size for nearly 100 km in any direction, and yet the earthquake apparently targeted it precisely. Is this a case of extreme bad luck, or is there more to it?
With our modern technology and understanding of earthquake-related faulting, we now know exactly what happened (Berberian et al. 2000; Parsons et al. 2006). Analysis of the earthquake seismograms shows that the fault which moved was a reverse fault (or thrust) inclined at about 45° to the horizontal, in which one block is pushed towards, and on top of, another. More detailed information comes from analysis of space-based radar measurements, fixing precisely the location of the fault (a few kilometres south of Sefidabeh), its length (approx. 12 km), how deep it goes (approx. 10 km), and how much it moved (approx. 2 m). In particular, the radar analysis shows that the top of the fault was about 4 km below the ground surface; in other words, that the slip on the fault was entirely confined below ground and did not reach the surface at all.
Instead of a fault rupture or scarp at the surface, what formed instead was a fold. As a useful analogy, imagine sliding the top half of a telephone directory over the bottom half towards the binding; the slip surface would be the fault but, because of the binding, a fold develops at the end of the fault. A fault of this type, on which slip fails to reach the surface is called a ‘blind’ fault. Slip in a single earthquake is only a metre or two, but repeated earthquakes over hundreds of thousands of years cause the fold to grow into a ridge. The ridge adjacent to Sefidabeh is about 100 m high and obvious (figure 2). Furthermore, there is clear evidence in the landscape itself that the ridge is young in origin and had been growing as a result of previous earthquakes. Sefidabeh is built on an old alluvial fan, formed where an ephemeral river that used to flow through the ridge discharged its waters onto the desert plain. But as repeated earthquakes caused the ridge to grow, not just in height but also by increasing its length towards the northwest, the river had to incise a gorge through it, eventually becoming blocked, and forming a lake. Finally, the river abandoned this course altogether, and now flows round the northwest tip of the ridge instead. The old lake beds remain (figure 2a), now dry and elevated 70 m above the desert plain. From the age of the sediments within them, we can date the switch in the river course to about 100 000 years ago (Parsons et al. 2006).
Thus, well before the earthquake, all the signals were there in the landscape that Sefidabeh was in a vulnerable location—if people had only seen them and known how to read them (they had not; Sefidabeh was too remote for anyone to have noticed). Blind thrusts are very common in Iran, as the whole country is being squashed in a north–south direction between the converging Arabian and Eurasian plates at about 25 mm yr−1 (Vernant et al. 2004). But our ability to recognize the folds created at the surface by blind thrusts dates only from 1980, when one moved in the El Asnam earthquake in Algeria. Earlier devastating Iranian earthquakes of modern times that occurred on blind thrusts include those at Ferdows in 1968 (Mw 6.3, approx. 1000 killed) and Tabas in 1978 (Mw 7.3, approx. 20 000 killed). In neither of these cases were the causative faults recognized at the time, though, in retrospect, they are clear in the landscape (Walker et al. 2003). The city of Bam, destroyed in 2003 (Mw 6.8, approx. 40 000 killed) was also located on a blind thrust (which was recognized beforehand; Berberian 1976; Walker & Jackson 2002), though the faulting in that earthquake was more complicated (Talebian et al. 2004; Funning et al. 2005). In each of these cases, at Sefidabeh, Ferdows, Tabas and Bam, although we know about the causative faults, we have not yet explained why the earthquakes apparently scored bull's-eye hits on the only substantial habitations in the desert for tens of kilometres in any direction.
2. The link with water
The answer is water. Sefidabeh means ‘white water’, and the village obtains its water from the white lake beds in the uplifted ridge (figure 2c,d), which leak in a series of springs at its base. In this case, the fault is responsible for the sub-surface aquifer of the lake beds, and ensures their continual uplift and elevation above the plain, causing the formation of springs. Sefidabeh is the only place it is possible to live and attempt a meagre agricultural existence in this extremely inhospitable environment, as it is the only place with water. It is the fault that provides the water, but the fault may kill you when it moves.
This situation is common in Iran. The country is mountainous except for flat regions in the interior, which are barren salt flats. The mountains provide the water, so habitations are common at the foot of range fronts, many of which exist because they are elevated by movement on thrust faults. For centuries, the indigenous people have exploited this situation. Some horticulture is possible on the toes of alluvial fans coming off the ranges, on the finer-grained material away from the coarse debris adjacent to the steep slopes, but only if water is available. The water table is usually elevated at the range front, sometimes exaggeratedly so if there is an active fault, because repeated grinding of rocks on the fault creates a very fine, impermeable clay (called ‘fault gouge’) that can act as an underground dam to the water table, elevating it still further. Tunnels are dug, by hand, back to the range through the semi-consolidated fan material, to tap the elevated water table at the range front (figures 3 and 4). In Iran, these tunnels are called ‘qanats’, and are one of the glories of the ancient Persian civilization. They can be several tens of kilometres long, up to 100 m deep at the range front, and are marked at the surface by lines of circular craters, where vertical shafts are sunk down to the tunnels to provide access, ventilation and removal of excavated material during construction. Many have been in continuous use for centuries, and the oldest are thought to have been dug more than 2000 years ago. More details of qanat history and construction are given by Smith (1953) and Wulff (1968).
Qanats provide fresh, continuous supplies of water, with little evaporitic loss, to thousands of villages in the deserts of Iran and other countries in the Middle East and Central Asia. They are engineering wonders, and the very lifelines by which existence is possible for many. For example, the oasis city of Bam, destroyed in 2003, uses water from qanats that tap the nearby aquifer above a blind thrust fault to feed the date-growing region of Baravat; one of the most famous date-producing regions in the Middle East (figure 4a). It is this fault-controlled water supply that determines where Bam-Baravat is located, and why it was destroyed. Elsewhere in the desert, qanats bring water to fantasy pleasure-gardens, with cascading pools of water surrounded by trees and pavilions, such as at Mahan (figure 4d) and Tabas (destroyed in the 1978 earthquake).
3. Living with earthquakes
For centuries, Iranian civilization and desert existence has lived with, and exploited, this link between mountains, faulting and water supply. Three short examples serve to illustrate this relationship.
The beautiful desert oasis of Tabas, properly known as Tabas-e-Golshan (lit. ‘Tabas the flower garden’), visited by Marco Polo in the thirteenth century, was destroyed in 1978 by movement on a series of blind thrusts, whose folds are clearly visible in satellite imagery (figure 5a). The water supply for Tabas comes from the adjacent range front by qanats, which penetrate the nearby fold. The fold is also cut by the ephemeral Sardar river, which has incised a deep gorge through the rising ridge. In the past, the river caused problems of its own, when flash summer thunderstorms in the mountains produced great volumes of water that were trapped within the Sardar gorge and then discharged when the river emerged through the fold, to flood Tabas. The local response, attributed to Shah Abbas (seventeenth century), was to build a dam or water-gate, where the river leaves the mountains (figure 5b); but one in which there is a vaulted arch at the base to allow the bed-load of the river through, while limiting the head of water to a height that was manageable. This ingenious, maintenance-free solution has stood for 350 years, and is still effective today.
After the devastating 1968 earthquake at Dasht-e-Bayaz in eastern Iran (Mw 7.1, approx. 12 000 killed), several qanats were cut and offset by horizontal movement on the causative fault, which slipped up to 4 m in places. These qanats were then either abandoned or repaired by reconnecting the offset channels. There is clear evidence on the ground, and in air photos (figure 6) for earlier generations of qanats that had been offset and abandoned in previous earthquakes. Even more remarkable, are subsidiary, minor, qanat tunnels that had been dug long ago so as to feed into the main channels, and which followed precisely the line of the 1968 fault rupture. These side-tunnels exploit a change in water-table level across the fault, caused by the impermeable clay fault gouge, to tap and increase the water flow into the main tunnels. Thus the local tunnel-builders were aware, and had exploited, the fault-related hydrology for a considerable time before the modern earthquake (and before seismologists or geologists understood any of this).
Figure 7a shows a view over a desert village near the town of Ferdows (destroyed in 1968, 1000 killed), towards a fault-bounded range front. The houses are built of adobe-brick walls, with roofs either of mud-brick domes or of poplar logs laid horizontally and sealed with mud. Most of the roofs have a distinctive blueish hue, caused by the clay that is used to seal against the winter rains. The clay that is used comes from a quarry at the foot of the range front, and is in fact the fault gouge itself, made from ground volcanic rocks (figure 7b). The material is suitable for this purpose, being fine and relatively impermeable. Such blueish mud roofs are a common sight at range-front villages in eastern Iran.
4. From villages to megacities
The point of the examples cited so far is to illustrate that, for centuries, desert-rim existence in Iran had established a way of living with earthquakes. Earthquake faulting, and the topography it produces, is largely responsible for the water resources and for the locations of habitations and agriculture, as well as of some building materials. Occasionally, earthquakes moved the faults, and villages were destroyed, but the repeat times of earthquakes on individual faults are likely to be measured in thousands of years and they are most unlikely to recur on a timescale relevant for human memory. When earthquakes do occur, the destruction, and particularly the mortality, can be shocking, because of the vulnerable local building styles. Thus in the town of Tabas in 1978, more than 80% of the population (11 000 out of 13 000) were killed outright; at Bam in 2003, the figure was nearer 30%. Most places are, nonetheless, rebuilt and resettled because their location is, at the end, determined by where water is available and agriculture is possible. In the past, when rural populations were relatively small and dispersed, the frequent earthquakes of Mw 6–7 that occur in Iran would kill typically a few hundred or thousand people. A modern example is the earthquake near Zarand in February 2005 (Mw 6.4), which destroyed two villages, killing 500 (Talebian et al. 2006).
The problem is that villages grow, and have grown, rapidly, while building quality remains equally vulnerable, though it may change from weak adobe houses to poorly built multi-storey apartment blocks, and so mortality rates remain appallingly high. Thus, the village of Sefidabeh (six killed in 1978) can become the large rural town of Tabas (11 000 killed in the town in 1978; 20 000 including other villages of the oasis), or the small cities of Bam (40 000 killed in 2003; figure 8a) or Rudbar (40 000 killed in 1990), or the megacity of Tehran, which now has a daytime population of 10–12 million (figure 8b). The case of Tehran is instructive. It is situated at the base of the Alborz mountain range front (figure 8b), which is elevated by movement on an active thrust fault. Several other active faults are also situated nearby. In former times, the site was occupied by relatively small towns on a major trade route. These predecessors of modern Tehran were damaged or destroyed completely in earthquakes of probable Mw∼7 in the fourth century BC, 855, 958, 1177 and 1830 (Ambraseys & Melville 1982; Berberian & Yeats 1999), but the number killed was probably quite small by modern standards, perhaps measured in hundreds or thousands. The modern Tehran is a megacity that grew rapidly on the same site in the later twentieth century. While the Tehran site was occupied by relatively small towns, the city of Tabriz (figure 1b) was always bigger, more prosperous and far more important as trade-route crossroads. Tabriz was devastated by major earthquakes on its nearby faults in 1721 (more than 40 000 killed) and 1780 (more than 50 000 killed), at a time when the population was a small fraction of today's.
The message is clear: there is no sign that the concentration of population into large towns and cities in Iran is accompanied by a decrease in the mortality rate during earthquakes. Many major towns and cities are situated adjacent to range fronts and faults, in places that made sense when they were initiated as agricultural settlements, and they retain that vulnerability to earthquakes. In such places, earthquakes that in the past killed a few hundred or thousand people will now kill tens or hundreds of thousands, or more. The situation is similar throughout much of the Mediterranean-Middle East-Central Asia earthquake belt.
5. Earthquake vulnerability in the modern world
The link between how and where people live and earthquakes is particularly dramatic in Iran, but for many other parts of the great earthquake and mountain belts that run from Italy to China, the situation is similar. Throughout this region the topography is largely created by fault movement in earthquakes, ultimately the result of the ongoing collision between the Eurasian plate and the African, Arabian and Indian plates to the south. Large tracts of this area are either low, barren, inhospitable deserts, or high, inaccessible and also inhospitable plateaus, such as Tibet. Habitations concentrate around the edges of these regions, at the range fronts, because their locations are on trade routes, are of strategic importance controlling access, or are near water supplies. Many have been destroyed in past earthquakes when their populations were relatively small, but have now grown into very large, and very vulnerable, cities.
The relentless rise in global population and in the growth of megacities in the developing world is discussed in a number of papers by Bilham (1995, 1998, 2004). As he points out, the existence of earthquake building codes has had little effect in many of the countries concerned, and mortality rates remain shockingly high; for example, 20–35% of the population (between 240 000 and 500 000) died in the 1976 Tangshan earthquake in China. Half the world's megacities of more than 10 million inhabitants are in locations vulnerable to earthquakes (figure 9), and the reason we have not yet had an extreme catastrophe in one of them is probably because they have only been exposed for a short time (about 50 years) compared with typical recurrence times of earthquakes (hundreds or thousands of years).
Figure 9b shows the most damaging earthquakes, in terms of lives lost, of the last 1000 years. Accurate fatality figures for historical earthquakes, and even for modern extreme events like the 1976 Tangshan and 2004 Sumatra earthquakes, are notoriously difficult to obtain. Figure 9b is an updated and amended version of an earlier figure from Bilham (1995), with most of the fatality data from Munich Re (1991), updated with my own research. The map simply distinguishes earthquakes that killed more than 10 000 people from those that killed more than 100 000. At that level of discrimination, it is probably accurate, in spite of the uncertainty in the precise figures. Figure 9 is alarming because it highlights the vulnerability of cities in the Mediterranean-Middle East-Central Asia earthquake belt, as well as the coastal regions of South and Central America and Indonesia, for which tsunamis pose an additional risk, as demonstrated in Sumatra in 2004. Many of the red dots indicate earthquakes that killed more than 10 000 at times when the cities involved had populations far lower than they have today. Of the 113 earthquakes included in figure 9b over the last 1000 years, 34 occurred in the last 100 years alone.
Another way of viewing the situation is with the histogram in figure 9c, which shows the number of earthquakes killing more than 10 000 (grey) and 50 000 (red) per century for the last 1000 years. The histogram shows the relentless increase in fatalities with the expanding global population after 1600, and it is quite clear where the trend is heading (see also Bilham 2004). We are only 5 years into this century, but already we have had the catastrophies of the earthquakes in Gujarat (2001; 19 000 dead), Bam (2003; 40 000 dead), Sumatra (2004; greater than 200 000 dead, though many of these perished in the tsunami), and Pakistan (2005; 73 000 dead).
The other point to note is the great success of the developed nations in reducing the earthquake hazard of their urban populations, at least for moderate-sized earthquakes. In California, and increasingly in Japan, earthquakes of Mw 6–7, which can routinely kill tens of thousands in rural areas of the Middle East and Asia, are now principally stories about economic loss. The earthquakes at Loma Prieta in northern California (1989, Mw 7.1, 64 killed) and Northridge in southern California (1994, Mw 6.8, 50 killed) occurred in regions that would be considered urban or suburban compared with Bam (2003, Mw 6.8, 40 000 killed) or Tabas (1978, Mw 7.3, 20 000 killed in total) in Iran. Expressed as proportions of the population, the mortality figures of the Californian earthquakes are insignificant compared to those of the two in Iran (roughly 30 and 85%). This comparison is not to lessen the importance of economic loss, as that can also be devastating for developing countries. As Tucker (2004) points out, the costs of the 1989 and 1994 Californian earthquakes were 0.2 and 1% of the regional (not national) gross domestic product (GDP), while the cost of the 1972 Nicaragua and 1986 El Salvador earthquakes were 40 and 30% of those countries' entire GDP. The experiences of California and Japan are testaments to good building design codes, sensibly enforced, though whether those same designs will prove effective in earthquakes much bigger than Mw 7, in which shaking durations and ground displacements are much larger, is uncertain (Heaton et al. 1995).
6. What can be done?
An article of this sort should not end without at least some comment on what can or should be done about the earthquake vulnerability of large megacities in the developing world. The problem is immense and urgent, and most of it is both beyond the scope of this article and the competence of its author. It is a problem that is being addressed by committed individuals, agencies and professional organizations of many kinds (Tucker 2004). It involves not just steps that need to be taken to prepare for the inevitable extreme catastrophe, so as at least to attempt to cope with its consequences. As Tucker (2004) points out, it also needs a sustained determined effort at hazard mitigation and reduction before such a catastrophe occurs, and one that must embrace a large number of cities. The problem of what to do about populations that are already housed in poorly constructed apartment blocks that accompanied the rapid growth of these megacities is particularly difficult. One thing at least is clear; we can expect 2 billion people to be added to the cities of developing countries over the next 20 years (Bilham 2004; Tucker 2004). This will be the biggest construction boom in history, and for those people at least, we should try and ensure that this construction conforms to good building practise and good land-use planning. The purpose of this article is to add to the efforts of Bilham, Tucker and others to alert and inform people of the situation we now face.
G. R. Conway (Chief Scientific Adviser, Department for International Development, UK). I want to ask about siting of buildings and communities. In the case of Sefidabeh, its siting near a spring is logical. How far away would it have to be sited to reduce casualties: a mile, 10 miles?
J. Jackson. The way to reduce casualties is to improve the quality of the buildings. Bad buildings will fall down, whether they are 1 or 10 km from the fault; as shown by Tabas (1978), which was destroyed 10 km in front of the fault (figure 5a) and Hotkan (2005), destroyed 10 km behind the fault (Talebian et al. 2006). Villages have to be situated near the range-bounding faults because that is where the alluvial fans produce fine-grained sediments for agriculture. The desert areas away from the range-front faults are often barren salt flats, and no-one can live there anyway.
I take this opportunity to express my admiration for, and thanks to, Roger Bilham of the University of Colorado at Boulder and Brian Tucker of GeoHazards International (www.geohaz.org); two academic seismologists who have devoted some or all of their later careers to alerting the world to the issues only touched on in this paper, and to transferring professional knowledge of earthquake behaviour towards the practical task of addressing the appalling earthquake vulnerability of megacities in the developing world. Their grim prognoses, persistently and clearly articulated over many years, are proving all too correct. Our ability to work on the tragedy of earthquakes in Iran owes much to the vision and dedication of M. Ghorashi and M. Korehie of the Geological Survey of Iran, and to the pioneering work of N. Ambraseys and M. Berberian, with all of whom it is an honour to have been associated. This work is also supported by a NERC grant to COMET (http://comet.nerc.ac.uk). Cambridge Earth Sciences contribution ES 8511.
One contribution of 20 to a Discussion Meeting Issue ‘Extreme natural hazards’.
- © 2006 The Royal Society