Dams, failures and earthquakes

Dam failures

It is only recently that we have started building large dams. Our experience is, thus, largely with small ones. Even those, however, have not proved to be particularly reliable; one percent of them 'fail' every year. [1] The consequence of such failures have often been serious despite the small size of the dams involved. Thus, although the Teton dam was only 95 metres high, its collapse in December 1976 nevertheless caused the death of 14 people together with a billion dollars worth of damage; while the failure of the 23 metres high Johnstown dam in Pennsylvania led to the death of over 2,000 people. [2]

The incidence of dam failures is, for a number of reasons, likely to increase in the years to come. To begin with, as Ferdinand Budweg, a noted Brazilian engineer, points out,

"The number of new dams in countries with little or no experience in the design, construction and operation of dams, increases from year to year, and lack of experience may lead to repetition of errors and serious mistakes." [3]

Secondly, as appropriate sites for dams run out - and such sites are strictly limited - dams will increasingly be built in less and less suitable places. A case in point is the Malpasset dam near Frejus in Southern France. Consultants pointed out how unsuitable the site was and recommended that the dam be built elsewhere; however, for reasons of engineering convenience, that advice was disregarded - with terrible consequences, as it turned out, since the dam failed on 2 December 1959, causing the death of 421 people. [4]

Peru's Tablachaca dam provides another example. The dam, which produces a quarter of the country's electricity, is seriously threatened by a landslide. Indeed, 20 miles upstream of the dam is a fast moving mass with a volume of one million cubic metres. Over the last ten years, it has moved at the rate of 1mm / day, with the speed increasing to between 2mm - 4mm/day in the wet season. In 1983, however, the movement increased to 70mm/day, causing serious concern in Lima. All sorts of solutions are at present being considered, one of them involving the complete excavation of about 10 million cubic metres of earth above the dam at a cost of approximately a billion dollars. [5]

A third cause of dam failure, possibly the most common, is 'overtopping' during floods. Such 'over-topping' occurred with the Machau II dam in India in 1979 and caused the death of 1,500 people downstream. In that case, the malfunctioning of equipment contributed to the failure, as the spillway gates could not be opened in time. [6] The failure of spillways to function properly also led to the near-failure of the 400 foot Tarbela dam in Pakistan in 1975-6. In this case, design errors and possible poor construction materials were also involved.

Shoddy workmanship is fairly common in constructions put up by foreign companies in the Third World. Many of the buildings put up in the oil-rich gulf states by western contractors are already beginning to fall apart. Dr Carl Widstrand of the Scandinavian Institute of African Studies quotes two studies - one done by the government of Ethiopia, another by the Republic of Kenya - which suggest that the engineering and construction of water supply schemes has been of low quality.

"Many consultants," he writes, "have sold shoddy workmanship, second rate material and third rate engineering capacity." [7] The same has also been true in the USA. The failure of St. Francis dam in California, which led to the death of 300 people, has been attributed to faulty foundations. Design errors were apparently also largely responsible for the collapse of the Teton dam.

A further factor to be taken into account is the terrible lack of cooperation between the various organisations involved in putting up a dam. According to Dr. E. G. Giglioli, for instance, the construction of the Mwea water scheme in Kenya, "proved a fertile ground for bureaucratic antagonisms." Thus,

"the Department of Agriculture had the financial responsibility and looked after the day-to-day control and contractors and Provincial Administration dealt with settlement, settlers and labour. All the departments had different objectives: maximum agricultural production, or design for design's sake of irrigation installations or security. A constant struggle existed between the departments to achieve managerial control of the scheme." [8]

To make matters worse, governments usually insist that the construction of dams be compressed into the shortest period possible - the main reason being that the politicians involved want to ensure that it is they rather than their successors who obtain the credit. Such 'empire-building' was apparently rife during the construction of the Mwea scheme, and largely explains why it was so badly built. "The crash nature of the programme," writes Giglioli, "gave all concerned an accelerated course in the wrong ways of going about the job." [9]

So too, with Sri Lanka's Mahaweli scheme. The original programme envisaged six dams being built over 30 years: the present government, however, decided to 'telescope' the time-scale and complete the scheme within six years. As a result, the British contractors building the victoria dam are reported to have cut a lot of corners in order to construct the dam in time. Widstrand points out that such corner cutting is "a common feature of water (development) programmes."

Sabotage is also a factor to take into account. During a civil war, rebel forces can cause great embarrassment to the government by demolishing its hydropower installations. Thus, during the civil war that led to the independence of Mozambique, the rebels made various attempts to sabotage the Cabora Bassa dam which was, at that time, under construction. The rebels in San Salvador are today apparently aiming to sabotage the country's hydro-electric installations. [11] Enemy action is another relevant consideration. The US air force, for instance, destroyed hydro-electric dams in North Korea in 1953. [12]

Finally, many dams fail as a result of what Widstrand calls "the pilot project syndrome". Thus, engineers assume that the technology used to build small-scale dams can be used, with little or no modification, for putting up large dams. As the hydrologist Philip Williams points out, "The new technology of large dams is only imperfectly understood and largely relies on the extrapolation from the design of smaller dams." [14]

Similar problems of 'scaling up' have been encountered in the nuclear industry. In fact, it is interesting that Williams regards the technology of large dams as being, in many ways, comparable to that of nuclear power plants.

"Both require massive capital expenditures; both are new technologies with limited operating experience; and, for both, the consequences of catastrophic failure are large-scale devastation."

Although the hazards associated with nuclear power are now generally accepted (though this has rarely been allowed to interfere with governmental nuclear policies), those associated with the building of large dams still tend to be ignored - this despite our knowledge that the "failure of a large dam could cause the loss of hundreds of thousands of lives and billions of dollars worth of damage."

As a result, the safety of large dams is nowhere near as intensively examined as is that of nuclear power plants, for which, as Williams notes, "comprehensive risk-analysis identifying all possible failure modes are routinely undertaken." For large dams, on the other hand, "if a safety analysis is carried out at all, it usually focuses solely on the dam embankment." [15]

Earthquakes and dams

It has only recently been recognised that the pressure applied to often fragile geological structures by the vast mass of water impounded by a big dam can - and often does - give rise to earthquakes.

The first time that seismic activity was imputed to a reservoir was in California in the late 1930s. The reservoir in question was Lake Mead, which was impounded by the Boulder dam when it was closed in 1935. The main shock occurred four years afterwards, although it was preceeded by a considerable number of smaller shocks.

The incident sparked a heated debate as to whether or not there was any connection between the reservoir and the seismic activity; eventually, however, the connection was generally accepted. The Lake Mead earthquake, as Dr. David Simpson of the Lamont-Doherty Geological Observatory at Columbia University points out, thus became "the first recognised case of reservoir-induced seismicity." [16]

During the next 20 years or so there were a few isolated cases of earthquakes occurring after the impoundment of a reservoir, but those earthquakes do not seem to have been regarded as of any real significance. So much so, that, as late as 1958, Professor Richter felt able to claim that Lake Mead "represents a local condition: similar shocks were not observed in tests at other large reservoirs." [17] Thus, the earthquakes at Lake Mead were seen as being caused by freak conditions which were very unlikely to recur.

Ten years later, the situation had changed completely. Major earthquakes had occurred at four large reservoirs; at Hsinfengkiang in China in 1962, (magnitude 6.1); at Kariba, in Rhodesia in 1963 (magnitude 5.8); at Kremasta in Greece in 1966 (magnitude 6.3); and at Koyna, in India in 1967 (magnitude 6.5). [18]

At least two of those earthquakes caused deaths, injuries and a vast amount of damage to houses and other structures. At Koyna and Hsinfengkiang, the dams themselves were damaged. In addition, a flood caused by a landslide at the Vaiont dam in Italy in 1963, which was probably triggered off by seismic activity, killed 2,000 people.

By 1969 the relationship between large dams and earthquakes came under renewed scrutiny, especially after the Fourth World Congress on Earthquake Engineering, in Santiago, Chile, in January of that year. At the conference a French seismologist, Professor Jean Pierre Rothe, at that time Secretary-General of the International Association of Seismology and Physics of the Earth's Interior, presented a paper entitled 'Man-Made Earthquakes' in which he showed that the earthquakes referred to above - and a few less important ones as well - were definitely caused by the impoundment of reservoirs. [19]

If many geologists and geophysicists refused to accept the connection, argued Rothe, it was because the different incidents were considered in isolation from each other. If they were studied together, then it would become clear that the occurrence of an earthquake under a large man-made lake or in its immediate vicinity was not purely fortuitous.

It is worth considering a few of the case studies presented by Rothe.

The Hoover Dam

The Hoover Dam (originally called The Boulder Dam) is 142 metres high and the reservoir impounded by it contains a maximum of 35 Gm3 of water. Filling began in 1935.

The first shocks were felt in September, 1936. In the following year, as the water height of the lake reached 120 metres, 100 shocks were felt. In 1938, seismological stations set up in the area recorded several thousand shocks that would not otherwise be perceptible by man. On May 4th, 1939 - some 10 months after the reservoir had risen to a height of 145 metres, and when the water volume had reached its normal capacity of 35 Gm3 - a serious shock (with a magnitude of 5) occurred. Seismic activity further increased in the following years.

In all, 6,000 shocks were felt over an area of 8,000 square kilometres within a ten year period after the start of filling. In August and September 1972, two other serious shocks occurred in the area around Lake Mead. Both were of a magnitude of 4 and occurred during short periods when the volume of water stored in the lake was nearly 40 Gm3.

Significantly, there had been no reports of earthquakes in the area for 15 years prior to the filling of the lake - although the area is geologically complex (being composed of granite and gness, pre-Cambrian schists, Paleozoic formations, and Tertiary volcanic rocks) and several faults had in fact been identified bordering the lake. [20]

The Kariba Dam

The Kariba dam is 125m high, and the reservoir impounded by it covers an area of 6,649 square kilometres and contains 175 Gm3 of water. The lake overlies a region mainly formed from sediments of the Karoo and of volcanic lava dating from the upper Carboniferous and Jurassic. At the same time, numerous faults dating back to the Mesozoic era have been identified and mapped. The filling of the lake started in December 1958, and was completed in August 1963. Twenty-two shocks occurred in 1959 and 15 in 1961 - one of which attained a magnitude of 4 on the Richter scale. Thereafter, seismic activity increased rapidly; 63 shocks were registered in March, 1962, and 61 were felt in the first seven months of 1963. Indeed, as the lake rose, "the frequency and energy of the shocks increased." [21]

When the lake was eventually filled in 1963, a series of particularly strong shocks occurred. Ten epicentres were calculated by the U.S. Coast and Geodetic Survey: all were situated in the deepest part of the lake, the strongest having a magnitude of 6.1, and one of its after-shocks having a magnitude of 6. Several hundred tremors occurred in September 1963, and seismic activity then decreased - although 50 shocks occurred in 1963; 39 in 1968; and several in 1969 and 1970.

Again, it is significant that, prior to the construction of the dam, the Zambezi valley was considered aseismic. Not a single epicentre for the region appears in the relevant UNESCO catalogue, although a few weak shocks did occur upstream of the Victoria Falls before the filling of the reservoir.

The Koyna Dam, India

The Koyna Dam is 103 metres high and its reservoir has a maximum volume of 2,780 million cubic metres (Mm3 ). Filling started in 1962 and ended in 1964 when the reservoir was less than half filled. In 1963, the frequency of the shocks increased greatly, and their epicentres were all found to be either in the neighbourhood of the dam or under the reservoir. In 1964, the volume of the water in the reservoir was brought up to 2 Gm3. In the next few years, there was little seismic activity. Indeed, by the summer of 1967, it was assumed that everything had settled down and that tremors would no longer occur.

Thus, P.M. Mane, the Chief Engineer at Koyna, argued at the Ninth Congress on Large Dams, held that year in Istanbul, that the tremors were probably due to "crustal adjustments" taking place in and around the lake. [22] He went on to tell delegates to the conference: "It is gathered that such tremors gradually decrease over a period of some years and stop completely. It is hoped that it will be so (at Koyna) also."

However, on 13th September, 1967, two important shocks were felt, the first of which caused great damage to the village of Koynanagar, killing 177 people and injuring 2,300 others. There were numerous after-shocks - one of which attained a magnitude of 5.4, its epicentre being calculated to have been either in the vicinity of the dam or directly under the reservoir itself. On December 10th, 1967, another shock - of magnitude 6.4 - occurred.

Could that series of shocks and, in particular, the two more serious ones - have occurred naturally? Rothe does not think so. The area in which the Koyna dam was built, the Deccan plateau, is uniformly covered by Basaltic rocks. According to Rothe, "this 'shield' is one of the least seismic of the pre-Cambrian areas of the world, there are no known faults." [23]

In spite of that, a number of geologists and geophysicists insist that the earthquakes were unrelated to the building of the dam and reservoir. A UNESCO study also denied that the filling of the reservoir was responsible for the two major shocks of September and December 1967. Other experts maintained that although the small shocks could be attributed to the reservoir, the two important shocks could not - an argument which does not appear very convincing.

The Vaiont Dam, Italy

The Vaiont Dam is 261m high and the volume of the water contained in its reservoir is 150 Mm3. Filling of the reservoir was started in February, 1960, and the water reached its maximum height in August, 1963. The reservoir was partly emptied in 1961 and seismic activity fell to almost zero. It was then filled again, the water reaching 155m in April, 1962. Fifteen shocks then occurred, all between April and May of that year.

In September 1963, the lake level reached a height of 180m; 60 shocks were registered in the first 15 days of that month and, at the same time, an earth movement started along the slope of Mont Toc above the lake. That movement accelerated in October and caused a landslide which gave rise to a giant wave that flooded the valley beneath, wiping out several villages and killing more than 2,000 people. As in the other cases cited by Rothe, there was "a clear relation between the frequency of shocks and the progress of filling the reservoir."

Other examples

Other dams suspected of causing earthquakes include the Monteynard dam in the French Alps and the Kremasta dam in Greece. In the former case, an earthquake of magnitude 5 on the Richter scale occurred after the reservoir was filled in April 1963. At Kremasta the filling of the reservoir in April, 1965 was followed by a series of tremors culminating in a violent earthquake with a magnitude of 6.2. That earthquake destroyed 480 houses, killed one person and injured 60 others.

Dams and earthquakes: recent research

By the late 1960s, Rothe and others had documented an impressive body of evidence which linked the incidence of earthquakes to the building of large reservoirs. Indeed, in 1969, the Joint Committee on Seismology and Earthquake Engineering of the International Association of Earthquake Engineering recommended that:

"UNESCO convene as soon as possible a working group of experts, to review existing information on the seismic phenomena that have been observed to accompany, in some cases, the filling of large reservoirs, and to recommend what action, if any, should be taken by UNESCO in this matter."

As a result, a Working Group on Seismic Phenomena Associated with Large Reservoirs was set up. It met for the first time on 14th-16th December, 1970. After a lengthy discussion, its members unanimously approved the following statement:

"During the past few years, the impounding of certain reservoirs has been found to be responsible for triggering seismic phenomena, irrespective of the seismicity of the region. Characteristic examples are associated not only with recent tectonics and high seismicity but also with older and more stable masses of very early tectonics.

Up to the present time, a small number of these events, some of them with magnitudes close to 6 (Richter), have been strong enough, to cause not only widespread concern but also damage to structures, including in at least one case damage to the dam itself. Nevertheless, in most cases the filling of reservoirs has not been accompanied by any significant increase in local or regional seismicity. It is believed, therefore, that special geotechnic and/or hydrogeological conditions are required for the triggering of earthquakes of engineering importance." [24]

The working party met for a second time in December 1971. At that meeting, it was concluded:

"In the present state of knowledge regarding the seismic effects associated with reservoir loading, it is impossible to predict with certainty whether hazardous earthquakes are likely to be triggered by the filling of a large reservoir." [25]

The working party acknowledged that it had no method for estimating the stress to which the earth's crust is subjected: that it did not know the effects on underlying rock masses of injecting fluids under pressure into deep wells: and that it still needed to establish the potential activity of faults. Other questions, too, remained unanswered:

• How should the ability of dams to resist earthquakes be judged?
• What is the seismicity of those ancient geological platforms which are normally regarded as stable?
• And what is the ability of the slopes of a reservoir to resist earthquakes?

To answer these questions, the working party recommended that a research programme be undertaken. By 1974, however - after just two more meetings and a conference. [One was held in London in March, 1973 at the same time as The Royal Society's 'Colloquium on Seismic Effects of Reservoirs Impounding' (COSERI), and another in September, 1975 at Banff (Canada) at the same time as the 'First International Symposium on Induced Seismicity'.] - the working party was wound up. One can only speculate on the reasons for its demise: certainly the research programme was far from finished; certainly, too, the incidence of reservoir-related earthquakes had not abated. It appears, therefore, that UNESCO simply tired of the project.

In the meantime, further evidence linking reservoirs to earthquakes has come to light. It was originally thought that seismic activity could only occur whilst a reservoir was being filled, or immediately after it reached its maximum height. It has now been found that earthquakes also occur after a reservoir has been emptied and re-filled.

A case in point is France's 110m high Vouglans dam whose reservoir has a maximum volume of 605 Mm3 of water. Filling began in April 1968, and was completed by November 1969. The reservoir was partially emptied from December, 1970 to March, 1971, and was refilled very rapidly, reaching maximum capacity in June, 1971. Almost immediately - on 21 June - an earthquake occurred with a magnitude of 4.5. It was followed by 20 or so tremors between 21 June and 2 July. Significantly, the epicentre was situated 5 km to the south east of the reservoir. Previously, no seismic activity has been known in the region. [26]

So too, seismic activity increased after the refilling of the reservoir behind Corsica's Alensani dam. The reservoir - whose maximum capacity is 11 Mm3 - was closed in 1971 and, on 29th September of that year, there was an earthquake with a magnitude of 2.9. Six and a half years later, in April, 1978, there was renewed seismic activity which was still occurring at the end of 1980. The main shock, in 1978, was much stronger than those which occurred in 1971 and was preceeded and followed by more than 150 shocks.

Most seismologists at the time failed to relate them to the seismic activity that occurred in 1971. [27] Rothe pointed out the relationship and showed that the seismic activity of 1978 occurred after the lake had been emptied, allowed to remain empty for several months and then rapidly refilled. Significantly, the Alensani dam was only 60m high. Indeed, prior to the Alensani earthquake, it was always argued - not least by Rothe himself - that only dams over 100m in height were likely to cause seismic shocks.

Alensani disproved that theory - a point now fully admitted by Rothe. In fact, as can be seen from Table 5 (p.114), a number of other earthquakes have also been associated with dams whose height is well under 100 metres - notably Marathon, in Greece; Bajina Basta, in Yugoslovia; Clark Hill, in the United States; and Grandval, in France.

Recently, it has emerged that earthquakes can also be caused when the water level in a reservoir is lowered. The implications are clear. As David Simpson notes, "One of the obvious ways of decreasing danger downstream from the dam - the rapid emptying of the reservoir - may, in fact, increase the danger by triggering a further increase in the level of activity."

A case is point is California's Oroville dam. Since the occurrence of an earthquake in 1975, there has been regular seismic activity within a 20 km radius of the dam. That activity as P. W. Morrison of the California Department of Water Resources, observes in a paper written jointly with T.R. Toppozada of the California Division of Mines and Geology, "decreased markedly during winter and spring filling of the lake and increased during summer and fall drawdown." [28]

Interestingly enough, the 1975 earthquake also occurred during the summer drawdown which followed the refilling of the reservoir. Commenting on that association, Morrison and Toppozada argue:

"These observations suggest that filling Lake Oroville results in fault stability, but that during drawdown, instability occurs when the decrease in load stress significantly exceeds the slower decrease in subsurface pore pressure. Seismicity accompanying the summer drawdowns has decreased steadily since the August 1975 earthquake, suggesting that the rupture zone of this earthquake has been largely relieved of stress." [29]

[The relationship between the Oroville earthquake and the filling of the reservoir is not as clear as for the other major induced earthquakes. See further comment.]

Mono Lake in California provides another example of the phenomenon. Earthquakes - which tend to be of magnitude 4 or less - occur within 15 kilometres of the lake, mainly in the late summer and autumn. In a study of the relationship between those earthquakes and the water levels in the lake, Toppozada and his colleagues at the Californian Division of Mines note:

"We found that the seismicity shows a striking correlation with the seasonal depletion of Mono Lake. Seismicity is minimal when the lake level is stable or recharged slightly by inflows during the winter and spring, but increases markedly during evaporative drawdown of the lake in the summer and fall." [30]

They go on to comment:

"The relation of the seismicity at Mono Lake to the variation in lake level is remarkably similar to that at Lake Oroville, where earthquakes occur during the annual drawdown of the lake but not during the annual refilling. During the latter half of 1980, six earthquakes of approximately magnitude 5 occurred some kilometres east of Mono Lake. However, the relation of those larger and more distant earthquakes to the seismicity near Mono Lake is not known."

New information has also shown that earthquakes can be triggered off some years after the filling of a reservoir when the water level is allowed to remain relatively stable. Such an earthquake has occurred at Lake Nasser, the reservoir behind Egypt's Aswan dam. Filling started in 1964 and the lake reached its maximum water level of 177.8m in 1978: since then it has fluctuated between 171 and 177 metres. On 14 November 1981, an earthquake with a magnitude of 5.6 occurred. It was preceded by three main foreshocks and followed by a "tremendous number of after-shocks." [31]

The intensity of the earthquake was estimated at VIII [*3] near the epicentre and dropped to VI at Aswan where it caused minor damage to old buildings. The earthquake was attributed by Dr. Kebeasey and his colleagues from the Helwan Institute of Astronomy and Geophysics in Egypt, to tectonic activity. However, they admit that the lake's effect in triggering the earthquake cannot be excluded for a number of reasons, among them the fact that the area has throughout history been considered to be aseismic, while the epicentre of the earthquake borders a very wide area of the lake. [32]

New information has also cast some light - however hazy - on the nature of the geological conditions in which the building of a dam is likely to trigger off seismic activity. Rothe originally suggested that seismicity would only occur when large dams were built in specific geological conditions - such as 'diclastic' formations, where a high water loss occurs (as in the case of the Hoover dam, the Kariba dam, the Kremasta and Monteynard) or where there is 'heterogeneity' of the underlying strata. Such conditions favour the circulation of water under pressure and this tends to trigger off the shocks. [33]

Today, he considers that it is extremely difficult to establish the geological conditions under which induced-earthquakes will occur. Simpson agrees. Nonetheless, he argues that - on the basis of our limited experience - a general pattern has emerged. Thus, the potential for induced seismicity appears to be highest "in areas of strike-slip or normal faulting"; induced activity has proved to be most common "in areas of high to moderate strain accumulation"; and "areas of low-strain accumulation" such as "stable interiors or pre-Cambrian shields" appear to carry the lowest risk of induced seismicity. [34]

That said, however, it should be noted that the Aswan dam is situated in just such an area of low-strain accumulation - as are the Akosombo and Bratsk dams, both of which have also experienced induced earthquakes. Indeed, such is the paucity of our knowledge of induced seismicity that Simpson concludes:

"Since no diagnostic criteria appear to be presently available for determining the risk of triggering induced earthquakes, all 'large reservoirs' must to some extent be considered potential sources of induced activity." [35]

The actual mechanism whereby reservoirs trigger off earthquakes is not well known. Rothe suggests that

"the weight of impounded water may, in some cases, be enough to explain the triggering of the stored strain energy. Such triggering action will be favoured by the existence of layers having different deformabilities."

On the other hand;

"The raising of the water level in a reservoir may change the field of effective stresses in the rock mass, as a result of the increase in the pore pressures, and failure may occur. Such change will occur especially along joints, faults or other weaknesses allowing flow of the pore fluid. As a result of the increase of pore pressures, the normal effective stress decreases and this may trigger earthquakes; in such cases the difference between the water level reached in the reservoir and the natural water table level will be an important factor ... In both cases (action of water weight or of pore pressures), an increase in the surface area loaded by a reservoir raises the probability of the occurrence of shocks, by increasing the rock mass subjected to a given condition of stress." [36]

Rothe considers pore-pressure changes to be a more important factor in the triggering off of earthquakes than the weight of impounded water. Certainly, where earthquakes have been induced by high-pressure fluid injection into deep wells (a process often undertaken to get rid of toxic wastes) the role played by pore-pressure changes was found to be significant. Such earthquakes have occurred near Denver, at Rangely, at Matsushiro, and at Dale.

For his part, however, Simpson (along with the British seismologists D. I. Gough and W. I. Gough) argues that pore-pressure changes played only a small part in reservoir-induced earthquakes caused by high-pressure fluid injection. Thus he writes: "It should be noted ... that the increases in pore pressure involved in the case of fluid injection are very much higher than those created by a deep reservoir." [37] He also notes that those earthquakes caused by fluid injection all took place at, or very near, a fault zone.

A further unanswered question is whether reservoir-induced earthquakes actually create stresses or simply serve to release existing ones. The British seismologist, Dr. R. D. Adams, regards it as

"generally accepted that induced seismicity only releases strains already stored in the region, perhaps bringing forward in time earthquakes which would have occurred in the future. In no cases should reservoirs increase the long-term seismic energy release, and it may be that an episode of induced seismicity will be followed by a compensating period of quiescence." [38]

However, there is no reason to suppose that this is so. Simpson, for example, does not consider it at all clear "whether the reservoir changes only the timescale for the release of stress, triggering earthquakes which would eventually have occurred anyway, or whether it can modify the magnitude of possible earthquakes too." [39]

Whatever the answer to that question, it now seems clear beyond any reasonable doubt that reservoirs can trigger off earthquakes - sometimes serious ones - even in areas where there has been no previous seismic activity. As Rothe puts it, when he builds dams,

"Man plays the role of the sorcerer's apprentice: in trying to control the energy of the rivers, he brings about stresses whose energy can be suddenly and disastrously released." [40]

That knowledge, however, does not seem to have had any influence on current dam building programmes. The Indian government, for instance, is at present constructing a large dam near Tehri on the Bhagirathi river in the mid-Himalayas, an area which has been marked by considerable seismic activity. Indeed, such activity appears to be on the increase: between 1971 and 1973 an average of one or two earthquakes occurred a year; in 1974, 5 earthquakes occurred; and in 1975, there were seven.

There also appears to be heavy cracking in the rocks of the river gorge where the Tehri dam is to be built. Those rocks, according to V. D. Saklani, President of the Tehri Bandh Virodhi Sangharsh Samati, are "most unlikely to be able to bear the weight of (the) 2.62 million acre feet of water to be impounded in the lake." [41]

In the light of our present knowledge of reservoir-induced seismicity, it is difficult to see how the government of India can justify the construction of the Tehri dam. Nor is Tehri the only dam under construction which is likely to give rise to seismic activity. Worldwide, many other dams are being built - or planned - in areas known to be seismically active. It is surely only a matter of time before one of those dams causes a truly serious earthquake. If that earthquake also destroys the dam structure - thus releasing the massive volume of water impounded in the reservoir behind - it could kill tens, if not hundreds, of thousands of people in the surrounding area.

Table 5: Reservoir-induced changes in seismicity

Dam name	Location	Height of dam (m)	Volume of reservoir (Mm3)	Year of impounding	Year of largest earthquake	Magnitude or intensity
Major induced earthquakes
Koyna	India	103	2780	1964	1967	6.5
Kremasta	Greece	165	4750	1965	1966	6.3
Hsinfengkiang	China	105	10500	1959	1962	6.1
Oroville *	USA (Calif.)	236	4295	1968	1975	5.9
Kariba	Rhodesia	128	160368	1959	1963	5.8
Hoover	USA (Ariz.)	221	36703	1936	1939	5.0
Marathon	Greece	63	41	1930	1938	5.0
Minor induced earthquakes
Benmore	New Zealand	118	2100	1965	1966	5.0
Monteynard	France	155	240	1962	1963	4.9
Kurobe	Japan	186	199	1960	1961	4.9
Bajina-Basta	Yugoslavia	89	340	1966	1967	4.5 - 5.0
Nurek	USSR	317	10400	1969	1972	4.5
Clark Hill	USA (S.C.)	67	2500	1952	1974	4.3
Talbingo	Australia	162	921	1971	1972	3.5
Keban	Turkey	207	31000	1973	1974	3.5
Jocassee	USA (S.C.)	133	1430	1972	1975	3.2
Vajont	Italy	261	61	1961	1963
Grandval	France	88	292	1959	1963	V
Canalles	Spain	150	678	1960	1962	V
Changes in micro-earthquake activity
Kamafusa	Japan	46	45	1970		2.5
Pieve de Cadore	Italy	112	68	1949		2.0
Grancarevo	Yugoslavia	123	1280	1967		1.0-2.0
Hendrik-Verwoerd	S. Africa	88	5954	1970		2.0
Schlegeis	Austria	130	129	1971		0.0
Transient changes in seismicity
Oued Fodda	Algeria	101	228	1932
Camarilles	Spain	44	40	1960	1961	3.5
Piasta	Italy	93	13	1965	1966	VI-VII
Vouglans	France	130	605	1968	1971	4.5
Contra	Switzerland	220	86	1965	1965
Decreased activity
Tarbela	Pakistan	143	13687	1974
Flaming Gorge	USA (Utah)	153	4647	1964
Glen Canyon	USA (Ariz.)	216	33305	1964
Anderson	USA (Calif.)	72	110	1950

Other possible cases

Height in meters follows dam name (n.a. = not available),
USA - Shasta (183), Calif.; San Luis (116), Calif.; Palisades (82), Utah; Clark Canyon (40), Mont.; Kerr (n.a.), Mont.; Cabin Creek (n.a.), Colo.; Rocky Reach (n.a.), Wash.
Australia - Eucumbene (116); Warragamba (137).
Pakistan - Mangla (116).
Spain - EIGrado (130).
India - Kinnersani, Parambikulam, Sharavathi, Ukai, Ghirni, Mula (all n.a.).

References

1. David Henry, 'Designing for Development: What is appropriate technology for Rural Water and Sanitation', in Carl Widstrand, The Social and Ecological Effects of Water Development in Developing Countries, Pergamon, Oxford, 1978, p.365.
2. Philip Williams, 'Dam Design; is the technology faulty?' New Scientist, 2 February 1978.
3. Ferdinand Budweg, USCOLD newsletter, November 1982. Quoted by Philip Williams, 'Damming the World', Not Man Apart, October 1983, p.11.
4. J. P. Rothe, personal communication to Edward Goldsmith, 1983.
5. 'Landslide Threat to Lima's Power Dam', WorldWater, Vol. 5, No. 6, June 1982, pp.7-8.
6. 'India's worst dam disaster', Water Power and Dam Construction, November 1979. Quoted by Philip Williams, op.cit. 1983, p.11.
7. Carl Widstrand, 'Conflicts over Water', in Carl Widstrand (Ed), Water Conflicts and Research Priorities, Pergamon, Oxford, 1980, p.147.
8. E. G. Giglioli, 'The National Organization of Irrigation (Kenya)', in R. Chambers and J. Morris (Eds), Mwea, an Irrigated Rice Settlement Scheme in Kenya, Weltforum Verlag, Munich, 1973. Quoted by Carl Widstrand, op.cit. 1980, p.131.
9. Carl Widstrand, op.cit. 1980, p.132.
10. Ibid, pp.147-8.
11. 'Salvador Rebels aim for hydro dams', USA Today, April 12, 1983.
12. Philip Williams, op.cit. 1983, p.11.
13. Carl Widstrand, op.cit. 1980, p.138.
14. Philip Williams, op.cit. 1983, p.11.
15. Philip Williams, Damming the World, Philip Williams and Associates, Pier 33 North, The Embarcadero, San Francisco, 1983, p.13.
16. David W. Simpson, 'Seismicity Changes associated with reservoir loading', Engineering Geology 10 (1976) 123-150, p.123.
17. Ibid, p.123.
18. Ibid, p.124.
19. J. P. Rothe, 'Man-Made Earthquakes', Tectonophysics, 1970, pp.215-238.
20. J. P. Rothe, 'Fill a Lake, start an Earthquake', New Scientist, Vol. 39 No. 605, 11 July 1978, p.78.
21. J. P. Rothe, 'Summary: Geophysics Report' in William C. Ackermann et. al. (Eds), Man-Made Lakes, Their Problems and Environmental Effects, American Geophysical Union, Washington DC, 1973, pp.441-442.
22. P. M. Mane, 'Earth Tremors in Koyna Project area', Ninth Congress on Large Dams, Istanbul, op.cit. 1973, pp.445-447. Quoted by J. P. Rothe, op.cit. 1973, pp.446-447.
23. J. P. Rothe, op.cit. 1973, p.446.
24. UNESCO Working Group on Seismic Phenomena associated with Large Reservoirs, Report of First Meeting, UNESCO, 14-16 December 1970, SC/CONF. 200 4, paris, 6 March 1971, p.3.
25. UNESCO Working group on Seismic Phenomena associated with Large Reservoirs, Report of Second Meeting, UNESCO, 14-17 December 1971, SC-71/CONF. 42/3, p.4.
26. J. P. Rothe, Note sur les seismes de Vouglons, unpublished paper, June-July 1971.
27. J. P. Rothe, op.cit. 1973, p.445.
28. T. R. Toppozoda and P. W. Morrison, 'Earthquakes and Lake Levels at Oroville Butte Cao. California', Earthquake Notes, January-March 1981. Vol. 52 No. 1, p.27.
29. Ibid, p.28.
30. T. R. Topozoda, J. H. Bennett and C. H. Kramer, 'Earthquakes and Water Levels at Mono Lake, Mono County, California, Earthquake Notes, Jan-March 1981, Vol. 52, No. 1, p.28.
31. Jean Coulomb, 'Sismologies - Un nouvel exemple de sismicite provoqueremarques sue une note de Jean Delannay, Rene Guirand et Christian Weber', Note du 2 November 1981 de J. P. Rothe, CR Acad. Sc. \lang2057 paris. t293 C 7, December 1981, Serie II, p.953.
32. R. M. Kebeasy, M. Maamour and E. M. Ibrahim, 'Aswan Lake induced earthquakes', Bulletin of the International Institute of Seismology and Earthquake Engineering, Vol. 19, 1981, pp.155-160.
33. J. P. Rothe, op.cit. 1973, p.450.
34. David W. Simpson, op.cit. 1976, p.147.
35. Ibid, p.130.
36. J. P. Rothe, op.cit. 1973, p.452.
37. David W. Simpson, op.cit. 1976, p.141.
38. R. D. Adams, 'Incident at the Aswan Dam', Nature, Vol. 301, 6 January 1983, p.14.
39. David W. Simpson, op.cit. 1976, p.146.
40. J. P. Rothe, op.cit. 1978, p.78.
41. V. D. Saklani, 'Tehri Dam Project that spells disaster', Tehri Bandh Virodhi Sangharsh Samiti, Tehri Garhwal (undated), p.viii.