Published as Chapter 9 of The Social and Environmental Effects of Large Dams: Volume 1. Overview. Wadebridge Ecological Centre, Worthyvale Manor Camelford, Cornwall PL32 9TT, UK, 1984. By Edward Goldsmith and Nicholas Hildyard.
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.  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. 
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.” 
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. 
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. 
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.  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.”  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.” 
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.” 
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.  Enemy action is another relevant consideration. The US air force, for instance, destroyed hydro-electric dams in North Korea in 1953. 
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.” 
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.” 
The scale of magnitude proposed by Richter measures the severity of earthquakes. The magnitude is defined as “the logarithm of the maximum amplitude, measured in microns, recorded on a standard Wood-Anderson seismograph 100 km from the epicentre of the disturbance”.
This original definition has been extended for use by different seismographs operating at different distances from the epicentre, and any seismological station can now attribute a magnitude to a disturbance recorded on its instruments. This magnitude (M) is related to the energy (E) in ergs developed at the focus of the disturbance by the equation:
log E = 11.4+1.5M
A disturbance that can barely be felt at the epicentre will have a magnitude of about 3 (E18 X 10  ) can bring about a small amount of damage, while the very large earthquakes, that cause extensive damage, have magnitudes between 7 and 8.6. The large earthquakes that occurred in Alaska in March, 1964, had a magnitude of 8.6, which corresponds to a value of E12 X 10  . This is some 250 million times as energetic as the small earthquake.
[see ‘Les Effets des Tremblements de terrain': Geophysique, Encyclopedie de la Picardie pp. 175-185, Paris, 1971].Back to top
The intensity of an earthquake is a measure of its visible effects on the surface. There is a macroseismic intensity scale which runs from zero to 12. The macroseismic intensity of a specific earthquake will obviously vary at different distances from its epicentre. The 12 degrees of the macroseismic scale are characterised by the following events:
- 1. An imperceptible disturbance
- 2. A disturbance noticed by only a very few people.
- 3. A disturbance perceptible to a number of people, and sufficiently strong for them to determine the direction and duration.
- 4. A disturbance felt by a number of people indoors.
- 5. A disturbance felt by all the inhabitants of the district; at night, sleepers are wakened.
- 6. People are sufficiently frightened to leave their houses; slight falls of pebble and plaster.
- 7. Chimneys fall; cracks develop in the walls of houses.
- 8. Partial distraction of some buildings.
- 9-12. Severe damage; total destruction of buildings.”
[‘Fill a lake, start an earthquake’, Professor J. P. Rothe, New Scientist 11 July 1968, vol. 39, no. 605].
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.” 
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.”  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). 
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. 
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.Back to top
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. Back to top
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.” 
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.Back to top