Paris, June 11th, 1999

A lot of data that has become available in the last 10 to 20 years provides a foundation for reviewing the set ideas currently used in risk analysis.

This paper addresses risks- and more particularly all human risks -associated with the construction, operation, and failure of dams, excluding indirect environmental or resettlement problems. The study is mainly based on analysis of the circumstances, causes and consequences of past accidents, and on analysis of the measures which could have avoided those accidents or reduced their gravity. Its purpose is to identify the main real risks associated with each type and height of dam, for all circumstances (construction, first filling, floods, etc.). It may be of help for several aspects:

in extensive risk analysis of very large dams-for reliably substantiating the probabilities chosen in event trees;

in simplified risk analysis of smaller dams-for focusing low-cost risk analysis on a few main risks;

for identifying possibilities for reducing these risks through low-cost structural or non-structural measures.

As there is little data available on accidents involving the 20,000 Chinese dams higher than 15 m or the 100,000 dams 10-15 metres high, this analysis is essentially based on existing data for the 17,000 dams (excluding tailings dams) more than 15 m high outside China: some 200 failures concerning these dams have been reported, but it is likely that a number of additional failures causing few or no victims have not been reported. The number of victims of failures is about the same as the number of dams built; it is lower than the number of lives lost due to work accidents during construction.

The consequences of failure vary considerably, depending on the dam type and causes of failure, because the extent of the breach and the time it takes for breaching to occur may be very different :

Eighty per cent of masonry or concrete dam failures were caused by excessive stresses in the dam (masonry) or foundation (sliding); breaches were sudden and often as much as five times wider than they were deep: the discharge of a failed 15 m high dam may be several thousand cubic metres per second. 15% of these dam failures were due to piping in the foundation (piping holes may grow quickly). Globally, masonry and concrete dams represent 30% of dams, and have caused 20% of failures but over 30% of victims.

Only 10% of fill dam failures have been sudden, due to sliding within the dam or its foundation or to liquefaction. Over 80% were caused by internal erosion (piping in dam or foundation) or external erosion by overtopping by flood or dam-break wave from upstream. In most cases, such failure took hours to occur: breaches may be narrow for cohesive fill but may become more than ten times as wide as they are deep for non­ cohesive materials (including rockfill) and large reservoirs. Failure flow is usually between 500 and 10,000 m³/s, but was over 50,000 m³/s for Teton (1976, 90 m. high), which caused few victims, and for Banquiao (1975, 25 m high) and Machu (1979, 25 m high), both of which had very wide breaches and caused thousands of victims despite their relatively moderate height.

The percentage of victims among the total populations in areas inundated by failures has varied from 0 to over 50%; 90% of victims have been due to 10% of failures. Progress in emergency planning and early-waming systems is essential for improved safety; this is facilitated by modem telecommunications, but effectiveness varies with dam type and circumstances.

Night failures and cold water increase considerably the risk of victims.

For 17,200 dams totalling about 800,000 dam-years, the 200 reported failures are dassified hereunder according to dam type and failure circumstances:

Exceptional circumstances: floods, upstream dam-break waves, earthquakes, war.

Normal circumstances: first filling, ageing (after 2 years of operation), construction.

FAILURES OF DAMS >15 m high (OUTSIDE CHINA)

Half of these failures occurred before 1950 when only 5,000 dams had been built.

The possibilities of risk assessment for these different circumstances are studied hereunder.

< 1 2 3 4 > Next