Tuesday, December 21, 2010

WHY DAMS FAIL?

Keeping Dams Safe
India ranks third in the world after China and USA in terms of number of dams.  The country has about 4711 completed large dams with 390 under construction. These dams play a very important role in the water resources development. In India most dams are constructed and maintained by the State Governments with a few manned by organizations like Bhakra Beas Management Board (BBMB), Damodar Valley Corporation (DVC) and National Hydro Electric Power Corporation (NHPC). Private sector is also poised to own and operate dams.

India follows International Commission on Large Dams (ICOLD) norms, where large dams are those with maximum height of more than 15 metres (from its deepest foundation to the crest).  Those between 10 and 15 metres height also are large dams if they comply with one of the following: 
  • Length of crest not less than 500 metres
  • Reservoir capacity not less than one million cubic metres
  • Maximum flood discharge not less than 2000 cubic metres/second 
  • Has specially difficult foundation problems, or 
  • Is of unusual design.
Being a critical infrastructure facility with wide ramifications, information needs for dam safety have to be available in an understandable format with all concerned. Several of dams are half-a-century old with the Centre planning legislation for Dam safety. The Mullapperiyar dam is more than 100 years old and concern about its safety is not misplaced. A proposed bill on safety of large dams called "the dam safety bill 2010" is slated for introduction during current budget session of parliament. It will "provide for proper surveillance, inspection and maintenance of dams to ensure their safe functioning"

"We have prepared a legislation which aims to establish an institutional system to regularly monitor the safety of large dams," Water Resources Secretary U N Panjiyar said. He said while some states like Bihar have their own dam safety laws, West Bengal and Andhra Pradesh have authorised Parliament to enact a law in this regard.

"As per constitutional provisions, if two or more states authorise Parliament to make a law on a state subject, the Centre can pass a law," he said, adding that "other states can pass a resolution in their legislatures to adopt the Central law. Water is a state subject".

With several private players now involved in operating dams, they would also be covered under the proposed law. Under the new bill, a Dam Safety Cell will be set up at the project level, a Dam Safety Organisation will be constituted at the state level and the Centre will carry out the overall monitoring. The owner could either be government or a private operator with obligation to allow inspections by experts to ascertain safety of the structure.

The Backdrop: In the 20th century, around 200 notable dam failures have occurred in the World, killing about 8000 people. It is notable that dam failures do occur in developed countries too. The biggest catastrophe recorded in this century had occurred in Vaiont in Italy in 1963. The accident killed about 2600 people. Another accident of nearer proportions occurred in India in 1979 about 2000 persons lost their lives when the Machhu II dam gave way. Other dam failures in the country included Ashti in Maharastra (this dam gave way twice-- in 1883 and 1933), Tigra in Madhya Pradesh (1970), Panchat (1961), Kadakwasala (1961), Nanak Sagar (1967) and Chikkahole (1972). The failure of the Malpasset dam in France in 1959 killed 421 persons and the Baffalo Creek dam in USA in 1972 claimed 125 lives.

These failures have caused severe devastation in the valleys downstream both in terms of lives lost and widespread damage to infrastructure and property.  The most common cause is extreme inflow events, exceeding the capacity of the spillway, but structural failures have also occurred at inflows less than the design flood. 50 per cent of dam failures in the World are from inadequate capacity of the spillways.

Why Dams Fail: Dams are artificial barriers designed to be capable of impounding water (wastewater, or any liquid) for the purpose of storage or control of flow. Either one or a combination of the reasons Cited hereunder can be attributed to Dams failure:
  • Overtopping meaning water levels that exceed the capacity of the dam. Due to,
  • Inadequate Spillway Design
  • Debris Blockage of Spillway
  • Settlement of Dam Crest
  • Structural failure of materials used in dam construction.
  • Foundation Defects Movement and/or failure of the foundation supporting the dam due to,
  • Differential Settlement
  • Sliding and Slope Instability
  • High Uplift Pressures
  • Uncontrolled Foundation Seepage
  • Settlement and cracking of concrete or embankment dams.
  • Piping and internal erosion of soil in embankment dams due to,
  • Piping and Seepage
  • Internal Erosion Through Dam Caused by Seepage-"Piping"
  • Seepage and Erosion Along Hydraulic Structures Such as Outlet
  • Conduits or Spillways, or Leakage Through Animal Burrows
  • Cracks in Dam
  • Conduits and Valves
  • Piping of Embankment Material Into Conduit Through Joints or Cracks
  • Inadequate maintenance and upkeep.
  • Deliberate acts of sabotage.
  • Any Other
Global Estimations: On apportioning principal causes of dam failure seen worldwide the findings are quite revealing. See below:

Overtopping: 1/3 of all dam failures globally Overtopping occurs when the level of a reservoir exceeds the capacity or height of the dam. It could be an inadequate or dysfunctional spillway or due to settlement of the dam crest. This is a direct consequence of water levels rising rapidly, without enough reaction time for operatives to take corrective action. Flashfloods, incessant heavy rains, landslide in the reservoir leading to a tsunami, or the upstream of dam collapsing can be causes. The net effect is structural failure, or   land erosion on either side of the dam, dam losing connection with its river slope embankments.

Failure by overtopping or a failed spillway is prevalent in earthen dams as seen in the Banqiao Dam. Landslides falling into storage reservoir Sending a wave of water over the top of the dam may cause dam failure. Concrete dams may survive but result in floods downstream. The Vaiont Dam in Italy in 1963 had 1900 people killed.

Foundation defects: 1/3 of all dam failures Foundation Defects due inadequacy in design might cause differential settlement of the soil underneath leading to failure. Dams built on slopes should avoid issues of instability and take care of landslides. The point  is, compromising foundation's integrity is invitation to failure. An earthquake may cause the movement of a foundation, which could trigger off failure.

Geological problems with the dam foundation Post failure findings in a case found some of the foundation rock, a conglomerate, disintegrated when immersed in water showing rock lost all its strength when saturated. This is exactly what happened as a new dam filled with water for the first time failed. The Malpasset Dam in France, which failed in 1959, is an example of foundation failure making it the first collapse of a modern, thin concrete arch dam.

Foundation failure is the main cause of concrete dam failure. A high uplift pressures and foundation seepage that goes uncontrolled may lead to dam's foundation failure. St. Francis Dam is an example. Piping and seepage: 1/5 of all dam failures Embankment dams - are generally semi-permeable. This means that high and therefore can be compromised when too much water seeps or leaks through the structure. Dam failure can occur when the structure becomes weakened from internal erosion, an effect referred to as piping. This can occur along hydraulic structures, spillways, conduits, or cracks. An animal burrowing in and around earthen dams can even cause such seepage or leakage. (Example of dam failure due to piping and seepage: Kelly Barnes Dam)

Other Reasons: Dams, which are improperly maintained or built with inadequate materials or unsound design, can result in structural weaknesses that lead to catastrophic dam failure. (Dam failure due to improper maintenance and structural weakness: Val di Stava Dam) Faults in construction methods:(Eg inadequate compaction of fill) or use of the wrong type of construction materials (eg silt) may lead to internal erosion or piping failures of embankment dams. An example is the Teton Dam failure in Idaho, USA in 1976.

Failures - Bad Dreams of the Past?
It would be worthwhile to stop and glance at these facilities that we built. Starting from the industrialized economies that have been ahead of all in most events in history. We can take a close look at the U.S.A. As on date, the National Performance of Dams Program (NPDP) stats show that, "at least 85% of the more than 75,000 dams in the US will be in excess of 50 years old by 2020." Further it adds that, "Perhaps more significantly, most of the large dams throughout the US are also approaching old age. Other nations are not far behind even if exposure scale is much lower. The message is loud and clear. Dam safety needs to be addressed on an urgent basis.

Dam Safety: Dam safety is the art of scientifically applying the resources to ensure the operation, maintenance, modification in a safe and techno-economically viable manner while meeting the social and environmental needs. Risks to the general public, property, and the environment being on stake,  dam safety is not a matter that can be compromised. This puts a great responsibility on all interfaces in the hierarchy of dams' managements as well as the governments. Thus it is imperative to put in place collective application of engineering principles, experience, right risk management practices. While doing so the realization that should dawn on each individual and organization finding berth anywhere in this hierarchical chain is that,

Dam is a structure whose safe function is not explicitly determined by its original design and construction.Dams safety and life is dependent on it's "As on date status" which is determined by a host of factors post construction too. This includes punctual and planned actions taken towards identification or prediction of deficiencies and consequences related to failure, documentation of these, sharing information with all stakeholders in order to mitigate to extent possible, unacceptable risks.

Dam Safety Programs 
: The purposes of any dam safety program are to protect life, property, and the environment by ensuring that all dams are designed, constructed, operated, and maintained as safely and as effectively as is reasonably possible. Accomplishing these purposes requires commitments to continually inspect, evaluate, and document the design, construction, operation, maintenance, rehabilitation, and emergency preparedness of each dam and the associated public. It also requires the archiving of documents on the inspections and histories of dams and the training of personnel who inspect, evaluate, operate, and maintain them.

Programs must instill an awareness of dams and the hazards that they may present in the owners, the users, the public, and the local and national decision-makers. At every forum periodic reporting on the progress or conformance to plan should figure alongwith remedies for non-conformities. The only way to realize these objectives is to have the resource in terms of adequate:
  • Manpower proficient in dam design.
  • Equipment & instrumentation for monitoring & maintenance
  • Financial allocation
  • Accountability at all levels of hierarchy within Technical, Bureaucratic and political regimes.
Dam Break Risks: Dam owners world wide are strengthening disaster preparedness for potential dam failures. The need for modernized emergency action plans, revised dam operation &Maintenance strategies etc has gained momentum in light of today's realities like terrorist strikes. Real time evaluation mechanism is the order of the day. To instantly apprise stakeholder of his risks exposure in case of effected modifications, or effect of meteorological changes etc. letting him to assess the consequences of possible dam break in terms of the affected areas, the time available to evacuate people, and the damage, which the flood wave will cause. For this we need possible models that can simulate perceived conditions and instantly come back to the stakeholder with not only the effects but also with plausible alternative solutions (eg: MIKE FLOOD of DHI used for Machhu II study). Flood mapping empowers professionals to devote more time in thought process than mundane calculations, which can be off loaded to the machine.

A wealth of intellectual capital is available in the country and if requisite monitoring in the right way with properly calibrated instrumentation located at appropriate locations is made part of dam safety programs it would be a great service to the experts to use their intellect and experience to be able to make right predictions on dam safety measures to be taken from time to time.

DAMS AND FLOODS

L. Berga,
Professor of the Polytechnical University of Barcelona, Spain.
Chairman of ICOLD Committee on Dams and Floods.

1. FLOODS AS A NATURAL HAZARD.
The natural hazards suppose an important impact for human life and produce serious social effects and grave economic losses. The natural disasters constitute a curb on the sustainable development affecting its three basic mainstays: economics, social and environmental. In spite of the efforts made by the International Decade for Natural Hazard Reduction of the United Nations ( 1,2,3 ), the natural disasters in the world, have experienced an increasing evolution during the last decades of the XXth.Century, producing at present a mean of some 40,000 victims per year and mean economic losses of more than 50 Billion $ per year. The number of major natural disasters during the period between 1963 and 1992 have been multiplied by 3.5 with relation to the affected people, and by 2.3 in relation to the number of victims (4). Likewise, the economic losses are increasing with an exponential tendency, having duplicated during the last decade of the 1990`s, with an actual evaluation of more than 50 B$ per year, as is shown in Fig. 1 (5). 
Fig.1. NATURAL DISASTERS. ECONOMIC AND INSURED LOSSES WITH
TRENDS.
The greatest natural disasters of the last decades which have produced economic damages greater than 10 B$ are shown in Fig 2. It can be observed that of the 18 disasters, 7 are big floods, such as those of 1993 in the United States, 1991, 1996 and 1998 in China, those of North Korea in 1995, and those of the year 1999 in Venezuela, with more than 20,000 victims.

Fig.2. MAJOR NATURAL DISASTERS.
Within the natural disasters, the greater number correspond to floods (Fig. 3) which suppose about 30% of the socio-economic impacts ( 32% in relation to the significant damage and to the affected people, and 26% in relation to the number of deaths) (4).


fig. 3. major disasters around the world 1963 - 1992. percentageof significant disasters by type, based on damage, personsaffected, deaths.
In the last decades the impacts caused by the floods have been very important and the
Table nº 1 refers to the most catastrophic floods that have occurred in the last ten years.

In order to analyse with more detail the significance and importance of the floods, ICOLD has carried out a survey on the social and economic impacts of the floods in the 20 most important countries in large dams, and which represent about 90% of the existing large dams (6). It was found that the floods are the most important natural hazard in 65% of the countries and that the floods constitute in 90% of the cases the first or second most important natural hazard.  On the other hand, the floods present a high recurrence, with an average incidence of 7.2 years, and in the majority of cases the number of years between important floods ranges from 5 to 10 years. The “mean” number of victims per year produced by the floods is shown in the Table nº 2.

The majority of victims are produced in the Asian countries, 200 victims per year in Bangladesh ( without including cyclones, nor storm surges), 250 in South Korea, 1,500 in India and more than 2,500 victims per year in China. Nevertheless, the cases also stand out of U.S.A., with 94 victims per year and Japan with 120 victims per year, in which exist an intense occupation of the flood plains, which together with the presence of flash floods give rise to these values. In most of the countries the mean number of victims per year is lower than 20.
 In relation to the economic impact produced by the floods, Table Nº 3 shows the evaluation of the “mean” annual damages.

TABLE Nº 3. MEAN ANNUAL DAMAGES PRODUCED BY FLOODS.
 It can be observed that in South Korea they represent some 500M$, Spain 600M$, China 3,000M$, but the most important damages are produced in very developed countries, United States with 3,400M$, and Japan with 7,200M$ per year. Also, it is necessary to quote the importance of the impacts and economic damages produced by the floods in several developing countries, where the amount of damages and social disruptions which floods produce, could become the cause of limiting their development. So then, the experience shows that the damages caused by the floods continue to increase progressively, and in many countries constitute a veritable restraint to the economic and sustainable development. For this the UN decided in the year 1987 to create the International Decade for Natural Disaster Reduction ( IDNDR ) for the ten years 1990 - 2000 with the objective of reducing by way of concerted international action, especially in the developing countries, the loss of lives, material damages and the social and economic disorders caused by the natural hazards. Among the essential elements of the activities of the IDNDR the following points stand out (3,7):
1.- Greater emphasis in planification and preventive measures.
2.- Adoption of integrated actions (structural and non-structural) for the reduction of the disasters.
3.- Establishment of forecasting and alarm systems compatible with the technology and culture
      of the countries.
4.- Development of a social conscience of the necessity of the reduction of the impacts.

2. EXTREME FLOODS IN THE WORLD.
The studies of extreme floods at world level date from the year 1984, when the International Association of Hydrological Sciences ( IAHS ), published the “ World catalogue of maximum observed floods “ (8). These studies show that the maximum floods observed, are limited by an envelope curve, which adapted well to the equation given by Francou-Rodier (9):

  The studies of the IAHS of the year 1984, selected the data of 41 extreme floods, which had values of K comprised between 5.19 and 6.76, corresponding this last value to the flood of the year 1953 of the River Amazon, in the Obidos station gauge. These extreme floods present an envelope curve with a value of the coefficient of Francou- Rodier of about 6, which for the upper enveloped reached the maximum value of 6.4, except for the data of the River Amazon which reached an extreme value of 6.76.
Recently, the Committee on Dams and Floods of the International Commission on Large Dams (ICOLD) has carried out a new study of extreme floods at a world level, with the objective of updating the data of the extreme floods and analyse the maximum observed floods on the dams (6). The measures of peak floods on the dams has the advantage of greater precision, with which it lessens the errors of measures of big floods in the gauging stations, which can reach a ± 15%. The work of ICOLD collects 340 new data of the big floods, which have permitted the selection of 21 data of the extreme floods, which have a coefficient K of Francou-Rodier greater than 6, and which are shown in the Table nº 4.

TABLE Nº 4. EXTREME FLOODS IN THE WORLD.
 It must be indicated that the data of the floods of the River Amazon have been revised which has permitted, with the experience of the floods of the year 1989, a better estimation of the flood of the year 1953, evaluating it in some 320,000m3/sec. Among the extreme floods, stand out those of New Caledonia, those of the River Ruhe in China in the year 1975, which gave rise to the catastrophic failures of the Banqiao and Shimantan Dams, those of the River Amazon in the years 1953 and 1985 and those of the River Narmada in India.
All the data of the extreme floods is well adapted to the envelope curve, the relationship between the peak flood and the catchment area, with a value of coefficient of Francou- Rodier of 6.4 ( Fig.4 ).

FIG. 4. EXTREME FLOODS IN THE WORLD ENVELOPE CURVE.
 Equally the Fig. 5 shows the values of the specific flow, peak flow per unit of area, in relation with the catchment area.




 FIG. 5. EXTREME FLOODS. SPECIFIC FLOW.

Also the study of the variation of the coefficient K during the last decades, shows that the value of the coefficient K of 6.4 has not been surpassed, and so no upward movement is observed of the extreme floods, with the available data (6). Nevertheless, the analysis of the data of the extreme floods and of the envelope curve
shows that there is a change of the general behaviour for the basins of area less than 300
km2., and the values of the peak flow are less than those obtained from the
extrapolation of the envelope curve. So, for catchment areas of less than 300 Km2. the
extreme floods adapt better to a new envelope curve, defined as:


The Fig. 6 shows the good adaptation of the new envelope curve for the small basins, and the Fig. 7 corresponds to the envelope curve of the specific flood, in which it can be observed that, for small basins the specific flows are less than 100 m3/sec/km2.
Fig. 6 ENVELOPE CURVES OF THE EXTREME FLOODS. (missed in article)
-----------------------------------------------------------------------------------------------------------
REFERENCES
1. IDNHR. 1987.”Confronting natural disasters”. National Academy Press. Washington. DC.
2. HOUSNER, G.W. 1989 “An International Decade of Natural Disaster Reduction: 1900-2000”. Natural Hazards
    2,45-75.
3. IDNHR, 1994. “Yokohama Strategy and Plan of action for a safer world”. World Conference
    on Natural Disaster Reduction. Yokohama.
4. ZUPKA, D. 1988. “Economic impact of disasters”. Undro News. Jan-Feb.
5. MUNICH RE.1998. “Topics. Annual review of natural catastrophes”.
6. ICOLD. 2002.” Dams and Floods”. Icold Bulletin. Paris
7. UN. 1987. “General Assembly. International Decade for Natural Disaster Reduction”   A/Res/42/169.
8. INTERNATIONAL ASSOCIATION OF HYDROLOGICAL SCIENCES. IAHS. RODIER,  J.A., ROCHE, M.
    1984. “World Catalogue of maximum observed floods”. IAHS Publication  Nº.143.
9. FRANCOU, J., RODIER, J.A. 1967. “Essai de classification de crues maximales observées dams le mond”
    Chaiers ORSTOM. Série Hydrologie. Vol IV, nº 3. pp 19-46. ORSTOM Bondy.
10. BERGA, L. 2000. “Benefits of dams in flood control”. R35. Q77. 20 Th.Int.  Congress on Large Dams. Beijing.
11. MORGAN, A.E., 1951. “The Miami Conservancy District”. McGraw-Hill, First edition, New York.
12. BERGA, L. 1995. “Dams in river flood hazard reduction”. In: Reservoirs in River Basin
     development”. L. Santbergen, C.J. Van Weston (Eds). Vol 1, 119-128. A.A.Balkema.
     Rotterdam.
--------------------------------------------------------------------------------------------------------------------------------------------

Fig. 7 SPECIFIC FLOWS OF THE EXTREME FLOODS.

In conclusion, the extreme floods in the world have envelope curves, which limit the relation between peak flows and catchment areas. For basins with areas greater than 300 km2, the envelope curve of Francou-Rodier with a coefficient K = 6.4 is valid, but for areas less than 300 km2, it adapts better to the world data, a new envelope curve with a new coefficient R.

3. THE ROLE OF DAMS AND RESERVOIRS IN FLOOD MITIGATION.
The dams and the floods present a mutual interrelation. On the one hand the floods suppose a danger for the integrity of the dams and for their safety, and on the other hand the dams and reservoirs could play an important role in flood routing, and are one of the most efficient structural measures to mitigate the damages produced by the floods. Analysing the natural history of the floods, the measures to prevent and reduce the damages produced by the floods, could be classified in two large groups (10).
A) STRUCTURAL ACTIONS:
They are measures to interfere in the phenomena of flood formation and routing.
A-1 Soil conservation and correction of the basins.
A-2 Dams. Flood control and regulating reservoirs.
A-3 Hydraulic works in the rivers (Levees and dikes, diversions, channel improvements, etc..)

B) NON-STRUCTURAL ACTIONS.
They are measures to mitigate or reduce the damages produced by the floods:
B-1 Risk maps.
B-2 Flood plains: Zoning. Land-use patterns.
B-3 System of insurances.
B-4 General legal regulation. Building regulations.

C) Other types of NON-STRUCTURAL measures are actions in order to foresee, and thus be able to reduce the damages produced by the floods.
C-1 Flood forecasting and flood warning systems.
C-2 Emergency Action Plans

In order to reach a greater effectiveness in the reduction of damages produced by the  floods it is necessary to assess the flood control by way of a holistic vision. With this the problem is posed with a more critical vision and less optimistic, and with an integral approach as regards the basin and of alternatives.
The planning of the flood hazard reduction measures should be carried out as regards the basin, with a vision of the whole of the basin, and analysing the incidence that each one of the measures has and the relations between them, as also their effects downstream on the flood routing. On the other hand, the actions as a whole should be considered as a system of integrated measures, developing in each case the implanting of combined measures which contemplate the joint application of structural and nonstructural measures, it being necessary in many cases the development of zonings and land-use patterns downstream of the dam, and also the implantation of flood forecasting and flood warning systems, which are essential for the emergency action plans (Fig. 8).
FIGURE 8. MEASURES IN THE FACE OF FLOODS.

Within the measures in the floods fighting, the role of dams and reservoirs should be emphasized, since the dams constitute a very efficient structural measure, as they are the only solution which permits the storage of large quantities of flood volumes, modifying significantly the flood routing, and being able to reduce the peak flood in an important manner.

Usually the dams and reservoirs that have the main or single purpose of the flood routing are referred to as flood control dams, a denomination which induces to think that they are capable of controlling all the floods and therefore avoid any damage downstream. Evidently, this is not possible and less still in the uncertain subject of floods, in which the absolute zero risk cannot be attained, with the actual physical and technical knowledge. For this reason, it would be better to refer to Flood Mitigation dams or reservoirs, in the sense of indicating the capacity of these structures in the reduction of the damages produced by floods. With all that it is necessary to learn to live with the floods, of course, reducing as much as possible their important impact.
The dams and the reservoirs can be classified in four categories according to their purpose of the flood mitigation:
1. Reservoirs with a single purpose of regulation ( water supply, irrigation or hydropower ), in
     which the incidence in the flood mitigation usually is small.
2. Multipurpose - reservoirs with a principal purpose of water storage, but in those in which the
    flood mitigation is also an important objective.
3. Multipurpose - reservoirs with a principal objective of flood mitigation, combined with other
    objectives of regulation of water.
4. Reservoirs with a single purpose of flood mitigation and the reduction of downstream
    damages. Flood mitigation dams.
Furthermore in diverse situations, principally in the cases of large dams on important rivers, the effect of reservoir routing is designed in order to be operative only in a seasonable manner, during the flood season, combining the flood seasonable control purpose with other multipurposes, generally irrigation, hydropower or water supply.
These dams could be referred to as “ Flood season mitigation dams”. In general , the effects of the regulating reservoirs on the floods are more notorious in the low and medium return period floods, in which the reduction in the peak floods and the volumes retained by the reservoirs could be very important and in consequence the mitigation of the downstream damages could be very significant. Their effects on the extreme floods can be less spectacular, although almost always positive. All dams, if they are well designed and operated correctly, present flood mitigation benefits, but the maximum benefits are obtained in the flood control dams in which the routing effects of the floods and the reduction of the downstream damages is the main purpose.   It must also be taken into account that in many countries of the world, large populations and important cities have been established over the years on the banks of the rivers which form the backbones of the country. With this, the application of some nonstructural measures is non-viable (resettlement, land-use patterns, etc.) , and the only possible measure for the reduction of the damages of the floods is to reduce the frequency and flows of the constant and repeated floods, a role that can only be carried out by the flood mitigation dams, which although not with a total protection, reduce in a very significant manner the grave impacts due to the almost “annual “ floods.
In the studies of flood mitigation dams there arises on numerous occasions, the alternative of constructing a larger dam on the main river close to the area to be protected, or various small dams located in the headwaters or middle stretch of the basin and on the tributaries of the river. In general, the smaller dams scattered over the basin although numerous, give lesser protection than one single large dam situated immediately upstream of the zone to be protected. So, the Miami Conservatory District in the valley of Miami showed that the realization of five large retention reservoirs give a much greater protection and with less cost than the construction of numerous small dams on the tributaries, and the Corps of Engineers in the basin of the River Merrimak showed that the construction of 13 small dams presented an efficiency of only 52% in
comparison with two large dams located on the main river (11).
In Spain, on the River Onyar in Girona, it has also been seen the effectiveness in flood mitigation decreases in a very important manner as soon as the dams were moved away from the zone to be protected, or were situated on the tributaries (12). So, then, in general, technically a greater protection is obtained with reservoirs situated upstream of the area where flood damages have to be mitigated, but on several occasions the economic, social and environmental aspects present a problem for the construction of dams in the immediate area upstream of the township to be protected.
The dams in the world have supposed enormous benefits in flood mitigation, and their operation during flood events have reduced, in the greater part of the cases, in a significant manner the damages produced by the floods.
The ICOLD Committee on Dams and Floods has studied and analysed diverse significant cases, which show with quantative values, the important role played by dams in the flood mitigation (6). The cases analysed refer to the flood control in ample areas with important flood problems, in Japan, USA, Brazil, China, Korea, Norway, Spain, etc., in which, in general, are combined the effects of the reservoirs, dams, levees and river canalisations, together with the operation in real time of flood forecasting and warning systems.
The flood control plan of the Tone River in Japan began in the year 1900 for the control of the floods in the area of the Bay of Tokyo. After successive revisions its actual formulation is of the year 1980. In the plan are included 10 flood control dams with a reservoir capacity of 229 Hm3. Of these 6 dams are in operation, and the 4 remaining dams are in construction. The combined project flood discharge of dams and dikes is of 22,000 m3/sec, corresponding to a protection for the flood of 200 years return period. In recent years no serious floods in the Tone River have occurred, partially due to the flood control dams and the improvement of river channels. The greatest flood of the last forty years, was that of the typhoon of August of 1982, in which the dams were operated to reduce the flood damage. From among these the Shimokubo Dam greatly contributed to mitigate the damages with a peak discharge reduction of 62%.
In the USA there was in 1993 an extreme flood, the Great Midwest Flood, principally in the upper Mississippi River basin. Seventy-six reservoirs have been developed by the U.S. Army Corps of Engineers ( USACE) in the upper Mississippi for the purpose of flood damage reduction, with a large capacity to store floods with more than 49 billion cubic metres, and a controlled drainage area of 956.000Km2. The USACE estimates that flood damage reduction facilities, (reservoirs, flood walls and levees ) prevented 19.1 Billion$ in damages. Of this quantity 7.4 Billion$ were attributed to the effects of the utilization of the flood storage reservoirs.
A case also to be pointed out in the USA is the Flood Control Miami District, developed on the Miami River, a tributary of the Ohio River, after a catastrophic flood in 1913, in which were produced 360 victims and some damages superior to 100 M$. The plan was implemented with the construction of five detention dams and river channel improvements in nine urban areas. Since completion the dams have stored water on more than 1,000 times providing some substantial benefits, and in the flood of the year 1959, with rainfall close to that of 1913, detention storage utilized was only 32% of the total storage available.
In Spain there exist numerous real cases of the beneficial effects of the dams and reservoirs in the mitigation of damages due to floods. One very significant case corresponds to the flood of November 1982 in the Ebro basin. The global effects of the reservoirs on the mouth of the River Ebro was of 57% of the reduction, with a peak flow discharged in the Ribarroja Dam of 3,200m3/sec, almost of the limit of capacity of the river channel in order not to produce important damages in the downstream townships, as contrasted with the 7,400m3/sec, that were estimated without the existence of the dam in the basin.
In general, in the cases analysed by the ICOLD Committee on Dams and Floods, the effects in the flood mitigation were very significant, with values varying between 25% to 85% in the reduction of the peak flow, surpassing the reduction in numerous flood situations the figure of 50%. The reduction of the flood volume varied from 10% to 73%, with greater values in the cases in which the flood reservoir capacity was high in relation with the flood volume, and in the cases in which the main purpose of the dam was the flood mitigation.
The hydrologic criteria for the design of flood mitigation dams is based on two design floods:
1. An “ Inflow Design Flood “ or “ Safety Check Flood “ to assure the hydrologic dam safety.
2. The protection design flood, which is the flood that the dam is capable of routing without
     producing damages downstream.
In general, and apart from the specific analysis in each case, the protection design flood recommended are :
· In rural areas return periods of between 20 and 50 years.
· In urban areas return periods of between 50 and 200 years. In cases of protection of important cities, and if the economic, social and environments aspects are favourable, return periods of 500 years or even 1,000 years may be considered.

In the real cases studied by ICOLD the design flood protection varied between 35 and 200 years of return period, being able to reach in singular situations, in which there exists an important occupation of the flood plains and large cities downstream, values as high as 500 or 1,000 years. At the present time close to 20% of the total of the existing large dams have as a purpose that of flood control, be it a single purpose ( 8%), or as one of its principal objectives.
In the future it has been indicated that due to the exponential growth of the damages produced by the floods, it will be necessary to increase the measures of prevention and reduction of damages, for which the implantation and construction of new flood control dams will be necessary, together with measures which control the progressive occupation of the flood plains and the improvement of the reliability of the flood forecasting systems. For this, an increase of the flood mitigation dams is to foreseen in the future, with extensive flood mitigation plans, as are the cases of Japan, China, Spain and some areas of the USA.
For example, in Japan, the Flood Control dams have a very relevant part to play in the reduction of the damages produced by the floods, protecting the population in a range between the 50 years and the 200 years of return period, for which they can count on some 500 large dams for flood control, providing a total flood control capacity of some 3,700 Hm3. In the future the construction is foreseen of some 400 new large dams, which will contribute some additional 2,400 Hm3, to the flood control reservoir capacity.
In China, the construction of large dams in which the role of flood control plays an important part is also experimenting a strong increase. In the last decade only the floods with a return period of between 10 and 20 years were relatively controlled in the major rivers, and the flood control standards for medium and small rivers are even lower.
Actually with the increase in population and with the economic development, there is a higher requirement for flood control. On the other hand, because of sediment deposition, the flow capacity of the rivers and the flood storage of reservoirs are gradually degrading. Also, the flood storage of the lakes used as additional inundation areas for the extreme flood is also decreasing, due to the deposition of sediments. So, the danger of inundation in extensive areas is increasing especially in the Yellow and Yangtze Rivers. From that, the large dams constructed on these rivers have an important part to play in the flood control, as are the cases of the dams of Longyangxia, Liujiaxia and Xialangdi, on the Yellow River, or the Three Gorges on the Yangtze River.
In Spain, due to serious impacts produced by floods, a considerable increase has been produced in the latter years in the number and importance of reservoirs dedicated exclusively to flood routing, or whose principal purpose is flood mitigation. Actually, today, there are 30 Flood Control dams, which represent about 3% of the existing dams, and in the Hydrologic Plan foresees the construction of some 40 new reservoirs for flood mitigation.
An interesting and illustrated case is that of the American River Basin in California, USA, with a grave problem of floods in the city of Sacramento. For this the U.S. Corps of Engineers carried out in the year 1991 a Feasibility Report and Environmental Impact Statement, with the proposal of a Detention Dam Plan, with the Auburn Dam. The theme has been very problematic and it is in continuous discussion, not having reached yet the implantation of the solution contemplated in the Plan. In similar cases it has not been until after catastrophic floods, when the actions foreseen have been carried out.

DAM FAILURES AND REMEDIES

 (Abstracted from the article by International expert Lempererie of France)
The increased safety now required for thousands of dams may be very costly if reached only by structural improvements. Various low cost non structural measures may reduce or avoid structural expenses and may be implemented in a short lime. They are even more attractive in developing countries.
The Committee on costs of ICOLD has prepared the bulletin E 102 on the subject: " Non structural risk reduction measures: benefits and costs for dams".
Based upon many comments and data from industrialized countries it has been written by M.Smart (U.S. Committee). This bulletin is freely available on the Net through: www.icold­cigb.org (Publications. Abstract of recent publications, Bulletins, E 02).
The present paper refers to many findings of this bulletin and analyses the efficiency of measures according to dam types and height and for various circumstances (conditions like first filling, floods, ageing...). It studies also specific data and advantages for developing countries.
This paper focuses on human risk which is actually the main basis of most criteria and decisions. For instance a number of countries require now to design or uprate (upgrade)  the spillways for the probable maximum flood if the dam failure may cause fatalities. Applying this worldwide could modify dozens of thousands of dams and costing hundreds of billons of $: this is unrealistic. But it seems possible to identify the true main risks, to reduce them on a number of dams by structural measures as far as they are cost effective and to insure a reasonable degree of safety by non structural measures applying to all dams at risk.
There are presently 45.000 large dams (39 000 being fill dams). Their failure would usually cause a (flood) wave flow of some thousands m3/s, and for 10.000 of them a flow over 10.000m3/s. Some past failures reached flows between 50 and 100.000 m3/s and possible failures could cause much higher flows. Moreover there are 100.000 small dams with storage between 0,1 and 3 millions m3 : their failure could cause waves in the range of 500 to 1.000 m3/s which may be dangerous locally in populated areas along small rivers.
As the world yearly rate of failure is in the range of 10-4 for the large dams and seems higher (possibly 5 x 10-4) for the small ones, some failures every year may have an impact on dozens
of km² and possibly over 100 km² : 50 or 100 may have an impact on few km² within a range of 10 or 20 km downstream of medium or small dams. The organization and cost of non structural measures may be different according to density of population (500/km² in most of Asia and 20/km² in most of America) and to failure hydro gram.
The efficiency of not structural measures may also vary according to dam type and failure circumstances because the time available for waming may be minutes or hours (or even days) and the percentage of fatalities in inundated areas may thus be less than 0,1 % or over 50 %. Night failures and cold water are evidently more dangerous.
Non structural measures include risk analysis, monitoring, training, maintenance, emergency planning and warning systems and modified operation.

I)   COST OF NON STRUCTURAL MEASURES

http://www.hydrocoop.org/images/puce1.gifRisk analysis reviews possible risks with rough evaluation of relevant probability and consequences (hazard) and may study various solutions to mitigate these risks.
For rather large dams in industrialized countries, typically costs range from 10.000 to 100.00 $ per dam but may be up to 500.000 $ in some cases. But studies made by a same team of engineers for many smaller dams, if focus on main risks, may require only two staff days for basic study and 10 or 20 additional staff days for some dams appearing most at risk from the basic study. Relevant costs may then be between 1.000 and 10.000 $ in industrialized countries, ten times less in countries such as China or India where cost of equally efficient staff in much lower.
Risk analysis should thus not be limited to few thousands of large hydroelectric schemes but should be adapted to dozens of thousands of water storage or irrigation dams: as relevant owners have low income and no technical knowledge, such rick analysis should be organized by authorities as well as advices and rules for structural or non structural measures. Risk analysis should take in account the fact that storage of usual floods in reservoir favours occupation of the river bed by people at risk for exceptional floods even when these floods do not endanger the dam.
http://www.hydrocoop.org/images/puce1.gifThe purpose of monitoring, beyond checking design data, is to give informations in order to take measures avoiding failures, or at least to give time for efficiency of warning systems. Monitoring includes instrumentation, visual inspections and periodic assessment performance. Cost of instrumentation may be high, in the range of one per cent of the dam cost but may be fully justified, particulary for arch dams or difficult foundations. However its efficiency is very questionable if the operating staff is not well trained.
The choice of reliable equipment with easy maintenance is more important than the value of investment or the theoretical precision of measures. For 80 % of dams at risk which are medium or small earth fill dams, low cost instrumentation based mainly upon simple piezometers and accurate leakage measurement may be reasonable. For all dams, visual inspection and simple measurements, if well organized and reported, are not costly and essential for safety : their cost is essentially bound with staff cost ; frequency of inspections may be reduced after some years for most dams in industrialized country but may be kept higher in low cost countries where standard of construction may have been lower thirty years ago.
http://www.hydrocoop.org/images/puce1.gifFor emergency planning and early warning system, timing and quality of warning are the keys of success, it is necessary to win as much time as necessary but "the clarity of warning messages in term of who, what, when and where is an important condition of successful waming" .
In 1916, many people were drowned by a dam failure when they came seeing views of the inundated area : as often an artificial dam created by an embankment downstream of the main dam failed suddenly.
When a dam failure may endanger large cities, initial cost of emergency planning may he over 100.000 $ and the yearly cost in the range of 20.000 $ in industrialized countries, much less in developing countries as costs are essentially staff costs.
But most future failures shall be medium or small fill dams endangering some hundreds or thousands of people within 10 or 20 km downstream of the dam most often by flood failures. Simple emergency planning may be implemented at low cosy and based essentially on local organization and mobile phones. This may apply to dozen of thousands of dams at a yearly unit cost in the range of 1.000 $ in developing countries where is most of the risk.
The need of permanent watching in based upon the number of people in the inundated area and the failure probability which may vary considerably along the year.
Beyond floods, the average daily risk of failure of a large dam is less than 10-7. During first
filling and during 10 or 20 days per years when weather forecasting foresees a risk of heavy rains, this average probability may be about 10-5. For floods during construction or during some days after an earthquake it may be close to 10-3, during 10 or 20 days per year the daily probability of an exceptional flood not endangering the dam but inundating large areas downstream may be 10-3 and the impact of dams on such flood is not always favourable. Permanent watching may thus be justified for thousands of dams and temporary watching for dozens of thousands.
http://www.hydrocoop.org/images/puce1.gifTraining may apply to all safety problems or focus on some key points such as spillway gates operation or visual inspections or warning systems. Training efficiency may be much improved by computer based simulators
It is advisable to organize the training for many dams and operators : training time of an operator may request few days and unit cost is low specially in developing countries. Quality of training is the key of success.
http://www.hydrocoop.org/images/puce1.gifMaintenance and more specially gates maintenance is an essential point. Budgets may not be high but informations about the many details to check and about past accidents are very important: addition of several minor defects has caused past disasters.
http://www.hydrocoop.org/images/puce1.gifModified operation may include lowering of the reservoir level. This may be permanent before improvement of a structural defect or limited to the flood season in order to reduce the probability of flood failure. In this later solution a human risk is often overlooked : when a part of the flood is stored before spilling the downstream flow at the time of spilling may raise much more quickly than with the natural flood. This is true either by gates operations of for many free flow spillways for with the time to peak of the downstream flow became in the range of :
T (in hours) = 200 s / L
"s" being the reservoir area in km² and "L" the free spillway length in m. for dozens of thousands of dams, "s" is between 0,05 and 1 km² and the spillway length between 20 and 100 m and the downstream flow may raise from nil to hundreds of m3/s in less than one hour : this risk may be substantial for thousands of dams even for not exceptional floods.
Many more data about costs are available in the ICOLD bulletin referred to hereabove.
II)  FAILURES OF DAMS
Risk analysis is mainly based upon reported past failures. It should mainly refer to recent ones, for instance since 1970. moreover many failures have not be reported for small dams or when failures caused few or no fatalities.
http://www.hydrocoop.org/images/puce1.gifBeyond China there are 17.000 large fill dams. Before 1970, 7000 were built and 100 failures reported. Since 1970, 10.000 more were built and 44 failures reported ; 7 flood failures during construction, 7 piping failures at first filing and 30 in operation including 22 by floods 6 by piping and 2 by earthquake. Rate of reported failures in operation after 1970 was thus 0,8 x 10-4 about the same for old or new dams. But all failures were not reported and the rate has been much higher within small dams. In China 20.000 large fill dams and 60.000 small ones (over 100.000 m3 storage) were built after 1950. 3% failed (including 2% flood failures) but the rate of failures after 1980 was reduced (in the range of 3 x 10-4 ?)
http://www.hydrocoop.org/images/puce1.gifBeyond China there are some hundreds of masonry dams, often very old : 18 failed before 1970 and 2 later.
http://www.hydrocoop.org/images/puce1.gifThere are 4500 large concrete dams beyond China: most built before 1970. 12 failures have been reported before 1970 and 3 later. Most failures happened at first filling and were bound with foundations.
http://www.hydrocoop.org/images/puce1.gifSudden failure are much more dangerous than progressive ones : beyond China 10% of the fill dams failures and 40% of the masonry or concrete dams failures caused more than 100 fatalities each : half of the relevant reservoirs stored less then 20 millions m3 and over half of these dams were lower than 30 meters.
III) EXISTING EARTHFILL DAMS
Over 80% of large dams and 90% of small dams are earthfill dams.
a)       Flood failures
The main failure risk worldwide is their overtopping by floods. It may be mitigated by low cost structural or not structural measures.
"Most floods failures have been caused by under-dimensioned spillways : risk assessment for this aspects is quite easy and effective if it is focused on the real problem :
Which dams are most at risk and what is the probability of actual failure, corresponding to imminent failure flow (IFF), and not merely of exceeding a regulatory high water level ? Because of their free-board, many ungated dams where the 500 year flood exceeds the regulatory high water level may well withstand the PMF, but some large gated dams with a 1.000 year design flood have little safety margin. The catchment area of the great majority of dams is less than 500 km², and simple regional flow formula will apply to most reservoirs in a given climatic area. For instance, the 10.000 year peak discharge, Q, can be evaluated simply by the formula Q = K S0,75, S being the catchment area and K a regional coefficient. Some simple adjustments, taking account of the shape of the catchment area and yearly local rainfall, can be made. Comparing the imminent failure flow with this calculated 10.000 year discharge can help do two things : it can identify the dams most at risk, and it can estimate a range of failure probabilities. All the factors used are easily determined. The impact that storage has on peak flow should be taken into account ; it may be important if the reservoir area of ungated dams is more than 1 or 2 per cent of the catchment area".
Scientific evaluation of the area inundated by a dam failure is difficult and precision of the result far from guaranteed.
But it is possible to give a range of value of the peak flow by a formula such as:
Q (flow in m3/s) = k v 0,5 h1,5 , v being the stored volume at failure time in m3, h the breach depth in m, and k a coefficient between 0,01 for cohesive dam body and 0,03 if uncohesive. It is then possible to know the range of number of people at risk and of value of possible damages and to have elements justifying structural or not structural measures.
Increasing the capacity of spillway is economically justified if its cost is lower than the actualised value of the risk of damages. If q (in m3/s) is the flow initiating overtopping of the dam crest, 1/T the yearly probability of such flow, D the amount of damages and c the cost for increasing the spillage capacity by 1 m3/s the improvement is justified as far as :
Cq<k D/T
Value of k varies with local and financial conditions but is most often in the range of 50.
Various low cost structural solutions may be used for increasing spillage capacity : parapet walls, lowering free flow spillways and placing fuse devices, downstream slope RCC lining : c may be in the range of 100 $ in not industrialized countries and 500 $ in industrialized countries.
It may be justified to improve the capacity of small spillways for instance in a developing country, if T = 1.000, c = 100 and q = 200, the improvement by 50 % costs 10.000 $ and is justified if the amount of damages is higher than 400.000 $.
It may be unjustified to improve the capacity of a large one. For instance, in an industrial country, if T = 5.000, q = 2.000, c = 500, improvement by 50 % costs 500.000 $ and is not justified if the cost of damages D is under 100 millions $.
Emergency planning and warning systems are usually much less costly than structural measures and may be implemented before or beyond them. They have also the key advantage of being used also for exceptional floods not endangering the dam. Floods of yearly probability 10-2 or 10-4 inundate areas in the range of 10 to 50 % of the areas inundated by a dam failure and this overall human risk may be higher than the failure risk. Downstream level may raise quickly ; specially if a part of the flood is stored in the reservoir and river bed may have been occupied since the dam construction. Usefulness of warning systems may thus be important for all exceptional floods.
Another important risk is jamming of all gates. Relevant risk analysis, proper maintenance, training of operators for emergency conditions, redundancy of operating devices are efficient measures.
Analysis of overtopping risk should include for a number of sites the problems of upstream dam failures in construction or operation, possible breaches of natural reservoirs and landslides in the reservoir.
b)      Earthfill dams piping
Piping has been the main cause of fill dams failures at first filling and is the second main cause in operation. The failure flow is usually substantially less than flood failure flow as reservoir level and volume are lower and there is no incoming flow.
The relevant yearly probability of failure seems to be in average between 10-4 and 10-5 but varies considerably and is difficult to assess for each dam. It is higher for old dams which have been built without proper drainage or filters systems and for long dams where foundation conditions may not be well known. A serious risk is also bound with embedded pipes and junction to concrete structures.
But the probability of failure is closely linked with quality of inspection and monitoring. "For dams in operation, when internal erosion occurs, it is at the beginning a slow process whose speed increases with time. Most often the phenomenon can be identified by surveillance.
Visual surveillance must focus on downstream face and toe of the dam, paying attention to wet zones and hydrophilic vegetation."
Instrumentation is the second aspect of surveillance. Priority must be given to measurement of leakage, with observation of possible fine materiel deposits. Measurement of piezometry of downstream dam slope and base is a good complement.
Warning systems in case offailure have been very efficient for large or small reservoirs.
c) Earthquakes
Few relevant failures have been reported for large dams and the average yearly probability of failures appears lower than 10-5. Actually this risk is quite nil for most dams and rather high for some of them. The risk is evidently bound with the probability of earthquake in the area but even more with the nature of dam body and/or foundation. Some fill dams, and particulary those built by hydraulic fill, are subject to liquefaction and may be partly or completely destroyed in few minutes; the failure is then very dangerous even for rather small reservoirs. Efficiency of warning systems is not evident. Hundreds of small dams have been destroyed by earthquakes and the failure of the 40 m high Van Norman dam in U.S. in 1971 was close to a disaster.
Risk analysis should then focus mainly on dams subject to liquefaction or sometimes to sliding due to their construction methods and materials. The wide experience of China, mainly in Huanghe basin and of Japan, may be very useful Some dams should be decommissioned.
Beyond sudden failures earthquakes have caused settlement and cracks of many dams. These cracks may extend within few days, specially for old dams not properly equipped with drains and filters ; monitoring and warning systems are thus essential after an earthquake.
c)      War
Breaching the dykes has been a military tool in China for over 1.000 years and Huanghe river banks breaches in 1938 caused the losses of 800.000 lives.
No complete failure of large fill dam due to war has been reported but in fact the number of
large dams x years during past wars is limited worldwide to few thousands. The war risk is also increased by the fact that the main targets may be the highest dams and largest reservoirs of which the failure probability during wars may be high. War risk analysis is then very specific, may anyway study possible breach waves, efficiency of warning systems and impact on safety of partial lowering of reservoir.
IV)   EXISTING ROCKFILL DAMS
There are 2.000 large rock fill dams, most of them higher than 30 m. Waterproofing can be clay core, bituminous screen or concrete facing. Risks of piping or earthquakes are lower than for earth fill dams and risk analysis should focus on overtopping as one per cent of rock fill dams so failed during construction or operation, along the 20th century. Failure breach is caused by overtopping depth in the range of 1 m over the crest ; the breach may widen more quickly and extensively than with a cohesive fill ; as relevant storage is often large, emergency planning and warning systems are essential.
V) TAILING DAMS
Water storage is small but worldwide fill volumes is globally larger than for all other fill dams.
Risks may be very high (sliding, piping, earthquake). The risk of overtopping is usually not caused by floods but by wrong operation. Failures may be sudden and very dangerous for close populations and detrimental to environment in large areas. Risk analysis and remedial measures are very specific and are not studied in this paper.
VI) EXISTING CONCRETE DAMS
The probability of failure in operation has been lower then for fill dams (in the range of 2 x 10-5 ?) but failures breaches may be sudden and wide (over 4 times the breach depth) and failure of a 15 m high dam may cause a flow of thousands m3/s. Monitoring and particularly instrumentation have avoided many failures and justified structural improvements or decommissioning of a number of arch dams. As many concrete dams are older than 40 years ageing may become a serious problem for thin structures. Earthquake risk may be serious for buttress dams or multiarches dams.
The risk of sliding on foundations such as shale's may be much increased during floods for low gravity dams. An increase of the reservoir depth by few meters and of the downstream level which creates up lift reduces considerably the margin of stability : failure may happen before or after overtopping of the crest.
VII) EXISTING MASONRY DAMS
There are over 1.000 large masonry dams. They are more dangerous than concrete dams:
http://www.hydrocoop.org/images/puce1.gifMany have been built before 1930, often with thin profile or poor foundation.
http://www.hydrocoop.org/images/puce1.gifBasic strength of masonry is much lower and it is more difficult to avoid poor workmanship. The failure may happen in dam body where cracks due to various reasons may extend downstream.
http://www.hydrocoop.org/images/puce1.gifAgeing due to leakage reduces strength and density.
Monitoring is not always easy. Structural improvements are often justified. Warning systems, at least during floods, may be justified even for rather small dams. Will full damage should not be overlooked, a well as for a number of thin concrete structures.
VIII) GATES
A number of gates accidents, at first filling or during operation or due to ageing has been reported.   Failures are sudden and may be very dangerous as the flow may raise suddenly from Nil to 1.000 m3/s.  Risk analysis and maintenance are essential.
IX)  FUTURE DAMS
When these dams will be in operation, the opportunities of failure will usually be the sane as for existing dams. But designs based upon improved experience and criteria will reduce the relevant probability. The risks analysed hereunder are thus 3 specific risks during construction and first filling (excluding resettlement problem)
a) First filing failures
Beyond China, 42 such failures of large dams have been reported in the 20th century : 34 from 10.000 built before 1970 and 8 from the 11.000 built later : they were two failures of masonry dams, 8 of buttress dams or multi arches (1,6%) 3 from arch dams (0,5%),26 from fill dams (0,2%) and 3 from concrete gravit y dams (0,07%).
Many old failure were due to a general lack of knowledge about dams or foundations behaviour but most failures and specially recent ones were caused by human behaviour : lack of experience of designers or contractors or over confidence of competent ones, reduced expenses for foundation studies, lack of communication or ill defined responsibilities, unclear specifications, control looking to many details and not focusing on key points. A number of failures were caused by poor workmanship for instance in masonry ; poor quality of a full lift of an earth fill or RCC dam cannot be excluded and care of embedded pipes in earth fill is difficult.
Risk assessment could refer to these human problems as much as to physical data: foundation failures are not caused by geology : they are caused by a lack of knowledge or ill adapted design or treatment.
First filling may happen earlier than foreseen by flood during construction and many piping failures happened some time after first filling.
Many failures have been bound with foundations : this increases the risk of very long dams when foundation is not homogeneous.
Sudden and wide breaches of concrete or masonry dams were more dangerous then progressive piping failures of fill dams.
Two special risks may be important for some large reservoirs and difficult to assess exactly : sudden slope sliding in the reservoir and induced earthquakes which may be more dangerous for neighbouring than for the dam.
Risk analysis for first filling is thus very specific and difficult : probability of failure is low but cannot be excluded entirely because there are many possible reasons. Monitoring and instrumentation are essential and have avoided many accidents. Warning systems are advisable for most large dams at least during the few monthes which are usually necessary.
b)   Floods during construction
Few masonry or concrete dams failed during construction and their failure was usually not due to the special conditions during construction. Failure would have probably happened at first filling. (Tigra in India in 1917).
Few construction failures of fill dams lower than 30 m. have been reported because construction in the river may often be completed in a dry season or because the failure caused little damage for an embankment of reduced height at failure time. But from the fill dams higher than 30 m., three per cent ofthose built before 1930 and 0,5 % of the recent ones failed by floods during construction. Failure was caused by floods exceeding the capacity of temporary diversion structures or by delays in construction. Consequences of Panshet (1961 in India) and Sempor (1967 in Indonesia) failures were heavy. Over half of failures were rock fill dams; possibly they were wrongly supposed to withstand limited overtopping.
Risk analysis is specific but can be rather precise. Risk may be reduced by increased temporary flood control facilities (including auxiliary tunnel higher than the basic diversion tunnel) and close analysis of delays consequences. It should be too expensive to avoid completely the probability of failure but it is essential to study and implement during few months warning systems which may be very cost effective. Over 100.000 people at risk were evacuated for Oros dam in Brazil in 1961. Such failures may cause downstream dams failures.
Fill dams higher than 30 m. may represent over 20 % of all future large dams. This risk is consequently a serious one. Failure of high fill cofferdams is a similar risk.
c)  Work accidents
Work accidents during construction are a main cause of fatalities bound with dams. And, according to ICOLD Bulletin n° 80 : "Dam construction sites: accident prevention" the corresponding direct and indirect cost has been in the range of three per cent of construction cost, i. e. more than the cost of all failures combined.
Most future dams in Asia or Africa will be built with heavy plant but the number of workers shall be kept higher than in industrialised countries because of their low cost. These construction sites associating heavy equipment with many workers may cause more victims of accidents than past entirely hand built dams. In a number of large schemes over some years, up to one per cent of workers were killed by accidents and as average five per cent were absent from work through injuries.
In industrialised countries, rates of accidents have been divided by about 3 in 15 years. This may be obtained if devoting 0,2 to 0,5 of construction cost to safety measures and beyond human target, this may reduce by 2% the cost of construction.
Efficiency of safety measures is the responsibility of the contractor and is essentially bound with the site organization and management. However, owners (and laws) may very usefully impose at the tender time safety rules to be applied by contractors and control them during construction. Statistics have shown that the frequency rate of accidents was higher for medium size dams than for very large ones, probably because greater care of safety was taken on largest sites. Consequently efforts on safety should not be limited to the largest dams.
Relevant risk analysis and suggested measures for improvement are detailed in ICOLD Bulletin 80 "Dam construction sites: accident prevention" which is easily available.
CONCLUSION
Over 100 large dams and 1 000 small ones failed worldwide since 30 years. Similar rates of  accidents may be avoided in the future if associating structural expenses where cost effective with not structural measures. : cost of these measures may be low if they are well adapted to various problems and extremely low in many developing countries where cost of efficient corresponding staff is low. Consequences of failures may also be greatly reduced.
It is also possible to reduce the human risks bound with all exceptional floods not  endangering the dams and the workers accidents during construction: there two risks are more important than human risks from failures and are often overlooked.
Implementation of these measures can be made in few years and is more a problem of organization and clear responsibilities than a problem of cost. ln many countries most dam owners have little technical knowledge and the authorities should organize these improvements.
BIBLIOGRAPHY

ICOLD :