F. LEMPÉRIÈRE and J.P. VIGNY
Published by « Hydropower & Dams » Issue 2 (2005)
The discharge of extreme floods (such as the Probable Maximum Flood: P.M.F.) is in the range of 3 times the likely maximum discharge during the dams life.
The failure of dams by flood is caused by a small overtopping of embankment dams and a huge overtoping of high concrete dams.
For most existing and many new dams the overall design has been based upon a “Design Flood” for which the reservoir level is kept well under the dam crest level. The yearly probability of this “design flood” was usually chosen between 1/500 and 1/5,000. The corresponding true probability of failure is unclear and this design method overlooks many low cost technical solutions for spilling extreme floods.
A more realistic approach, advocated by ICOLD bulletins, refers to a “Safety Check Flood” of very low probability (often chosen as the P.M.F.), for which are accepted a reservoir level close to the dam crest and some limited damages. In this case the meaning and the return period of the “Design Flood” are questionable and its name confusing.
An overall review of these design methods and of the alternative spillway solutions may help increase the safety and reduce the costs.
A key question for new or existing dams is the flood which the dam may withstand without failure. Should the true yearly probability of this flood be 10-3, 10-5 or quite nil? The criteria for answering to this question and the design methods are often the same as 50 years ago and have not been adapted to the present knowledge and conditions. The official regulations vary considerably between countries and are hardly well adapted to the extreme variety of dams data; most overlook any cost optimization and are based upon the always wrong hypothesis that the return period of floods may be evaluated precisely. An in depth review of these problems may be justified by the huge changes worldwide since 1980 in safety requirements, technical knowledge and available solutions.
1) Key changes since 25 years
– The safety requirements are now much higher. The acceptable yearly probability of failure was in the range of 10-3; it is now closer to 10-5 but some limited damages or extra costs are accepted in between.
– There are now much more data on extreme rains and floods which were considerably underestimated 30 years ago. One key point is that the flow and volume of an extreme flood (P.M.F.) are in the range of 3 times (most often between 2 and 5) the flow and volume of the maximum flood likely to happen during the dam life, i.e. over 100 years. Another key point is the uncertainty in the evaluation of floods corresponding to various probabilities. The true return period of an estimated “1,000 years flood” used as “design flood” may well be 200 years or 5,000 years.
– The analysis of incidents and failures caused by floods brought essential informations for safety optimization: it is possible now to evaluate the reservoir level corresponding to a probable failure according to the material of embankments and to the structural design of concrete dams.
– Two expensive traditional solutions were used for quite all spillways: fully gated spillways and long free flow spillways of low specific flow. Low cost alternatives are now available: they include much more efficient free flow spillways (labyrinth weirs which multiply by 3 the specific flow) and various fuse devices (earth or concrete fuseplugs, fusegates, …). Such fuse devices, much less expensive than gates, open for exceptional floods and are replaced in few weeks or months. All these solutions may be combined with gates and / or with an optimization of embankments crest; but the traditional design methods prevent or underestimate many of the relevant possibilities.
– Most future dams and dams upgrading will be in developing countries: the cost efficiency is much more important than it appears in present criteria or regulations.
2) Traditional “Design Flood” or “safety Check Flood”?
The design of most existing dams and of many dams under construction has been based upon a “design flood” which can be spilled (and possibly partly stored) without damage. For most gated dams the corresponding reservoir level, called High Water Level, is the same as the normal operating level. The impact of earthquakes or long term seepage is evaluated for this reservoir level which is well under the dam crest; the gap between the High Water Level and the dam crest is called Freeboard and is designed for avoiding any disturbance from possible waves, i.e. according to the possible wind speed and to the reservoir shape; it is usualy 2 to 6 m for the large embankment dams.
The yearly probability of this “design flood” is chosen by the Owner and the Designer; it may be fixed by law and varies according to the countries, to the dam type or the reservoir size and to the possible damages. This design method was rather logical when the chosen “design floods” were close to the likely floods during the dam life. It seems to have now two very serious drawbacks:
– This method gives a poor idea of the true safety because it does not take in account the fact that the failure happens for a reservoir level higher than the High Water Level nor the uncertainty in evaluating the 1000 years flood. Many small ungated embankment dams may withstand the P.M.F.; but very large gated dams with a design flood of yearly probability 10-3 may fail for a 10-4 flood with all gates open or for an yearly flood in case of all gates jamming.
– This method prevents or reduces the utilization of low cost solutions that may increase the true safety and reduce the cost.
Consequently additions to this old design method have been made in various countries. When upgrading existing dams, reference is often made to the overtopping of the dam crest and not to the High Water Level. This should better take care of the true reservoir level corresponding to the failure: this level is in the range of 0,5 m over the dam crest for most earthfill dams, 1 m for rockfill dams, well over the dam crest for most arch dams or high gravity concrete dams but it may be under the crest for a very pervious crest of earthfill dams, for masonry dams or for some low undrained concrete gravity dams. Some countries are refering accordingly to the Imminent Failure Flood (I.F.F.), i.e. to the levels hereabove. It is however difficult to evaluate these levels precisely and the present tendency is to refer to a “Safety check flood” (S.F.) which is advocated by ICOLD Bulletin 82 published (1992 pg 161 and 169) and Bulletin 125 (2003 pg 173 and 191).
This “Safety check flood” as defined by Bulletin 125 (p.191) “represents the most extreme flood conditions to which the dam could be subjected without failure, but also with a low safety margin (scenario limit). In this case, a limited overtopping may be permitted for concrete dams, but not for embankment dams”. For instance in China, for this flood, the water level for very large embankment dams is most often 0 to 1 m under the dam crest. For this very exceptional flood some limited extra costs and damages to waterways or loss of fuse devices are acceptable. And the exceptional loads such as seismics or long term seepage impact should not be added to the impact of the increased water level: usually such extraordinary water level increase will have thus little or no impact on the main dam structure.
If using the “Safety Check Flood” for extreme floods, the purpose, the value and even the name of the traditional “design flood” may be questioned. A reservoir level well under the dam crest as for the traditional design flood should be kept normaly all along the dam life and no damage would be then anticipated; but the return period of the corresponding flood may be 1/100 or 1/500. It is confusing to keep for this flood the name of “design flood” because the design will be mainly based upon the Safety Check Flood and it could be better to choose an other name such as “Operational Flood” (O.F.).
If accepting this logical basis of two floods for the designs, the S.F. and the O.F., the choice of their return period is examined below. But the impact of the costs and of floods evaluation is first underlined.
3) Cost data
Most regulations and design criteria refer to ethical, philosophical or political views and not at all to costs; the only reference is that the risk should be kept “As Low As Reasonably Possible” (A.L.A.R.P.)”. This approach may cause extremely high expenses, with a low efficiency on human safety improvement. The Peak Flow to spill for the Safety Check Flood (S.F.), may be close to the incoming flow or reduced by storage in the reservoir. Anyway it includes two parts: – The peak flow to spill for the Operational Flow (O.F.) herebelow called Basic Spilled Flow. – The gap, herebelow called Additional Spilled Flow. The best relevant solutions may not be the same and the cost per m3/s for the Additional Spilled Flow may be much lower than for the Basic Spilled Flow because an higher reservoir level and some damages or losses are accepted.. Some examples are given below.
3.1) For Basic Spilled Flows under 1,000 m3/s the spillways are usually long ungated weirs. A nappe depth of 2 m for such flood is typical of many of them, with an earthfill dam crest about 3,5 m above the weir sill (1,5 freeboard). The structure is simple but the indirect cost of the Basic Spilled Flow is high; the main extracost is usually the loss of storage (or extra dam height) corresponding to the 2 m nappe depth. As most of these dams are lower than 25 m, the loss of storage is over 30 % of the live storage. And the relevant total cost per m3/s is thus usually thousands of U.S. $.
If the reservoir level for the Safety Check Flood is 0,50 m under the dam crest, the corresponding nappe depth is increased from 2 to 3 m, the relevant flow is about 80 % more than the Basic Spilled Flow, at no extra cost. There are three ways for increasing further the acceptable flow at the same place:
– Lowering the sill by H and placing a fixed labyrinth weir (such as a Piano keys Weir) of same wall height. The extra flow per m shall be in m3/s close to 2 H1,5 and the quantity of reinforced concrete will be about 0,4 m3 per m3/s in addition to the cost of the basic weir. For existing dams, the cost of lowering the sill should be added. The total cost will be few hundreds U.S. $ per m3/s.
– Lowering the sill and placing concrete fuseplugs which may tilt progressively for nappe depths of 2,5 to 3 m. They can be simple, requiring less than 1 m3 of ordinary concrete per extra m3/s , or more sophisticated and possibly labyrinth shaped, with low quantities of reinforced concrete (fusegates).
– Raising the dam crest by a parapet: if L is the length of the crest and l the spillway length (in m), and c the cost per m for increasing the height by 1 m, the extra cost for heightening by 1 m will be Lc and the flow increase in m3/s, about 5 l. The cost per extra m3/s is c x L/5l i.e. close to c if the dam length is 5 times the spillway length. c is the cost of about 1 m3 of ordinary concrete and is in the range of 100 U.S. $.
The cost of several solutions is thus in the range of a few hundreds U.S. $ per extra m3/s for additional Spilled Flow and not thousands as for the Basic Spilled Flow. With Labyrinth Weirs it is easy to reach 50 m3/s/m, with fuse devices over 100 m3/s/m. it is thus possible to reach rather high Safety Check Floods at a low cost per m3/s even for large spilling capacities.
3.2) For larger flows the traditional spillways are usually fully gated, most often with large tainter gates. The direct cost is high; beyond the mechanical and electrical equipment it requires for the piers and sill 1 or 2 m3 of reinforced concrete per m3/s. The total cost per m3/s is thus thousands U.S. $.
There are usualy at least 3 or 4 gates for taking care of maintenance and repairs. As an example consider a spillway of 4 tainter gates of 12 x 12 m spilling 4 x 1,000 m3/s as Basic Spilled Flow with a freeboard of 5 m. If the Safety Check Flood is 12,000 m3/s, and the corresponding reservoir level 1 m under the crest, the flow of gates is increased proportionally to: 
i.e. by about 50 % i.e. 2,000 m3/s at no extra cost.
The balance of 6,000 m3/s may be reached by 4 additional similar gates or for a new dam by a total of 6 larger gates. The extra cost would be high and the risk of total gates jamming would not be avoided. An increase of crest level by 1 m would only increase the flow by about 500 m3/s. But there are various solutions of additional spillway at low cost:
– A labyrinth weir may spill 40 m3/s/m using the freeboard as a 4 m nappe depth requiring less than 0,5 m3 of reinforced concrete per m3/s. The corresponding spillway length is 150 m.
– Fusegates 10 m high, tilting progressively for exceptional floods; the specific flow under 10 + 4 = 14 m of nappe depth without piers will be over 100 m3/s/m and the spillway length about the same as with gates. It requires less reinforced concrete than the piers of gates and its total cost is thus much lower than the total cost for gates.
– Earthfill fuse plugs, whose cost may be close to the cost of fusegates, with however less precision and less possibility of progressive opening according to the flood.
– For dams including long low dykes, it is often possible where the dam is lower than 10 m to line the downstream slope and toe in Roller Compacted Concrete (2 or 3 m horizontal width) and to spill 10 to 15 m3/s/m. Such solution requires about 2 m3 of R.C.C. per m3/s but needs specific sites because the spillway length is in the range of 500 m. It is also possible to line the slope with a geomembrane, protected by 1 m erodible fill.
The cost per m3/s of extra spilling capacity will thus vary considerably with the dam site but will be usually much lower than for the Basic Spilled Flow. The cost of waterways (chutes, tunnels, stilling basins) are also usually much lower because important damages are acceptable for instance for floods of yearly probability 10-3 or 10-4.
For many embankment dams which cannot be upgraded easily up to the P.M.F., it is at least possible at low cost to reduce the dambreak flow peak. This flow peak varies considerably with the depth of the breach (or breaches). Placing a parapet about 1 m high where the dam is higher than 20 or 50 % of the maximum dam height will reduce the possible breach depth and at least halves the dambreak flow: such solution would have considerably reduced the damages and fatalities of most past embankment flood accidents.
4) Floods evaluation
The evaluation methods are not the same for the “Safety Check Flood” which is close to the extreme floods and for the “Operational Flood” which is close to the 100 years Flood.
4.1) Evaluation of Extreme Floods
There are now much more worldwide data about the maximum reported floods according to the catchment areas (ICOLD Bulletin 125, page 75). They are close to the following: Extreme flows Reported Worldwide 
and may be roughly represented by 2 formulae:
For S < 300 km² q = 10,000 (S/300)0,8
S > 300 km² q = 10,000 (S/300)0,4
The corresponding water depth over catchments areas under 1,000 km² has been about 0,5 m, which corresponds to the maximum rain depths registered : over 0,7 m in 4 hours on few km² or in 12 hours for 1000 km² (some higher rain depths have actually been registered but were limited to some mountainous islands in warm seas).
Usually an exceptional storm event corresponding to extreme floods hits fully the whole catchment areas of 1,000 km² but only a part of very large catchment areas.
For extreme floods, (P.M.F.) on catchment areas under 1000 km², the rain depth involved is usually between 0,20 m and 0,60 m according to the climate. The water retention in few hours by soil or vegetation is a rather small part of such rain depth, especially if there have been previous rains. The flood duration may be evaluated according to S with some limited adjustment for the slope and shape of the catchment area (ref. Floods and Reservoir Safety Institution of Civil Engineers London – p.27).
A flood evaluation by a detailed method may thus be checked by simple regional formulae; and the comparizon between dams of a same region is reliable because the impact of the different soil and vegetation conditions is small; the catchement area, its shape and slope are well known and may be easily taken in account for comparing dams with same methods of flood evaluation for such catchment areas. It is thus possible to identify the dams most at risk even if there is uncertainty upon the regional value of extreme rains.
The evaluation is much more difficult for catchment areas over 10,000 km² of which a part only is concerned by an exceptional storm, and where the impact of the retention on most of the catchment area may be high and variable; the evaluation of their exceptional floods may vary considerably with the designer. And there are few worldwide data for very exceptional floods on very large catchment areas.
4.2) Evaluation of the 100 years flood
The water retention by soil and vegetation which may be close to the rain depth corresponding to such flood varies considerably with each catchment area. In a same climatic area, the 100 years flood may thus be very different for 2 dams of same catchment area.
The existing data of past floods may cover half a century for catchment areas of 1000 km² and several centuries for very large catchment areas. For catchment areas under 100 km² there may be few or no precise data of past floods. Even in populated areas, data may be limited to 20 years with some data about few rare flood events.
Based upon floods data it is thus possible to get reliable evaluations of floods of yearly probability 1/50 for small catchment areas and 1/500 for very large ones. But changes of climate or vegetation may seriously modify them.
4.3) Evaluation of the 1000 years flood
This has been the choice for many “design floods” and the basis of many dam designs before 1970, i.e. for most existing large dams. The most usual floods evaluations were based upon statistical methods relying only on floods data.
These methods are now much questioned because many floods data are not precise or reliable and may refer to different climatic events; and many existing spillways require huge upgrading based upon modern evaluation methods.
The modern methods for such floods combine a statistical analysis of regional and local rain data with an evaluation of water retention based upon floods data.
The reliability varies with the reliability of floods data; and it is higher where the probable rain depth is well over the retention value.
For most catchment areas the reliability of the modern evaluation of such floods is not better than the evaluation of extreme floods which is less impacted by questionable flood data. However, for very large catchment areas, such evaluation may be better because there are historical floods data along centuries.
4.4) Dambreak floods and natural floods
The dam failure may multiply by 5 or 10 the natural flow or have little impact on it.
– The floods caused by embankment dam failures vary considerably with the reservoir volume and with the dam height or more exactly with the place of breach (or breaches) and with the nature of dam and foundation materials because the breaches may widen slowly or quickly.
The reservoir volume may be higher than or close to the volume of extreme floods, specialy for irrigation dams on rather small catchment areas, but it may be under 1 % of the flood volume for large catchment areas.
The dambreak flow may thus for many dams be not much higher than the natural flow of a flood of yearly probability 1/1000 or 1/10,000 and it may not be justified to design the spillway for more extreme floods. Avoiding overtopping where the dam is high may also reduce considerably the dambreak flow because practically a part of the dams is thus acting as an earthfill fuse plug.
– For concrete or masonry dams the failure is sudden and difficult to foresee precisely. The breach is usually wide, easily five times the depth; the flow may thus be 10,000 m3/s in minutes for a dam 20 m high. Such failures will then be quite always much more dangerous than natural floods.
5) Choice of the return period of the Operational Flood (O.F.) and of the Safety Check Flood (S.F)
5.1) Operational Flood (O.F.)
– As the safety is essentially linked with the Safety Check Flood, the return period of the operational flood may be closer to the dam life than it was for the traditional “design flood”. It may be in the range of 100 years or few hundreds years. An uncertainty in its evaluation has no impact on the safety. The relevant design conditions may not be the same for gated or ungated dams; for gated dams the reservoir level will be most often at the top of the gates; extreme waves and the possibility of extreme earthquake should be considered. For ungated dams such extra loads of waves or earthquakes could be combined with a small flood and could be reduced for the higher water level reached by the 100 years flood. Some damages (waterways, fuse devices loss) may be accepted for floods between the O.F. and the S.F. For instance the actualized value of a damage of 1 million $ for a 1,000 years flood is only about 20,000 $.
5.2) Safety Check Flood (S.F.) – Political choice
For the safety check flood, the reservoir level is close to the embankments dams crest, higher than the crest for most concrete dams. The structural calculations for this exceptional level should not take in account earthquakes or long term extra seepage. The waves have no impact on the structural stability and the probable waves in few hours will have little impact on the embankments breach opening (no failure report refers to waves impact); and anyway there is usually for embankment dams a gap of 1 m between the S.F. reservoir level and the probable level of failure.
The choice of the return period may be based upon ethical or political reasons.
The presently reported worldwide rate of failure by floods is in the range of 10-4 for large dam, 10-3 for small ones. This situation is not considered as disastrous, but as requiring improvement.
The legal return periods are often between 10-4 and the P.M.F. (i.e. possibly 10-6). It is not always precised if they apply to the “design flood” or to the “safety check flood”.
The usualy acceptable yearly risk of fatality for downstream people is in the range of 10-6, the same as the probability of death from lightning or from one hour driving (this means about 1/10,000 along a man life). As the probability of fatalities may be under 1/1000 or over 1/10 in a dam break inundated area according to dam type and alarm systems, the relevant dam failure return period is between 10-5 and 10-3. Even with few fatalities risks, a failure probability of 10-3 appears unacceptable upstream of populated areas because during their life, people downstream would have a global probability of 1/10 to be subject to a dambreak flood.
– The political choice may be linked with public acceptance and vary with the countries. There is anyway a general view that the true yearly probability of failure should be less than 10-4 for dams of which the failure is likely to cause over 10 fatalities, an event that media call disaster.
All considered, it appears that the future acceptable yearly probability of failure should be as average in the range of 10-5 and anyway less than 10-4 if there are more than 100 people downstream. Same rates are also advisable for existing dams. P.M.F. will be required probably for very large dams.
It is often considered that these requirements are well beyond cost optimization and that the relevant expenses should better be used else where. This may be true with the traditional “design flood” method and with traditional technical solutions. But designs based upon the “Safety Check Flood” and new technical solutions may meet as well cost optimization as future safety targets. Examples are given below for various catchment areas and for new or existing dams.
5.3) Cost optimization for the Safety Check Flood
It may be based upon sophisticated calculations: a simpler presentation is given hereunder.
5.3.1) Ungated embankment dams
Most large dams are ungated earthfill dams for irrigation; built on catchment areas under 100 km²; they store most of the yearly flow.
Such a typical dam for a 50 km² catchment area, would have an average flow of 0,5 m3/s, an yearly flow of 15 millions m3 (world average for 50 km²) and a storage of 10 million m3 for a construction cost of some million U.S. $. Relevant floods data are realistic:
The extra cost for dividing by 10 the failure probability i.e. for 200 m3/s, (according to chapter 3 above few hundreds U.S. $ per m3/s) is in the range of 50,000 $.
If the yearly probability of failure is 1/T and if D is the cost of relevant damages (or more exactly the extra cost over the damages of the natural flood), the actualized value of this failure risk is in the range of 20 x D/T. For a storage of 10 million m3 the dambreak flow may be 5 or 10,000 m3/s inundating probably 5 or 10 km² and hundreds houses. The value D of damages may thus be 10 million U.S. $ or more.
The total cost over the cost of a dam designed for the 100 years flood (operational flood) includes the cost of the extra spillage capacity and the actualized value of the failure risk. It is given below according to return periods.
The total cost appears lowest for T close to 10,000 years. In fact the optimum of T is close to: 
The design can be made for this return period, i.e. for a Safety check Flood of 850 m3/s; choosing a return period of 100,000 years, i.e. 1,000 m3/s would need an extra cost less than 50,000 $, about 1 % of the dam cost and could take care of the uncertainty in floods evaluation, the increasing cost of houses and the impact of possible climatic change.
The figures here above apply as well to new or existing dams. For damages limited to a part of the dam cost, a return period of 5 or 10,000 years should be advisable. For smaller catchment areas, for instance 10 km², the floods (and spillway costs) should be divided by about 4 but the storage and probable damages would be also seriously reduced. The optimal return periods should not be much different.
For ungated spillways the low cost of solutions for the additional spilling capacity between the operational flood and the safety check flood gives the possibility to meet the political advisable safety level and to be however close to the cost optimization. This cannot be reached with the traditional “design flood” method and relevant traditional expensive solutions.
5.3.2.) New gated embankment dams
An example may be an hydroelectric dam on a catchment area of 10,000 km²; with an average flow of 100 m3/s.
Relevant flood data are:
The cost C for increasing the discharge by 2,000 m3/s may be in the range of 1 or 2 million U.S. $ (2,000 x 500 or 1,000). The yearly flow is 3 billion m3. It is unlikely than the storage be so high but a storage of hundreds million m3 is possible with a dam cost well over one hundred million U.S. $. The failure of such dam may cause a flood over 50,000 m3/s with damages of hundred million $. Optimization may be studied for C = 2 million $ and D = 500 million $.
The optimized value of T is close to 20,000. The choice may be T = 100,000 to take in account some uncertainties on occupation and climate. Choosing the P.M.F. or T = 1 000,000 is not justified economically but the extra cost is only few per cent of the dam cost.
But many dams on such catchment areas may be rather low and store only a small part of the extreme floods volume. The damages over the natural flood may be low. The cost optimization is examined below for D = 20 millions $ and C = 1 million $
In this case the “safety check flood” could be taken for 1,000 years and it is better to invest in an alarm system used also for natural floods than to spend million $ for avoiding unlikely damages of 20 million. Adapting such dams to the P.M.F. is a waste of money.
5.3.3) Existing gated embankment dams
The risks of damages are the same as for new dams. The cost for upgrading spillways may be similar to the cost for new ones, especially for long dams where it is easy to place an additional free flow spillway or fuse devices. For short length dams, additional spillways may require tunnelling for the waterway: the cost may be reduced if using tunnels controled by fuse devices, at high level, with a slope in the range of 5 %, possibly unlined, accepting exceptionally serious damages downstream of the tunnels.
5.3.4) Concrete or masonry dams
The traditional approach and regulations suggest to accept overtoping for return periods of 1,000 to 10,000 years. This seems very questionable: the true risk of failure may be for huge overtoping of most arch dams or high concrete gravity dams but it may be caused by a small overtoping if foundation may be eroded (Lake Lanier Arch in U.S. in 1926), for overtoping by few m for low gravity dams (Zerbino in Italy in 1935) or with little overtoping or even before overtoping for masonry gravity dams (Tigra and Chikkalobe in India, Xuriguera in Spain).
Anyway the failure of such dams is much more dangerous than from embankment dams because the breach is sudden and much wider and more difficult to forecast exactly. A 15 m high dam may cause a sudden flow over 5,000 m3/s with a storage under 5 millions m3. The safety check flood of concrete and masonry dams should thus be the Probable Maximum Flood or a flood of 10-6 probability with a reservoir under the dam crest or much over it according to the dam structure and to the possibility of foundation erosion. The cost for upgrading concrete or masonry dams is usualy lower than for embankment dams because long spillways or emergency spillways are easier: this favours the choice of a very unprobable safety check flood for all concrete and masonry dams.
5.4) Overall choice of the Safety Check Flood
If using this design method and the low cost relevant solutions, the best choice for the S.F. may be often a return period in the range of 10-5 close to the cost optimization, with a reservoir level adapted to the dam structure, often close to the dam crest. However:
– For very large reservoirs it may be politically justified to choose the P.M.F. at an extra cost of few percent of the dam cost.
– For concrete or masonry dams the P.M.F. or a yearly probability of 10-6 may be advisable to avoid fatalities, with a reservoir level adapted to the structural data.
– For earth dams storing a small part of the large floods, or for dams without downstream population, a return period under 1,000 years may be reasonable. Some adjustments or precisions of existing regulations may be justified accordingly.
6) Additional comments
The comments hereabove refer only to the dams safety. The choices hereabove could be adjusted for taking care of storage optimization, and of floods or sedimentation mitigation.
– The flow of existing free flow spillways may be increased by labyrinth weirs or concrete fuse plugs or fusegates. These solutions may be used also very efficiently for raising the permanent reservoir level. Both targets may be reached and paid by the corresponding increase of storage.
– Free flow spillways may use labyrinth weirs for most of the extreme floods but may be combined: . Either with low gates for sluicing or flushing sediments . Or with surface gates for mitigating the 100 years floods down to a 10 years flood. Such gates would not be used often and would not require permanent operators.
Conclusion
The traditional design method based upon a “design flood” of about 1,000 years return period with an important freeboard is costly and may however be unsafe. The designs should preferably be based upon :
– A “safety check flood”(S.F.) of very low probability with a reservoir level close to the embankment dams crest (or over it for high concrete dams).
– An “operating flood” (O.P.) of return period 100 to 500 years (slightly over the dam life) with an important freeboard and no damages. Some damages are acceptable for floods between the two floods hereabove. The name of ‘design flood” is confusing and should be preferably avoided.
This solution is also better adapted to the uncertainty in floods evaluation. The evaluation of floods is mainly based on rain data for the S.F., on floods data for the O.F. This solution favours many low cost solutions for spilling the flows over the operational flood.
It is thus possible to associate the cost optimization with a high safety level for quite all new or existing dams upstream of populated areas. This applies as well to rather small dams as to very large ones and may be implemented worldwide in few years for upgrading existing dams.
References :
– ICOLD Bulletins
82 (1992) Selection of Design Flood
109 (1997) Dams lower than 30 m (appendix 1)
117 (2000) The Gravity Dam: a dam for the future
125 (2003) Dams and Floods
1989 Floods and Reservoir Safety: Institution of Civil Engineers (London)
1993 Unusual Storm Events: C.B.I.P. New Delhi
1999 International Symposium on Flood Control Beijing
F. LEMPÉRIÈRE has been involved in the construction and/or design of fifteen hydraulic schemes on large rivers including Rhone, Rhine, Nile and Zambezi. He is honorary chairman of the French Committee on Large Dams. He has been chairman of the ICOLD Committee on Cost of Dams (1991-2001). He is chairman of Hydrocoop, a not profit international association for technical exchanges.
J.-P. VIGNY has been involved in the design or construction of large civil engineering schemes in rivers or at sea, including Cabora Bassa Dam on the Zambezi, and Mudhiq Dam in Saudi Arabia. He is general manager of Hydrocoop.