- Dams, Piano Keys Weirs, Tidal Energy & Energy Storage - http://www.hydrocoop.org -

New methods and criteria for designing spillways could reduce risks and costs significantly

 F. Lempérière and J-P. Vigny, Hydrocoop, France L. Deroo, ISL, Franc e

Present methods for designing spillways are generally based on traditional criteria, which are not well adapted to present knowledge of actual risks. A review of facts and costs suggests that better adapted criteria and methods of design could reduce risks and costs. After an analysis of dam failures, the authors review present methods of designing spillways, and propose possibilities for reducing costs, while increasing discharge capacity. The suggested methods and design criteria could apply to both new dams and upgrades.

Floods have been and are the main cause of dam failures. The criteria and methods for designing spillways vary from country to country, but are usually based on traditional practice established when the knowledge of risks was not the same as it is today. They have the following serious drawbacks:

• they overlook the serious risk of gates completely jamming
• they do not focus enough on costs, and thus overlook or prevent low-cost solutions which may be the key to safety (halving the cost for spilling 1 m3/s makes it possible, for the same cost, to double the spillway capacity, and thus to divide by 100 the failure probability).

This paper reviews facts and costs, and proposes new criteria and methods for designing spillways: they may lead to cost savings as well as safety improvement.

1. Causes of flood failures and resulting damage

Floods have been the main cause of dam failures and of 90 per cent of related fatalities. Most fatalities resulted from failures of 20 to 30 m-high dams, and few have resulted from the thousands of failures of dams lower than 15 m.

Most failures have been caused by floods which exceeded the spillway capacity, but a significant number of gated dams have failed as a result of total jamming of the gates for various reasons, such as mechanical or electrical problems, lack of operators or operational errors. Some failures were caused by excessive floating debris reducing the discharge capacity or from failures of upstream reservoirs (artificial or natural).

The failure conditions are very different for embankments and for gravity dams, and are analysed separately below (see also ICOLD Bulletin 109 pages 24 to 39 and appendix 1, Bulletin 117 pages 40 to 45 and Bulletin 82: cases studies).

Spitskop dam (South Africa) overtopping in 1988

Spitskop dam, one hour later

Euclides da Cunha dam, Brazil, after failure in 1977

1.1 Embankments

More than 90 per cent of flood-related failures of dams have been at embankment dams. Many failures occurred in the USA before 1930 and in Asia up to around 1980. The rate of failures has greatly reduced in the past 30 years for large dams, but hundreds of dams in the height range of 5 to 15 m fail by overtopping every year (usually without fatalities).

Virtually all embankment failures by floods have been caused by overtopping of the crest, and some by erosion along the spillway or failure of the spillway.

The specific discharge over the crest for a nappe depth d (in m) is about (m3/s/m) 2 d1.5; that means, with a quite low water speed on the crest of 2 √d. The water speed along the downstream slope is much higher, and can reach about 10 √d, that means, 5 m/s for a nappe depth of 1 m. The erosion is thus usually initiated on the downstream slope, especially at the dam toe, for a crest nappe depth which varies a lot according to the dam materials and the vegetation if any.

The speed in the breach, when it is open, is in the range (in m/s) of p0.5, p being the dam height (in m) of the dam at the breach point. It is thus less than 5 m/s for p < 25 m. The time for opening and widening the breach can thus be very different according to the dam body or dam core quality, and to the reservoir volume.

The breach has tended to open widely in old, long dams lower than 30 m: Banquiao dam (China, 1975) and the Machu dam (India, 1979) failed with a 80 000 m3/s downstream discharge and breach lengths of about 500 m (Banquiao) and more than 1000 m (Machu). Good compaction of the clay core, or the dam body itself, may limit and delay the extent of a breach, as was the case at the 40 m-high Euclides da Cunha dam (Brazil) or the 100 m-high Teton dam (USA), where the breach remained almost triangular. The flow (m3/s) through such breach may be limited to about p2.5, but several breaches may occur if the reservoir storage is very large.

Erosion in rockfill may be more significant than in well compacted clay materials, and the widening of a breach in a rockfill dam is mainly limited by the width and quality of the clay core. Erosion in gravel and sand may be very extensive.

1.2. Concrete and masonry dams

The data for masonry and concrete dams are quite different. The rate of flood failures has been quite high for masonry gravity dams, either at the foundation level or in the dam body. This may be caused by a lower density than expected, uplift in upstream cracks or low mechanical strength, especially for very old dams. The breach may be instantaneous and its length can be about five times the dam height. As the reservoir level corresponding to the failure may not be well known, alarms are less efficient than for embankment dams, and such old dams require great care, as the risk of fatalities is high.

Few failures from floods have been reported for concrete gravity dams, and these have generally been limited to dams lower than 20 m: the impact of an exceptional increase by a few metres in the reservoir level during floods is much more important for a low gravity dam than a high one, and an increase in the downstream water level impacts the stability by uplift.

There have been very few failures in floods for arch dams, including more than 500 masonry arch dams in China. A 20 m-high arch dam failed in the USA 80 years ago as a result of overtopping and erosion of poor quality bank rock.

Fig. 1. Typical discharges, q from failures.

1.3 Downstream damage and human risk

These are based on the downstream discharge curve, which is very roughly represented in Fig. 1 for a flood volume close to the reservoir storage volume, and for three different dams of the same height and storage capacity, with the following types/conditions:

• concrete or masonry dam;
• earthfill dam with poor compaction; and,
• earthfill dam with efficient compaction.

According to ICOLD Bulletin 109, Appendix 1, referring to all failures (except China) before 1990:

• Of 4500 masonry or concrete dams, 35 failures have been reported: 22 caused fatalities, including 10 failures with more than 100 fatalities each.
• Of 12 000 embankment dams, 140  failures have been reported: 33 caused fatalities, including 14 failures with more than 100 fatalities each.

The failure of a concrete or masonry dam is the most dangerous: the discharge can be very high, even for low dams with a limited storage (more than 1000 m3/s for a 10 m high dam). The ‘check flood’ for such dams should have a very low probability.

The risk for embankment dams varies significantly with the quality of the earthfill, the reservoir volume and, for long dams, with the point of the first breach, which may possibly be predicted in advance.

The time taken for the breach to widen may be high compared with the flood duration: this may reduce the risk for catchment areas smaller than 100 km2 (the time for floods to peak is usually less than 5 hours).

Incremental damage and fatalities may be low for embankment failures caused by floods with an annual probability of less than 1/10 000, because the area inundated naturally is already very broad. They may be much higher for a failure associated with a total jamming of the gates during a flood with an annual probability of 1/10 or 1/100.

2. Flood evaluations and present relevant

There has been, is, and will be great uncertainty in evaluating floods of various probabilities and their impact on dams. This uncertainty could increase with anticipated climate change.

Design criteria are often based on a ‘design flood’ of theoretical annual probability of 10-3 to 10-4, or the PMF, corresponding to a reservoir level lower than the dam crest by a freeboard adapted to possible wave height. Criteria are similar for gated or ungated spillways, and thus favour the choice of gated spillways, and do not take into account the risk of complete jamming of the gates.

It is remarkable that design floods, even when based on modern standards, have been exceeded, or re-evaluated, by a factor of 2 or more, during dam operation. This was the case, for instance, at the following dams.

• Bagré reservoir in Burkina Faso was impounded in 1992; its design flood was re-evaluated from 1950 m3/s to 5500 m3/s, because of observations during the first years of operation,

• Yatédam in New Caledonia was designed for a 4500 m3/s flood; cyclone Anne in 1992 brought a flood with a peak of 8000 m3/s and the newly computed 1:1000 flood is now 12 000 m3/s,

 

• The La Rouvière, Conqueyrac and Ceyrac reservoirs in the south of France were designed in the 1960s for a flood with runoff of 100 to 140 mm; this was re-evaluated to 175 to 210 mm in 1985 , and surpassed by the 2002 flood, which brought runoff of 220 to 450 mm.

• Kissir dam in Algeria was completed in 2011; floods occurring during construction resulted in a re-evaluation of the 1:10 000 years design flood, from 940 m3/s to more than 2000 m3/s.

These dams did not fail, either because the flood they underwent did not exceed the design flood, or because the initial reservoir elevation was below maximum, or because the dam withstood overflowing, with only little erosion.

2.1 National regulations regarding floods

National regulations for floods vary a lot, even in cases when they have recently been revised. A short comparison is given below between regulations or standards from the Australia, USA, Switzerland, Italy, UK, Canada (Québec), Sweden, and Norway.

Most of these countries adopt a design combining a design flood and a freeboard. Some also consider verification for a check flood, with no freeboard, see Table 1.

Furthermore, methods for flood computation also vary a lot, see Table 2.

Tests were made to compare how various regulations and methods would impact the computation of the dam crest elevation, given a specific spillway. Three 30 mhigh French dams, each using a different spillway, were compared. Regulations or standards used were from: Australia, USA, Switzerland, Italy, UK, Canada (Québec) and France.

These tests led to the differences shown in Table 3.

Consensus is far from being reached on the issue of spillways design.

The surprising figures in Table 3 result from the great uncertainties about the impact of floods on dam safety.

Uncertainty in flood evaluations includes the high degree of uncertainty regarding past flood estimates, and the uncertainty related to flood computation methods.

Flood routing computations are mainly based on rainfall-runoff methods. It has been shown that these methods systematically underestimate peak discharge and flood volume. Estimates of extreme floods are always altered by some uncertainty, and extreme flood estimates which do not take into account uncertainty are mathematically underestimated.

In the case of a medium sized catchment with few available data, the incidence of uncertainties is much larger.

There is also uncertainty about the real discharge capacity of gated spillways. spillways do not always provide their nominal discharge capacity; it can happen that gates do not open.

There is quite a high probability that one gate will remain closed during a flood, either because it is undergoing maintenance (stoplogs) or because of defects.

This ‘one-gate-not-opening’ probability depends greatly the condition of the dam. For reasonably well maintained dams, the probability of one gate not opening during the largest annual flood is likely to be of a magnitude of 10-1 to 10-2, or in exceptional cases it could be 10-3.

There is also a probability that, during a flood, all gates will remain closed. The probability of such an event is highly dependant on all the site conditions, including the following.

• the possibility to detect flood events soon enough (which depends on: the flood duration, the existence of automatic control-command, and human presence at dam);

• once the flood event has been detected, the ability to give the gate opening order (which depends on: the existence of strict gate operating procedures, the existence of an automatic control-command system, and, the organisation and skill of the operations team); and,

• once the order to operate is given, the possibility to operate effectively (which depends on: access conditions and energy availability).

Even for a well designed and well managed spillway, this risk cannot be ruled out, except in the case of very large catchments, when flood events are slow enough to allow several days for these operations. For a well designed and well managed dam in a medium size catchment, the probability of all the gates remaining closed during a flood event is of a magnitude of not less than 10-4 to 10-5. This figure takes into account that many ‘non-opening’ events have already occurred, for various reasons, in several countries. Some of them resulted in dam failures.

This probability may be increased by the various impacts of exceptional rainfall, by poor management, or by civil war or similar disorders.

Of eight accidents analysed in detail in ICOLD Bulletin 82, four were associated with partially or totally reduced discharge capacity of the gated spillway.

Other uncertainties are:

• uncertainty about the impact on floods of climate change;

• uncertainty about the precise reservoir level for which a failure is probable, especially in the case of masonary and concrete dams; and,

• uncertainty about downstream damage and human risks, depending on the dam and the failure point.

However there are some certainties:

• Increasing by 20 or 50 per cent the true discharge capacity can reduce by between 5 and 20 times the probability of failure: reducing the cost of spillways, and increasing their capacity accordingly may be the best solution; this should be favoured by design criteria. The present design criteria based on the ‘design flood’ overlook most of the uncertainties mentioned above, favour fully gated spillways, are too costly for many free-flow spillways and virtually prevent often the attractive solution of combining a gated and an ungated spillway.

• For failures of low probability at many dams, the incremental downstream damage and human risk may represent only a small part of the risks of the natural flood: this may limit the desirable discharge of the spillways.

• Reducing the probability of the ‘design flood’ will increase the cost; it may improve the safety but it could also reduce it, by preventing the choice of more cost-effective solutions.

In any case, increasing the cost increases the amount of work and the associated risk of accidents to the workforce during construction. For 30 years there have been more fatalities during dam construction than have been caused by dam failures.

3. Possible cost savings

3.1. Optimization of the dam crest

3.1.1 Embankments

The usual criteria for embankment crest design address two issues: freeboard for waves and limiting the development and the consequences of clay material cracking.

Wave protection could be investigated, to account for the probability of the combined occurrence of wind and flooding. It is clear that, excluding the case of reservoirs without surcharge flood (maximum water elevation = normal water elevation), and assuming the plausible combined probability of flood and wind, standard criteria are conservative. In fact, freeboard above MWL is probably mainly useful for flood safety, in the case of floods which are greater than the design flood, or in cases of unavailability of the spillway.

Above the MWL, special care should be taken with clay core earthfill dams. Material issues (such as cracks in the clay fill) are not well dealt with by increasing either the crest freeboard or width. The safety is instead ensured by an adequate design (regarding shape, materials and methods of construction) which will limit cracking, and filters potential leakage, and also a design for the parapet wall and dam crest which can resist very high water levels and some submergence by waves. This would safely handle reservoir elevations above the MWL.

3.1.2 Gravity dams                 

Criteria for the dam crest should be very different for gravity and embankments dams. It may be efficient to increase the freeboard of an embankment; but an increase in a gravity dam’s freeboard increases the reservoir level in the case of overtopping, and this reduces the dam’s stability, especially for low dams. It may be better to reduce the gravity dam’s freeboard and to protect the dam toe in case of overtopping. Waves have no impact on the safety of gravity dams, and the freeboard requirement is mainly associated with the risk of waves for dam operators or possible traffic on a road across the crest.

Fig. 2 Layout of Piano Keys Weir A

Fig. 3 Layout of Piano Keys Weir A

Fig. 4 Layout of Piano Keys Weir A

Fig. 5 Layout of Piano Keys Weir A

 

3.2. Increasing the discharge of free-flow spillways

The drawback of traditional free-flow spillways is their rather low specific discharge. For the usual shape of a Creager weir and for 1 m of spillway length, the discharge is (in m3/s) 2.15 H1.5, H being the nappe depth in metres. For large discharges, the spillway is therefore long and costly, and/or the nappe depth will be large, causing a significant storage loss.

Labyrinth spillways have been used since 1950; these are vertical walls with a long trapezoidal layout, doubling or tripling the discharge of a Creager weir. But their use has been limited to about 100 dams, for two main reasons: a limited cost saving because the design of the structure has not been optimized, and an excessively sized base, preventing their use on most spillway structures.

Since 2000, many studies and model tests have demonstrated the possibility for improved discharges, lower structural costs and a reduced base, favouring the adoption of labyrinths on most spillway structures. The efficiency also remains high for very large nappe depths.

More data are given in Hydropower & Dams Issue 2, 2012 and in Appendix 2 of the ICOLD Bulletin 144 (Cost Savings in Dams). Many shapes are under study in various laboratories, to optimize the hydraulic efficiency more.

Labyrinth spillways have a rectangular layout (and hence the name of Piano Keys Weirs or P.K. Weirs) and inclined overhangs. The best ratio between the developed wall length and the spillway length seems to be 4 to 6. Various designs have an upstream or a downstream overhang. A reference, called model A, has both, and is represented below: its specific discharge as compared with a traditional Creager weir discharge for a same usual nappe depth H (in m and m3/s/m) is close to Qs × 4.3 H√Pm instead of Qs × 2.15 H1.5 (Pm is the maximum wall height).

For an average use of a wall height Pm double the nappe depth H, the discharge capacity of a P.K. Weir is equal to 2√2 (about three times) the discharge of a Creager weir. The nappe depth saving is close to 0.5 Pm. The discharge of P.K. Weir per m2 of nappe is about 6 m3/s for small structures, and 100 m3/s may be discharged by a 1.5 nappe depth along a 10 m spillway. The discharge is more than 10 m3/s per m2 for high P.K. Weirs and 5000 m3/s can be discharged by a 100 m-long spillway and a 4 or 5 m nappe depth.

For very large discharges, P.K. Weirs may also be combined with a gated spillway. This may apply also to run-of-river schemes, because the efficiency of P.K. Weirs remains high when the downstream water level is over the P.K. Weir structure.

For small spillways, it is possible to avoid overhangs, and to keep vertical walls with a rectangular layout, see Figs. 4 and 5. This design model E is much more cost efficient than traditional trapezoidal layouts; it requires a longer base than model A (one-third more) but still half of the base of a traditional trapezoidal labyrinth.

3.3. Fuse devices

Various solutions for fuse elements, which open for exceptional floods (and can be rebuilt within a few weeks or months) have been used, based on earthfill, steel or concrete elements.

For some 30 years, earthfill fuseplugs have been built, mainly in China, the USA and Australia, for discharges of around 1000 m3/s up to several thousand m3/s. As they are designed to open during exceptional floods, there is little experience of the operation of these fuseplugs. They require favourable dam sites, with quite a flat area, and they can only be used in combination with another spillway, usually gated. They may thus be of interest essentially for large discharges and specific dam sites.

The best application of the principle of ‘earth fuseplugs’ could be for long earthfill dams, to raise the crest level (possibly by a parapet) along almost the entire length of the dam, except in places of moderate height (10 to 15 m) where only a limited breach could be anticipated in extreme floods.

Many quite small spillways in the USA have used flashboards for some time; that means, vertical walls made of wood, standing against steel pipes anchored into the concrete sill of free-flow spillways. These can be overtopped by ordinary floods, and will bend for exceptional floods. They may be damaged by floating debris, and are hardly used except at many small dams in the USA.

A more promising solution may well be simple concrete fuseplugs, as shown in Fig. 6. They are presented in Appendix 3 of ICOLD Bulletin 144.

Model tests have helped to establish the ratio between the width of a block and the water level at the time of tilting. For preliminary purposes, the simplified formula: E= h + 0.4 P

can be used, where:
 h = upstream water depth over the fuseplug for tilting; E = upstream to downstream base width of the fuse plug; and,
 P = height of the fuseplug.

This relationship can be used for blocks with the same general shape as described above, with a hollow section under the block, of 10 per cent of the block height P, and with a concrete density in the range of 2.3 t/m3. Concrete density variations of 5 per cent will lead to variations of about 10 per cent in the nappe depth at the time of tilting.

They may be used for small or large discharges, and they can be overtopped by ordinary floods and tilt, for instance, in sequence, for floods such as the 100 year events. They can, at a very low cost, double the capacity of a free-flow spillway, that means, decrease by a factor of 100 the failure probability. It could thus be possible to design most future dams in catchment areas smaller than 100 km2 for very exceptional floods without a significant cost increase.

Fig. 6 Concrete Fuseplugs; cross section

3.4. Combining labyrinth shapes and fuse devices

Over the past 20 years, more than 50 large dams have been equipped with fusegates, which are usually fuse elements with a labyrinth shape. Their design is more complex than that of simple concrete fuseplugs or P.K. Weirs, but their performance is higher and is cost effective, for instance for improving existing free-flow spillways, or in addition to gated spillways for very large discharges (10 m-high fusegates combined with a freeboard of 5 m can discharge more than 100 m3/s/m after tilting, which is about the same as gated spillways).

It may also be possible to increase the storage of dams using P.K. Weirs, by adding a fuse wall (about 1 m high) to the walls of P.K. Weirs, which would be overtopped by ordinary floods and fail for the 100 year flood.

3.5. Optimized combination of gated and
 ungated spillways

Most existing dams or dams under construction spill floods through either:

• a free-flow Creager spillway, discharging 20 m3/s/m from a 4m nappe depth, or 6m3/s/m from a 2m nappe depth; or,

• a gated spillway with surface sector gates requiring permanent operators, spilling up to about 100 m3/s/m of spillway.

Two new solutions may be much more cost-effective for future dams:

• A free-flow spillway with a modern labyrinth shape or fuse devices: they can increase by a factor of three, at low cost, the Creager weir discharge (for example, 50 m3/s/m for a 4 m nappe depth). Such a solution could be used for extreme floods of up to 5000 m3/s; the spillway could be complemented by one or two bottom gates, able to discharge the annual flood, either for sluicing sediments or for improving flood mitigation (which does not require permanent operators).

• Combining, for very large discharges, surface gates for about half of the extreme flood, and a modern freeflow spillway with the sill level close to the upper level of the gates. Such a free-flow spillway (P.K. Weir or fusegates) will be able to discharge up to 50 or 100 m3/s/m if a freeboard of 4 or 5 m is used for the check flood. The number of gates can be reduced, and permanent operators are probably not necessary for most of the year.

But these two attractive solutions may be prevented by traditional design criteria or existing regulations, which favour fully gated spillways.

Some recent designs are described below.

At the High Grand Falls RCC and rockfill dam project in Kenya, where the design flood is 12 000 m3/s, the design takes into account large uncertainties in the flood derivation (short data sets, strong impact of climate change in this area), possible jamming of gates, and combines:

• a gated free-flow spillway, the best option to maximize energy production at an optimum cost for the dam.

• a fuse saddle dam, with fuseplugs over an earthfill dam which would be washed away once the fuseplugs have tilted.

At the old masonry Pont dam, in France, where the design flood is 400 m3/s, one option for the design would be to replace the present gated spillway with an upgraded system, incorporating flap gates and fusegates. The design criteria were in accordance with the criteria listed above (including possible non-opening of gates), plus a criterion concerning the maximum discharge downstream in case of accidental opening of one gate.

At the Tessa RCC dam project in Tunisia, the design incorporates various issues of flood protection downstream, and safety in the event of very severe flash floods (11 000 m3/s). The most appropriate design was considered to be one combining: two bottom gates with a large discharge capacity (which can handle up to the 1:10 flood), and a spillway with one gated span and several spans equipped with fusegates.

At the Rassisse dam in France, the spillway upgrading project will combine a P.K. Weir spillway and possibly also fuseplugs. The design in this case takes into account the very low available margin between the NWL and MWL, and also the owner’s requirement not to adopt gated spillways. The design discharge is estimated at 600 m3/s. The P.K. Weir option proves to be a suitable solution, because of the large increase in the discharge capacity between the MWL and crest.

The Inga scheme in Congo is designed with gates to discharge 60 000 m3/s and a 1 km P.K. Weir which could discharge 70 000 m3/s in the event of ‘check flood’ or if the gates jammed.

Shongweni Fusegates in South Africa

3.6. Structural improvements

These are very different for gravity dams and for embankments dams.

3.6.1. Concrete dams

Failures of gravity dams are much more dangerous than those of earthfill dams with much higher incremental damages and human risks. The failure of a 15 m-high dam may cause an instantaneous additional flow of up to 5000 m3/s and a reduced efficiency of alarm systems. The ‘check flood’ should then be of very low probability.

The best quality of a gravity dam is the gravity, and an optimization of the cross section may be more useful than an increase in the concrete quality. And a gravity dam may be designed for withstanding an exceptional overtopping along the full crest. Various improvements may thus be foreseen accordingly:

• Reduction in the freeboard to reduce the water level during overtopping.

• Design of fuse elements of the upper part of the crest along several tens of metres.

• Low-cost protection of the downstream dam toe for the case of overtopping.

• Stability for such overtopping, through arching of the layout, which saved some masonry gravity dams, or by improving cross section. For instance, a symmetrical cross section with 0.7/1 slopes may actually be able to withstand any overtopping; it may use low cost concrete (hardfill) and could be founded on medium quality rock: this may be the safest dam solution.

3.6.2. Earthfill dams

Improvements may be made to reduce the failure probability or the damage which would result from failure.

• Efficiency improvements could be made to the crest, increasing the freeboard by a steeper slope, and/or a parapet and protecting the crest against wave action and piping.

• A search for an erosion-resistant design might be attractive for small dams: erosion in the case of overtopping may be delayed or limited by grassing the downstream slope, and by adopting a 1 or 2 m-wide concrete protection of the downstream toe.

• Choosing the point of a possible failure by adopting a lower crest (for example, by 0.50 m) or a fuse device, where the dam height is much lower than the maximum dam height. This could reduce the incremental damages significantly, for instance by selecting a saddle dam for ‘first failure’.

3.6.3. Dams in narrow valleys

It may be difficult to avoid tunnel spillways which are costly. It may be possible to use costly gated and lined tunnels for the flood of 1/100 or 1/1000 probability, and to use, in addition, unlined virtually horizontal tunnels with upstream fuse elements for exceptional discharges. This may reduce costs and operate also in the event of gates jamming.

4. Proposed solutions and relevant design criteria and methods

The uncertainty about flood evaluation and about the impacts on floods of climate change favours an increase in the spillway capacity.

Gated spillways have two drawbacks: they require skilled and continuous maintenance, and the possibility of total gate jamming cannot be overlooked: if the complete jamming probability is 1/1000 or 1/10 000, a flood with a probability of 1/100 or 1/10 should be able to be discharged with all gates closed, and no dam failure.

Beyond structural improvements, the principle of proposed solutions is that at least half of the extreme discharge should be through a low-cost free-flow spillway: it is thus possible to achieve, at an acceptable cost, discharges of floods for which the incremental downstream damages of a failure should be low, for instance floods with a probability of 10-5 or 10-6. This is possible with the solutions of free-flow spillways discussed in 4.2 and 4.3 above. This is difficult with the specific discharges of traditional free-flow spillways.

• For a small earthfill dam in a small catchment area of only a few km2, an extreme flood of 100 m3/s, bringing the reservoir up to the crest level would require, with a traditional Creager weir, a spillway length of 15 m with a gap of 2 m between the dam crest and the weir, or a length of 10 m and a gap of 3 m. With a P.K. Weir sill, it would require only 1.50 m of the length of 10 m. With concrete fuse plugs, it would require 2 m of 10 m.

• For an extreme flood of 1000 m3/s, a traditional ungated weir would require 60 m and a 4 m gap. P.K. Weirs would require 30 m and a 3 m gap. For many dams, such a weir could be reduced by 15 per cent and be completed by bottom gates discharging up to 150 m3/s. These bottom gates may be used for sluicing sediments at the beginning of flood seasons or for mitigating floods. They do not require permanent operators.

• An extreme flood of 10 000 m3/s is traditionally discharged by a gated spillway about 100 m long, with a freeboard of 4 or 5 m above the gates. In the case of total gate jamming, the discharge above the gates is limited to a nappe depth which is about 25 per cent of the open sluices, which means to less than 15 per cent of the ‘check flood’ discharge, or less than the annual flood discharge. This risk is by far the main risk of many gated dams, especially for embankment dams with a reduced storage between the dam crest and the gates. A less expensive and safer solution combines a 50 m-long gated spillway with an 80 m-long PK Weir spillway or a 50 m-long fusegated spillway.

These solutions should be favoured or at least allowed for in the design criteria.

4.2. Proposed design criteria

The basic target in all countries is that the real annual probability of a dangerous failure (i.e. involving more than ten fatalities) should be very low, in the range of 10-6, and that should be obtained at a reasonable cost. This choice of 10-6 may be safer than the economical optimum, but the extra cost may be low with proposed ungated solutions, and this margin of safety may cover uncertainties.

The technical solutions proposed here may meet this target, but the traditional design criteria do not favour them and may even prevent the best ones.

This 10-6 criterion might be used to cover several types of situations:

• extreme flood;

• very large flood and spillway remaining partially closed during the flood; and,

• moderate to large flood, and spillway remaining completely closed during the flood.

Every dam should be assessed individually.

Standard conditions for a well designed and managed dam in a medium sized catchment may, for instance, use the range of probabilities of table 5 (taking in account the uncertainty in floods evaluation, analyse in section 3 above):

Combining these probabilities would lead to the following design criteria, all of them of about 10-6 probability:

• The ‘check flood’ (adjusted PMF) will pass under the crest level with the spillway completely available, and under ‘probable instability’ level with one gate remaining closed.

• The adjusted ‘1:100’ to ‘1:1000’ flood will pass under the MWL with the spillway available.

• The annual flood will pass under the MWL with one gate remaining closed, and below the crest level, with all the gates remaining closed.

• The 1:10 year flood will pass under the probable instability level with all the gates remaining closed.

These criteria should of course be adapted for other situations: the criteria should somehow be calibrated to downstream risk, to dam and spillway conditions, to the gates’ operational reliability, and to the range of uncertainty associated with flood estimates.

These analyses may also draw attention to some overlooked risks, and help to remedy them.

This approach, based on the ‘10-6 criteria’ (check flood and other situations), often remains associated at present with a study based on the traditional ‘design flood’, with design flood being associated with the MWL.

A key point should be underlined. If the discharge of the chosen ‘design flood’ is more than about half of the discharge of the chosen ‘check flood’ the overall design may actually be based on the ‘design flood’, with all its drawbacks, and especially the wrong choice of a fully gated spillway, instead of the association of a gated spillway and a free-flow spillway. The 10-6 criterion’ approach tends to assume that the traditional design flood criterion is not very well adapted, being too conservative and not covering all situations.

This choice of very low probability of failure risk is favoured by new low cost solutions. It is, however, reasonable to accept a higher probability in the following cases:

• for dams where the incremental damages and human risks of a failure are a very small part of the risks from the natural flood: this is the case for many low dams even on very large rivers;

• for small or medium dams without significant human risks. However the low-cost solutions for freeflow spillways will often economically justify keeping their annual failure probability to less than 10-4.

It is also possible to reduce the magnitude and impact of failures by technical choices and alarm systems.

The solution proposed above should adapt the choices and designs to the actual risks, should correctly favour low-cost free-flow spillways and/or the association of gated and ungated spillways, and should consider the risk of all the gates jamming. It should also favour a high-water resistant design of dam crest, which is often overlooked.

5. Upgrading of existing dams

Most solutions for improving new dams can also be used for upgrading many existing dams: however, they deserve some specific comments:

• The costs of some solutions may be much higher for existing dams than for new ones, and the preferred solutions for upgrading existing dams may thus be different from those for new dams.

• The key point is to reduce, at an acceptable cost, the probability of failure. Reducing the probability of the ‘design flood’ is usually unnecessary, and may prevent the adoption of low-cost solutions improving the true safety.

• If the existing spillway is gated, an additional spillway should preferably be ungated, to mitigate the risk linked with total gates jamming.

• The works for upgrading may cause a temporary additional risk of failure. This increased failure risk during upgrading is sometimes too high; it may then be preferable to reduce the impact of a failure.

6. Conclusion

Most present methods and criteria for designing spillways appear not to be adequately adapted to the present knowledge of failure data and dam costs, and to the possibilities offered by new technical solutions.

They are essentially based on a ‘design flood’ and a significant freeboard, adapted according to probable waves. While much preference is given to gated spillways, it is not possible to avoid the risk of total jamming of the gates. The attractive solution of combining a gated spillway with an ungated one may be overlooked.

Design methods based essentially on a ‘check flood’ of very low probability associated with a higher reservoir level appear to be more justified, and may also reduce both risks and costs. This principle was already advocated in ICOLD Bulletin 82 (1992) and 144 (2011). ◊

F. Lempérière has been involved in the construction and/or design of 15 hydraulic structures on large rivers including: Cabora Bassa in Zambezi, redesign and excavation for the Jonglei canal in Sudan, the hydropower plant at the Old Aswan Dam and studies for Niger Energy. He is Chairman of the ICOLD Committee on Cost Savings in Dams and of HydroCoop, a not-profit association providing advice on dams and floods.

J.-P. Vigny has been involved in the design or construction of large civil engineering schemes on rivers or at sea, including Cahora Bassa dam on the Zambezi, and the Mudhiq dam in Saudi Arabia. He is General Manager of Hydrocoop.

Hydro Coop, 4 Cité Duplan, Paris 75116 France.

L. Deroo has been involved in the detailed design or appraisal of many dams in Europe (France, Belgium, Portugal, Greece) and Africa (Algeria, Tunisia, Burkina Faso, Ethiopia and Kenya), including masonry, concrete, earth and rockfill dams. He is currently the CEO of ISL.

ISL, 75 Boulevard MacDonald, 75019 Paris, France.