The implications of movements caused by active or potentially active faults passing through the foundation of a large dam, and movements of blocks in the dam foundation caused by strong earthquakes at a fault close to a dam site, are discussed by M. Wieland, A. Bozovic and R.P. Brenner with respect to the design of dams and their safety

Faults_Fig03a

Earthquakes are multiple hazards, which can affect a storage dam through the following scenarios: ground shaking causing vibrations in dams, appurtenant structures and equipment, and their foundations; fault movements in the dam foundation causing structural distortions; fault displacement in the reservoir bottom causing seiches in the reservoir or loss of freeboard and consequent dam overtopping; and mass movements into the reservoir causing impulse waves in the reservoir. Other effects such as wind-generated water waves and reservoir oscillations are of lesser importance for the earthquake safety of a dam.

The main tectonic phenomenon, which has been and still is, a major concern for dam designers, is the seismic shaking and related consequences for the safety of dams. Accordingly, the main hazard, which is addressed in codes and regulations, is the earthquake ground shaking. It causes stresses, deformations, cracking, sliding, overturning, liquefaction etc. Very significant advances were made in this field, i.e. (i) in monitoring earthquakes and quantifying the input parameters, and (ii) in analyzing the dynamic response of both earth and concrete structures.

A hazard which is often underestimated is the large number of rockfalls in mountainous regions. For example, during the 1990 Manjil earthquake in Iran and the 1999 Chi-Chi earthquake in Taiwan, more than 10,000 rockfalls and slides were observed. These mass movements can block access to dam sites and the reservoir, or may sometimes form landslide dams (e.g. the lake created by a 20Mm3 slide during the 1959 Hebgen Lake earthquake, Montana, US; Lake Sarez, created by a 2200Mm3 slide permanently blocking the Murghab River as a result of a magnitude 7.4 earthquake in Tajikistan in 1911).

If a major earthquake occurs, which can cause damage to a well constructed dam that has been designed to withstand the ground motions caused by the maximum credible earthquake (MCE), then it has to be expected that the buildings and infrastructure in the dam and reservoir region are severely damaged and that access to the (remote) dam site and the reservoir may be obstructed due to landslides, rockfalls, debris on roads, cracks in the road surface, soil deformations, damaged bridges, local flooding etc. Access to remote dam sites may only be possible by helicopter. During the 1990 Manjil earthquake in Iran, some 40,000 people were killed – many of them close to the Sefid Rud dam. Most of the buildings at the dam site were destroyed or heavily damaged. Access to the dam site was blocked by numerous rockfalls and the damaged dam was virtually deserted immediately after the earthquake. A similar scenario has to be expected at other dam sites after a major earthquake.

Overview of Tectonic phenomena at dam sites

The most severe condition for a dam is when it is subjected to both ground shaking and movement of faults and other discontinuities in the footprint of the dam during strong earthquakes.

However, as mentioned in the previous section, the main issue for dam engineers has been the ground shaking while the possibility of surface fault breaking or block movements in the dam foundation was usually disregarded. Surface fault breaking, or more precisely surface slip along an identified fault zone under the dam, was always understood to be the most dangerous tectonic manifestation that can influence a dam. The growing number of dams, the tendency to make use of less favourable dam sites, and the growing awareness that the hazard of foundation movements is about the most severe process that could affect the structural integrity of dams, gave rise to the conclusion that such possibilities are not to be treated as negligible, but should be considered and analysed each time when indicated by prevailing tectonic conditions.

The tectonic movements, active at present, result generally in the formation of fault breaks and in creep movements. In connection with fault breaks block movements also have to be considered as a possible feature, especially in the near-field of major faults, which are capable of producing large earthquakes. Such blocks can be formed by joints, bedding planes, shear zones, and higher order faults. Both fault breaking and creep movements constitute the crustal mobility, which can directly affect dam sites and structures. In the subsequent part the effect of creep movements on the behaviour and the design of dams is not discussed as this effect is of quasi-static nature and less severe than the sudden slip along a fault or the movement of blocks during a strong earthquake. From the viewpoint of dam engineering, the objective is to collect and evaluate the related information on crustal mobility and to consider it in the design of dams and in evaluations of their safety in a reasonably conservative manner.

Faults with surface breaking capability crossing the dam site and potential block movements are the main point of interest, considering the safety of dams. Consequently, recognizing such features is the decisive step in the problem. If the existence of such features has been determined in a dam foundation, the best policy is to look for a tectonically more stable alternative site. If this is not possible, then a conservatively designed embankment dam might be an acceptable solution if the expected differential movements can be absorbed by its sealing and drainage zones, without provoking a dam failure.

Therefore, the definition of an active (or potentially active) fault is very important and quite sensitive. Such definition is given, for example, in icold Bulletin 72 (1989):

‘A fault, reasonably identified and located, known to have produced historical fault movements or showing geologic evidence of displacement and which, because of its present tectonic setting, can undergo movement during the anticipated life of the dam.’

However, Allen & Cluff (2000) stated that there is no sharp division into ‘active’ and ‘inactive’ faults. Modern geologic studies in combination with improved age-dating capabilities have shown that there are in fact all degrees of fault activity and any categorization is arbitrary.

It is quite obvious that a time mark is needed for identifying fault movements. This is usually a soil or rock surface, for example in a trench, offset by a fault. Dating of fault gauge has not been reliable so far. The consequences of dam failure may also play an important role in selecting potentially active faults. It is also important to note that in view of new information and assessment techniques, faults, which in the past were considered as inactive, are now considered as active or potentially active. In the past, faults with no clear evidence of (instrumentally recorded) activity were considered as inactive. Today, however, faults where there exist no proof that they are inactive, are considered as active or potentially active. The first approach might have been too optimistic and the latter might be too pessimistic especially in regions of high seismicity, where most faults have to be assumed as potentially active. Minor faults are not of concern as they are not capable of causing important earthquakes or surface rupture with significant movements.

Tracing and dating the active tectonic phenomena into the geologically recent past is, in a number of cases, a difficult and elusive problem. On the other hand, and in each particular case, the question whether the considered fault is active or not needs a clear answer: yes or no. The answer to this question has serious consequences for the design and type of the dam being considered.

Possible block movements caused by surface breaking of a main fault are shown in Figure 1. Second order fractures (or branch faults) may propagate along existing discontinuities in the dam foundation if the main fault is relatively close to the dam site. The likelihood of displacement along such a fault can be judged from the condition of the Quaternary strata overlying the branch fault, i.e. whether or not these have been broken sympathetically with the main fault over a sufficiently long period of time. In addition, the geometrical relationship between branch and main fault and the prevailing stress fields controlling these faults should be evaluated (Allen & Cluff, 2000).

Figure 1

Figure 1: Sub-faults and block movements close to surface due to reverse (or thrust) (left) and normal faults (right)

Fault activity

The fault studies related to a dam site have various levels depending on the area considered and on the stage of studies. They range from regional investigations (150-200km radius) outlining the general tectonic set-up, to local investigations (approx. 50km radius) to identify faults likely to be significant for the dam in question, and to faults intersecting the dam site. From the viewpoint of geodynamic hazard to dams, the late Quaternary evidence of faulting is most important. Tectonic features are, as a general rule, inherited from the distant past and their importance for the surface breaking capability at the present time decreases as the chronological position of the considered features recedes into the geological past. According to Allen & Cluff (2000) the most critical parameter in the fault activity assessment is the long-term fault slip rate, i.e. the average rate of slip over a sufficiently long time period (expressed in mm/year)

The existence of historical or instrumental data on surface fault breaking is very important. But the time span for which such data are available is generally far too short. Large surface breaking events might be separated by thousands of years. It means that in general the dating of surface breaking evidence has to be attempted prior to the historical period. Hence, the related investigations have to cover the full field of geological, seismological, geophysical and geodetic information, including remote sensing, trenching, adits, exploratory borings and geochronological dating, as required by the prevailing circumstances.

Of specific interest are data gathered by monitoring the dam region at the microseismic level (important for reservoir-triggered seismicity), which should be undertaken for each important dam project. If the low magnitude events cluster along linear features corresponding to surface fault traces, it is an important indication of geodynamic activity along such features. But there are also some major earthquakes that have occurred on faults with very little, if any, precursory micro-seismicity within the preceding few years.

Creep movements can best be verified by repeated geodetic measurements.

Predictability of fault movements

When a sufficiently large fault considered capable of producing a surface break exists at a dam site (length and estimated magnitude being its parameters), dating of past movements has to be attempted, using all available techniques. Structural and stratigraphic superposition and geomorphic studies are to be undertaken. If Holocene strata are disturbed by fault movements, this is considered a decisive evidence for fault activity. When the disturbed Quaternary strata are of unidentified age, then geochronological isotopic dating should be attempted. If cosmogene radioactive carbon or other radio isotopes can be found, dating of the fault disturbance can be undertaken.

Seismicity triggered by impounding is a good indicator of fault activity, the disadvantage being that it comes usually as “a posteriori”, i.e. when the dam is already completed.

Evaluation of effects of fault movements on dams

Fault slips or block movements under the dam during strong earthquakes are considered to be the most dangerous manifestations for the structural integrity of dams. Creep movements, rather slow and uniform, can be monitored and may be amenable to mitigating measures.

If a fault crossing the dam site is evaluated as potentially active, the next question is what size of fault movements is to be expected, as the size expresses its damage potential. Correlation of the length of faults with earthquake magnitude and displacement along the fault, supplies such information. The coefficients for the linear least square regressions are derived from best fit procedures.

Several authors have performed such analyses and rather significant correlations were derived. A comprehensive elaboration is given by Wells and Coppersmith (1994), based on analyzing world-wide data including hundreds of earthquakes, developing regressions of magnitude on fault rupture parameters. Quite credible correlations are obtained so that the order of magnitude of fault break displacement can be estimated.

Effects of active tectonic features on selection of sites and types of dams

The tectonic activity assessment plays a key role in selecting sites and types of dams. Adequate studies and investigations have to be performed to this end. When comparing dam sites, a reduced level of tectonic uncertainties should be rated as a significant advantage. Hence, evaluating the safety of any large dam with respect to fault slip and block movement hazards should be introduced as a regular practice.

Still, there are cases where no geodynamically favourable alternative is available so that a decision might be taken to build a dam facing a fault slip hazard. In such a case, concrete dams should be avoided and preference be given to a conservatively designed embankment dam, built with ample filter and transition zones, on both sides of a rather wide core, displaying ductile properties. There is considerable confidence that such a structure can withstand, without failure, significant fault offsets. Such a dam should be even more effective against fault creep and other foundation displacements.

Potentially active faults in dam foundations – case studies

A comprehensive discussion of potentially active faults in dam foundations was given in the landmark paper by Sherard et al. (1974). It contains information on existing dams founded on active faults, a summary of lessons learnt from historic dam breaks and fault mechanisms, and opinions by the authors regarding the design of dams on active faults. The authors state that (i) concrete dams on active faults, or near some major active faults, are not advisable, and (ii) if a site with fault movements cannot be avoided then it is reasonable practice to construct a conservatively designed embankment dam.

The above statements still hold true today as well as their introductory statements of the paper:

‘In earthquake regions faults are frequently found to exist in dam foundations. Evaluation of the likelihood that displacement could occur along the fault during the lifetime of the dam, and the selection of the design details to ensure safety against such possible fault displacement, are difficult problems for which there is (still) little guidance in the literature.’

Several case studies are given in ICOLD Bulletin 112 (1998), whose main author is A. Bozovic. This bulletin represents the international state-of-the-practice for dams on potentially active faults. Selected case studies described in the recent literature are summarised below:

• The 27m high San Andreas Dam, which in 1906 was hit by a surface fault break along the San Andreas fault during the M = 8.6 San Francisco earthquake. The fault break line passed through a small ridge between the main dam and the adjacent saddle dam, producing an off-set of 2.5m in the crest line. There was no damage to the dam.

• The Upper Crystal Springs Dam first served as a dam and was then submerged by a downstream structure. The 1906 San Andreas fault break has sheared this dam and offset it by 2.5m. (Note: The 43m high Lower Crystal Springs curved gravity dam, located about 200m from the fault, successfully resisted the extreme ground shaking of the San Francisco earthquake.)

• The 27m high Hebgen Dam is an embankment situated near the surface fault break which was activated during the Yellowstone earthquake (M=7.1) in 1951. The nearest fault break was 200m from the dam and had a vertical movement of approx. 5m.

• The 143m high Tarbela Dam, completed in 1974 on the river Indus in Pakistan, is a good example of the significance of active faults studies. The dam was designed without consideration to a fault (Darband fault) crossing the dam foundation which was revealed during the construction stage. A detailed seismotectonic study at a later stage proved the fault to be active. The seismic design parameters had to be revised by taking into consideration the active fault passing below the dam body. The effect of surface rupture on the Darband fault beneath the embankment was also reviewed. It was estimated that movement of 1m to 1.5m could occur on this fault. It was believed that if such a movement did occur no catastrophic failure of the embankment would result, since the core of the embankment is constructed of self-healing material with a transition zone and a wide chimney drain on the downstream.

• The Matahina Dam in New Zealand is a rockfill embankment 82m high and 400m long, with a central core. The dam leaked after first filling in 1967 due to core cracking, and was consequently repaired. In 1987 the dam was exposed to strong seismic shaking (peak horizontal crest acceleration 0.42 g) due to a magnitude ML6.3 earthquake, located on the Edgecumbe Fault. Significant rehabilitation work was performed in 1988 due to internal erosion of the core at the left abutment. The dam is sited across a part of the Waiohau Fault Zone, 80km long. The fault is active with proven surface fault breaking during the Holocene. Several prominent strands of the Waiohau fault intersect the dam site and are considered capable of rupturing during the lifetime of the dam. Investigations in trenches showed that movement on the fault had occurred four times during the past 11,300 years. It was assumed that movement could occur on any of the fault traces present within the dam site. The design criteria, estimated at the 84th percentile level, assumed the magnitude of the Safety Evaluation Earthquake (SEE) as Mw7.2 with corresponding horizontal and vertical peak ground accelerations of 1.25 g and 1.35 g respectively and 3m of oblique slip on the fault (i.e. 2.7m horizontal and 1.3m vertical). Such displacements would surely result in major cracking of the dam body inducing piping and internal erosion (as evidenced by observed core erosion during the 1987 Edgecumbe earthquake). Therefore, it was decided to strengthen the dam with a leakage resistant buttress (Gillon et al. 1997; McMorran & Berryman, 2001).

• Aviemore Dam in New Zealand is a composite structure consisting of a 364m long and 56m high concrete gravity dam and a 430m long and 49m high zoned earth embankment with upstream sloping core. There is a 1m thick filter separating the core from the downstream pervious shoulder, while serving as a drainage blanket on the bottom of the core trench under the shoulder. The Waitangi Fault was recognized in the foundation excavation. From trenches it could be observed that this fault had a late Quaternary displacement of 1-2m and that the most recent displacement took place about 14,000 years ago. The fault crosses the footprint of the embankment dam and strikes normal to the dam axis. Waitangi Fault was assessed as being capable of a magnitude 7.0 event and with a peak ground acceleration of 1.0 g. For the fault a vertical separation of 1.2m and a horizontal to vertical ratio of 1H:3V were adopted for the SEE fault surface displacement. To analyze the performance of the embankment under the SEE fault break, a three-dimensional numerical model was created. It showed that the zone of embankment deformation would be several tens of meters long. Cracking of the core is expected to occur over a 40m long zone, mainly in transverse direction. Cracking would take place from the crest downwards to a depth between 5-10m. The filter is expected to remain continuous across the zone of faulting. The fault in the foundation is of low permeability and it would be expected that no concentrated leaks would develop as a result of fault displacement. However, in the post-earthquake phase, the fault rupture may provide a mechanism for continued erosion (Mejia et al., 2005).

Case Study: Shih-Kang Dam, Taiwan

Shih-Kang weir, located at the Da-Jia river in Taiwan, comprises two sluiceways and 18 spillway gates. The concrete gravity dam has a height of 25m and a crest length of 357m. On September 21, 1999 the weir was severely damaged during the magnitude 7.3 Chi-Chi earthquake and the reservoir with a volume of 2.7Mm3 was released through spillway gates 17 and 18. The spillway was designed for a total discharge of 8,000m3/sec. The uncontrolled release of the reservoir through two openings did not cause any flooding in the downstream area of the dam as the total discharge was much less than the design discharge of the whole spillway (Wieland et al. 2003).

The weir was damaged (i) by the rupture of segments of the Chelungpu fault, (ii) by surface movements, and (iii) by strong ground shaking. The most spectacular damage occurred at spillway bays 16 to 18 near the right abutment and was due to fault movements (reverse faulting) of several metres mainly in vertical direction (Figures 2 and 3). However, there was widespread cracking on the whole structure. In addition, the irregular ground movements caused separation of most of the blocks from the foundation rock, which consists of layers of mudstone, siltstone and sandstone. Thus, five of the spillway gates and one sluice gate were inoperable after the earthquake. Moreover, all the simply supported spans of the bridge across the weir fell off the bearings.

Figure 2

Figure 2: Shih-Kang spillway weir damaged by 1999 Chi-Chi earthquake: Spillway openings 17 and 18 destroyed by fault movement

Figure 3

Figure 3: Effect of fault movement on Shih-Kang dam: Detail of destroyed spillway openings 17 and 18

The dam, which was completed in 1977, was designed against earthquakes using a seismic coefficient of 0.15. There exists no direct information about the peak ground acceleration (PGA) at the dam site. However, the strong motion station closest to the dam recorded PGA-values of 0.51 g and 0.53 g in horizontal and vertical directions, respectively. During excavation of the foundation no clear evidence of a fault in the dam foundation was found.

The earthquake induced ground movements at Shih-Kang weir are rather complex as the Chelungpu fault splits up into different sub-faults in this area, one of them crossing the weir and destroying the spillway gates at the right abutment. Another fault branch crossed the intake tunnel at the left abutment and sheared it off.

Except for the spillway bays 16 to 18, the weir was still in a repairable condition. In order to protect the dam from further damage due to floods, the following emergency rehabilitation works were carried out: repair of cracked intake structure and sheared off intake tunnel (steel lining); reconstruction of overturned wing wall at left abutment; repair of sluiceways, spillways and cracked stilling basin; repair of tainter gates; construction of an upstream cofferdam for spillway bays 16 to 18; construction of a downstream cofferdam (tail dam); rehabilitation of bridge across weir.

The cracks were repaired by epoxy cement and the fissures in the foundation were grouted with cement.

Due to the tectonic movements in the reservoir, the storage capacity of the weir has been reduced from 2.7 to 0.4Mm3. In view of the active fault crossing the right part of the dam and in view of the greatly reduced reservoir capacity, which is needed for water supply, the long-term solution is the construction of a new dam further upstream.

The Shih-Kang dam is an important case study as it is the first large concrete dam exposed to substantial fault movements. The lessons learnt are expected to have an impact on similar projects.

Selection of dam types

Discussion of movements in dam foundations

In the subsequent part the implications of (mainly horizontal) foundation movements on different dam types and allowable displacements are given:

1. Earth core rockfill dams (ECRD): After a displacement, caused by the fault slip, the remaining overlapping filter zone should at least be 2m. This means that with a design displacement of 1 to 2m, a filter of at least 3m thickness would be required. A safety factor of 1.5 may be added to cover all the uncertainties and one would end up with a filter thickness of about 4.0 to 4.5m. The filter will have no cracks, because its material has to be perfectly cohesionless. This means that any opening that will be created during the displacement process will collapse when saturated. The clay material of the core should be more ductile over the fault zone, so that when it is sheared during the displacement it will not form open cracks which could provoke internal erosion. A horizontal (or strike-slip) displacement of 1 to 2m in an embankment dam foundation can therefore be tolerated without problems. Bray et al. (1994) reported from experimental evidence that ductile materials with large failure strains can accommodate significant fault movements without actually breaking. The rupture would then not propagate to the top of the core and the zone of shearing would remain narrow.

2. Concrete face rockfill dam (CFRD): Displacement in the rock foundation will shear the plinth and rupture the perimetric joint sealing. Leakage will start through cracks in the plinth, the foundation (if the grout curtain is not wide enough) and through joints or cracks in the face slabs. This does not imply that the dam will collapse because the dam body, if properly zoned, can sustain seepage without washing out of the finer particles. The rockfill zones on the downstream side provide sufficient stability against the additional forces of seepage flow. These zones are free-draining and excess pore water pressures cannot develop. Still, the stability of the upstream slope under the changed hydraulic conditions (partial submergence and saturation of the transition zones in the dam body under the cracked face slab) has to be checked and verified.

The plinth will require repairing which means that the reservoir has to be emptied (a bottom outlet is required) and the fill over the plinth has to be removed. If the horizontal displacement is less than about one half of the width of the plinth there is still a sufficiently wide continuous plinth slab, although cracked, which can be repaired. The tolerable horizontal displacement in the foundation below the plinth of a CFRD should not be larger than about one fourth of the width of the plinth slab. Usually, the minimum width of the plinth (or toe) slab is 3m, so that a displacement of 0.7 to 0.8m could be accommodated. In vertical direction the displacement should not exceed one fourth of the thickness of the plinth slab.

The grout curtain will also have to be repaired. Seepage through the foundation can be minimized if at the location of the fault zone the grout curtain is made wider, i.e. consisting of several rows. The continuity of the grout curtain is then ensured even after the displacement.

The situation is different with certain older dams which may not have sufficient drainage capacity below the impervious face, for example, older asphalt-faced embankments. If the impervious face suffers cracking and the water penetrates through the cracks into the dam fill with inadequate means for fast drainage, the upstream slope will become hydraulically unstable. Such cases should be investigated by an analysis modeling transient water ingress through the cracks in the facing. For such embankments the tolerable displacements in the foundation may be on the order of 5 to 20cm only.

3. Concrete gravity dam (RCC or conventional concrete): For rigid materials such as concrete, small displacements in the foundation will cause cracking. The extent of the cracking is very difficult to predict and also depends on the topography of the foundation. Cracks can, in general, be repaired if their extent is within limits. However, in the case of an earthquake the dynamic behaviour of the cracked dam has to be predicted, which is beyond the currently available commercial analysis tools (no generally accepted or verified software is known to the authors). Again cracking in a gravity dam does not mean that the dam will collapse, since each block acts independently. The limit for fault displacements in the foundation may be set at not more than 0.10m.

4. Concrete arch and arch-gravity dams: The same applies as for the gravity dam. Here a three-dimensional analysis of the cracking and the dynamic behaviour of the cracked dam is required. The limit for displacements depends on the valley shape and the geometry of the dam. In the case of an arch dam, the limiting displacements can be less than 5cm. Much depends on the nature of the displacement. This is true for all concrete dams. Fault movements will cause cracks in almost all concrete dams, whereas differential foundation movements resulting, for example, from the lowering of the ground water table are less critical and larger displacements can be accepted. A typical case for such deformations is the Zeuzier arch dam in Switzerland, which experienced substantial cracking due to the lowering of the ground water table and the resulting deformations of the dam abutments (Amberg and Lombardi, 1982).

Based on this discussion it can be concluded that the dam which could best resist foundation movements is a zoned embankment dam.

The estimates of acceptable movements along discontinuities in the dam foundations given above are very rough and cannot be generalized as all dams are prototypes located at sites with specific local conditions.

Concrete dams can only be considered when the location of any fault(s) in the foundation and the direction of future movements are reliably known. In this case it would be possible to design a slip joint to cope with these deformations. The best known concrete gravity dam which was built with such a joint is Clyde dam in New Zealand (Hatton 1991, see Figure 4). As no fault movements have occurred at Clyde dam there exists no verification of the adequacy of this type of unique design.

Figure 4

Figure 4: River channel fault joint of Clyde gravity dam in New Zealand (Hatton 1991) (Explanations: 1. Contact surfaces; 2. Wedge plug; 3. Filter zone;
4. Drainage zone; 5. Drainage pipe; 6. Riprap; 7. Gravel blanket; 8. Low permeability blanket; 9. Cutoff shaft)

For the proposed 185m high Steno double-curvature arch dam in Greece, which is located at a site with a potentially active fault in the foundation, a special design was developed, which included a peripheral joint and a horizontal slip joint as shown in Figure 5 (Gilg et al. 1987). To check the viability of the proposed joint system, special model tests were carried out. It was concluded, based on these tests, that the dam can withstand horizontal movements in the order of 5-10cm without damage. Larger movements up to 100cm caused significant damage in the dam. Additional investigations will be needed before such a dam can ever be built. Also at this location, despite economical disadvantages, an embankment dam would be a safer and less problematic solution.

Figure 5

Figure 5 – Proposed Steno arch dam with joint system to cope with potentially active fault at the base of the dam (Gilg et al., 1987)

Main features of embankments with potentially active fault in their foundation

The basic ingredients of an embankment dam, which can resist both differential ground movements and strong earthquake ground shaking, are as follows:

• 1) Impervious core made of ductile material with a high failure strain to minimize the propagation of the rupture zone; prevention of internal erosion if core is cracked.

• 2) Thick filter and transition zones: about 50% shall still be available after faulting and slip movements.

• 3) Wide dam crest.

• 4) Flat slopes.

• 5) Generous freeboard: to prevent overtopping due to impulsive waves in reservoir and settlement of the dam.

• 6) Material selection and compaction of rockfill, etc.

Other fill dams would also be feasible, which include the features discussed in the paper by Sherard et al. (1974).

The main concern of any embankment dam with impervious core is the erosion resistance of the core material. According to Sherard (1967) ‘the filter and transition zones provide the first line of defense against earthquake-induced concentrated leaks through the dam. If thick, adequately graded, cohesionless transitions are provided, a leak can only get out of control in extreme cases of embankment distortion caused by foundation movement.’

‘Where there is a choice between several types of materials for the core of a dam, which may be subject to an earthquake, it seems apparent that the resistance to concentrated leakage should be the main factor in the decision.’

An approximate classification of core materials on the basis of resistance to concentrated leaks (materials with self-healing properties) was also made by Sherard as shown below:

• 1) Very good materials: Very well-graded coarse mixtures of sand, gravel, and fines.

• 2) Good materials: (i) Well-graded mixtures of sand, gravel, and clayey fines; (ii) highly plastic tough clay (CH) with plasticity index greater than 20.

• 3) Fair materials: (i) Fairly well-graded gravelly, medium to coarse sand with cohesionless fines; (ii) clay of medium plasticity (CL) with plasticity index greater than 12, (iii) coarse mixtures of sand, gravel, and fines

Very poor materials are (i) fine, uniform, cohesionless silty sand; (ii) silt from medium plasticity to cohesionless (ML) (plasticity index less than 10).

Conclusions

The active tectonics being displayed at present constitutes a specific hazard, which has to be taken into consideration during design and safety evaluation of dams. The most dramatic scenario is the surface fault break through a dam foundation. Fault creep can also endanger the structural integrity of dams although in a less violent manner.

In spite of the fact that such active phenomena are rather infrequent, the existing risk is not negligible. When an active or potentially active fault in a dam foundation is recognized, far reaching consequences for the dam design must follow. This crucial evaluation must be formulated by the dam geologist and endorsed by the responsible dam design engineer who bears the individual responsibility for his dam.

The accumulated experience and evidence on fault movements indicate that it is necessary to define the engineering strategy. The possibility of surface fault breaks should as a rule be considered while designing dams or evaluating their safety.

In cases where recent tectonic activity of a fault crossing the dam site is recognized and it is not possible to find an alternative site, a conservatively designed embankment dam (with large filter and transition zones of non-cohesive materials) is the type which offers best chances to survive the fault break effects. In general, concrete dams should not be accepted for sites affected by active tectonic features.

The experience in evaluating the geodynamic hazard is accumulating and monitoring of seismic phenomena is steadily developing. In general, seismic monitoring should be provided and kept active on and around large dam sites, preceding by a couple of years the impounding of the storage reservoir in question.

As a general guideline, if significant movement along a fault crossing the dam site is accepted as a reasonable possibility during the lifetime of the dam, the best advice is to select an alternative site, less exposed to geodynamic hazard. Such standpoint is supported by the fact that no dam, foreseen to successfully survive the shearing action of a fault slip in its foundation, has ever been exposed to actual test under such event.

Due to the cumulative nature of fault movements and due to the vulnerability of dams to foundation movements, conservative estimates of the maximum possible fault movements are necessary for design and safety checks.

Martin Wieland, Chairman, ICOLD Committee on Seismic Aspects of Dam Design, Poyry Energy Ltd., Zurich, Switzerland, martin.wieland@poyry.com

A. Bozovic, Former Chairman, ICOLD Committee on Seismic Aspects of Dam Design, Consultant, Belgrade, Serbia, abozovic@EUnet.yu

R.P. Brenner, Past Chairman, ICOLD Committee on Dam Foundations, Consultant, Weinfelden, Switzerland, brenner.gde@sunrise.ch