In an update to a 1979 paper on the response of concrete dams to earthquakes, Kenneth D Hansen and Larry K Nuss present details on the earthquake performance of six dams, and discuss their significance for the dam engineering profession
An important paper was published in the April 1979 issue of International Water Power and Dam Construction entitled Response of Concrete Dams to Earthquakes. Authored by Kenneth D Hansen and Louis H. Roehm, the paper reported on the performance of 17 concrete dams in nine countries that had been subjected to ground shaking in excess of 0.10 g. Specific case histories were presented for the following six dams that were shaken severely up to that time and the performance noted:
• Lower Crystal Springs Dam, USA (1906*) – curved gravity dam,
• Blackbrook Dam, England (1957) – straight gravity dam,
• Hsinfengkiang Dam, China (1962) – concrete buttress,
• Koyna Dam, India (1967) – straight gravity dam,
• Pacoima Dam, USA (1971) – arch dam,
• Ambiesta Dam, Italy (1976) – arch dam.
From the recorded performance of these and the other dams listed in a table, it was concluded, in general, concrete dams had performed extremely well when subjected to earthquake motions, even when shaken by forces far in excess of their design loading.
Thirty-two years later, many more earthquakes have occurred which have shaken concrete dams of all types worldwide to a greater extent. Now 19 dams are listed in Table 1 that were shaken by peak horizontal ground accelerations (PHGA) greater than 0.3g. The new list includes five dams from the original table.
Case histories are now presented for the following dams:
• Sefid Rud Dam, Iran (1990) – concrete buttress,
• Bear Valley Dam, USA (1992) – straight gravity (after modification),
• Pacoima Dam, USA (1994) – arch,
• Shih Kang Dam, Taiwan (1999) – multiple bay gravity spillway,
• Kasho Dam, Japan (2000) – straight gravity,
• Shapai Dam, China (2008) – RCC arch.
These dams, similar to the previously reported case histories, were selected based on a number of factors which led to their significance to the dam engineering profession. These factors include importance of the dam, severity of the ground motion, occurrence or lack of observed damage, and availability of quality strong motion records at or near the dam. In addition, a report on the performance of concrete dams subjected to the 2011 Tohoku Earthquake is presented.
Sefid Rud Dam
The performance of Sefid Rud Dam in Iran as shown in Figure 1 is an important case study as the 106m (348ft) high concrete buttress structure suffered appreciable damage due to severe shaking from the M 7.7 Manjil Earthquake of 21 June 1990. Completed in 1962, the dam was designed to withstand a 0.25 g PHGA and is located in the north central region of Iran between the towns of Manjil and Rudbar. The exact location of the epicenter of the earthquake was not determined, but it was close to the location of the dam. The estimated PHGA at the site was back calculated to be 0.714 g based on the nearest record 40km (25 miles) away where a PHGA of 0.56 g was recorded.
Two strong aftershocks having magnitudes in the 6.2 to 6.5 magnitude range occurred several hours after the main shock. In addition, more than 400 aftershocks with magnitudes up to 5.9 were reported in the following weeks by the Geophysical Center of Tehran University. An estimated 40,000 people died due to the earthquake and another 60,000 people were reported as injured. The towns of Manjil and Rudbar were almost completely destroyed.
The dam suffered several types of damage, including horizontal cracks about 18m (59ft) below the dam crest where the inclined buttress intersected the vertical “chimney” section. The cracks were due to a sudden change in stiffness of the structure at this reentrant corner (see Figure 2). No significant displacement occurred in these cracked joints probably due to the high frictional resistance in the horizontal concrete joints between the buttresses. Minor displacement both in the upstream and downstream direction occurred in a few of the 23 blocks. The accumulated displacement in blocks 10 through 20 was 10mm (0.4 inch). Leakage was reported through some of the cracks. Other damage included minor damage and displacement of all the gates, varying types of damage at the dam crest, and damage in the switchyard and to transformers. Repairs were made in 1991 by grouting the cracks and installing prestressed anchors in the cracked areas. The repaired dam remains in service at this time.
Bear Valley Dam
Bear Valley Dam located in southern California, US, as shown in Figure 3 is significant in that it was shaken by two distinct earthquakes one day apart. Bear Valley Dam is a 28m (92ft) high concrete multiple arch dam that was completed in 1912 and modified in 1988. The modification was due to the concern with adequacy of the dam when subjected to the design earthquake or overtopping by large floods. The structural upgrade consisted of converting the multiple arch to basically a gravity structure by partially infilling the arch bays with conventional concrete.
The strengthening of the dam considered two maximum credible earthquakes (MCE). The controlling seismic shaking was an M 8.3 earthquake on the San Andreas Fault 16m (10 miles) away with a PHGA of 0.45 g and 35 seconds of strong shaking.
On 28 June 1992, the fault rupture on the Landers earthquake (M 7.4) located 45km (28 miles) away shook the dam. Then on 29 June 1992 the M 6.6 Big Bear Earthquake occurred about 14.5m (9 miles) from the dam site on a unnamed fault in response to the rupture on the Landers Fault.
Thorough post earthquake investigations indicated Bear Valley Dam had not been damaged. The only indication of shaking at the site was a slight displacement of girders on the highway bridge located at the dam crest.
At the Big Bear Lake Civic Center located about 4km (2.5 miles) from the dam, PGA of 0.18 g horizontal and 0.08 g vertical was recorded during the Landers Earthquake and 0.57 g horizontal and 0.21 g vertical due to the closer Big Bear Earthquake.
Pacoima Dam
Pacoima Dam, also located in southern California, US as shown in Figure 4 is of interest because it has been shaken by two major earthquakes, the 9 February 1971 San Fernando Earthquake of M 6.6 and then the 17 January 1994 Northridge Earthquake of M 6.8. The performance of the 113m (372ft) concrete arch dam was described in the 1979 Hansen-Roehm article. At that time a number of engineers questioned the validity of the peak accelerations of 1.25 g horizontal and 0.70 g vertical recorded by accelerometers located 16m (52ft) above the dam crest on the left abutment.
Since the 1971 seismic event, an extensive seismic instrumentation system was installed and the upper rock mass of the left abutment was secured to more competent rock below through the use of 35 rock anchors.
The epicenter of the 1994 Northridge Earthquake was determined to be 18km (11.4 miles) southwest of the dam at a focal depth of 17km (10.5 miles). Strong motion records indicate a peak horizontal ground acceleration of 0.53 g at the base of the dam. The acceleration records from the instrument at the upper left abutment measured a peak acceleration of 1.58 g due to the topographic amplification in the canyon. Peak accelerations of greater than 2.3 g horizontal were recorded at the dam crest. The records from the 1971 San Fernando Earthquake are no longer being questioned.
Despite being subjected to high accelerations, the arch dam survived the earthquake well with the main damage being an opening of the contraction joint between the arch dam and the thrust block at the left abutment of approximately 50mm (2 inches) (see Figure 5). A downward movement of about 13mm (0.5 inches) indicates that the thrust block and underlying rock mass may have moved away from the dam. The water level at the time of the 1994 earthquake was approximately two-thirds the maximum depth of the dam and 4.1m (13.5ft) higher than the reservoir level during the 1971 earthquake.
Mojtahedi and Fenves (1995) studied the response of Pacoima Dam using recorded ground motions obtained at the dam site. The analyses indicated opening-closing of contraction and horizontal joints and non-uniform seismic input. A reasonable agreement was obtained between the accelerations recorded on the dam body and the computed accelerations.
Shih Kang Dam
Shih Kang Dam (see Figure 6) in Taiwan is especially relevant to the study of performance of concrete dams subjected to earthquakes as it 1) is the first concrete reported to have failed due to an earthquake and 2) the dam is located directly over a branch of the fault caused by the M 7.6 Chi Chi (also called 921) earthquake on 21 September 1999. The 21.4m (70ft) high gravity dam is essentially an 18-bay gated spillway.
The main Che-Lung-Pu fault located about 300m (1000ft) away from the structure was known at the time of the design of the dam, but not the branch that ruptured under the dam. The fault rupture extended both upstream and downstream of the dam and caused extensive damage to bays 16 to 18 on the right side of the structure. The ground movement raised the left part of these bays about 11m (36ft) and the right side by about 2m (7ft). Thus, there was a vertical differential movement of about 9m (29ft) in this area. There was also a diagonal horizontal offset through the dam of about 7m (23ft). A PHGA of 0.51 g and a peak vertical acceleration of 0.53 g were recorded 500m (0.31 miles) from the dam site.
The remaining portion of the dam adjacent to the damaged area also experienced high accelerations and separated from its foundation bedrock. There was some cracking in the piers closest to the fault rupture, but no structural damage. Simply supported reinforced concrete bridges spanning the spillways all came off their bearings. Six gates were inoperable after the earthquake.
Considering the magnitude of the displacement, the performance of the dam was quite remarkable. It did not fail as indicated by the definition of failure as a complete loss of the reservoir water. Due to the upstream topography as well as the gates and piers falling into the passageway the flow through this area was limited to an estimated 100 to 200m3/sec (3,500 to 7,000ft3/sec.)
The performance of Shih Kang Dam seems to confirm what had been postulated by Zienkiewicz, Clough, and Seed in icold Bulletin 52 (1986) “Earthquake Analysis Procedures for Dams – State of the Art”. The authors noted “we shall not consider the possibility of a large differential movement occurring in a fault transversing the dam foundation. In particular the possibility of a movement on a fault zone should at all times be avoided by a suitable geological study. Often, it is speculated that in the event of such a movement, ‘soft’ structures – e.g. earth dams – are safer than more rigid concrete ones. This prediction is however beyond the realm of calculations which are feasible at present and it certainly is possible that a concrete gravity dam is safer than an earth dam in such a fault movement due to its inherent stability after damage.”
Kasho Dam
Kasho Dam as shown in Figure 7 is significant in that it is the second concrete dam, the first being Pacoima Dam, in which an acceleration greater than 2.0 g was recorded at the dam crest during an earthquake. Kasho, a 46.4m (152ft) high concrete gravity dam was shaken by the 6 October 2000 Western Tottori Earthquake (M 7.3) in Japan. The epicenter of the earthquake was about 3km (1.9 miles) from the dam. Numerous aftershocks were recorded with one having a magnitude greater than 7. At the time of the earthquake, the reservoir was nearly 5.8m (19ft) below normal pool.
Peak accelerations of 0.54 g horizontal and 0.49 g vertical was measured in the lower inspection gallery located nearly 9m (30ft) above the base of the dam and 2.09 g in an elevator shaft at the crest of the dam.
Using accelerations measured in the lower inspection gallery from the main shock as well as aftershocks, the natural periods of vibration were 0.84, 0.96, 0.92, and 0.87 seconds. The first period first increased and then decreased. It was theorized with a high degree of probability that the change in period was due to the nonlinearity of the hydrodynamic pressure acting on the upstream face of the dam. Measurements indicated the water level in the upper portion of the reservoir dropped about 200mm (8 inches) and nearly 60mm (2.4 inches) at the dam. From plumb line readings in the dam the main shock produced a relative displacement of 28mm (1.1 inches) toward the right abutment and 0.7mm (0.28 inches) in the upstream direction. The maximum displacement was a little more than 29mm (1.125 inches). There was basically no damage to the concrete gravity dam. The walls at the base of the control room, a reinforced concrete structure cantilevering upstream from the crest of the dam, were cracked (see Figure 8).
Two other concrete gravity dams, the 73.5m (341ft) high Sugesawa Dam and the 14m (46ft) high Uh Dams were also located close to the earthquake epicenter. There was insignificant damage at Sugesawa Dam with PHGA of 0.16 g consisting of a small 1m by 0.3m (3.3ft by 1ft) concrete spall on the downstream face of the dam. Uh Dam is located about 1.0km (0.6 miles) from the earthquake epicenter. Seismographs recorded peak accelerations at the surface of 1.16 g and at a depth of 100m (330ft) of 0.62 g. Although Uh Dam was subjected to severe shaking, the only damage to the dam was cracking 10 to 30mm (0.4 to 1.2 inch) wide on the spillway channel near the base of the downstream face.
Shapai Dam
Shapai Dam as shown in Figure 9 is the first roller-compacted concrete (RCC) dam shaken by a major earthquake. The dam located in Sichuan Province, China is a 132m (433ft) high three-centered RCC arch dam completed in 2003. It is located 12km (7.8 miles) from the Wenshuan (also called Sichuan) earthquake that occurred on May 12, 2008. The M 8.0 earthquake killed more than 80,000 people and about 675,000 were injured.
The PHGA at the site was estimated to be about 0.8 g compared to the design acceleration of 0.13 g. With a nearly full reservoir at the time of the earthquake, the dam was undamaged. One of the spillway gantries was slightly damaged. However, all gates could be opened after the shaking. The power station was badly damaged when large rocks rolled down the steep mountain site knocking holes in the building walls.
The Tohoku Earthquake
As this article was being prepared, the M 9.0 Tohoku earthquake of 11 March 2011 occurred in the Pacific Ocean 131km (81 miles) east of Sendai, Japan. The earthquake triggered tsunami waves up to 10m (33ft) high when it hit shore that caused extreme destruction.
Following the earthquake, more than 400 dams were inspected. No damage was reported to any concrete dam except a minor slope failure at one reservoir formed by a concrete dam. Many of these dams are concrete dams as about 37% of all dams and 60% of dams greater than 30m (100ft) high in Japan are concrete structures.
Aydan (2011) published a map of the location of maximum ground accelerations color coded from 50 to 800 gal (0.05 to 0.8 g). The area of maximum ground acceleration is located about 48 km (30 miles) north of Sendai.
The authors determined that Miyatoko Dam (see Figure 10), a 48m (157ft) high RCC dam was located approximately 13km (8 miles) north of Sendai. From Aydan’s map, this would place this concrete dam in an area where the PHGA was greater than 0.7 g. There are many other concrete dams located in Miyagi, Fukashima and Iwate Prefectures which apparently were subjected to PHGAs in excess of 0.3 g. All were not damaged by this event which has been reported as the fifth greatest magnitude earthquake on record. A recent paper from Japan indicates the 77m (252ft) high Takou concrete gravity dam was shaken by a PHGA of about 0.4 g using Aydan’s map. Takou Dam suffered no damage except cracking in the walls of the gate house located on the crest of the dam. The PHGA’s for Miyatoko and Takou Dams, and possibly others in Japan, will need to be confirmed at a later date.
Conclusions
Since the 1979 article was published, many large magnitude earthquakes have occurred as expected. With a greater number of higher quality strong motion instruments located at or near dams, our base of knowledge of the magnitude of shaking to which concrete dams have been subjected has increased. Thus, our knowledge of the performance of severely shaken concrete dams has increased and this knowledge can be applied in a positive and beneficial manner to the design of future dams.
While many previous reports on the performance of dams subjected to major earthquakes focused on the magnitude of the earthquake and a not so precise distance to the dam, it has become apparent the most significant factor in determining the response of concrete dams is the PHGA and probably the spectral acceleration at the natural frequency of the dam.
While the 1979 article provides a list of 17 dams shaken by an acceleration measured or estimated to be in excess of 0.1 g, Table 1 lists 19 dams shaken by PHGAs greater than 0.3 g. In general, it can be concluded that concrete dams have performed very well when subjected to high intensity accelerations. The threshold of no damage is project specific, but can quite probably be significantly higher than 0.3 g for properly designed and constructed concrete gravity and arch dams.
Concrete buttress dams when subjected to severe shaking have developed horizontal cracks at the elevation high in the dams where the downstream buttresses intersect the vertical “chimney” section. This is an area where the stiffness of the concrete structures significantly changes.
Some other specific conclusions are:
• While a fault located directly below Shih Kang Dam (Taiwan) caused a rupture and relative vertical displacement of 9m (29ft), the remaining damaged concrete limited an immediate total and sudden release of the reservoir.
• PHGAs are amplified from the base of the dam to the crest. In two cases, this amplification produced measured peak accelerations at the crest in excess of 2.0 g (Pacoima arch dam – USA (2.3 g) and Kasho gravity dam – Japan (2.05 g)).
• Peak accelerations, as expected, at the crest are greater with full reservoirs.
• Several dams have been severely shaken on two occasions by separate major earthquakes – Bear Valley Dam – USA (one day apart) and Pacoima Dam – USA (23 years apart) with only minor damage. Many concrete dams have also been shaken by high intensity aftershocks that occurred after the main earthquake without any additional damage.
• Roller-compacted concrete dams – Shapai arch in China and Miyatoko Dam in Japan performed no differently to date than a dam built of conventionally placed concrete despite concern by some of less strength at the many lift joints.
• Where damage has been identified, it has been cracking high in the dam and where additional features such as curbs, railings, gates, or guard or control houses are located.
• Very little in the way of increased leakage has occurred in concrete dams subjected to major earthquakes. This can be attributed, in part, to the fact that any cracking caused by the earthquake has mainly been horizontal and located high in the dam together with the reservoir not being full in many cases. Some rock foundations have experienced a temporary increase in seepage following an earthquake.
There may be a number of reasons why concrete dams have performed well and invariably better than that predicted by design or analysis when shaken by an earthquake. The main reasons may be 1) the redundancy of the damaged structure to redistribute load, 2) the duration of strong shaking being too short to cause failure, 3) the increase in the tensile strength of the concrete during dynamic loading that increases resiliency, 4) an increase in the damping that reduces the seismic impact on the dam, 5) reduced seismic impact because the natural frequency of the dam does not match the frequency of the earthquake, and 6) three-dimensional effects of canyon confinement or dam geometry (curvature) that help prevent failure.
A generally accepted potential failure mode for concrete dams during an earthquake is cracking of the concrete, cracking through the dam that forms removable blocks, sliding of the blocks during or after the earthquake to cause failure. Severely shaken concrete dams to date have cracked at locations of change in geometry (reentrant corners), but have not formed removable concrete blocks. Thus the entire potential seismic failure mode has not been fully achieved or experienced for concrete dams.
Commentary
Earthquakes are natural phenomena that will continue to occur and cause extensive loss of life, damage to buildings and all types of infrastructure. Currently, there is no real way of predicting earthquakes. We can just design our structures including dams to withstand severe seismic shaking.
When a major earthquake occurs, there are all kinds of devastation as was recently recorded following the M 9.0 Tohoku earthquake off the coast of Japan. This resulted mainly from an earthquake induced tsunami that caused extensive damage after washing ashore for a maximum distance of nearly 10 km (6 miles).
Concrete dams have performed very well when subjected to high intensity earthquakes as noted in this paper. We do not need to have a concrete or any type of dam fail and release a wave of water rushing downstream that would add to the devastation already caused by the intense seismic shaking and other earthquake induced phenomena.
While concrete dams are designed to withstand a higher degree of seismic shaking than buildings and have performed well in the past, we should not become overconfident of their performance in the future. Great care should be taken in the design details and quality of construction. Particular attention should be given to possible faults located under the dam.
Kenneth D. Hansen, Consulting Engineer, 6050 Greenwood Plaza Blvd., Suite 100, Greenwood Village, Colorado, 80111, 303-695-6500, ken@ken-hansen.com.
Larry K. Nuss, Structural Engineer, Technical Specialist, Structural Analysis Group, Bureau of Reclamation, P.O. Box 25007, Denver, Colorado, 80123, 303-445-3231, lnuss@usbr.gov.
The authors would like to thank Louis Roehm, one of the co-authors of the original 1979 “The Response of Concrete Dams to Earthquakes” paper [1], for his Peer Review of this paper.
TablesTable 1