 In many areas of the world, the potential for an earthquake is an everyday fact of life. Earthquakes are caused by tectonic movement on geologic faults, or by volcanic activity, and may affect large geographic regions. Generally, the larger the earthquake, the more severe the shaking, and the bigger the area it impacts. Moderate to large earthquakes have caused numerous structural failures and ground instabilities, including landslides and subsidence. Our society has constructed an extensive, complex infrastructure critical to our everyday lives and economy. Buildings, highways and rail systems, power facilities, dams. Everything man has constructed is affected by earthquake shaking. For each project, engineers have tried their best to build safe, reliable structures, and in many areas they are exposed to earthquake shaking. It is fortunate that large, damaging earthquakes do not happen every day, but some information, essential for engineers, can't be gained without earthquakes happening. As more information becomes available with each new earthquake, the understanding of earthquake effects increases. Scientists and engineers learn facts from each major event, which show what needs to be done to improve the safety of existing structures and future designs. Every structure is built on soil or rock foundations, and some structures, like dams and highway or railway embankments, are built of soil and rock. The effect of earthquakes on these geologic materials is central to the safety of many structures and has been the subject of extensive investigation since the 1960s. Examples of earthquake-induced settlement or liquefaction have been evident in most large earthquakes, frequently resulting in structural damage. Being constructed of soil and rock, and often on soil foundations, embankment dams are susceptible to failure during large earthquakes. Although no catastrophic failures of major, modern embankment dam structures have occurred because of an earthquake, the potential for failure exists. Several embankment dams have been damaged to the point that their reservoirs would have been released had they been at normal operating levels. This presentation, entitled Behavior of Embankment Dam During Earthquakes, focuses on the significant effects that earthquakes can have on embankment dams. The Federal Interagency Committee on Dam Safety, also known as ICODS, asked Dr. I.M. Idris, an internationally recognized authority on the subject, to prepare and make this presentation. Professor Idris is a pioneer and recognized leader in developing and applying rational engineering procedures to evaluate the effects of earthquakes on structure foundations and embankment dams. He received his Doctor of Philosophy degree in civil engineering from the University of California Berkeley in 1966. His thesis advisor was a late Professor H. Bolton Seed, with whom he continued to interact and collaborate for some 25 years. He spent several years as a lecturer and research engineer at UC Berkeley and joined Woodward Clyde Consultants in 1969. Dr. Idris subsequently spent 20 years as a consultant in private practice, specializing in geotechnical and earthquake engineering. During this time, he continued to maintain a mix of academic pursuits and practice, serving as an adjunct or consulting professor at UC Berkeley, Stanford, and UCLA until 1986. In 1989, he joined the faculty at the University of California Davis and was appointed director of the Center for Centrifuge Modeling, a position he held from 1989 until 1996. He continues to serve as a professor of geotechnical engineering at UC Davis. Dr. Idris has been involved in post-earthquake investigations since the occurrence of the 1964 Great Alaska earthquake. He has done comprehensive post-earthquake investigations of some 17 earthquakes, from Anchorage and Nagata in 1964 to Turkey and Taiwan in 1999. During this time he has developed, or co-developed, many of the procedures currently used for evaluating liquefaction potential, site response, and the behavior of embankment dams during earthquakes. In recognition of his contributions, he was elected a member of the National Academy of Engineering in 1989. Dr. Idris's achievements have also been recognized with many awards. These include the following from the American Society of Civil Engineers, the Norman Medal, the Middlebrooks Award, the Walter L. Huber Research Prize, the J. James Crows Medal, and the 1st H. Bolton Seed Medal. After the 1989 Loma Prieta earthquake, he served on the California Governor's Board of Inquiry into the collapse of the Cypress Section of Interstate 880 and the damage to the San Francisco Oakland Bay Bridge. He has published over 150 papers in various professional journals worldwide and has lectured in many countries. A recognized international authority on earthquake engineering, he has consulted on seismic safety issues for the government of Italy, the International Atomic Energy Agency, UNESCO, several federal agencies in the United States, and a number of departments of the state of California. Dr. Idris is active in a number of professional societies. He is a fellow of the American Society of Civil Engineers, a member of the Earthquake Engineering Research Institute, and a member of the Structural Engineers Association of California. He served on the Seismology Committee of the Structural Engineers Association of Northern California from 1971 to 1981, chairing subcommittees on soil structure interaction and sliding and overturning. Professor Idris consults on problems and projects of many types, with special interest in earthquake engineering, seismic risk mitigation, and geotechnology. He has worked on many international projects on several Corps of Engineers and Bureau of Reclamation Dams in the United States, and with the Federal Energy Regulatory Commission for their dam safety and licensing activity. He recently completed assignments as a member of the peer review panels for Seven Oaks Dam and the Eastside Reservoir Project in Southern California. He is currently a member of the peer review panel at the new San Francisco Oakland Bay Bridge and serves as a member of the Technical Coordination Committee of the U.S.-Japan Cooperative Research Program for the National Science Foundation. His presentation for this video is divided into four parts. Part one will cover observed behavior of embankment dams in previous earthquakes. Part two reviews geologic, seismologic, and earthquake ground motion issues. Part three describes methods for analyzing and evaluating embankment dam performance and potential for deformations, including soil strength. And in part four, he covers instrumentation and makes concluding remarks. It is my pleasure to introduce Dr. I. M. Idris for Dr. Ed Idris as he is known to his friends and colleagues around the world. It's a real pleasure and a distinct honor for me to be invited by ICOTS to participate in this series on dam safety. Embankment dams of earth and rockfell are unique in that to design and evaluate such dams require a multitude of disciplines and a multitude of capabilities within each discipline. Almost every area of geotechnical engineering is involved in embankment dam design and evaluation. In this regard, I am pleased that ICOTS scheduled my part following those by Professor Peck who discussed seepage and piping that may lead to failure by Professor Deere who reminded us of the need for thorough investigations of dam sites and proper preparation of the foundations of dams to include the abutments to ensure stability and to control seepage by Mr. Lowe who described and illustrated the importance of filters in controlling seepage as well as the formations of sinkholes and presented a comprehensive discussion of sudden drawdown and its potential detrimental effect on the stability of dams. And by Professor Mitchell who addressed and illustrated ground improvement for dam safety. Each of the previous speakers included in his presentation a healthy dose of illustrations and case histories. Design and evaluation of embankment dams depends heavily on precedence, hence the importance of case histories. I encourage all those who wish to get seriously involved in embankment dams to study these case histories as thoroughly as possible. Today I'll be discussing observed behavior of embankment dams in previous earthquakes, then some geologic, seismologic and earthquake ground motion issues followed by method of analysis and evaluation including a brief discussion of soil strength. And finally instrumentation and some concluding remarks. It's not surprising that prior to the San Fernando earthquake in 1971 the prevailing belief within the embankment engineering community was that earth and rock-filled dams are inherently stable during earthquakes. This belief was probably heavily influenced by the observed behavior of earth dams during the 1906 San Francisco earthquake. The 1906 San Francisco earthquake is the strongest earthquake to occur in California in the 20th century. It occurred on the San Andreas Fault as shown in the figure and had a moment magnitude of about 8. Some 35 water retaining earth dams had about 5 to 50 kilometers from the ruptured part of the San Andreas Fault. Reports describing the construction of only 10 of these dams are available and showed that little or no compaction effort had been used in the construction of these dams. Yet only a handful of these dams suffered minor damage as a result of this very strong earthquake. The other 30 dams did not suffer any discernible distress. The embankment engineering professionals were therefore pleased to learn that even relatively poorly constructed embankments could withstand such strong shaking. They concluded that embankment dams constructed an improved method should then behave quite well during future earthquakes. The common element among the dams shaken by the 1906 earthquake was that each dam was composed of essentially clay soils. The Sheffield Dam, an 8 meter high embankment constructed of silty sand on a silty sand foundation with an impervious upstream face failed during the 1925 Santa Barbara earthquake. The location of the dam is in Central California as shown in the map. A schematic of the cross section of the dam is shown in this figure. Photographs taken before and after the earthquake show what happened as a result of the shaking. The failure of the Sheffield Dam in the saturated silty sand soils in the foundation causing the massive slide that resulted in the failure of the dam. Unbeknown to the embankment engineering profession was the performance of some 74 earth dams in Japan during and following the occurrence of the 1939 Ojika earthquake. Obviously there were other more pressing items on people's mind in 1939. Communication was not that readily available with Japan at the time and the reports prepared in 1939 describing the performance of these dams were written in Japanese. These reports were finally translated in 1970 and we became aware that 12 dams suffered complete failure and 62 dams were severely damaged. The primary causes of failure were washed out by excessive seepage due to slope failures, cracks and settlements, failure of the intake structures, failure of buried conduits, damage to spillways and overtopping to the failure of upper reservoir of a multi-level reservoir system. A study by Akiba and Semba in 1941 provided the following conclusion. One, most of the failures occurred either a few hours or up to 24 hours after the earthquake. Two, the damaged and failed embankments consisted of sandy soils even at some significant distances from the earthquake source. Three, no complete failures occurred in embankments constructed of clay soils even in dams at very close distances to the earthquake source. It's noteworthy that dams built of clay soils behaved equally well in 1906 San Francisco earthquake and in 1939 Oheka earthquake. 1971 San Fernando earthquake was a modern earthquake with a moment magnitude of 6.6. It probably had the most influence on the profession by raising concerns about the performance of embankment dams under earthquake loading conditions. Slow failures occurred in the lower and upper San Fernando dams both of which had been constructed using hydraulic fill methods resulting in relatively loose sandy shells. Nearby rolled earth dams suffered little or no effects in the same earthquake. The general location of the San Fernando dams are shown in the figure. The upper San Fernando dam moved some 1 to 2 meters downstream resulting in the cracks shown in the aerial photo taken shortly after the earthquake. Note that the bridge connecting the crest of the dam to the inlet outlet tower collapsed because of the movement. A close-up of the deformed crest of the dam is shown in this figure. An aerial view of the entire San Fernando dam complex is shown in this figure which came from a photograph taken shortly after the earthquake. As can be seen, an upstream slide occurred in the lower dam. Some noteworthy observations from the photograph are the inlet outlet tower near the right abutment of the lower dam was operational and was instrumental in providing the means to empty the reservoir. Some 80,000 people lived downstream of the lower dam and were in jeopardy at the dam being breached. In fact, the residents downstream of the dam were evacuated but were able to return once it was evident that the reservoir could be expeditiously emptied. If we take a look at a close-up of the remaining part of the dam we can see how precariously close the water came to overtopping and possibly eroding what was left of the downstream shell of the dam. The freeboard in this picture was no more than about 3 feet a little less than 1 meter. Indeed, it was very fortunate that the reservoir had been lowered while the dam was being evaluated for the effects of earthquakes. Refilling of the reservoir was initiated a few weeks before the earthquake occurred. Had the earthquake occurred a year earlier when the reservoir level was about 4 feet higher there would have been a significantly more catastrophic outcome. As the water was released from the reservoir the exposed surfaces of the dam were inspected and a detailed study was undertaken of this dam. The exposed upstream surface was shown in the next two figures. Shortly after the earthquake the Department of Water Resources of the State of California and the National Science Foundation supported a study of both the upper and the lower San Fernando dams. The lead investigators were Professor H. Bolton Seed, Professor Kenneth L. Lee and myself. Dr. Faiz McNissie became part of the investigation team as he was finishing his graduate studies at UC Berkeley. The laboratory tests and analytical studies and evaluations are included in the report published about two and a half years after the earthquake. The next figure shows the retrieval of the seismoscope record. The seismoscope had been installed a few years earlier at the crest of the dam based on recommendations by Professor George Hausner. Professor H. Bolton Seed and Professor Kenneth L. Lee are shown in this picture which I took. The pre-earthquake cross-section is shown in the figure. Both the upstream and downstream shells consist of sand which had been hydraulically filled. The sand was silty, finds contents of about 25% and not very dense, SPT low count and 160 of about 10. The soil blocks that slid into the reservoir were very carefully surveyed and measured and are depicted in the figure below the pre-earthquake cross-section. These blocks were then refitted to create the cross-section shown in the bottom of the figure. The zone of the upstream shell colored blue was judged to have suffered loss of strength leading to the instability of the upstream shell and the massive landslide that ensued. Several large earthquakes have occurred in California since the 1971 San Fernando earthquake. The motions generated during these earthquakes were felt by numerous embankment dams throughout the state. Unfortunately none resulted in the almost catastrophic performance of the lower San Fernando dam. A few suffered lateral and vertical movements. Some movements even reached a few feet resulting in visible cracks along the entire embankment. For example, the maximum movement of the crest of the Austrian dam in Northern California in the 1989 Loma Prieta earthquake are illustrated in this figure. The next figure shows the pattern of the cracks that developed in this dam due to the shaking during this earthquake. Based on the observations and conclusions derived from the case history we just reviewed from other observations in Japan and Russia, the following general conclusions can be developed. One, hydraulic fill dams are vulnerable to failure if the level of shaking is sufficiently strong. Two, many hydraulic fill dams, however, have performed well if they are built with reasonable slopes on stable foundations and the level of shaking is no more than about 0.2G caused by magnitude 6.5 earthquakes. Three, dams constructed at clay soils on clay or rock foundation have withstood extremely strong shaking from a magnitude 8 earthquake. Number four, virtually any well-built dam on a stable foundation will perform well during moderate earthquake shaking, say about 0.2G. Five, dams dams that suffered complete failure or significant slope movement were constructed primarily of saturated sandy shells or on saturated sandy foundations. These lessons emphasize the need for our concern with the effects of earthquakes on embankment dams composed of or supported on vulnerable soils. The lessons have also motivated studies aimed at understanding the behavior of these soils under cyclic loading conditions and at developing methods for evaluating these effects. Before we discuss evaluation of embankment dams it is necessary to go over some of the geologic seismologic and earthquake ground motion aspects necessary for such evaluations. This figure shows the major earthquake belts of the world where earthquakes of varying severity occur quite frequently. Earthquakes also occur outside these belts but far less frequently. In fact, Dr. Richter noted in his famous book published in 1958 it is probable that no large area of the world is permanently unaffected by earthquakes. The geologic seismologic and earthquake ground motion issues are addressed by conducting the needed regional as well as local geologic investigations. A regional geologic investigation provides information about the seismic sources that need to be considered in evaluating the performance of the embankment and its foundation during future earthquakes. This regional evaluation can cover an area extending several tens to possibly a few hundred kilometers from the dam site. The regional geologic investigation is completed to establish the seismic sources relevant to the dam site and to estimate the earthquake ground motions at the site. The local geologic investigation provides information about possible faults or other geologic features such as faults in the immediate vicinity of the embankment or even along or across the footprint of the embankment. Such a local geologic investigation complements the investigations that Professor Dier discussed and usually covers an area extending several hundred meters to possibly a few kilometers from the dam site. It's important to avoid building a dam over a fault that can have a major offset such as that shown in the photographs. The first photograph was taken a few days following the occurrence of the 1971 San Fernando earthquake. The next photograph taken shortly after the occurrence of the 1999 Chi Chi earthquake in Taiwan shows what can happen to a dam when subjected to a few meters of fault offset. It may be noted, however, that embankment dams can be proportioned to withstand a significant fault movement as discussed by Girard, Kluf and Allen in 1974 and updated by Allen and Kluf in 2000. To do so, it is necessary that the location and sense and amount of movement of the fault be fully evaluated for the proper design of the dam. Quite often, as noted by Dr. Dier, minor faults are encountered in dam foundations and require special treatment. For the purpose of this video, I will concentrate my comments about evaluating the earthquake ground motions that the embankment needs to withstand. These earthquake ground motions depend on the location of the dam in relation to known and suspected seismic sources. These seismic sources can be well-defined faults, hidden or blind faults, seismic zones, subduction zones or random sources. Random sources are typically included to incorporate known seismicity that cannot be ascribed to any specific source. For example, in West North America, particularly in California, seismic sources are well-defined faults, blind faults and random sources. The Cascadia subduction zone has generated earthquakes that have been felt in Northern California, Oregon, Washington and British Columbia. The seismic sources in East North America, on the other hand, are mostly seismic zones. Once the seismic sources that can affect the dam site have been identified, there are seismic hazard evaluation procedures that can be used to estimate the earthquake ground motions that the site may experience. These two procedures can be identified as a deterministic seismic hazard evaluation procedure or a probabilistic seismic hazard evaluation procedure. In a deterministic seismic hazard evaluation, the current practice consists of completing the following steps. One, a geologic and seismologic evaluation is conducted to identify and define the sources relevant to the site. Two, the maximum magnitude on each source is estimated and the closest distance to the site is determined. This earthquake has been typically designated as the maximum credible earthquake or MCE. Three, recurrence relationship for each source are derived. Historical seismicity as well as geologic data are used for this purpose. In a deterministic evaluation, the recurrence relationships are used to select an earthquake having a selected average recurrence and a magnitude lower than the maximum magnitude update in step two. An earthquake selected this way may be designated an operating basis earthquake or OVE. It may be noted that this event is typically not important for embankment dams. Four, the needed earthquake ground motion parameters, for example, accelerations, velocities, spectral ordnance, etc., are calculated using one or more attenuation relationship or an analytical procedure for the MCE and the OVE if applicable for each source. And five, the magnitude and distance producing the largest ground motion parameters are then used for analysis and design purposes. A probabilistic seismic hazard evaluation involves obtaining a mathematical process, the level of a ground motion parameter that has a selected probability of being exceeded during a specified time interval. Typically, the annual probability of this level of the ground motion parameter being exceeded is calculated. The inverse of this annual probability is the return period in years. Once this annual probability is obtained, the probability of this level of the ground motion parameter being exceeded or any specified time period, which is often defined as the exposure period or life of the structure, can be readily calculated by the equation shown. In this equation, P is the probability of this level of the ground motion parameter being exceeded in T years and lambda is the annual probability of being exceeded. Note that the inverse of lambda is return period in years. For example, if the ground motion parameter is selected to have an average return period of 100,000 years and the exposure time is 100 years, then the probability of this value of the parameter being exceeded over this exposure period would be approximately 9.5%. If the exposure time is only 50 years, the probability changes to approximately 5%. A probabilistic seismic hazard evaluation at a site due a particular source requires knowledge of the following three functions. One, the recurrence rate to calculate the probability that an earthquake of a particular magnitude will occur on this source during a specified time interval. Two, the probability that the earthquake would occur at a specified distance from the site is assessed by considering both source geometry and if applicable the magnitude rupture length or area relationship. Three, the probability that the ground motion from an earthquake of a certain magnitude occurring at a certain distance would exceed a specified level at the site is based on the selected attenuation relationship. By combining the three probability functions for each source the annual probability of exceeding a specified level of ground motion at a site is computed. This process is repeated for each source and the contributions are added to obtain the total seismic hazard at the site. The results of the probabilistic seismic hazard evaluation are then used to select the earthquake ground motion parameters corresponding to a desired average return period. The contributions of each magnitude range in each distance range to the hazard associated with the selected parameters are used to obtain an average magnitude at an average distance. These are usually designated as m bar and r bar. For some applications this pair of m and r is then used in the same way as the MCE and the closest distance To avoid making this video about earthquake ground motions I will illustrate with a simple deterministic example and refer the viewer to the list of references provided in the video too for more details about both deterministic as well as probabilistic seismic hazard evaluation procedures. In this regard it is useful to note that the probabilistic seismic hazard evaluation procedures are still in the early stages of being used in evaluating the seismic performance and embankment dams. It is hoped that they will receive more attention from engineers engaged in evaluating these dams because probabilistic procedures provide far greater insight about the potential seismic hazard to a dam than deterministic procedures. The example I am using today pertains to a site with some nearby seismic sources as well as a distance source and a random source. The maximum magnitude and the closest distance to the site are listed in the table and consist of random source magnitude 5.5 closest distance 5 km this earthquake is designated event A nearby source magnitude 6.5 closest distance 21 km this earthquake is designated event B distant source magnitude 8 closest distance 150 km this earthquake is designated event C The target spectra associated with each of these events are obtained using attenuation relationships that provide values of spectra acceleration for a number of periods depending on the magnitude and distance and the style of rupture on the seismic source. Attenuation curves for spectral ordnance for a period of one half second and three magnitudes are shown in the figure. Similar attenuation relationships are available for some 10 to 20 periods ranging from 0.05 to 5 seconds. The target spectrum for event A is shown in the figure. This target spectrum was developed using three available attenuation relationships as noted in the figure. The same attenuation relationships are used to obtain target spectra for events B and C. The target spectra for events A, B and C are shown in the figure. It's interesting to note that the spectral values for event A are the largest at short periods less than about half a second. While the spectral values for event C are the largest for periods longer than about one second. These trends are typical for many sites and hence more than one event needs to be considered in evaluating the performance of various sections of the dam. For most embankment dams spectral ordnance unnecessary but not sufficient to conduct an evaluation of the seismic performance of the dam. Time history are needed for this purpose. There are at least three approaches by which these time histories can be attained. One approach is to select a number of accelerograms recorded the distances comparable to that of the event under consideration and during earthquakes whose magnitude are the event under consideration. A second approach is to construct a spectrum-compatible accelerogram. That is, the spectral ordnance of the constructed accelerogram are very close to those of the target spectrum. The third approach is to generate one or more accelerograms using analytical procedures that account for source path site characteristics for the event under consideration. The first approach is illustrated in the figure for event B, 6.5 to 21 kilometers. There are five recorded accelerograms in this figure whose ordnance were adjusted to have the same peak acceleration as the target zero period acceleration for event B. Spectral ordnance for each of these accelerograms are shown in the next figure. Also shown in this figure are the target spectrum for event B and the average spectrum for the five accelerograms. The information in this figure indicates that the target spectrum for any individual accelerogram can be significantly different from the target spectrum, but on the average, the resulting spectrum for the selected accelerogram is comparable to the target spectrum. It should be noted that the fit could be improved by increasing the number of the selected recorded accelerograms. In this regard, the minimum number should be 5, the constructed accelerogram was adjusted so that its spectrum is almost identical to the target spectrum as is evidenced in the figure. The next three figures show the information that should be obtained to further judge the adequacy of any accelerogram. This figure shows the time histories of acceleration, which is the constructed accelerogram, velocity and displacement. The thing to watch for in this figure is the possible drift of time or displacement, time histories which would result in unrealistic velocities and displacements. The next figure shows what is called a HUSID plot, which represents the cumulative buildup of areas intensity, which is a measure of energy as a function of time. The ordinate of this plot is normalized with respect to the areas intensity and hence has a maximum value of 100%. The time from 5% on the HUSID plot has often been used as a measure of effective duration. Thus the duration for this accelerogram would be 16.5 minus 4.2 or 12.3 seconds. This plot is useful to check that the buildup of cumulative areas intensity is comparable to naturally recorded accelerograms and that the duration is neither too long nor too short. The next figure shows the Fourier amplitudes for the spectrum, compatible accelerogram. The unsmooth and the smooth amplitudes are shown in this figure. The important thing to check in this figure is that there are no significantly deep valleys in the Fourier amplitudes, especially in the frequency range of interest.