Lessons from the Teton Dam failure

Engineering Geology, 24 (1987) 239--256

239

Elsevier Science Publishers B.V., Amsterdam -- Printed in The Netherlands

LESSONS FROM THE TETON DAM FAILURE

JAMES L. SHERARD*

Consulting Engineer, P.O. Box 1416, San Ysidro, CA 92073 (U.S.A.) (Accepted for publication December 1986)

INTRODUCTION

The Teton Dam failure is one of the most important single events in the history of dam engineering. Teton Dam is the highest dam which has completely failed. It was designed and constructed under the supervision of the U.S. Bureau of Reclamation which had built many major dams over the previous 70 years, and which was widely reputed to be a leader in the field. This paper is one of a group on the same subject for the International Workshop on Dam Failures, Purdue University. The objective is to determine whether concentrated re-examination of the failure details after about 9 years have elapsed, and most of the emotional heat generated by the disaster has dissipated, will allow any better understanding of the "lessons learned" for the profession. No attempt is made here to describe all the project features or review the failure details. The reader is assumed to be already familiar with the main reports prepared by the two eminent review panels (Independent Panel, 1976; U.S. Dept. of the Interior Group, 1977, 1980). A good summary of the main failure details is given in Seed and Duncan (1981). Soon after the failure I spent several weeks studying the problem, for m y own edification. This effort included discussions with the responsible engineering staff of the USBR in the Denver headquarters. In Denver also I studied all the main documents, and had conferences with the geologists who worked on the job during the design stage, and others who had mapped the detailed geology exposed during the dam construction. Subsequently during September 1976 1 spent about ten days at the site. During this time the embankment remnant on the right abutment was being slowly and carefully excavated, exposing the rock foundation and rock walls of the cutoff trench as part of the failure cause investigation. The Resident Engineer and inspection staff which had controlled construction were still on the site and they spoke frankly about all the main points, although they were still somewhat shell-shocked. Later, during 1979--81, I was employed to study the subsequent events by a group of insurance companies who had paid flood damages and were petitioning the U.S. Government (unsuccessfully) for reimbursement. During *James L. Sherard died in July, 1987. 0013-7952/87/$03.50

© 1987 Elsevier Science Publishers B.V.

240

this period I studied in detail the results of the exploration of the left bank remnant and the discovery of the famous " w e t seams". RESULTS OF S H E R A R D ' S INVESTIGATIONS

My basic conclusions and opinions from the activities described above were essentially the same as those of the two main official groups of investigators. The official reports (mentioned above) included descriptions of all parts o f the project in exhaustive detail. These reports are so voluminous that, in spite of the panel's efforts to summarize briefly, they tend to give the impression that the situation was complex. In fact, in m y opinion the situation was clear and straightforward, with main elements as follows. (1) The failure was caused when a concentrated leak developed through the earth-filled trench excavated in rock or at the earth--rock interface at the b o t t o m o f the trench. The leak eroded the core material and carried the eroded material into large open cracks in the rock foundation. Progressive erosion of the highly erodible Zone 1 e m b a n k m e n t material led to complete failure within a few hours after the erosion started. (2) It was known from the original explorations and observations during construction that the rock foundation and rock walls of the c u t o f f trench had large open cracks. There was no provision in the design to seal these cracks at the surface and the sealing carried o u t during construction was inadequate and incomplete. The design provided no filter between the fine, erodible Zone 1 and the cracked rock and none was installed during construction. (3) Because of this absence of crack sealing and filters the Teton Dam design was not acceptable, as compared with general practice in the industry. The great majority of experienced dam engineers, knowing the nature o f the rock, would have rejected the design as unacceptable: if the dam were built according to the design, serious damage or failure b y erosion of the Zone 1 material into the rock cracks would have been considered not only possible, b u t probable. (4) The absence o f design provisions for sealing or filtering the rock cracks, and subsequent failure to m o d i f y the design during construction, when the extreme conditions of rock cracks under Zone 1 were exposed clearly to view, can only be explained as a monumental error in judgment. This error was made possible b y long term bureaucratic restrictions on the activities of the dam design group which had severely limited their experience and capability. (5) The initial erosive leak could have been caused b y any of the several mechanisms cited as most likely b y the main investigators. (6) There are no technical "lessons learned" from the Teton Dam failure which make it desirable to consider any change in current dam design or construction practice. (7) The general lesson learned is the reconfirmation of the old fundamental rule that no important dam should be left wholly in the hands of one engineer or close team without independent review b y other specialist engineers with

241 the p o w e r of veto. This was expressed eloquently in the verdict of the Los Angeles County Jury following the even more disastrous failure of the St. Francis Dam, California, in 1928: " A sound policy of p u b l i c . . , and engineering judgment demands that the construction and operation of a dam should never be left solely to the judgment of one man no matter how eminent, without check b y independent expert authority, for no one is free from e r r o r . . . " In addition to these main points, study of the failure raises several technical details of considerable interest to the dam specialist, some of which are discussed in the remainder of this paper. These include: "Where did the initial leak start? .... What caused the wet seams?", and "Was the lack of instrumentation an important factor?", etc. While these points have large intrinsic technical interest, t h e y have no significant influence on the main conclusions presented above. Reasons for these main conclusions are presented in more detail in the following sections. ZONE 1 : MATERIAL PROPERTIES During the time that 1 was at the site investigating the failure in September 1976 1 was still actively involved in the research then being carried out by the U.S. Soil Conservation Service on dispersive clays and erodibility of finegrained impervious soils (Sherard et al., 1976a, b; ASTM, 1977). I t o o k ten large samples o f the typical loessial soil used in the Teton Dam Zone 1 from various parts o f the dam remnants, left and right side, and the borrow area. These were thoroughly tested in the SCS National Soil Mechanics Laboratory, Lincoln, Nebraska, using all the main tests which had been developed for evaluation o f dispersive (highly erodible) fine-grained soils. Later these same samples were used as one of the soil types tested during the comprehensive research on filters for dams carried out b y the Soil Conservation Service during 1981--85 (Sherard and Dunnigan, 1985). The main results of these studies, pertinent to the Teton Dam, are as follows. (1) In gradation and visual appearance the Zone 1 material is remarkably uniform. All ten samples taken from different parts of the dam and borrow area had very similar gradation and plasticity limits. The soil deposit comprising the large volume borrow area had a smaller range of properties than is c o m m o n for the impervious sections of e m b a n k m e n t dams. (2) The Zone 1 material, silt of loessial origin ranging from slightly cohesive to cohesionless (plasticity index generally from 1 to 7), is a c o m m o n t y p e of soil in the midwestern USA. Many earth dams have been built with practically identical soil over wide geographic areas, such as large parts of western Nebraska, including some of the main USBR dams. (3) The material is not dispersive; that is, it does not have high content of dissolved sodium in the pore water, causing repulsive forces between clay particles. Nevertheless, the material is among the most erodible fine-grained soils in nature, because of its low plasticity and the high content of silt-sized particles which do not have high surface forces. Compacted specimens of the typical Teton Dam Zone 1 material erode in the pinhole test (Sherard et al.,

242

1976a) as readily as a highly dispersive clay, an unusual property for a nondisperswe soil. This means that a small concentrated leak with a velocity of only a few centimeters per second will erode the compacted material. (4) In filter tests on the typical material, sand filters with D,s size of 0.5 mm or smaller will prevent all erosion. Using sand filters with Dis size of 1.0 mm or larger, important erosion occurs in the filter tests using compacted base specimens of the typical Teton Zone 1 material. Since the diameter of critical pore channels in sand filters is roughly 10% o f the Dis size (Sherard et al., 1984), this shows that the typical Zone 1 material can be eroded and carried into rock cracks of widths only slightly larger than (0.1) (1.0) = 0.1 mm, and would easily be carried into cracks with widths of 0.2 m m or larger. (5) When compacted in the laboratory at water content near or slightly below Standard Proctor Optimum, the material is very stiff and brittle, compared to similarly compacted specimens of other fine-grained impervious soils. When well-compacted at or below o p t i m u m water content the material is highly dilative and has relatively high capillary stresses in the pore water, which actions give compacted specimens the appearance of a brittle, cemented material. (6) There are few impervious soil deposits in nature which are more uniform in visual appearance or have a smaller range in gradation and Atterberg Limits. Compared to borrow areas commonly used for impervious sections o f embankment dams, the Teton Zone 1 material was remarkably uniform in properties. In summary, the relevant properties of the Teton Dam Zone 1 material are: (1) highly erodible even b y small concentrated leak with low velocity; (2) very fine, and erodible material can be easily eroded into narrow open cracks in the rock foundation; (3) very rigid and brittle compacted embankment so that relatively small settlements in the lower part of the dam would be expected to generate significant transfer of stresses and arching of the portion o f the embankment above. " G R A V I T Y G R O U T I N G " OF THE RIGHT ABUTMENT SURFACE CRACKS

A major element of the Teton Dam story has to do with the sealing of open rock foundation cracks under Zone 1 on the right abutment. The wide open surface cracks were treated b y gravity grouting during the first part of the construction. However, this surface crack filling was abandoned near the location where the failure occurred (about Station 14+00). Subsequently the wide surface cracks under Zone 1 embankment were left open and untreated from about Station 14+00 to the right end o f the dam. These facts are documented in the report o f the U.S. Dept. of the Interior Group (1977), but they deserve more prominence than they have received. These facts support the conclusion that USBR bureaucratic restrictions had a major influence on the failure. When the excavation was made for the 70-ft. deep trench and the rock foundation surface was uncovered upstream and downstream of the trench

243 b y excavation of the colluvial soil overburden, many large cracks in the rock were exposed to view. These were c o m m o n l y several inches in width, frequently up to 1.0 ft. Some were open (empty), some were silt-filled and some partially filled. During construction o f the dam the USBR geologists made an excellent map, showing locations, widths and filling of these cracks, reproduced in fig.34 (and also p.D-101) of their report (op. cir. ). This map shows literally many dozens of wide open rock cracks exposed in the foundation excavation from Station 16+00 to the right end of the dam. These cracks in the foundation rock under the main Zone 1, many completely open, over several hundred feet of the dam length, were exposed for inspection b y aUparties for about 2 years before t h e y were covered b y the dam. Since there was no provision in the contract for sealing these surface cracks, the inspection forces devised a method of filling them b y "gravity" or "slurry" grouting in stages above the rising e m b a n k m e n t surface. This consisted of bringing in transit-mix concrete trucks filled with cement-water grout, and pouring the grout b y gravity into open cracks, working from the rising e m b a n k m e n t surface. No piping or grout pumps were used for this activity. This gravity grouting was evidently instigated b y the inspection staff and accepted b y the design group, although there was no written record or change order on the subject. A p a y m e n t item for miscellaneous concrete backfill was agreed with the Contractor. Approximately 1800 cu. yd. of grout were poured into cracks at about 265 individual locations on the right abutment, over an area of about 300 b y 500 ft. This gravity grouting was done in cracks in the excavated walls of the cutoff trench, as well as on the natural rock foundation upstream and downstream of the trench. The pours into individual open cracks ranged from less than 1.0 cu. yd. to more than 100 cu. yds., with many in the range from 5 to 30 cu. yrd. Locations and amounts of each pour are presented in the report (op. cir. ), fig.34, in tabular form (pp.G32--G39) and shown on the USBR geologists' "as built" map. When the embankment construction reached approximate El. 5200 on the right side, roughly at Station 14+00, this gravity grouting was abandoned. After this date no further sealing o f surface rock cracks on the right abutment was carried o u t during the remainder of the dam construction. During my site visit of September 1976 1 discussed this problem in detail with the responsible inspection staff, trying to understand how this vital piece o f the work could have been stopped. Later the Interior Dept. Review Group interviewed everyone involved about this point, and included summaries of the interviews in their main report (op. cit. ), pp.C-13 and C-19 and G99 to G102. Conflicting statements were obtained from the site inspection and Denver design staffs. The inspectors generally stated that the gravity grouting was stopped on orders from " a b o v e " even though there were still many open cracks in the foundation. The rephes b y the design staff members were not consistent. The record o f interview with Mr. Richard Bock contained the following: "Design was n o t involved in the decision to terminate the surface treatment at Elevation

244

5205. Mr. Bock did not know of the decision until after the failure. To Mr. Bock's knowledge, the geologists had no input into the problem." (Op. cir., p.C-18.) Later (p.G100) the designer's position was rephrased: "The designers did not make the decision to not treat the rock above approximately elevation 5200. The decision not to treat the rock above this elevation was dictated b y the rock conditions. Above this elevation, the openings were minimal, horizontally oriented, and rubble and silt infilled. This t y p e of rock and the joints were not conducive to gravity grouting." The latter statement is in complete disagreement with the inspectors' statements and with the USBR geologists' excellent "as built" map of the exposed bedrock surface, which shows large open cracks in the trench walls in the critical area of the failure, as follows (op. cir., fig.34): (1) In the upstream wall of the 70-ft. deep trench, there is a vertical crack shown at approximate Station 13+50, striking u p s t r e a m - d o w n s t r e a m , described as "0.5 to 1.0 ft. (wide) open with rubble and silt filling locally". (2) In the downstream wall of the trench directly opposite the above wide crack there are several vertical cracks shown with widths of 0.1 ft. striking in various directions and one larger, vertical crack striking about 45 ° with the dam axis described as " u p to 0.3 ft. (wide) open" at about Station 13+25. It is apparent from the documentation that the above wide, open cracks were not grouted. I have no d o u b t about this practice since in September 1976 I saw similar open cracks, nearby up to 0.5 ft. wide, in the trench walls being exposed as the e m b a n k m e n t section remaining in place on the right end o f the dam, which was being carefully excavated for exploration purposes. As far as I was able to understand the problem from personal discussion with the people involved and study of the documents, the inspection staff at the site decided themselves to stop the gravity grouting, and did not ask for approval from the design group. Probably the gravity grouting was stopped in 1975 because the e m b a n k m e n t construction was going very rapidly and delays created b y this gravity grouting,which was not in the contract, were an item o f significant controversy with the Contractor. THE BUREAUCRACYPROBLEM

Main elements of this experience are as follows. (1) Although it was realized during the design stage that the foundation rock had open cracks, a design was made in which a wide central Zone 1 o f highly erodible, fine soil was placed directly on the cracked rock with no design provision to prevent the fine soil from entering the cracks. (2) During construction the rock foundation surface with dozens of wide open cracks was exposed for inspection for about 2 years. (3) A decision was made to seal the larger surface cracks b y gravity grouting. (4) On the right abutment the gravity grouting was abandoned at approximate Station 14+00, after which the Zone 1 embankment was compacted

245 directly against rock with wide, open cracks, both in the trench and in the foundation upstream and downstream o f the trench. The decision to abandon the gravity grouting was surprising. The gravity grouting was not a very conservative m e t h o d of sealing the cracks but it was much better than doing nothing. I believe that there will be no strong argument with the opinion that it was completely unacceptable practice to stop the gravity grouting at Station 14. It is very difficult to conceive w h y the design staff did not insist on the continuation of the gravity grouting for the entire foundation. I believe that this blunder can only be explained as the long time result of bureaucratic restrictions on the USBR staff. The abandonment o f the gravity grouting could only have been permitted because the individuals in the design group who would have known to insist on the continuation o f the gravity grouting were separated from the decision. In the early 1950s I was employed as a young engineer by the USBR Earth Dams Section. In the period 1960--1975, I made periodic visits to the Earth Dam Section for the purpose of discussing problems and practice with the staff and also occasionally visited USBR dams under construction. Hence, I believe I have a reasonably reliable understanding of the USBR organization as it related to design and construction of earth dams at the time when the Teton Dam was being studied and built. While there were undoubtedly exceptions, I believe the following are generally valid. (1) The organization suffered from inbreeding. Engineers and geologists were hired at the junior level, and were trained by engineers who had been taught by USBR engineers. T h e y had little or no understanding of the structure and modes of practice of other engineering groups designing and building dams. (2) Travel to sites for design engineers was considered generally unnecessary and was frowned upon. The responsible design engineer for major USBR dams frequently never visited the site, either in the design stage or during construction. At periodic fairly regular intervals, usually following federal elections, all "unnecessary travel by federal employees" was banned by fiat from Washington. As a result, the designer was sheltered from problems and experience. (3) Consultants were not used, so that the designers were not exposed directly to the experiences of independent specialists. (4) Chimney drains and processed filters were not employed in USBR dams (except small toe drain filters). (5) Cooperation between the construction supervision staff and the design group was not encouraged and was held to a minimum. It was considered that the resident engineer in charge of the construction should have the ability to solve problems arising during construction. Frequent requests for assistance or advice from the design staff by the resident engineer would have had a negative influence on the resident engineer's advancement record. (6) In the design organization there was no qualified independent review group with veto power to challenge designs.

246

Decades of such bureaucratic restrictions and inbreeding led to the situation in which the Teton Dam was constructed according to a design which would have been considered unacceptable b y independent specialists. At the same time the general mode of practice prevented the recognition b y USBR designers during construction that there was a special problem at the site and that it was essential to do something besides the "gravity grouting" to assure that the erodible fine silt would not be carried into the open rock cracks. The practice allowed even the incredible decision b y the inspection staff to abandon the gravity grouting and allowed them to do it without the knowledge of the designers. This bureaucratic and organization problem has since been well recognized. Since the Teton Dam failure the USBR design organization has been completely changed. All the main bureaucratic problems listed above have been eliminated. They have a new design staff of experienced engineers and geologists recruited from outside the USBR at the senior level. Engineers travel frequently from the Denver design office to job sites. Independent consultants are used routinely. Chimney drains and processed filters are used. The organization includes a technical review staff, independent from the design group, with veto power over the design. T H E L O C A T I O N O F T H E I N I T I A L E R O S I V E L E A K IS N O T I M P O R T A N T

All investigators agreed that the failure was caused by progressive erosion of a concentrated leak which carried eroded Zone 1 material into open cracks in the rock foundation. It was also generally agreed that the design was inadequate because of the failure to provide for conservative rock surface treatment and filters. But there was a lot of speculation among different investigators on the probable path and cause of the initial erosive leak. A large part of the efforts of the two official review panels was devoted to examination o f the most likely path of the initial concentrated leak. This is also true for the efforts o f other investigators who presented opinions on the failure cause later (Seed and Duncan, 1981; Leonards and Davidson, 1984). Before the " w e t seams" were discovered there was general agreement that there were two principal alternate locations, both possible and likely, of the initial leak: (1) along the interface between the Zone 1 embankment and the rock b o t t o m o f the trench; and (2) directly through the e m b a n k m e n t in the trench, either through an open differential settlement crack or a crack opened b y hydraulic fracturing. For each of the above there were several subcategories considered possible. For example, for {1)above, the initial leak could have gone over or under the concrete cap, and it could have traveled at the rock embankment contact (possibly because of low compressive stress at the contact; i.e. hydraulic fracturing) or just below the contact in open rock cracks. In such a situation there is always a basic scientific interest in determing as well as possible the path o f the initial leak, and the mechanism which caused it to develop. From the standpoint of "lessons learned" from the failure to

247 guide future design practice, however, reliable knowledge of the origin and location of the initial erosive leak is of no significant importance, because: (1) the initial leak could easily have developed at any of the several paths considered likely b y the main investigating panels, as discussed in more detail below; and (2) regardless of the path or origin of the initial leak the most economical and reliable defense is the same; i.e., conservative rock surface treatment, filters or both. As an illustration for support of this point, let us assume a hypothetical situation in which b y exhaustive investigation it was proved conclusively that the initial leak leading to the Teton failure developed through an embankment layer which had been frozen during construction. The main lesson learned would not be that it is undesirable to include frozen layers in a dam. The main lesson is the same: the failure should and would have been avoided b y a design with conservative provisions to prevent eroded embankment material from entering the open rock cracks. Some investigators of the failure believed it was impossible for the initial erosive leak to have developed at the b o t t o m of the trench at the rock-e m b a n k m e n t interface, because the rock seen everywhere in the b o t t o m of the trench was essentially free of open cracks and there was no place for the eroded soil to be carried. I do not believe this was a valid conclusion because the Zone 1 soil was so fine that large quantities of eroded material could have been carried into cracks with open widths equal to a small fraction of a millimeter. Such cracks are difficult to identify with the naked eye and not normally considered important. Also at any point along the trench there could have been a much larger crack just below the trench b o t t o m , capable o f carrying away large quantities of eroded soil rapidly. At the time of the Teton Dam final design, about 12 to 15 years ago, it was not the c o m m o n practice for a designer to assume that a concentrated leak should be expected through the impervious section of an embankment dam. At that time there was already considerable evidence for the conclusion that concentrated leaks probably developed fairly c o m m o n l y b y hydraulic fracturing (Sherard et al., 1971; Sherard, 1977) but there was no widespread discussion of the p h e n o m e n o n in the profession. This situation has changed. There is now sufficient evidence available to conclude that concentrated leaks c o m m o n l y occur through the impervious sections of embankment dams b y hydraulic fracturing without being observed, even in dams which are not subjected to large differential settlement. I beheve that there is now essentially incontrovertible evidence that hydraulic fracturing probably develops unseen in most e m b a n k m e n t dams, as discussed in detail in Sherard (1985). Usually these leaks do not cause erosion, either because the velocity is t o o low or because the leaks discharge into an effective filter. In light o f this conclusion, based on the experience of many dams, it would be expected that a concentrated leak would probably develop through the Zone 1 in the narrow, steep-walled rock foundation trench, where there was obviously a very high stress transfer to be expected b y the natural hanging up of the Zone 1 backfill on the steep and incompressible trench walls:

248 the t e n d e n c y for stress transfer and hydraulic fracturing in this trench, and along the Zone 1--rock interface on the trench b o t t o m , was obviously much greater than that existing in m a n y dams which have had concentrated leaks by hydraulic fracturing (Sherard, 1985). I believe, therefore, that the Teton Dam failure must be considered another experience to support the general conclusion that concentrated leaks by hydraulic fracturing should be anticipated at least when there is more than a moderate t e n d e n c y for differential settlement and stress transfer. WOULD FAILURE HAVE OCCURRED WITHOUT AN INITIAL CONCENTRATED LEAK? There was no disagreement by the two official investigating panels that the failure must have resulted from the development of a concentrated leak. The time was probably too short and the Zone 1 e m b a n k m e n t too impervious for water to have seeped through the soil pores to the downstream edge and then progressively eroded back upstream by piping. I agree that in all probability there was at least a small initial concentrated leak for the same reason. From the standpoint of lessons learned it is necessary also to examine the question: " I f no concentrated leak developed when the reservoir was filled for the first time, would Teton Dam have failed?" It is impossible to form a strong opinion on this question. If the reservoir became filled without generating a concentrated leak anywhere through the Zone 1 embankment, it is easily possible that the dam would never have failed because the seepage discharge at the downstream exit point did not have sufficient energy to initiate backward erosion. But it is also quite possible that backward piping, starting from a small seepage discharge in one of the large open rock cracks in the downstream face of the foundation trench, would gradually lead to failure after several years. In this event, the final failure would be expected to have been just as sudden and disastrous as the 1976 failure. It is also possible that an initial erosive leak from hydraulic fracturing could have occurred during reservoir filling on some subsequent year, as the result o f continuing settlement and stress transfer, In any event, I believe t h a t few of our experienced colleagues, knowing the nature of the foundation and the way the dam was designed and built, would have been very surprised to learn t h a t it did not fail on the first reservoir filling, but failed some time later. A most fortunate aspect of the Teton Dam failure was that it started in the morning at a time when the USBR staff were still there to observe and take action. Because o f this it was possible to evacuate most of the people from the area downstream. If the dam had failed a year later in the middle of the night the failure consequences would have been very much more serious. WET SEAMS: HYPOTHESIS AND SIGNIFICANCE During the exploratory excavation made in the left end of the dam, " w e t seams" were encountered o f rather peculiar nature located near and above

249 the elevation of the construction surface during the 1974--75 winter shutd o w n (U.S. Dept. o f the Interior Group, 1980; Seed and Duncan, 1981). Various members o f the t w o official investigating panels suggested different possible hypotheses for the cause of these wet seams. These included the influence of a rainy period during construction, frost heave and ice segregation, and simply poorly compacted layers. Some of the official investigators had the opinion that the action that caused the wet seams probably was related to the failure and some were doubtful. I did not see the wet seams and am not convinced that I understand with certainty their origin. However, I was intrigued to read descriptions and see photographs o f free water "dribbling" o u t of these seams in the walls of newly excavated surfaces in the compacted Zone 1 embankment, and I did form definite opinions about this. The portion of the embankment comprising the main wet seam area was constructed in the Spring of 1975 and the wet seams were first discovered in October 1977 about 30 months later. This main wet seam area was roughly 200 feet below the t o p of dam, so that it was loaded under a pressure o f the order of 150 psi since the completion of the dam construction in November 1975, for about 23 months. It seems to me very strange that free water could dribble out, albeit in very small quantities, of a compacted layer in a silty e m b a n k m e n t which had been loaded for 23 months under the weight of about 200 feet of dam. During these months any conceivable amount of surplus water which should have been added to a given embankment layer during construction, such as b y rainfall penetration would have long since disappeared. It would have been pushed out b y consolidation and sucked up b y capillarity into the adjacent e m b a n k m e n t layers and there would have been no trace remaining. Therefore, I believe it is impossible that the wet seams could have been caused b y anything that happened during construction: the water had to come from the reservoir in 1976 after the water level raised above the wet seam location. The reservoir level reached the general elevation of the main wet seams (El. 5115) in December 1975 and rose to about El. 5300 on the day of failure (June 5, 1976). Therefore, the reservoir elevation was above the main wet seams for more than 5 months and reached a maximum height above the wet seams of the order of 185 ft. I believe it should be generally agreed that the wet seams could not have been caused simply b y water seeping into the embankment from the reservoir through the voids of the compacted Zone 1 embankment. Such seepage only increases the water content a few percent b y filling the portion of the voids in which a little air remains (most of the original air voids would have disappeared already because of the compression which has taken place under the load of the overlying embankment). The little added water caused b y general seepage, displacing air, could not possibly have accounted for free water dribbling o u t of the wet seams in October 1977, 15 months after the dam failure.

250

The only w a y which I can imagine a sufficiently large amount of water being injected into these wet seams to explain the observed p h e n o m e n o n is directly from the reservoir by hydraulic fracturing. In order to examine the mechanics b y which such hydraulic fracturing could occur, it is necessary to consider the relative settlement of various points at different times during construction. First, consider a point " c " inside the dam near the center o f the valley located directly over the b o t t o m o f the cutoff trench in the river alluvium, at about the initial riverbed elevation (El. 5030; Figs.1 and 2). Point " c " has about 100 ft. of compacted Zone 1 embankment under it and 300 ft. of dam above it. Assuming that the Zone 1 in the cutoff trench compresses about 5 or 6% of its height under the load of 300 ft. of dam, a reasonable compressibility for this material, a m o n u m e n t set during construction at point " c " would settle about 5 or 6 ft. before it reached equilibrium. This is an expectable estimated settlement. The cutoff trench was backfiUed to the river level b y the summer of 1974 and the embankment level was constructed to about El. 5130 b y November 1974 when work was stopped for the winter. Starting in the late spring of 1974, when dewatering of the cutoff trench excavation in the river alluvium was abandoned, the Zone 1 e m b a n k m e n t in the cutoff trench {below point

~

Primary arch prevents crest

fr0m settling whenPoint "c"

settles from Oct. 1975 to

..............

.

a

,

b. "

~

~ Scale

L_^

Fig.1. Longitudinal section (looking d o w n s t r e a m ) ing hypothesis for the origin of the " w e t seams".

f

Main "Wet seams"

~

A--A.

used to illustrate the hydraulic fractur-

Secondary arch assists in stress transfer

Alluvium

/'~"~/ ~ - ~ R o c k foundation OL~ I?0 200 ~00 Feet Scale

Fig.2. Section

Original riverbed " - - Rock foundation

Probable wet seams

(located in borings)

Original riverbed ~

Trench

251

" c " ) had ground water on b o t h sides and there was a small gradient from upstream to downstream, caused b y the cofferdam pushing the river into the diversion tunnel. This condition existed for about one year b y the time construction started again in May 1975. During this year the degree o f saturation of the submerged Zone 1 e m b a n k m e n t in the cutoff trench increased substantially and pore water pressures increased, probably approach: ing hydrostatic pressure distribution below point " c " b y May 1975. During May to October 1975 (6 months) the embankment was constructed from about El. 5130 to crest level (El. 5330); i.e., in about 6 months an additional 200 ft. of embankment weight was placed above point "c". Because o f this relatively rapid load application and the higher degree of saturation, probably a substantial part of the load added in 1975 to the Zone 1 e m b a n k m e n t in the cutoff trench below point " c " was carried temporarily b y "construction pore pressures". During the period of time between the end of construction (Nov. 1975) and the failure (June 1976), 7 months, the construction pore water pressures in Zone 1 in the trench progressively dissipates, causing progressive settlement of point "c". The magnitude o f the settlement of point " c " during this 7-month period caused b y the consolidation of the 100 ft. of Zone 1 below it could easily have been of the order of one to two feet. During this 7-month period, as point " c " continues to settle because of pore pressure dissipation, there is practically no continuing settlement of points " a " and " b " (Fig.l) and points " e " and " f " (Fig.2), because t h e y are underlain b y rock and relatively incompressible and pervious coarse alluvium, respectively. The continuing settlement of point " c " w i t h respect to these other four points at the same elevation creates differential settlement in two directions which must cause a substantial stress transfer within Zone 1 above point "c". Point " d " located about 100 ft. above point " c " settles less than point "c", primarily because the t o p of the dam above point " d " acts as a beam or arch in the longitudinal direction, a b o u t 200 ft. thick, spanning from rock abutment to rock abutment (Fig.l). There is also a significant secondary tendency of the t o p of the dam to arch in the upstream--downstream direction (Fig.2). Since point " c " settles more than point " d " during the winter and spring 1975--76, the distance between the two points increases and compressive stresses on horizontal planes between the two points are significantly reduced. Sometime during the spring o f 1976, such as March or April, as the reservoir level is rising, the water pressure acting on the upstream face of Zone 1 at the approximate elevation of the wet seams exceeds the reduced vertical stress acting on near horizontal planes in the e m b a n k m e n t and the water enters the e m b a n k m e n t in a thin concentrated leak b y hydraulic fracturing. This might have occurred, for example, when the reservoir level reached a b o u t El. 5200, or when there was about a water head of 85 ft., at the level o f the main wet seams located at El. 5115. At the time of the initial water penetration the water pressure is only slightly greater than the earth pressure on the plane of the crack, so that there is no large stress change or deformation

252 in the surrounding embankment, and the initial water-fiUed crack is very thin (a fraction o f a millimeter). In the subsequent weeks or months, as the reservoir continues rising, the water pressure in the thin crack increases, b y perhaps 70 ft. of head, or more, and the crack width is jacked open, possibly to a final width of many centimeters. Failure or erosion doesn't occur during this initial period of hydraulic fracture development, either because the crack does not extend all the way across the wide Zone 1 or because the leak discharges into the downstream Zone 2, which is a suitable filter. Hence, the water in the hydraulic fractures in the " w e t seams" during the weeks or months before the failure was not eroding the Zone 1: the water in the near horizontal cracks was only seeping o u t into the e m b a n k m e n t upward and downward. During this period the e m b a n k m e n t material forming the r o o f of the near horizontal water filled crack would inevitably collapse and pieces of softened e m b a n k m e n t would fall d o w n b y gravity into the water. Hence the crack width would be increased and the fracture would consist of loose pieces of softened e m b a n k m e n t surrounded with free water. Such near horizontal cracks filled with loose m u d d y material were caused b y hydraulic fracturing in the central core of sandy silt (cohesionless) of the Manicouagan 3 Dam, Quebec (Sherard, 1985). Under this hypothesis concentrated leaks of water could have been fed into the wet seems from the reservoir for weeks (or even for 2 or 3 months) before the failure. On the day of the failure the reservoir pressure dropped abruptly below the elevation of the wet seams (within a few hours). The free water in these near horizontal cracks would tend to be squeezed out of the interior portion of Zone 1 as the pressure dropped rapidly and the cracks closed. But the cracks would collapse erratically and close first at the exterior boundaries o f Zone 1, inevitably trapping a large portion of the free water in the cracks inside. By hydraulic fracturing action similar to the hypothesis described above concentrated leaks of water would have been fed into the wet seams for weeks or months before the failure and a substantial quantity of free water would have been trapped in the wet seams on the day of failure (June 5, 1976). This is the only w a y in which I can imagine that sufficient water could have been injected into the w e t seams to account for the fact that free water dribbled o u t o f the wet seams when they were uncovered in the exploratory excavation in October 1977, at depth of 200 ft. below the crest. Hence, while it is impossible to have a strong conviction o n t h e subject, I believe that hydraulic fracturing, not greatly different than described above, is the most likely origin of the wet seams. None o f the members of the two official investigating panels speculated that hydraulic fracturing m a y have caused the wet seams. I am sure also that m a n y readers of this paper will find the hypothesis difficult to accept, b u t I believe that the hydraulic fracturing hypothesis is consistent with the recent dam experience (Sherard, 1985} and is the only w a y to explain the free water seeping o u t of the wet seams in October 1977.

253

Before leaving the subject, I call the reader's attention to the experience o f the Wister Dam in Oklahoma, where I believe it has been demonstrated b e y o n d all reasonable d o u b t that a near horizontal leak, more than 700 ft. long, broke through a well constructed clay dam b y hydraulic fracturing (Sherard, 1985). The physical conditions at the Wister Dam were not more conducive to hydraulic fracturing than at Teton Dam. The Wister Dam settlements were not high and the clay embankment, compacted near Standard Proctor Optimum water content, was n o t particularly rigid or especially susceptible to arching and stress transfer. The settlements at Teton Dam were higher than at Wister Dam and the e m b a n k m e n t was very rigid and capable of arching and stress transfer. Everyone who had the opportunity to inspect the Teton site after failure was impressed with the near vertical cliff, 300 ft. high, of Zone 1 e m b a n k m e n t standing stable for months forming the left wall of the breach. It was apparent to me that the Zone 1 embankment had sufficient undrained strength and rigidity so that the 200 ft. thick t o p of the dam could span across the valley as shown in Fig.1. The fascinating series of photographs of the failure included a picture of an arch developed temporarily over the final erosion tunnel on the right abutment, at about 11.30 a.m. on the day of failure, just before the crest caved into the breach. There was some indication from the tests made later that the soil comprising the w e t seams had somewhat different properties than the average embankment material. Evidently the average tested wet seam material had a somewhat lower maximum density and higher o p t i m u m water content when tested in laboratory Proctor Compaction Tests. Permeability tests on compacted specimens indicated that the wet seam material may have had higher permeability than the average material. Also, mineralogical examination of two undisturbed samples (one cut from a wet seam and one from nearby ordinary embankment) made b y soil scientists of the U.S. Department o f Agriculture showed significantly different mineralogy (U.S. Dept. of the Interior Group, 1980, p.A18-4). I do not k n o w if these differences are real or only random differences resulting from averages of an inadequate number of samples. If, in fact, it is true that the wet seams were comprised of a significantly different soil mineral, the hydraulic fracturing hypothesis described above should still be valid; the lower density wet seam material probably offered a slightly easier entry point into the e m b a n k m e n t than the adjacent material for the hydraulic fracturing, because of the slightly higher permeability. MISCELLANEOUS POINTS

Left bank excavation The exploratory excavation on the left side was made for the primary purpose o f examing Zone 1 in the left bank trench. It was reasoned that the conditions on b o t h sides were very similar, so that it might be expected there could be some evidence uncovered on the left side of a leakage mechanism at w o r k that would help explain the failure.

254 The result o f the careful removal of all the embankment material from within the left bank trench was only to show nothing unusual. There was no indication found anywhere t h a t an initial erosive leak had started on the left side {in the trench). The discovery of the "wet seams" was a different matter, outside the trench backfill. This result led some investigators to the conclusion t h a t it was difficult to explain the development of the initial leak on the right side (Leonards and Davidson, 1984). They essentially concluded that since no evidence was found of incipient hydraulic fracturing or erosive leakage through the trench or along its b o t t o m on the left side, t h e n it was not likely such leaks developed o n the right side. This conclusion is not valid because it does not take into consideration the probability t h a t the time which elapsed between the development of the initial erosive concentrated leak and complete failure was only of the order of 24 hours, or perhaps 2 or 3 days at the most. If the failure had not occurred on June 5, 1976, on the right side, it could have occurred on the left side on June 10, 1976, when the reservoir was higher, as a result of an initial erosive leak which first appeared on June 8, 1976. The new left side leak could have been along the b o t t o m of the trench or directly through the Zone 1 embankment by hydraulic fracturing, and would probably have been caused by the higher water pressures. Therefore, the absence o f evidence of any kind of incipient erosion in the left bank trench means nothing.

Lack o f instrumentation The fact t h a t no instruments were provided inside the dam is not a significant point. The failure would have occurred in the same way at the same time regardless of the numbers and types of instruments which might have been installed. The lessons learned from the Teton Dam failure do not support the conclusion that more or better instruments are generally needed.

Rate o f reservoir filling The reservoir was filled at a somewhat faster rate than had originally been planned, and the main b o t t o m outlet was not ready for operation for use in controlling the reservoir level during the initial filling. These points received much scrutiny by the various investigators, with the general conclusion that t h e y had negligible influence on the failure. Conceivably if the reservoir had been filled much more slowly the initial erosive leak might not have developed, because the dam possibly would have had sufficient time to accommodate to the changing conditions w i t h o u t the development o f the leak. But even if this occurred, the risk of failure at later date,discussed above, would still have been high, and the consequences might have been much worse. Therefore, this experience cannot reasonably be used to support the general idea that more future restrictions are needed on rates o f first reservoir fillings.

255

Clearly our embankment dams should be designed to be completely safe, regardless of the rate of the first reservoir filling. SUMMARY

(1) There is no fundamental technical "lesson learned" from the Teton Dam failure which requires changes in current practice for dam design and construction. (2) The Teton Dam design was unacceptable compared with common general practice in the industry because it did not provide for the needed conservative rock surface crack sealing and filters. (3) The circumstances which permitted the Teton Dam to be built according to an inadequate design were created b y long-time USBR bureaucratic restrictions. These restrictions limited the capability of the USBR design and construction control staffs sufficiently so that t h e y did not recognize the inadequacy of the design. (The USBR organization has subsequently been completely changed to eliminate the problem.) (4) Failure was caused when an initial erosive concentrated leak developed through the erodible fine silt Zone 1 embankment, and progressively carried the eroded material into open rock cracks in the foundation. (5) There are two possible general locations along which the initial leak is most likely to have developed: (a) directly through the Zone 1 b y hydraulic fracturing, or (b) along the embankment rock interface at the b o t t o m of the trench. Both o f these causes are likely and leaks could have been expected from them; therefore, knowledge of the actual origin of the initial leak has no importance to the evaluation of the lessons learned from the failure. (6) Even if an initial erosive concentrated leak had not developed during the first reservoir filling, there is a strong probability that the dam would have failed later in about the same way. (7) The main general lesson learned is the reconfirmation of the old fundamental rule, that is, no important dam should be designed and built under direction of one man or close team without independent review b y other specialist engineers with veto power. If the Teton Dam design had been reviewed b y completely independent specialists there is almost no conceivable possibility that the design would have been accepted or that the failure would have occurred. (8) The origin for the wet seams considered most likely is near horizontal water-filled cracks which were jacked open to widths of many centimeters b y hydraulic fracturing during 1976 when the reservoir level was above the elevations of the wet seams. No other action conceivable to me can explain the existence of free water dribbling out of the wet seams when exposed in October 1977. (9) Whatever the origin of the wet seams, it is clear that a concentrated leak occurred through the impervious section of the dam on the right abutment, which was well constructed and founded on a relatively incompressible bedrock. This is another experience among many which supports the

256 c o n c l u s i o n (Sherard, 1 9 8 5 ) t h a t c o n c e n t r a t e d leaks o c c u r c o m m o n l y in wellc o n s t r u c t e d e m b a n k m e n t d a m s w i t h o u t being o b s e r v e d . REFERENCES A S T M , 1977. Dispersive Clays, Related Piping, and Erosion in Geotechnical Problems, A S T M Spec. Tech. Publ., 623, 486 pp. Independent Panel to Review Cause of Teton D a m Failure, 1976. Report on Failure of Teton Dam. U.S. Govt. Printing Office, Washington, D.C. 20402. Leonards, G.A. and Davidson, L.W., 1984. Reconsideration of failure initiatingmechanisms for Teton Dam. Proc. Int. Conf. Case Histories in Geotechnical Engineering, University of Missouri--Rolla, Vol. III,pp.1103--1113. Seed, H.B. and Duncan, J. M., 1981. The Teton D a m failure-- a retrospective review. Proc. 10th ICSMFE, Stockholm, Voh 4, pp.214--238. Sherard, J.L., 1977. Discussion of"Load transferand hydraulic fracturing in zoned dams", by G.H. Kulhawy and T.M. Gurtowsky. J. Geotech. Eng. Div., ASCE, July 1977,

pp.831--833. Sherard, J.L., 1985. Hydraulic fracturing in embankment dams. Proc., ASCE Symp. Seepage and Leakage from Dams and Impoundments, Denver, May 1985, pp.115--141. Sherard, J.L. and Dunnigan, L.P., 1985. Filters and leakage control in embankment dams. Proc., ASCE Symposium Seepage and Leakage from Dams and Impoundments, Denver, May 1985, pp.1--30. Sherard, J.L., Decker, R.S. and Ryker, N.L., 1972. Hydraulic fracturing in low dams of dispersive clay. Proc., Specialty Conf. Performance of Earth and Earth-Supported Structures, Purdue University, ASCE, June 1972, Part I, pp.563--590. Sherard, J.L., Dunnigan, L.P., Decker, R.S. and Steele, E.F., 1976a. Pinhole test for identifying dispersive soils. J. Geotech. Eng. Div., ASCE, January 1976, pp.69--85. Sherard, J. L., Dunnigan, L. P. and Decker, R.S., 1976b. Identification and nature of dispersive soils. J. Geotech. Eng. Div., ASCE, April 1976, pp.287--301. Sherard, J.L., Dunnigan, L.P. and Talbot, J.R., 1984. Basic properties of sand and gravel filters. J. Geotech. Eng. Div., ASCE, June 1984, pp.684--700. U.S. Dept. of Interior Teton Dam Failure Review Group, 1977. Failure of Teton Dam. A Report of Findings. U.S. Govt. Printing Office, Washington, D.C., 20402. U.S. Dept. of Interior Teton Dam Failure Review Group, 1980. Failure of Teton Dam, Final Report. U.S. Govt. Printing Office, Washington, D.C. 20402.