Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
4 History of Development of Present Practices DESIGN CRITERIA FOR EXTREME FLOODS Engineers designing dams have been and continue to be handicapped by lack of reliable bases for estimates of future extreme flood runoffs from the basins upstream from their dams. Myers (1967) has described the evolution in the United States of the cur- rent practices for estimating extreme flood-producing capabilities of water- sheds for use in spillway design. The main developments in that evolution took place in a 30-year period, from about 1910 until 1940. The central concepts in the current methods of estimating spillway requirements have changed little in the last few decades, but there have been many refinements in their application. Also, in the last 40 years, the concepts and methodology developed in the United States have become fairly well established through- out the world as the standard approach for sizing spillways of large dams where failure of the dams would result in hazards to life and property. As noted by Myers, the evolution in the United States of techniques for determining spillway capacity requirements can be divided into four stages, outlined below. Early Period Before 1900 the designer of a dam in the United States usually had little hydrologic data on which to base estimates of spillway requirements. The rainfall reporting system of the U.S. Weather Bureau was largely a widely 35
36 SAFETY OF DAMS scattered system of nonrecording gages that reported daily rainfall amounts primarily in the interest of agriculture. Systematic collection of stream stage data was being accomplished at a relatively small number of gages on major streams. Often the designer of a dam had little information except high- water marks on the stream he was damming or in adjacent watersheds to indicate the flood potentials at his dam site. The designer might estimate peak discharge rates of past floods based on such meager data and base his spillway design on such estimates with, perhaps, some added capacity pro- vided as a safety factor. Some spillways were designed for some multiple of the maximum known flood, e.g., twice the maximum known flood. Myers noted a tendency among engineers of this era to assume that nature had already demonstrated the maximum flood potential on each stream and that spillway designs could be based with safety on the records available of past floods. Some earth dams built during this period failed because of overtop- ping, but since most dams of the era were relatively low masonry or rock- filled timber-crib structures, overtopping could be experienced by many structures without failure. Regional Discharge Period As more stream discharge data became available, engineers began to recognize that looking at all past flood peaks in a region might give more reliable estimates of maximum flood-producing potentials than a limited record on a single stream. This idea recognized the random-chance nature of major storm rainfalls occurring over a specific watershed. It also introduced the concept that hydrologic data observed at one location, when appropri- ately transposed, could serve as a basis for estimates at other locations. A flood-frequency formula developed by W. E. Fuller (1914), based on analy- sis of flood peaks in hundreds of streams, sought to relate the peak discharges having various average return periods to the mean of the highest annual floods from the same watershed. Later formulas, notably one known as "Myers" rating and others developed by W. P. Creager and C. S. Jarvis, based on enveloping the maximum observed floods from many watersheds, sought to relate the "maximum flood" from a watershed to some function of the size of the area drained. Prior to about 1940, many spillway designs for large dams were based on such formulas developed from regional analyses of maximum flood peaks. Such formulas are still used to compare observed and computed flood peaks. Myers noted that, about the year 1930, formal statistical methods began to be applied to hydrologic problems, including estimating maximum floods. Some engineers felt that, given a reasonably long period of discharge record and the right type of frequency analysis, reliable estimates of rare
History of Present Practices 37 flood peaks could be obtained for determining spillway design require- ments. However, the occurrence of floods far exceeding any that could have been estimated by frequency analysis of past records of basin discharges has discouraged the use of frequency analysis as a tool for estimating the magni- tude of rare floods. With the longest flood records in the United States extending beck about 150 years, most American hydrologists are reluctant to give estimates of stages or discharges for floods with long average return periods. However, in recent years the National Flood Insurance Program has required numerous estimates of floods having annual probabilities of ex- ceedance of 0.01, or average return period of 100 years for the purpose of defining floodplains in connection with community planning activities and administration of the flood insurance program. Also, some state dam safety agencies call for similar floods for design of spillways for small, low-hazard dams. In contrast to American practice, the guide issued by the Institution of Civil Engineers in London calls for estimating 10,000-year floods in spillway design. Storm Transposition Period In the 1920s engineers in the Miami Conservancy District in Ohio under- took a program of studying past major flood-producing storms to develop rainfall duration-area-depth relationships for use in planning and design of the comprehensive flood control project for the Miami Valley. It was recog- nized that measured flood peaks are dependent on topography and size of individual watersheds and chance placement of storm centers over the wa- tershed. Also, within meteorologically similar areas, observed maximum rainfall values could provide a better general indication of maximum flood potentials than flood discharges from individual watersheds. In the 1930s, the work that had begun in the Miami Conservancy District was used by other engineers, notably those in the Tennessee Valley Authority and the U. S. Army Corps of Engineers, to develop a system for estimating what was termed "maximum possible" rainfalls. The independent development of the unit hydrograph concept in the early 1930s gave a way of converting the maximum possible rainfall values into streamflow to produce maximum possible spillway design floods. The method of transposing and enveloping past maximum storm rainfalls to obtain maximum possible estimates was illustrated graphically in some U.S. Army Corps of Engineers reports of the period by what was termed a "peg" model. Suppose it was desired to determine the maximum 2-day rainfall over a 500-square-mile drainage area above a proposed dam site. From duration-area-depth data available for past major storms in the mete- orologically similar area, the maximum rainfall depth over 500 square miles
38 SAFETY OF DAMS and 2-day duration for each relevant storm was determined. Then, on a map of the region, a vertical wooden peg whose height represented, to some convenient scale, the 500-square-mile 2-day rainfall value for each storm was placed at the geographic location where the rainfall was observed. Strings were stretched connecting the tops of pegs representing individual storms at their respective locations. The height of the network of strings over the map location of the basin at the proposed dam site represented the estimated maximum possible rainfall over the basin. Probable Maximum Precipitation (PMP) Period By the end of the 1930s it was recognizes] that the enveloping and transpos- ing of past major storms might not necessarily yield the upper limit of proba- ble rainfall over a basin. Concepts of air mass analysis then being introduced showed that in any storm system such factors as the humidity of the incoming air, the velocity of wind bringing moisture-laclen air into the basin, ant] the percent of the water vapor that can be precipitates] serve to place limits on the amount of precipitation from the storm. Thus, if in any major storm of record, any of the limiting factors was less than what could be expected in the basin being studied, an adjustment (i.e., increase) in the observed rainfall values would be inclicated if maximum probable rainfall values were sought. Such transposition, adjustment, and enveloping techniques are the bases for the PMP estimates now in use and which are more fully discussed in Chapter 5 and Appendix C. Composite Criteria From the summary of present practices in Chapter 3, it will be recognized that both PMP estimates and estimates baser] on frequency analysis of either rainfalls or streamflows are in use for determining sizes of spillways. Some agencies specify various percentages of the PMP or the probable maximum flood (PMF) derived from the PMP as bases for spillway design. Other criteria attempt to combine flow frequency concepts and PMF concepts. Considering the great differences in the basic concepts involved, the ration- ale for such combinations seems questionable. Risk-Based Analyses Other bases for determining spillway capacity requirements have been suggested. A 1973 report of an American Society of Civil Engineers (ASCE) committee advocated that spillway capacities of existing dams be reevalu- ated by economic analyses of costs of spillways of various capacities and the
History of Present Practices estimated long-term damages (or risk costs) associated with each spillway capacity (ASCE, 1973~. The committee proposed that costs associated with loss of life be evaluated on the basis of current practices of courts in awarding damages in cases involving accidental deaths. The spillway capacity selectee] would be that which gave lowest total costs (i.e., the lowest sum of risk costs plus structure costs) . The ASCE committee's proposals have not been widely adopted. It appears that the following factors have contributed to the lack of enthusiasm for such an approach: 39 · There is a great reluctance to attempt to place a value on a human life in such a decision process. · The results of such economic analysis of present-clay costs versus possi- ble future damages are very much dependent upon the interest rate selected for the analysis, and many engineers are reluctant to have design involving safety determined by the interest rate that happened to be applicable when the design was done. (This factor has less importance for governmental agencies in which the discount rate to be used for public sector benefit-cost studies is prescribed (based on some chosen social discount rate) and which is permitted to change very slowly from year to year) . · Estimates of average annual risk costs are dependent on the flood fre- quency curve acloptecI, and the persistent problem of estimating frequencies of rare flood events becomes a deterrent. As discussed in Chapter 5, another risk-based approach to determining spillway capacity that has been used involves estimating, by dam break analysis and flood toutings, the flood flow for which failure of the dam would] cease to create significant additional damage downstream over dam- age sustained without failure. That flood then becomes the design discharge for the spillway. This approach has merit because it avoids unneeded expense for added spillway capacity and does not worsen the situation downstream in case of failure. Other risk-base`] analysis methods similar to the method advocated by the 1973 ASCE committee report, but which do not attempt to place a value on human life, are being advocated in those agencies attempting to apply risk analysis techniques to decisions regarding dam safety. In such analyses, potentials for deaths resulting from dam failure become an element in the decision process but not part of the economic evaluation. EARTHQUAKE-RESISTANT DESIGN OF DAMS The possible effects of earthquakes on the safety of dams were first taken into account by the engineering profession as early as the midcIle 1920s. The
40 SAFETY OF DAMS classic paper by H. M. Westergaard entitled "Water Pressure on Dams Dur- ing Earthquakes" was published in 1933. There is every indication that many design agencies were at that time making some analytical studies to evaluate seismic safety, or incorporating simple defensive measures into design of projects to increase the safety of dams against earthquake-shaking effects. In the period 1930-1970, design practice usually considered earthquake effects by simply incorporating in the stability or stress analysis for a dam a static lateral force intended to represent the inertia force induced by the earthquake. Most often this force was expressed as the product of a lateral force coefficient and the force of gravity. The coefficient generally varied, depending on the seismicity of the area in which the dam was located and the judgment of the engineer involved, between values of about O.OS and 0.15. This method of approach was termed the pseudostatic analysis method, in recognition of the fact that the static lateral forces were only intended to represent the effects of the actual dynamic earthquake forces, which effects could well be substantially different from those of the static forces used in the analysis procedure. This approach was essentially similar to that used for building design in seismically active areas. For concrete dams it was customary to consider also the hydrodynamic pressures on the face of the dam using the approximate method proposed by Westergaard. For earth dams, studies by Zangar (1952) led to the conclusion that consideration of hydrodynamic pressures was generally unnecessary but might be required for dams with relatively steep slopes. Pseudostatic analyses, combined with engineering judgment, were the only methods used to assess the seismic stability of dams until the late 1960s, and these approaches were generally considered entirely adequate by dam engineers and regulatory agencies based on the fact that the field perfor- mance of dams designed by these procedures and subsequently subjected to strong earthquake shaking was generally found to be satisfactory. The ob- served performances lending support to this belief included: · The excellent performance of dams located close to the San Andreas fault in the 1906 San Francisco earthquake (M8.3) . At the time of that event there were 14 earth dams located within 5 miles of the fault on which the earthquake occurred, and none of these suffered any significant damage. There was also one concrete dam, the Crystal Springs Dam, located only a few hundred yards from the San Andreas fault that survived the earthquake with no apparent damage. · Several 200-foot-high concrete gravity dams in Japan were also sub- jected to earthquake-shaking intensities of M 8 with no damage.
History of Present Practices 41 · The Hebgen Dam, an earth dam located in Montana only a few hun- dred yards from the fault of the Hebgen Lake earthquake of 1959 (M 7.1), survived the shaking without failing. The adequate performance of these structures under strong shaking gave the profession a high level of confidence in the design procedures in use at the time. In the 1960s and early 1970s a number of events occurred that caused engineers to reevaluate the adequacy of this approach: · The observation was made that the pseudostatic method of analysis would not adequately predict the slope failures that occurred at many places in Alaska in the 1964 Alaska earthquake (M 8.3~. · Major cracking occurred in the Koyna Dam, in India, a 340-foot-high concrete gravity dam, subjected to a near-field M 6.4 earthquake in 1967. · Major cracking developed near the top of the Hsingfengkiang Dam, a 345-foot-high concrete buttress dam in China, as a result of a near-field M 6.1 earthquake in 1962. · A near failure occurred at the Lower Van Norman (San Fernando) Dam, and significant sliding occurred in the Upper Van Norman (San Fernando) Dam as a result of the San Fernando earthquake of 1971. In spite of the fact that both of these dams were judged to be adequately safe on the basis of pseudostatic analyses made before the earthquake, their perfor- mance cast doubt on the validity of this approach. · Accelograph records showed peak accelerations during earthquake shaking greater than 0.3 g. These events, and others, led to an increasing concern that the pseudo- static method of analysis could not always predict the safety of dams against earthquake shaking. At about the same time, new tools for making improved analyses of seismic response had become available (finite element methods and high-speed computers), ant] investigations were made of the ability of dynamic response analyses to provide better insights into probable field performance. The results of these studies were extremely encouraging and sufficiently convincing that by 1973 three major changes occurred. 1. In the late 1960s the California Department of Water Resources adopted methods of dynamic analysis for design of the State Water Project. 2. The Division of Safety of Dams of the California Department of Water Resources decided to require owners of many earth dams to reevaluate the seismic stability of the dams using dynamic analysis methods, regardless of the results indicated by pseudostatic methods of analyses.
42 SAFETY OF DAMS 3. The Earthquake Committee of the International Commission on Large Dams (ICOLD, 1975) recommended that: As regards low dams in remote areas, they may be designed by the conventional (pseudostatic ~ method for any type of dam. As regards high gravity or arch dams or embankment dams whose failure may cause loss-of-life or major damage, they should be designed by the conventional method at first and they should then be subjected to dynamic analyses in order to investigate the deficiencies which might be involved in the seismic design of the dam in case the conventional method is used. These requirements ant] recommendations led to a major review of design procedures by many design agencies, as a result of which increasing empha- sis was given to the use of dynamic analysis methods where appropriate. Thus, among federal agencies in the United States, methods of analysis currently (1984) in use are generally as follows. Concrete Dams · defensive design measures (studies to ensure foundation and abutment integrity, good geometrical configuration, effective quality control, etc.~; · pseudostatic analysis methods, commonly used to check sliding and overturning stability for gravity or buttress dams and in areas of low seismic- ity; and · dynamic analysis methods (assuming the dam to consist of linear elas- tic, homogeneous, isotropic material) to determine the structural response and induced stresses in areas of significant seismicity. Earth Dams etc.~; · defensive design measures (ample freeboard, wide transition zones, · for reasonably well-built dams on stable foundation soils, no analysis required if peak ground accelerations are less than about 0.2 g; · a dynamic deformation analysis for clams constructed of or on soils which do not lose strength as a result of earthquake shaking and located in areas where peak ground accelerations may exceed about 0.2g; and · a dynamic analysis for liquefaction potential or strain potential for dams involving embankment or foundation soils that may lose a significant portion of their strengths under the effects of earthquake shaking. Private engineering companies follow generally similar practice. These procedures necessarily require an evaluation of the seismic exposure of the dam and, in cases where dynamic analysis procedures must be used, a deter-
History of Present Practices mination of the earthquake motions (either in the form of a response spec- trum or in the form of time histories of accelerations) for which the safety of the dam must be evaluated and for which the safety of the dam must be assured. The selection of these motions involves considerations of seismic geology, seismicity, probability of occurrence, consequences of failure or damage, and public safety. 43
