the Inieriot’ when of Ground Motion .‘____"_‘W, , V,_ 7.”) I aper 15524)” Availability of Publications of the US. Geological Survey Order US. Geological Survey (USGS) publications from the offices listed below. Detailed ordering instructions, along with prices (if the last offerings, are given in the current-year issues ' of the catalog “New Publications of the US. Geological Survey.” Books, Maps, and Other Publications ' By Mail Books, maps, and other publications are available by mail from—: USGS Information Services Box 25286, Federal Center Denver, CO 80225 Publications include Professional Papers, Bulletins, Water- Supply Papers, Techniques of Water-Resources Investigations, Circulars, Fact Sheets, publications of general interest, single copies of permanent USGS catalogs, and topographic and thematic maps. Over the Counter Books, maps, and other publications of the US. Geological Survey are available over the counter at the following USGS Earth Science Information Centers (ESIC’s), all of which are authorized agents of the Superintendent of Documents: 0 Anchorage, Alaska—Rm. 101, 4230 University Dr. - Denver, Colorado—Bldg. 810, Federal Center - Menlo Park, California—Rm. 3128, Bldg. 3, 345 Middlefield Rd. 0 Reston, Virginia—Rm. 1C402, USGS National Center, 12201 Sunrise Valley Dr. , - Salt Lake City, Utah—2222 West, 2300 South (books and maps available for inspection only) 0 Spokane, Washington—Rm. 135, US. Post Office Building, 904 West Riverside Ave. - Washington, D.C.——Rm. 2650, Main Interior Bldg, 18th and C Sts., NW. Maps only may be purchased over the counter at the following USGS office: - Rolla, Missouri—1400 Independence Rd. Electronically Some USGS publications, including the catalog “New Publica— tions of the US. Geological Survey” are also available elec— tronically on the USGS’s World Wide Web home page at http://www.usgs.gov Preliminary Determination of Epicenters Subscriptions to the periodical “Preliminary Determination of Epicenters” can be obtained only from the Superintendent of Documents. Check or money order must be payable to the ' Superintendent of Documents. Order by mail from— Superintendent of Documents 1 Government Printing Office Washington, DC 20402 Information Periodicals 1 Many Information Periodicals products are available through the systems or formats listed below: Printed Products Printed copies of the Minerals Yearbook and the Mineral Com- modity Summaries can be ordered from the Superintendent of Documents, Government Printing Office (address above). Printed copies of Metal Industry Indicators and Mineral Indus- try Surveys can be ordered from the Center for Disease Control and Prevention, National Institute for Occupational Safety and Health, Pittsburgh Research Center, PO. Box 18070, Pitts— burgh, PA 15236—0070. Mines FaxBack: Return fax service 1. Use the touch—tone handset attached to your fax machine’s telephone jack. (ISDN [digital] telephones cannot be used with fax machines.) 2. Dial (703) 648—4999. 3. Listen to the menu options and punch in the number of your selection, using the touch-tone telephone. 4. After completing your selection, press the start button on your fax machine. CD-ROM A disc containing chapters of the Minerals Yearbook (1993— 95), the Mineral Commodity Summaries (1995—97), a statisti- cal compendium (1970—90), and other publications is updated three times a year and sold by the Superintendent of Docu— ments, Government Printing Office (address above). World Wide Web Minerals information is available electronically at http://minerals.enusgs.gov/minerals/ Subscription to the catalog “New Publications of the US. Geological Survey” Those wishing to be placed on a free subscription list for the catalog “New Publications of the US. Geological Survey” should write to— U.S. Geological Survey 903 National Center Reston, VA 20192 The Loma Prieta, California, Earthquake of October 17, 1989—Earth Structures and Engineering Characterization of Ground Motion THOMAS L. HOLZER, Editor PERFORMANCE OF THE BUILT ENVIRONMENT THOMAS L. HOLZER, Coordinator U.S. GEOLOGICAL SURVEY PROFESSIONAL PAPER 1552—D Prepared in cooperation with the National Science Foundation UNITED STATES GOVERNMENT PRINTING OFFICE, WASHINGTON : 1998 DEPARTMENT OF THE INTERIOR BRUCE BABBITT, Secretary U.S. GEOLOGICAL SURVEY Charles G. Groat, Director Any use of trade, product, or firm names in this publication is for descriptive purposes only and does not imply endorsement by the US. Government. Text and illustrations edited by Jeff Troll Manuscript approved for publication, January 24, 1997 Library of Congress catalog-card No. 92-32287 For sale by US. Geological Survey, Map Distribution Box 25286, MS 306, Federal Center Denver, CO 80225 @575 /7é M $79.71.: D E%RT” CONTENTS Page Introduction ----------------------------------------------------------------------- D1 By Thomas L. Holzer Performance of earth dams during the Lorna Prieta earthquake ----------- 3 By L.F. Harder, J .D. Bray, R.L. Volpe, and K.V. Rodda Analysis of soil-nailed excavations stability during the 1989 Lorna Prieta earthquake --------------------------------------------------------------- 27 By Mladen Vucetic, Mark R. Tufenkjian, Guy Y. Felio, Pirooz Barar and K. Ronald Chapman Empirical analysis of peak horizontal acceleration, peak horizontal velocity, and modified mercalli intensity ----------------------------------- 47 By Kenneth W. Campbell Attenuation of vertical and horizontal response spectra of the Lorna Prieta earthquake --------------------------------------------------------------- 69 By Yousef Bozorgnia and Mansour Niazi III THE LOMA PRIETA, CALIFORNIA, EARTHQUAKE OF OCTOBER 17, 1989: PERFORMANCE OF THE BUILT ENVIRONMENT EARTH STRUCTURES AND ENGINEERING CHARACTERIZATION OF GROUND MOTION INTRODUCTION By Thomas L. Holzer, US. Geological Survey This chapter contains two papers that summarize the per- formance of engineered earth structures—dams and stabi— lized excavations in soil—and two papers that characterize for engineering purposes the attenuation of ground motion with distance during the Lorna Prieta earthquake. Document- ing the field performance of engineered structures and con- firming empirically based predictions of ground motion are critical for safe and cost effective seismic design of future structures as well as the retrofitting of existing ones. Catastrophic flooding caused by dam failure is a persis— tent concern in areas with earthquake potential. Harder and others (this chapter) present a valuable compilation of the performance of earth dams during the 1989 earthquake. They estimate that 111 dams, the majority of which were homog- enous earth dams, were within 80 km of the seismogenic rupture. Thirty-six of these dams experienced free-field peak horizontal ground accelerations greater than 0.2 g, but only one major dam—Austrian—and one minor dam—Soda Lake—suffered moderate damage. The authors conclude that low reservoir levels, which reduced saturation of embank- ments, was a major contributor to the good performance of these dams. The field performance and strong-motion data provide an opportunity to learn more about how earth and rockfill dams behave during earthquakes at low reservoir levels. Failures of embankments in soils during earthquakes both disrupt traffic and damage nearby buildings. Vucetic and others (this chapter) evaluate the performance of nine exca- vations that were stabilized by soil nailing, an in—situ tech- nique used to reinforce existing soil masses during excava- tion by horizontally drilling and installing passive inclusions. The inclusions, called nails, cause the soil to behave as a composite unit. The 1989 earthquake provided the first seis- mic loading of earth structures of this type and thus presents a special opportunity to evaluate the performance of nailed walls. None of the nine soil-nailed excavations evaluated by the authors showed any signs of distress or movement even though one in the epicentral regional was strongly shaken. - The authors conclude that conservative design and construc- tion are the primary reasons for their good performance. The most common methods for estimating ground motion in future earthquakes in California are empirical and rely on syntheses of observations from previous earthquakes. Pa- pers by Campbell (this chapter) and Bozorgnia and Niazi (this chapter) build on this tradition. For further discussion of other aspects of strong ground motion during the earth- quake, the reader is referred to the 16 papers in Borcherdt (1994) as well as selected papers in Spudich (1996). Campbell (this chapter) examines the dependence of re- corded peak horizontal acceleration, peak horizontal veloc- ity, and Modified Mercalli Intensity on distance from the seismogenic rupture zone, source—to-site azimuth, and site geology. He develops empirical fits to these parameters. He concludes that peak accelerations recorded on alluvium at sites more than 50 km from the seismogenic rupture zone were significantly higher than would have been predicted by existing attenuation relations available at the time of the earthquake. He also notes a strong directional or azimuthal dependence of ground motion and intensity. Significantly higher amplitudes corresponded to azimuths pointed at Santa Cruz, San Francisco, and Oakland than at azimuths to the northeast and east. Local site geology strongly affected ground motion with sites underlain by Holocene San Fran— cisco bay mud typically recording higher accelerations. Bozorgnia and Niazi (this chapter) study the frequency dependence of the attenuation of ground motion by examin- ing vertical and horizontal response spectra for spectral or- dinates at periods ranging from 0.05 to 2 s. They observe that the shape of the response spectra for both vertical and horizontal components of ground motion and their ratio de— pends on both distance and magnitude. The vertical-to-hori- zontal spectral ratio at higher frequencies is substantially higher than the two-thirds value commonly used in engi- neering practice. The ratio is highest near the earthquake source. However, at low frequencies and farther from the earthquake source, the spectral ratio decreases significantly, which implies current design practice is conservative. REFERENCES CITED Borcherdt, R.B., ed., 1994, The Loma Prieta, California, earthquake of October 17, 1989—Strong ground motion: U.S. Geologi- cal Professional Paper 1551—A, 272 p. Spudich, RA., ed., 1996, The Loma Prieta, California, earthquake of October 17, 1989—Main—shock characteristics: U.S. Geo- logical Professional Paper 1550-A, 297 p. D] / 1 . , . . ‘ y, . ;/, , , ‘ . . ,_ . x /. V on.» a H; THE LOMA PRIETA, CALIFORNIA, EARTHQUAKE OF OCTOBER 17, 1989: PERFORMANCE OF THE BUILT ENVIRONMENT EARTH STRUCTURES AND ENGINEERING CHARACTERIZATION OF GROUND MOTION PERFORMANCE OF EARTH DAMS DURING THE LOMA PRIETA EARTHQUAKE By L.F. Harder, Jr., California Department of Water Resources, Sacramento, Calif; J .D. Bray, University of California, Berkeley, Calif; R.L. Volpe, R.L. Volpe & Associates, Los Gatos, Calif; and K.V. Rodda, Wahler Associates, Palo Alto, Calif. CONTENTS Page Abstract D3 Introduction 3 Characteristics of earth dams considered in this study ——————————————— 4 Pre-1906 earth darn“ 4 Hydraulic fill dams 6 Rockfill dams 6 Selected case histories of performance --------------------------------- 6 Austrian Dam 8 Lexington Dam l3 Guadalupe Dam 13 Newell Dam 15 Elmer J. Chesbro Dam --------------------------------------------- 15 Vasona Percolation Dam ------------------------------------------- 17 Soda Lake Dam 17 Mill Creek Dam 18 Leroy Anderson Dam 18 San Justo Darn 21 Hawkins Darn 21 Performance of dams at greater distances ------------------------------ 21 Calaveras Dam 21 San Luis Dani 21 O’Neill Dam 23 Costs of major repairs for earthquake damage ------------------------ 23 Strong motions recorded on embankment dams ---------------------- 24 Conclusions 25 Acknowledgments 25 References 25 ABSTRACT The earthquake shook a large number of earth and rockfill dams. There were more than 100 dams within 50 miles of the fault rupture associated with this event. Although more than half of the dams were less than 60 feet in height, a num- ber of major dams were strongly shaken. In general, the dams performed satisfactory with one major dam and one minor dam developing moderate damage. A small number also de- veloped minor to moderate cracking and required repairs. The great majority, however, sustained no significant dam- age. Although this result was very encouraging, it should be tempered by the fact that the reservoirs retained by many of these dams were low at the time of the earthquake. In addi— tion, the duration of strong shaking during this earthquake was relatively short. Consequently, the earthquake was not the most critical test of these structures. INTRODUCTION The earthquake resulted from a slip on a segment of the San Andreas fault near Santa Cruz, Calif. As summarized by Spudich (1996), slip occurred on a dipping fault surface approximately 22 miles long and was generally confined to depths between 4 and 12 miles. From the hypocenter, slip propagated in both the northwest and southeast directions over a relatively short source duration of approximately 6 t015 s. Most studies suggest that slip in the southeast portion was predominantly strike-slip, whereas slip in the northwest portion also had a significant thrust component. Most of the slip associated with the main shock is consid- ered to have stopped within 4 miles of the ground surface, although surface cracking observed in the area near the fault may have been associated with zones of compression and extension generated by a tectonically offset basement (Spudich, 1996). The event represents the largest earthquake in the San Francisco/Santa Cruz area since the great 1906 San Francisco earthquake of magnitude 8+. The Loma Prieta fault rupture occurred along a portion of the San Andreas fault segment which ruptured during the 1906 earthquake. The duration of strong shaking for this earthquake was generally between 7 and 10 s, about half of that usually as- sociated with a magnitude 7 event. It has been speculated that this short duration was a result of the central location of the earthquake’s focal point and its bidirectional rupture pat- tern (see Seed and others, 1990). The highest horizontal ground acceleration recorded was 0.64 g and was measured at the Corralitos station located adjacent to the surface pro— jection of the fault rupture and only a few miles from the epicenter. Seismographs in the epicentral area also recorded relatively high vertical accelerations that were comparable to those recorded in the horizontal direction. For example, at D3 D4 PERFORMANCE OF THE BUILT ENVIRONMENT the Capitola recording station, located approximately 9 miles southwest of the epicenter, the peak vertical acceleration was 0.60 g whereas the peak horizontal acceleration was 0.54 g. CHARACTERISTICS OF EARTH DAMS CONSIDERED IN THIS STUDY Presented in figure 1 is a map showing the epicenter and locations of major aftershocks associated with the earthquake. Also shown in this figure are the locations of 111 earth dams situated within 50 miles of the fault rupture zone. The ma- jority of these dams are essentially homogeneous earth dams. The heights and completion dates for these dams are sum- marized in tables 1 and 2. Table 3 presents a listing of the number of dams subjected to various levels of ground shaking during the earthquake. The estimates of peak ground acceleration were developed by interpolating between values measured at nearby seismo— graphic stations. Most of the reservoirs in the affected area were relatively low at the time of the earthquake, with many at less than half their normal heights. Some reservoirs were essentially empty. This was a result of the fact that the earthquake occurred immediately after the irrigation season following 3 years of lower-than-average rainfall. PRE-1906 EARTH DAMS Of the 21 dams built prior to the 1906 earthquake, none experienced significant damage as a result of the Loma Prieta earthquake. This is not unexpected, as these dams generally performed satisfactorily during the 1906 earthquake. Because the 1906 earthquake involved a much greater fault rupture and release of energy, all of these older dams had probably been subjected to much greater amplitudes and durations of shaking during that event than during Lorna Prieta. Many of these older dams were located relatively close to the 1906 fault rupture and have been estimated to have been shaken by motions having peak accelerations between 0.6 and 0.8 g (Seed and others, 1978). In their examination of the performance of earth dams during earthquakes, Seed and others (1978) attributed the good performance of these older dams during the 1906 earth- quake to the fact that they were constructed mostly of clayey soils and built on either clayey or rock foundations. Such materials are generally thought to retain most of their strengths during strong shaking. Two of these older dams, University Mound N. Basin and Piedmont Dams, were built primarily of sandy soils but were not thought to have been saturated at the time of the 1906 earth- quake. These two dams are located more than 45 miles from the Loma Prieta fault—rupture zone and experienced peak ground accelerations of only about 0.1 g with no reported damage. Table 1.—Maximum heights of earth dams considered in this study Maximum height (ft) Number of dams < 10 1 11 - 20 7 20 - 40 31 41 - 60 24 61 - 80 16 81 - 100 8 101 — 150 14 151 - 200 5 201 - 250 4 251 - 300 0 301 - 350 1 Total 1 1 1 dams Table 2.——Dates of completion for earth dams considered in this study Completion dates Number of dams 1861 —1906 21 1906-1920 2 1921 —1930 9 1931 - 1940 14 1941 —1950 10 1951 -1960 23 1961 - 1970 27 1971 - 1980 4 1981 - 1989 1 Total 111 dams Table 3.—Estimates of peak ground accelerations for earth dams considered in this study Estimated peak ground Number of dams acceleration (g) 0.05 - 0.10 34 0.11 - 0.20 41 0.21 - 0.30 10 0.31 - 0.40 11 0.41 - 0.50 14 0.51 - 0.60 1 Total 1 l 1 dams PERFORMANCE OF EARTH DAMS DURING THE LOMA PRIETA EARTHQUAKE D5 1: i STOCKTON \ 37- 30‘- -37° 30' O'NEILL DAM .\ Emu-1mm o . 0 SAN LUIS DAM \ ~ 37° 00' o . 37° o'o'— I .. . I gamma: "- o l ‘. o ‘. O " . HOLLISTER o [I FIGURE 2 . I. d... 0 / v¢'_ SALINAS 0.95. / ‘fi- / 0 v0}. / ($2 kg / ~36° 30' / . u / 36 30 .2 / /'-_ o / ~. / \’\\l‘\; '.. GE / 50-mile SO / SCALE 0 5 10 I5 20 25mlles I22'130' [22:00' 121: 30‘ IZIIOO' Figure l.—Locations of 111 earth and rockfill dams within 50 miles of the Lorna Pn'eta fault rupture zone (diagonal line pattern). D6 PERFORMANCE OF THE BUILT ENVIRONMENT HYDRAULIC FILL DAMS There are five hydraulic fill dams within 50 miles of the fault rupture associated with Loma Prieta: Mill Creek, Chabot, Calaveras, Hawkins, and Old Upper San Leandro. Chabot Dam is one of the predominantly clayey dams which performed well during the 1906 earthquake. This darn was located about 37 miles from the Loma Prieta fault—rup- ture zone and experienced a peak ground acceleration of about 0.1 g without sustaining significant damage. The Old Upper San Leandro Dam was replaced in 1977 with a new dam immediately downstream of the previous one. Both structures were located about 41 miles from the Lorna Prieta fault-rupture zone and experienced peak ground accel- erations of about 0.1 g without sustaining significant damage. The performance of the Calaveras, Mill Creek, and Hawkins Dams is discussed in subsequent sections. ROCKFILL DAMS No dams are composed completely of rockfill within 50 miles of the Loma Prieta fault— rupture zone. However, many earth dams in this area were built with substantial zones of rockfill. Even Calaveras Dam, known as a hydraulic fill, was completed with substantial zones of rockfill. The performance Table 4.—Hydmulic fill dams located within 50 miles of the Loma Prieta fault rupture zone Darn Year Max. height completed (feet) Mill Creek 1889 76 Chabot 1892 142 Calaveras 1925 210 Hawkins 1928 72 Old Upper San Leandro -- -- of some prominent dams built with zones of rockfill is dis- cussed in later sections. SELECTED CASE HISTORIES OF PERFORMANCE Information regarding the performance of 35 earth dams located relatively close to the fault rupture is presented in table 5. The locations of these dams are shown in figure 2. Additional details for selected dams are presented in follow- ing sections. A PERCOLATION . LAKE .ORANCH . LOS GATOS IO miles SCALE 0:: SEM.PERVIRENS . .LEXINGTON .GUADALUPE . LOWER-C CALERO LEROY MURRY HOWELL-UPPER HOWELL ALMLgDEN ° ANDERSON o ALMADEN T 0A: SITE /wSTRIAN 0C0, . ELMER J. . OKELLY CABIN MILL CREEK NEV‘ELL ' CHESBRO COYOTE CANYON 0R. swom ma ouwxs . PEABODY . EPICENTER SELVAGE No.2 nag? DEBELL :5. 0 +4wale ' - SODA '9 LAKE WATSONVILLE ’ OVESSEY LAUREL SPRING CLUB ‘ . HOLLISTERO 1% o o -_ . HOLLISTER '.SAN JUSTO Figure 2.—Locations of earth and rockfill dams in close proximity to the Lorna Prieta fault rupture zone. PERFORMANCE OF EARTH DAMS DURING THE LOMA PRIETA EARTHQUAKE D7 Table 5.—Summary of performance for selected case histories during the 1989 Lama Prieta earthquake PM I TYPE 1 DATE OF 1 MAx. NEIGNT 1 APPROX. SOJRCE | ESTIMATED 1 DAMAGE 1 1 COMPLETION 1 (feet) 1 DISTANCE (miles) 1 RCA (9) I AUSTRIAN 1 ERTN 1 1950 1 185 | < 1 1 0.55 . 0.6 1 MODERATE SETTLEMENT, IRANSVERSE AND LONGITUDINAL CRACKING. SPILLNAV DAMAGE CWELL 1 ERTN | 1890 1 50 1 11. 1 0.45 - 0.5 1 NO REPORTED DAMAGE, RESERVOIR EMPIV BAV STREET 1 ERTN 1 1924 1 27 1 10. 1 0.45 - 0.5 1 MINOR SURFICIAL CRACKING 0N CREST DEBELL 1 ERIN 1 1952 1 53 1 4.5 1 0.45 - 0.5 1 No REPORTED DAMAGE LEXINGTON | ERIN 1 1953 1 205 1 2. 1 0.45 1 MODERAIE LONGIIUDINAL AND TRANSVERSE CRACKING, 0.9 FEET OF SETTLEMENT LONER NONELL 1 ERIN 1 1877 1 39 1 1.5 1 0.4 - 0.45 1 1/4 » INCN LONGITUDINAL CRACK ON CREST UPPER NONELL 1 ERTN | 1878 1 36 1 1.5 1 0.4 - 0.45 1 ND REPORTED DAMAGE ALMADEN 1 ERTN 1 1936 | 110 1 6. 1 0.4 - 0.45 1 MINOR LONGIIUDINAL CRACK AT JUNCTION 0F DAM CREST AND U/S CONCRETE FACING UVAS | ERTN 1 1957 1 118 1 5.5 1 0.4 ‘ 0.45 1 NO REPORTED DAMAGE R SIMONI IRRIGATION 1 ERIN 1 1961 1 44 1 6. 1 0.4 ~ 0.45 1 NO REPORTED DAMAGE, RESERVOIR VIRTUALLY EMPTV GUADALUPE 1 ERIN 1 1935 1 142 1 6. 1 0.4 - 0.45 1 MODERATE CRACKING AT TOP OF u/S BERN, SPALLING 0F u/S CONCRETE PANELS LAKE RANCN | ERIN 1 1877 | 38 1 4. 1 0.4 - 0.45 | 2 EMBANKMENTS NITN MINOR LONGITUDINAL AND TRANSVERSE CRACKING NENELL 1 ERRK 1 1960 1 182 1 6. 1 0.4 - 0.45 1 1 To 9-INCN NIDE LONGITUDINAL CRACKS IN UPPER U/s SLOPE, INCREASED SEEPAGE ALMADEN VALLEV 1 ERIN 1 1965 1 38 1 7. 1 0.4 - 0.45 1 No REPORTED DAMAGE ELMER .I. CNESDRO 1 ERIN | 1955 1 95 1 8. | 0.4 - 0.45 | MODERATE LONGITUDINAL CREST CRACKING, MINOR TRANSVERSE CRACKING AT ASUTMENT CALERO 1 ERIN | 1935 1 90 | 8. 1 0.35 - 0.4 1 NO REPORTED DAMAGE VASONA PERCOLATION 1 ERIN 1 1935 1 34 | 5.5 1 0.35 < 0.4 1 MINOR LONGITUDINAL CRESI CRACKING, MINOR IRANSVERSE CRACKING NEAR SPILLNAY SELVAGE No. 2 1 ERTN 1 1948 1 42 1 7. 1 0.35 - 0.4 1 NO REPORTED DAMAGE RINCONADA | ERIN 1 1969 1 40 1 6. 1 0.35 - 0.4 1 4 TEARS IN RUBBER LINER SEAMS, CRACKED CONCRETE INTAKE VAULT VESSEV 1 ERIN 1 1945 1 20 1 13. 1 0 3 - 0.35 | MINOR LONGITUDINAL CRACKING ON CREST, RESERVOIR EMPIV OAK SITE | ERIN 1 1969 | 43 1 9.5 1 0.3 - 0.35 1 NO REPORTED DAMAGE SODA LAKE 1 ERTN 1 1978 1 35 1 5.5 1 0.3 - 0.35 1 4 IAILINGS DAMS - LIOUEFACTION 0F IAILINGS, SLUMPING 0F 10-FT NEST DAM MILL CREEK 1 HYDF | 1889 1 76 1 12. 1 0.25 - 0.3 | MINOR LONGITUDINAL CRACKS ON CREST SEMPERVIRENS | ERTN | 1951 1 42 1 12. 1 0.25 - 0.3 1 NO REPORTED DAMAGE NOLLISTER w PD 1 ERTN 1 1972 1 13 1 16. 1 0.25 - 0.3 1 MODERATE CRACKING, RESERVOIR EMPTv SAN .IUSTO | ERRK | 1986 1 135 | 17. 1 0.26 1 NO REPORTED DAMAGE LEROY ANDERSON 1 ERRK 1 1950 1 235 1 13. 1 0.26 1 MINOR LONGITUDINAL CRACKING ON NEN CREST FILL NANKINS 1 NYDF | 1931 1 72 1 21. | 0.2 - 0.25 1 MINOR LONGITUDINAL CRACKS ON CREST AND U/s SLOPE, RESERVOIR EMPTV PEABODV 1 ERTN 1 1950 1 63 1 12. 1 0.2 - 0.25 1 No REPORTED DAMAGE COYOTE 1 ERRK 1 1936 1 140 1 14. 1 0.19 1 NO SIGNIFICANI DAMAGE KELLY CASIN CANVON 1 ERIN 1 1955 1 32 1 18. 1 0.15 - 0.2 1 NO REPORIED DAMAGE COIT | ERIN 1 1956 1 54 1 19. 1 0.15 - 0.2 1 NO REPORIED DAMAGE LAUREL SPRING CLUB | ERTN 1 1968 1 28 1 20. 1 0.15 - 0.2 1 N0 REPORIED DAMAGE NURRV | ERTN 1 1957 1 54 1 22. 1 0.1 - 0.15 1 NO REPORIED DAMAGE NORTN FORK 1 ERIN 1 1939 1 100 1 23. 1 0.1 - 0.15 1 No REPORTED DAMAGE, RESERVOIR VIRTUALLY EMPIV NOTES: ERTH = EARTH, ERRK = EARTH AND ROCKFILL, HYDF = HVDRAULIC FILL D8 PERFORMANCE OF THE BUILT ENVIRONMENT AUSTRIAN DAM Austrian Dam was the earth dam most heavily damaged by the Loma Prieta earthquake. This dam is located directly above the projected northern segment of the fault rupture and about 7 miles from the epicenter (fig. 2). Because of its proximity to the earthquake, it is also thought to have been the dam to have experienced the largest shaking with peak ground accelerations estimated to have been between 0.55 and 0.6 g. Fortunately, at the time of the earthquake, the reservoir elevation was about 100 feet below the dam crest. In fact, the reservoir level had been depressed during a 3 to 4 year drought prior to the earthquake, and the upper embankment materials were not fully saturated. The damage sustained by the embankment included moderate settlement, downstream movement, and moderate longitudinal and transverse crack- I AREA OF LONGITUDINAL CRACKING AREA OF TRANSVERSE AND OBLIOUE CRACKING . LOCATION OF PRE-EARTHOUAKE PIEZOMETERS ing (fig. 3). The concrete spillway located on the right abut- ment was also heavily damaged. Details of the investiga- tions and remedial construction are described in a report by Wahler Associates (1990). Previous summaries of damage were presented by Bureau and others (1989), Rodda and oth- ers (1990), and Seed and others (1990). Austrian Dam is a 185—foot—high rolled earth fill dam and was completed in 1950. The dam site is situated between the San Andreas and the Sargent faults. The dam is about 4,000 feet southeast of the main intersection of these fault zones, with the trace of the 1906 movement on the San Andreas fault located only 1,700 feet south of the dam. The Sargent fault is located less than 700 feet north of the dam. The dam is founded on rocks typical of the Franciscan Com- plex, including highly fractured sandstone, graywacke, cobble conglomerate, shale, and serpentine. “ ‘~ 0 50 I00 I50 Zoofeel SCALEEE Figure 3.—Plan View of Austrian Dam locating cracks induced by the earthquake—section A-A’shown in figure 4 (adapted from Wahler Associates, 1990). PERFORMANCE OF EARTH DAMS DURING THE LOMA PRIETA EARTHQUAKE D9 -IZOO A P-6 +4 8 3" —| 100 Z RESERVOIR ELEVATION 9 ON ocr. l7,1989 =1024mv 32‘ V E IOOO =-_—:\ FTTLFTF‘LTT‘?“ E 3' "IMPERVIOUS" ZONE _1 LU -900 CUTOFF TRE NCHES ELEV. I125 feeI 1200 - HOO- . 900 ' WITH GROUT CURTAINS Figure 4,—Maximum cross section of Austrian Dam (section A—A’ adapted from Wahler Associates, 1990). The embankment was built by selective borrowing in an attempt to create a more impervious upstream zone in com- parison to the downstream half of the embankment. Gravel strip drains were also placed beneath the downstream “per- vious” zone. Figure 4 presents a View of the maximum cross section, together with the approximate reservoir and piezom- eter levels on the day of the earthquake. Sampling of the embankment materials during earlier stud- ies of the dam and the piezometer readings together indicate that there is not an appreciable difference between the up- stream “impervious” and downstream “pervious” zones, and that the gravel drains are not completely effective in reliev— ing downstream seepage pressures. Hence, the dam can be considered to be nearly homogeneous. Selected embankment material properties are summarized in table 6. This table presents gradation, compacted density, and placement moisture content values from control tests performed during the original construction in 1950, together with dry density results measured in the dam following the earthquake. Table 6 also presents plasticity, shear strength, and K2max values from laboratory tests performed on undis- turbed tube samples taken in 1979. During the remedial con— struction in 1989, the average relative compaction measured in the preexisting upstream shell of the dam (based on a compactive effort of 20,000 ft.-lb/ft3) was 93 percent. As shown in table 6, the in-situ dry densities of the fill materials in those tests were generally in the upper half of the range of dry densities determined during the 1950 construction con- trol tests. Significant ground movement in the vicinity of the dam was observed on the downstream right abutment above the spillway chute, on the upstream left abutment between the dam crest and the inclined inlet, and along the right side of the reservoir upstream of the dam (fig. 3). On the right abutment, a nearly continuous arcuate scarp was observed, extending up to about 200 feet above the dam crest and about 1,000 feet downstream, with maximum ver- tical and horizontal displacements of about 3 feet. The crack- ing and movements observed in this area were coincident with topographic features associated with landsliding. Ob— servations in exploration trenches indicated that these cracks split bedrock materials and that the fissures existed prior to the earthquake. On the left abutment, nearly continuous cracking was ob- served about 400 feet along an access road from the dam crest to the upstream inlet structure. The cracking generally paralleled the slope, with vertical scarps up to 14 inches in height, and was subsequently determined by exploratory trenching to be the expression of a shallow landslide in loose material overlying graywacke and shale bedrock that had experienced previous, lesser movements. This landslide did not threaten the inclined inlet. In addition to several small slumps and landslides around the rim of the reservoir, a set of aligned fissures was observed after the earthquake on the right side of the reservoir. These fissures developed along a topographic bench for a distance of about 1,500 feet upstream of the dam (fig. 3). In general, the fissures appeared to correspond to slumping of the ground toward the reservoir, with vertical scarps up to 3 feet being measured. The dam owner’s geotechnical consultant excavated borrow and exploration trenches in this area and concluded that the aligned cracking resulted from shaking- Table 6,—Characterizati0n of Austrian Dam fill materials Engineering property Range Mean USCS Classification SC, GC, CL Gradation: >No. 4 (%) 26.0 - 71.5 46.3 Gradation: < No. 200 (%) 16.0 - 43.7 31.8 Specific Gravity, GS 2.60 - 2.78 2.70 Liquid Limit 28 - 32 31 Plasticity Index 11 - 15 13 WC (as compacted ~1950, %) 9.5 - 19.5 14.5 Yd (as compacted ~1950, pcf) 107.5 - 132.0 121.] yd (in situ—l989, pcf) 121.3 - 131.6 126.6 C (psf) 0 q? (degrees) 44 C (psf) 290 45 (degrees) 21 K2,,” 106 - 128 122 D10 PERFORMANCE OF THE BUILT ENVIRONMENT induced settlement of loose, clayey fill overlying a steep bedrock surface and that the linear orientation of the fissures was most likely a result of excavation and/0r shaping of the ground during the initial dam construction (Wahler Associates, 1990). There remains, however, some disagreement as to the cause of the aligned fissures. In a study byAydin and others (1992), another set of cracks, located approximately 500 feet upslope of the aligned fissures, was mapped. These higher cracks had offsets commonly between 2 and 8 inches, and the study concluded that they were associated with sympathetic tectonic movements on the primary strand of the Sargent fault during the earthquake. That study went on to suggest that the aligned fissures on the bench along the reservoir might be associated with movements along another strand of the Sargent fault. Figure 5 presents horizontal and vertical movements for crest monuments measured just prior and after the earthquake. These measurements indicated that the dam crest settled over 2.5 feet along most of its length. In addition, the right end of the darn appeared to move downstream horizontally 1.5 feet relative to the left end. Since the survey monuments were not tied into a stationary benchmark, absolute displacement vectors cannot be calculated. Movements and damage near the toe of the dam suggest that the dam moved primarily downstream. The strong shaking and ground movements produced extensive cracking in the crest of the dam near both abutments. Near the left end of the crest, predominantly transverse cracks in the embankment were up to about 8 inches in width and 10 feet in depth. However, the most pronounced cracking near the left end of the dam occurred along the embankment-foundation contact, where vertical scarps up to 16 inches were observed at the surface. Subsequent explorations traced open cracks varying from 1/ 8 inch to as much as 1 1/2 feet in width along this contact to depths of up to about 27 feet. The explorations suggested that the cracks were within old landslide deposits that had been left in place on the upper part of the abutment. The possibility that some undisclosed separations might have occurred at even greater depths led to a subsequent remedial grouting program to further examine the contact at greater depths. Particularly severe cracking occurred where the right end of the dam abuts onto the concrete spillway. This area of the spillway includes an entrance section from the unlined approach channel, the weir, and a transition section which converges to the chute section. The entrance section includes a wing wall extending upstream along the left side of the approach channel, and a “return” wall coming back in a downstream direction, presumably intended to prevent scour around the upstream end of the wing wall. The two walls thus form a “U” pointing downstream, with the right end of the dam abuting on the outside of the return wall and also backfilling the space between the two walls. Up to 9-inch- wide diagonal cracks occurred in the embankment between the walls. Separations up to 10 inches also developed between the inner faces of the walls and the enclosed embankment. A separation along the outside face of the return wall extended to a depth of about 23 feet (nearly the base of the wall). This would have been particularly dangerous had the reservoir been full, as the normal amount of freeboard was only about 15 feet. Immediately downstream of the spillway entrance, where the dam abuts the left wall of the transition section, several cutoff collars on the outer (embankment) side of the wall were sheared-off due to upstream-downstream movement of the adjoining embankment relative to the wall. Extensive damage also occurred in the transition and chute sections of the spillway walls. Up to 3/4-inch-wide tension cracks were observed in the walls and slab of both sections. The cracking in the slab of the chute section was somewhat regularly spaced at about 2 to 4 feet, normal to the axis of the spillway. The 80 feet of chute and transition section farthest upstream was found to have elongated a total of about 7 inches. In the transition section, tension cracks generally paralleled the trend of cracking which occurred in the natural slope above the spillway. Up to l-inch-wide separations were observed between the bottom of the slab and its highly weathered bedrock foundation along the left wall of the transition section. Subsequent grouting through the slab indicated that these voids developed existed primarily beneath the walls, but not beneath the central part of the slab. This suggested that the spillway walls had been subjected to some sort of rocking action. At the downstream end of the chute, the concrete wingwall on the left side was rotated and torn away from the left wall, while the right wing wall could not be found. In addition to the damage to the crest and spillway, a series of roughly parallel longitudinal cracks formed in the upstream and downstream faces of the dam. The initial widths of the cracks were relatively small, but ultimately widened within a few weeks to as much as a foot in places. These time—dependent movements may have resulted from major aftershocks, rainfall during this period, soil creep, and/or pore-pressure dissipation and consolidation. Seven days after the main shock, the four longitudinal cracks which formed in the upper 50 feet of the upstream face were approximately 5 to 15 feet deep and 1 to 4 inches wide. A number of longitudinal cracks which were 3 to 8 feet deep and 2 to 6 inches wide formed in the downstream face. The majority of the longitudinal cracks in the downstream face were located near the crest, although some limited cracking also occurred near the toe of the dam. In addition, there was some minor bulging of the downstream toe. Exploration trenches indicated that the longitudinal cracks generally dipped steeply toward the dam crest on both embankment slopes. Hence, the cracks did not appear to result from slope instability, but rather from settlement and rearrangement of the earth embankment. PERFORMANCE OF EARTH DAMS DURING THE LOMA PRIETA EARTHQUAKE 1.5 _ I I I I I I I I I I I I I | I | _ : Mm Original alignment :- 1 0 "_ MQH Oct 16, 1989 alignment ’1' E J ' - w»... Oct 19, 1989 alignment x g j A _ X (D L". ; ,xv-T-w’ E 1 0.5 *- X D ._ Z : (’1 A : v _ k *— * A _g7‘§_‘ r \a\ _ [—- -— ... u ‘ “L af- u ’i/ .3 .. .- .. 3.- n: 2 0.0 _ \‘\/r ,x’ *4! . Lu T \\ "l,’ ‘ E : \L ,’)" : Z ~05 _ - C9 *- E ' g : § 1 - 5 - —1.0 _— A 3“- _1'5 - I I I i L I I I l I I I I I I I - 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 L STATION (IN 100 FT) R 1134.0 I I I I I I I I I I I I I I I I M Original alignment - 1132'0 “I?“ Oct 16, 1989 alignment 5 — r- H...” Oct 19, 1989 alignment 2 - ’I: 1130.0 - g - LL _ \ 8 - Z 1128.0 — K \ - = _ x ‘K _ \\ A \ 2 1126.0 '— \ Thr+~nhfi _ E F “x 0r- ~10 .. <1: 1124.0 — “"‘~. .. > - ¥~‘~~'+"’l-— -1 LLI *-~*'*—»- -I‘"'¥ E .J _ ‘*" .- m 1122.0 g — Q -1 D 1120.0 - B 9 1118.0 I I I I l I I I I I I I I I I I 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 L STATION (IN 100 FT) R Figure 5,—Horizonta1 (A) and vertical (B) movements of crest monuments at Austrian Dam (from Seed and others, 1990). D11 D12 The movements of the embankment were concluded to be due to general settlement and spreading of the fill during the strong earthquake shaking, followed by subsequent downslope creep. Figure 6 presents piezometer measurements made for 5 standpipe piezometers installed prior to the earth— quake. These instruments recorded pore-pressure increases between 12 and 54 feet Of water. Possibly related to the pore- pressure increases is the fact that some of the trenches exca- vated across the longitudinal cracks near the crest encoun- tered free water at elevations considerably higher than reser- voir levels at the time of the investigation (Wahler Associ- ates, 1990). The largest pore-pressure increase, 54 feet of water, oc- curred in piezometer P- 1, with its tip located within the down- stream “pervious” zone near the bedrock contact. The standpipe for this instrument was found to be significantly deformed between elevations 955 and 960 (about 25 to 30 feet above the bedrock contact). The standpipe for piezom- eter P-6, which measured an increase in hydraulic head of about 49 feet of water, was also found to be deformed be- PERFORMANCE OF THE BUILT ENVIRONMENT tween elevations 1,017 and 1,041 feet (about mid-height of the embankment). These deformations were suggestive of earthquake-induced internal movements corresponding to lateral spreading of the embankment. The repair of the earthquake damage consisted Of (1) ex- cavating and recompacting the fill in the areas of extensive cracking, (2) placing a zoned fill with chimney and blanket drains in the crest fill at both embankment ends, (3) excavat- ing and recompacting the upstream face Of the fill to create an impervious blanket, (4) epoxy grouting the cracking in the spillway and cement grouting voids beneath the spill- way slabs, and (5) grouting of the rock at the left abutment contact with the fill. This repair successfully remediated the earthquake—induced damage to the embankment and was accomplished within 8 weeks following the earthquake. Details of the repair can be found inWahler Associates (1990). Due to the severity of dam— age to the concrete spillway and concern about potential fu- ture landslides in the right abutment, a new spillway was later constructed on the left abutment of the dam in 1993-1994. |080 rLOMA PRIETA EARTHQUAKE 1060 - 1\ Surface water due to ‘ 1?. construction activities I' \ on crest i . ./-\ !-\.“,/~.\.‘. P—6 IO4O ,- V " ‘ i A .' 23 I020 :.___ .' 8 ' 1» ~' 1 Z l o \\ ' ---~_ _ DR ; 1000- ‘\:'....__.._____.___"_.‘_‘_:.~...__=__4_T____E'I “ ‘>‘ :‘ LL] l d .' .......... e .-...2 FIN ...................... “ \___\. 4 _ 960 =________‘::~_~J V\.-,_-_. ~~_-——/ \Jii__ — 94{) SEPTEMBER l OCTOBER | NOVEMBER | DECEMBER 1 JANUARY | FEBRUARY 1989 1990 Figure 6.—Piezometer measurements for preearthquake piezometers at Austrian Dam (from Wahler Associates, 1990). PERFORMANCE OF EARTH DAMS DURING THE LOMA PRIETA EARTHQUAKE D13 LEXINGTON DAM Lexington Dam is a 205-foot-high darn located about 6 miles downstream of Austrian Dam and about 2 miles from the fault rupture associated with the earthquake (fig. 2). The dam was completed in 1953 as a zoned earth structure hav— ing a relatively thick sandy and gravelly clay core that is supported by upstream and downstream random shell zones of clayey sands and gravels. The dam also has relatively flat upstream (5.521) and downstream (3:1) slopes. A plan view and cross section are shown in figure 7. The embankment material properties are summarized in table 7. At the time of the earthquake, the reservoir was about 100 feet below the crest of the dam. Previous summaries of damage were pre- sented by Bureau and others (1989), Seed and others (1990), and in the studies by R. L. Volpe & Associates (1990a). Lexington Dam was instrumented with strong-motion in— struments on the left abutment, left crest, and right crest. These accelerographs recorded transverse peak accelerations of 0.45, 0.39, and 0.45 g, respectively. This shaking was com- posed of about 6 to 7 s of relatively strong long- period mo- tion. The left abutment or “bedrock” peak acceleration is within the range predicted by appropriate strong-motion at- tenuation relationships for a site approximately 2 miles from the nearest point on the fault rupture surface for a M, = 7.1 event, but is a bit lower than the mean or expected value based on such relationships. In addition, there appears to be some spectral acceleration amplification at lower frequen- cies (0.9 to 1.2 Hz). This low frequency amplification may indicate that the recorded “bedrock” motion may have been affected by local topographic or geologic conditions. The strong ground shaking produced transverse cracking on both the upstream and downstream sides of both abut- ments, oblique cracking on the crest about 150 feet in from the left abutment, longitudinal cracking on both the upstream and downstream slopes of the dam, and cracking of an ac- cess road on the right abutment upstream of the dam. The cracks, which were fairly isolated, were commonly less than 3/4 of an inch wide, and trenching indicated that they only extended to depths generally between 2 and 7 feet (R. L. Volpe & Associates, 1990a). The maximum earthquake-in- duced crest deformations were approximately 0.85 feet of vertical settlement, and 0.25 feet of lateral displacement in the downstream direction (R. L. Volpe & Associates, 1990a). An old slope indicator casing was found to have raised from beneath the crest to over 3 inches above the crest due to the embankment settling around it. The earthquake shaking and ground movements produced extensive cracking in the bridge abutment at the left abutment and ruptured a buried water line near the crest of the dam. About 6 weeks after the earthquake, a relatively large seep- age area developed high up on the downstream face of the dam. The seepage area was about 170 feet long and 35 feet wide and oriented at an oblique angle with the axis of the darn. This seepage area was really more of a wet or damp area and never really flowed water. Although the cause of the seepage area is not definitively known, one explanation that has been offered is the fact that old exploration holes extending into the rock foundation lie within the area and that these old borings could have been acting as relief wells for earthquake-induced pore pressures within the lower por— tions of the embankment and bedrock (R. L. Volpe & Associ— ates, 1990a). Another possible explanation is that the fill is rela- tively impervious at this elevation and that any surface water that infiltrates the dam becomes perched at this level. The repairs made to the dam consisted of trenching the cracked areas to depths ranging between 3 and 7 feet and compacting the excavated soil back into the trenches (R. L. Volpe & Associates, 1990b). GUADALUPE DAM Guadalupe Dam is a 142-foot-high dam located about 6 miles from the Lorna Prieta fault—rupture zone, and it prob- ably experienced peak ground accelerations between 0.4 and 0.45 g (fig. 2). The dam was completed in 1935 as a rolled earth structure with an upstream facing of concrete panels for erosion protection. In a manner similar to that described for Austrian Dam, the embankment is apparently nearly ho— mogeneous, as the selective borrowing to create upstream “impervious” and downstream “pervious” zones did not ap— pear to be completely successful in creating distinctly dif— ferent zones. In 1972, an upstream buttress was added to the dam to improve drawdown stability. A plan view and cross section are shown in figure 8. At the time of the earthquake, the reservoir was about 78 feet below the crest of the darn; however, the reservoir had been full up to about 3 months before the earthquake, and it is assumed that the upstream shell materials were nearly saturated at the time of the earth- quake. Previous summaries of damage were presented by Bureau and others (1989), Seed and others (1990), and in the studies by R. L. Volpe & Associates (1990a). The earthquake induced up to 0.64 feet of settlement and 0.15 feet of lateral displacement in the upstream direction as measured on the crest. Minor transverse cracking developed at the crest at both abutment contacts along with minor longitudi- nal cracking on the crest. The principal damage was to the up- stream slope, where the upper portion of the buttress fill devel- oped longitudinal cracking. Shortly after the earthquake, these cracks were observed to have a maximum width of less than 1 inch and extended across the entire face of the dam. About 5 weeks later, the cracks had widened to about 4 inches and extended to a depth of about 5 feet (R. L. Volpe & Associates, 1990a). These cracks may have been caused by concentrations of dynamic stresses induced by the change in slope geometry. Alternatively, they may have resulted from possible past settlements caused by the placement of the berm. These past settlements may have created preexisting cracks which sur- faced only after the development of strong ground motion. D14 PERFORMANCE OF THE BUILT ENVIRONMENT \ //‘\\ \_ fl \f‘\ DOWNSTREAM TOE / i 810’ CREST UPSTREAM TOE \ SCALE (feet) 0 50 100 200 Figure 7.—Plan View (A) and cross section (B) of Lexington Dam. PERFORMANCE OF EARTH DAMS DURING THE LOMA PRIETA EARTHQUAKE D15 Table 7.—Characterizati0n of Lexington Dam fill materials Engineering CORE SHELL Property (Zone 2) (Zones 1 & 4) Depth (ft) 0-80 >80 Classification SC, CL-CH SC GC, CL Gradation: > No. 4 (%) 13—30 0.2 i0-50 Gradation: < No. 200 (%) 29-52 86-97 218-97 Specific Gravity, Gs 2.67 - 2.73 2.73 Liquid Limit 31 - 37 33 - 39 Plasticity Index 14 - 18 14 - 24 We (in situ range 11.2-17.7 21.5-30.6 9.4-26.5 and avg, %) (14.4) (25.6) (15.3) Yd (in situ range 117.5-l3l.5 92.5-102.2 395.2-134.8 and avg, pcf) (120.9) (96.6) (119.8) Ydmax’ Wc0 t 131.7 pcf (20,000 ft-lb/ft3) 8 % C' (psi) 400 0 0 45' (degrees) 36 25 36 C (psi) 1200 0 0 ¢ (degrees) 18 17 22 1(me 100-150 ~50 ~100 Vsmax (ft/sec) 1200-1600 1400-1600 1400—2200 Notes: I Only one sample had less than 13 percent gravel. For 70 percent of the samples, <35 percent finer than No. 200. Average dry unit weight was about 6 pcf lower at depths of 0-40‘, and 4 pcf higher at depths >40'. The earthquake also caused the concrete panels on the upstream face above the berm to pound against each other, resulting in cracking and spalling in about 10 percent of the panels (R. L. Volpe & Associates, 1990a). The repairs made to the darn consisted principally of exca- vating a 70—foot band of material (as measured along the slope, parallel to the crest) at the top of the upstream buttress to a depth of about 6 feet. The excavated material was temporarily stored in order to allow the material to dry to an acceptable water content and then recompacted into place. The cracks in the crest at the abutment contact were excavated to about 3 to 4 feet and the excavated material was recompacted back into place (R. L. Volpe & Associates, 1990b). NEWELL DAM Newell Dam is a 182—foot-high dam located about 6 miles from the Lorna Prieta fault-rupture zone, and it probably expe- rienced peak ground accelerations between 0.4 and 0.45 g (fig. 2). The dam was completed in 1960 as a zoned earth and rockfill darn generally composed of clayey zones except for an up- stream zone of dirty rockfill. At the time of the earthquake, the reservoir was about 49 feet below the crest of the dam. Previ— ous summaries of damage were presented by Bureau and oth- ers (1989), Seed and others (1990), and Creegan (1990). Although the earthquake did not induce significant crest movements, a longitudinal crack was found on the 3:1 upstream slope running the entire width of the dam face at about the spillway elevation. This crack was generally between 1 and 9 inches in width. Trenching explorations of the crack indicated that it was formed by tension within the zone 2 rockflll. The trenches, however, only extended to a maximum depth of 10 feet, and the crack, which was about 1 inch in width at the bottom of the trenches, extended farther to greater depths. There were also other minor cracks at various locations. Seepage through the dam and abutments, measured at the downstream toe, was also found to have increased from a normal 8 gpm to 41 gpm, but remained clear (Creegan, 1990). By early December 1989, the seepage had decreased back down to about 17 gpm. The development of the longitudinal cracking was theo- rized to have resulted from settlement of the dirty rockfill upstream shell relative to the rest of the dam. This material consisted of quarried sandstone and shale and was placed in 5-foot lifts and compacted by sluicing. The other clayey zones were placed in thin lifts and compacted to about the maxi- mum Standard Proctor dry density. Consequently, the up- stream rockfill zone was relatively loose in comparison with the other zones in the embankment. There were indications that some of the cracking ran along the interface of the zone 2 rockfill and the clayey core. Further evidence of settle— ment of the zone 2 material was found at the bell toggle joints along the sloping intake tower where the embankment seemed to have pulled away from the structure by about 1 to 3 inches in a downstream direction (Creegan, 1990). The repair of the longitudinal cracks consisted of using a large backhoe to excavate to a depth of about 6 feet along the alignment of the crack. The trench was then backfilled in 18-inch lifts with each lift being compacted with a vibrating shoe on the backhoe. The upstream slope was then rolled with a Vibratory roller (Creegan, 1990). ELMER J. CHESBRO DAM The Elmer J. Chesbro Dam is a 95-f00t—high embankment located about 8 miles from the Lorna Prieta fault-rupture zone, and it probably experienced peak ground accelerations between 0.4 and 0.45 g (fig. 2). As for many of the embank- ment dams in this area, the embankment is a nearly homoge- neous compacted fill, as selective borrowing to create up- stream “impervious” and downstream “pervious” zones do not appear to be entirely successful in creating distinctly dif- ferent materials. The upstream slope varies between 2:1 and 3:1. The downstream slope is about 2:1, but is fitted with a 45-f00t-wide berm at about half the height of the darn. At the time of the earthquake, the reservoir was about 69 feet be— low the crest of the darn. Previous summaries of damage were presented by Bureau and others (1989), Seed and others (1990), and in the studies by R. L. Volpe & Associates (1990a). Surveys of crest monuments showed that the earthquake induced up to 0.37 feet of settlement and 0.05 feet of lateral displacement in the upstream direction. The main area of cracking that developed at this dam occurred as longitudinal cracking near the upstream edge of the crest. This cracking extended about 240 feet and had a 4-inch width together with D16 a 4-inch vertical offset, with the upstream side of the crack being on the down side. There was also minor transverse cracking on the right abutment at both the crest area and near the spillway entrance (R. L. Volpe & Associates, 1990a). Trenching explorations indicated that the longitudinal crack- ing on the upstream edge of the dam extended to a maximum depth of about 8 feet and was related to settlement of an up- stream slope-protection zone. The transverse cracking on the PERFORMANCE OF THE BUILT ENVIRONMENT crest at the right abutment was estimated to have a maximum depth of 4.5 feet (R. L. Volpe & Associates, 1990a). The repairs to Chesbro Dam consisted of excavating the upper 10 to 12 feet of the upstream slope and replacing this material with compacted, imported materials. The cracking on the right abutment was repaired by trenching to a depth of about 5 feet and recompacting the excavated material back into place (R. L. Volpe & Associates, 1990b). \"———\ ILLWAY / UPSTREAM TOE 670’ / l/ 640 2.5 g 500 1r“ 3 z 560 O u u Er: 520 .IMPERVIOUS. ZONE II PERVIOUS ZONE U>J l d 480 440 Q 50 Figure 8.-—Plan view (A) and cross section (B) of Guadalupe Dam. CONCRETE FACE DOWNSTREAM TOE A CREST SCALE (feel) 100 200 PERFORMANCE OF EARTH DAMS DURING THE LOMA PRIETA EARTHQUAKE VASONA PERCOLATION DAM The Vasona Percolation Dam is a 34-foot-high embank- ment located about 5.5 miles north of the Loma Prieta fault- rupture zone, and it probably experienced peak ground ac- celerations between 0.35 and 0.4 g (fig. 2). The dam is com- posed of two embankments flanking a central concrete spill- way. The two embankments were completed in 1935 and con- tain an upstream zone of rolled “fine” material together with downstream zones of coarse and random materials. Beneath the upstream “fine” material there is a puddled clay cutoff trench which extends 25 feet to an impervious foundation material. At the time of the earthquake, the reservoir was about 12 feet below the darn crest (R. L. Volpe & Associates, 1990a). The dam developed maximum settlements of about 0.16 feet and about 0.09 feet of lateral movement in the down~ stream direction (R. L. Volpe & Associates, 1990a). The major damage consisted of longitudinal cracking along the crest of the right embankment. This cracking was intermittent, but extended along the entire length of the embankment. These cracks were up to 3/4 of an inch in width, and trenching indicated a maximum depth of about 5.5 feet. There was also limited longitudinal cracking on the crest of the left crest embankment, generally between 1/8 to 1/2 inches in width. Minor transverse cracking, generally less than 1/8 of an inch in width, was found along the spillway contact from the crest centerline to the downstream toe. Cracking was also noted along the spillway wingwall at the downstream toe of the dam and along the parapet wall along the crest of the dam. Longitudinal cracking of the crest had previously been noted following the August 8, 1989 earthquake (M=5.1), now considered a possible foreshock of the Loma Prieta earth— quake. In some locations, the longitudinal cracks observed EARTHQUAKE-INDUCED CRACKS IN DAM FILL CREST OF DAM D17 following the Lorna Prieta earthquake were in the same 10— cations as those from the earlier earthquake. The repair principally consisted of removing the upper 4.2 feet of the right embankment and recompacting the material back into place (R. L. Volpe & Associates, 1990b). SODA LAKE DAM Soda Lake Dam consists of four embankments that are used to retain granite tailings from a nearby quarrying operation. The dam had been enlarged in 1978 to give a maximum height of 35 feet for the main or south embankment. The dam is lo- cated approximately 5.5 miles from the Loma Prieta fault—rup- ture zone, and it probably experienced peak ground accelera- tions between 0.3 and 0.35 g (fig. 2). Although the reservoir was nominally empty at the time of the earthquake, most of the reservoir volume had been filled with granite tailings. Although the reservoir may have been considered empty, portions of the granite tailings were evidently saturated, as they appeared to have liquefied in several places. Evidence of liquefaction was in the form of numerous sand boils de- veloped within the reservoir deposits and the fact that the reservoir sediments settled about 2 to 3 feet relative to the embankments and reservoir rim. One reconnaissance team found the sand boils still ejecting water and sand 72 hours following the main shock (J. Egan, oral c0mmun., 1993). No damage was found at the north, south, or east embank- ments. However, the 10-foot-high West Saddle Dam was found to have an extensive pattern of cracking at and below the crest. As shown in figure 9, the cracking was in the pat- tern of an arc and seemed to identify a wedge of material that had developed into a minor slump towards the reser- LIMIT OF TAILINGS PRIOR TO DAM CONSTRUCTION '- u '- 0. o ON OCTOBER I7 I9 SCALE I00 feet 0 50 Figure 9.—Plan view of the Soda Lake west saddle dam showing earthquake-induced cracking. D18 PERFORMANCE OF THE BUILT ENVIRONMENT voir. These cracks had typical widths of about 1 to 4 inches and could be probed to depths of at least 6 feet. The cracks also had vertical offsets of about 1 inch with the reservoir side of the crack being lower than the downstream side. The cause of the cracking and/or slumping of the West Saddle Dam was theorized by engineers within the Califor— nia Division of Safety of Dams to be a result of having left potentially liquefiable tailings deposits in the foundation beneath this embankment during its construction in 1978. As shown in figure 10, prior to construction of the West Saddle Dam, preexisting tailings extended from the reser- voir to about the centerline of the dam to be built. These tailings were supposed to have been removed. However, if they were not completely removed, then liquefaction of tail- ings remnants within the foundation might explain the crack- ing and movements observed following the earthquake. Sup- porting this theory is the fact that the shape and location of the cracked wedge appears to match the contour of the limits of preexisting tailings in the foundation (figure 9). As a result of the cracking/slumping in the West Saddle Dam, the dam was not considered safe to store water, and the owner was required to take the dam and reservoir out of service. MILL CREEK DAM Mill Creek Dam is a 76-foot-high dam with a compli— cated history. It was originally built in 1889 as a timber crib dam, with a substantial hydraulic fill zone placed upstream of the timber crib portion. There is apparently no informa- tion available concerning its status and performance follow- ing the 1906 earthquake. In 1932, additional fill was placed upstream to give the upstream face a 3:1 slope. In 1947, part of the timber crib dam burned. Part of the burned tim- ber crib zone was removed, leaving a downstream slope of about 1.5: 1. In 1957, a large sinkhole formed in the upstream face above corroded sections of the original 14—inch outlet pipe. The repair for the sinkhole consisted of excavating a large triangular trench to remove portions of the corroded pipe and then recompacting the excavated materials back into the trench. A new outlet consisting of a 12-inch pipe was also installed at this time. Figures 11 and 12 present sections illustrating the different materials present in the dam. Mill Creek Dam is located approximately 12 miles from the Lorna Prieta fault-rupture zone, and about 17 miles from the epicenter. The dam may have experienced peak ground accelerations of up to about 0.3 g (fig. 2). At the time of the earthquake, the reservoir was about 16 feet below the crest. Despite the questionable materials left within it, the dam performed well during the earthquake. The only damage observed was minor cracking at the crest, which became ob- scured during the rains that followed a few days after the earthquake. The good performance is particularly surprising due to the presence of the sandy and silty hydraulic fill left in place beneath the 1957 repair trench. Piezometers indicated that these soils were saturated prior to the earthquake. Penetra— tion tests made during investigations conducted in 1987 and 1992 indicated that these soils are very loose [Standard Pen- etration Test (N1)60vCS of the hydraulic fill is equal to about 10 blows per foot]. A Standard Penetration Test-based analysis of the dam was performed by Wahler Associates (1992) for the Loma Prieta earthquake using an estimated peak acceleration in bedrock of 0.17 g. The results indicated that liquefaction would be triggered in the hydraulic fill, resulting in a flow slide of the upstream slope during the Loma Prieta earthquake. A back- analysis was also performed to determine the value of re- sidual strength of the hydraulic fill which would have been needed to prevent the calculated onset of a flow slide and result in computed deformations in agreement with the ac— tual observed seismic displacements. It was determined that a residual strength of about 600 psf would have been needed. However, based on the Standard Penetration Test results, the estimated actual in-situ residual strength would only have been about 300 psf. A possible explanation for the discrepancies between ob- served and back-calculated performances is that the site ac- celeration could have been significantly less than the esti- mated 0. 17 g peak value. Low cyclic loadings might not have caused liquefaction of the hydraulic fill and would explain the lack of damage to the dam. There were no seismic in- struments at the site, and there were no physical indications that liquefaction had occurred. Although the dam performed well during the 1989 Lorna Prieta earthquake, design modifications to the existing spill- way were developed to restrict the reservoir surface to an operating level which would reduce the risk of sudden re— lease of the reservoir during a future large earthquake. LEROY ANDERSON DAM Leroy Anderson Dam is a 235-foot dam located about 13 miles from the Loma Prieta fault-rupture zone. The dam was completed in 1950 and is composed of central core zones of compacted sandy clay and clayey sand that are flanked by shell zones of rockfill. The rockfill was placed by end—dump- ing from trucks into 10 to 25—foot—high lifts that were com- pacted by sluicing with water (fig. 13). The dam is well instrumented with seismographs, which indicated a peak horizontal ground acceleration of about 0.26 g at the base together with a crest horizontal peak accelera- tion of about 0.43 g. There were several other instruments placed on the crest and downstream face which recorded peak horizontal accelerations of between 0.1 and 0.4 g. This dam also experienced strong shaking during the nearby 1984 Morgan Hill earthquake (ML=6.2), with a peak PERFORMANCE OF EARTH DAMS DURING THE LOMA PRIETA EARTHQUAKE D19 20 feet I70 r——bl I70 ELEVATION I65fee1 E LEVEL OF GRANITE TAILINGS —OCTOBER I7 I989 33 '50 .. ............................. I60 3 """""" COMPACTED FILL ; |50 PRECONSTRUCTION LEVE F TAILINGS 150 § _..__.__.________.._____._ ‘3 I40 I40 Lu | 3 O _. I 30 Figure lO.—Cross section of the Soda Lake west saddle dam. I500 - COMPACTED TRENCH SECTION ELEVATION I487 feet '3 _ ELEVATION , - “fin :13 ”'80 I989“ I47I feet fififit V 3% CONCRETE SACK 3 ”’60" m.- RIPRAP E: I440 — 0 ' . . .......... ; era/mm QWIQIQYC’IQYJRB-n LL, .................... 0.0- o . . .2 '.‘25)}?KfiIbY@Y@I@rJJ@mV(0f@XZI/Y@i. -’ ‘420 _ _— ‘ "' '.ffi L” WEATHERED A v / I400 D'OR'TE LAYER HYDRAULIC FILL wooo CRIB Figure 1 l.——-Cross section of Mill Creek Dam . CREST ELEVATION = I487 feet . mkfikfib \Y ..... \Th€§$§}\ f: .-: ........... nch V»: .. ..... ,. .. 20 40 feet SCALE 0 Figure 12,—Longitudinal section of Mill Creek Dam. D20 PERFORMANCE OF THE BUILT ENVIRONMENT UPSTREAM TOE SCALE (feet) I I O 50 100 200 4d DOWNSTREAM TOE CREST EL 64d W SCALE (feet) 0 50 100 200 3 Figure 13.—Plan view (A) and cross section (B) of Leroy Anderson Dam. acceleration at the downstream toe of about 0.41 g and a corresponding horizontal peak acceleration of 0.63 g at the crest. Although of larger peak amplitude, the duration of strong shaking was less during the 1984 earthquake than during the 1989 event. Both the 1984 and 1989 earthquakes produced similar patterns of distress in the embankment. In 1984, two paral— lel, and relatively shallow, longitudinal cracks formed in the compacted fill along the crest approximately overlying the buried contacts between the central clayey core and the up- stream and downstream rockfill shell zones. These cracks were generally less than an inch in width and are thought to have been the result of settlement of the shell zones relative to the core zones. Between 1984 and 1989, the embankment crest was raised about 5 feet and a small sliver fill was placed in a localized area on the downstream edge of the crest. The PERFORMANCE OF EARTH DAMS DURING THE LOMA PRIETA EARTHQUAKE D21 5-foot crest raise was placed to provide additional head for a modified spillway, and the sliver fill was placed to provide space for an instrumentation vault. The main damage following Loma Prieta was a longitudi— nal crack on the downstream half of the crest approximately above the contact between the shell and core zones. In addi— tion, the longitudinal cracking extended to and around the base of the vault on top of the sliver fill. These cracks were generally less than 3/4 of an inch in width. A second set of longitudinal cracks was also observed at both edges of the paved crest road along the bases of guard rail posts. This latter set of cracks were as much as 1.5 inches in width. The maximum settlements of crest monuments following Loma Prieta were less than 0.04 feet on the upstream edge and about 0.13 feet on the downstream edge. Only minor repairs were considered necessary for the crest cracking (R. L. Volpe & Associates, 1990a). SAN J USTO DAM San Justo Dam is a 135—foot dam located about 17 miles from the Loma Prieta fault rupture zone (fig. 2). The dam is relatively new and was completed in 1986 as a zoned earth and rockfill embankment (figs. 14 and 15). Although no sig— nificant damage was observed following the earthquake, this dam provides an outstanding opportunity to investigate the dynamic response characteristics of earth dams. This is be- cause the dam is heavily instrumented with strong—motion sensors. Furthermore, one of the sensors was embedded in a borehole approximately 62 feet below the crest of the dam, providing one of the few opportunities to calibrate response analyses with the internal motions of an embankment dam. Figures 14 and 15 illustrate the peak accelerations recorded at different locations at San Justo Dam. The transverse hori- zontal peak acceleration recorded at the downstream toe was about 0.26 g; whereas the transverse horizontal peak accel- eration recorded at the crest was about 0.40 g, an amplifica- tion of about a factor of 2. However, the interior sensor at about the mid-height of the dam indicated a transverse peak acceleration of only about 0.26 g. This would suggest that most of the transverse amplification took place within the upper half of the dam. The strong-motion sensors also suggest that longitudinal peak horizontal accelerations were amplified by about a fac- tor of 3, but that the longitudinal amplification took place in the lower half of the embankment. HAWKINS DAM Hawkins Dam is a 72-foot-high hydraulic fill that was built slowly between 1912 and 1931. It is located about 21 miles from the Loma Prieta fault-rupture zone, and it probably experienced peak horizontal ground accelerations of about 0.2 to 0.25 g (fig. 2). Although the reservoir was empty at the time of the earthquake, the dam experienced minor lon— gitudinal cracking on the crest and on the upstream slope. The dam was investigated by Lee and Roth (1977) during the performance of a seismic stability evaluation. This study found that the dam was generally composed of a fairly ho- mogeneous and impervious gravel-sand-clay fill, together with a small zone of homogeneous clay in the central por- tion of the embankment (fig. 16). Because it was built slowly, with only about 2 t0 3 feet of material added each year, there was time for consolidation and desiccation of the fill to take place. The study by Lee and Roth (1977) concluded that the dam would have adequate stability for even a magnitude 8+ earthquake on the San Andreas fault. Unfortunately, because the reservoir was empty at the time of the earthquake, an opportunity to verify this evaluation was lost.] PERFORMANCE OF EARTH DAMS AT GREATER DISTANCES In addition to the dams located in relatively close proxim- ity to the earthquake, the performance of a few dams located at greater distances is also of interest. CALAVERAS DAM Calaveras Dam is a 210-foot-high hydraulic fill dam that was completed in 1925. During construction, it developed a large slide due to static liquefaction. Although the larger portion of the embankment is composed of hydraulic fill, the upper portion was built with a rolled clayey core flanked by rockfill shell zones. In 1974, a rockfill buttress was added to the upstream slope to increase freeboard and to provide increased stability. The dam is located approximately 24 miles from the Loma Prieta fault-rupture zone, and it probably experienced peak ground accelerations between 0.1 and 0.15 g (fig. 1). At the time of the earthquake, the reservoir was about 70 feet be- low the crest of the dam. The dam experienced no significant damage. Reported incidents consisted of minor changes in piezometers, a tem- porary 25 to 30 percent increase in seepage, and the muddy- ing up of seepage water attributed to a stirring up of silt by the shaking of the earthquake. SAN LUIS DAM San Luis Dam is a 313-foot-high dam that was completed in 1967. It is composed of a homogeneous section of com- pacted clayey material with internal drains and is flanked with slope protection zones. In 1981, a portion of the up- stream slope developed a large slide following a severe draw- D22 PERFORMANCE OF THE BUILT ENVIRONMENT 0.26 0.3. OAOJfiOAQ UPSTREAM SLOPE UPSTREAM WASTE BERM 200 400 feet SCALE ' Figure 14.—Plan View of San Justo Dam. WASTE BERM CREST ELEVATION 508.5 feet 100 200 feet SCALE 0 Figure 15.—Maximum (A) and Left Quarter (B) sections of San Justo Dam. PERFORMANCE OF EARTH DAMS DURING THE LOMA PRIETA EARTHQUAKE D23 x/fl\ /’ CLAY \ GRAVEL-SAND-CLA‘Y ‘‘‘‘ " .. (BROKEN SHALE) o r c SCALE ;__5:° " Figure 16.-—Cross section of Hawkins Darn (from Lee and Roth, 1977). down of the reservoir. Although the slide began below the crest, the slide propagated far enough back to take out a por- tion of the crest. The slide was found to involve slopewash material left in place in the foundation. The dam was re— paired by leaving the slide material largely in place, but- tressing the slide with a large upstream berm, and rebuilding the crest section. San Luis Dam is located about 34 miles from the Loma Prieta fault-rupture zone (fig. 1). At the time of the earth- quake, the reservoir was less than half full. Strong motion sensors at two locations along the downstream toe recorded peak horizontal accelerations of 0.06 and 0.10 g. The corre- sponding locations on the crest recorded peak horizontal ac- celerations of 0.25 and 0.17 g. Despite experiencing moder- ate levels of ground motion, the dam experienced no signifi— cant damage. Two days after the earthquake, small longitu- dinal hairline cracks were found on the crest in the vicinity of the slide. These cracks appeared to be largely preexisting cracks which may have widened slightly following the earth- quake. Longitudinal hairline cracks were also found on the crest at other locations. The largest crack was longitudinal and had a width of less than 1/4 of an inch. It was located near the right abutment, where the height of fill was only about 5 feet. No repairs were needed at this dam as a result of the earthquake. O’NEILL DAM O’Neill Dam stores a reservoir that serves as the forebay to San Luis Dam. It was also completed in 1967 and is a RESERVOIR ELEVATION ON OCTOBER [7, I989=5 221 feet 2 ZONE 3 BLANKET Figure 17.—Maximum cross section of O’Neill Dam. LIOUEFIABLE SANDS homogeneous earth dam with an internal downstream blan— ket drain and exterior slope protection zones. It has a maxi- mum height of about 70 feet (fig. 17). Recent studies have indicated that the alluvial foundation contains in some locations a layer of potentially liquefiable sand lying immediately beneath a surface clay cap. Corrected Standard Penetration Test (SPT) blowcounts in this sand layer have been found to be as low as three in some areas. Because the dam may someday experience a peak acceleration as high as 0.55 g from a nearby M=6.5 earthquake, it had been con- cluded that these shallow layers could liquefy and result in a failure of the dam. As a result of these evaluations, suspect areas were treated in 1991 by removing and replacing the shal— low liquefiable sands downstream of the dam with compacted, imported material. Large downstream buttresses were also con- structed. O’Neill Dam is located about 37 miles from the Lorna Prieta fault-rupture zone (fig. 1). At the time of the earthquake, the reservoir was about 13 feet below the crest of the darn. Strong- motion sensors recorded a peak horizontal acceleration of 0. 10 g at the downstream toe and values of 0.11 g and 0.14 g at two locations on the crest. No damage was reported. Because of the planned repair to improve seismic stabil- ity, it was of interest to compare the predicted versus actual performance of the liquefiable sand layers during Lorna Prieta. To this end the downstream areas of two sites having the lowest SPT blowcounts, the station 100 and station 133 areas, were analyzed to determine the predicted factors of safety against the development of liquefaction. The result- ing calculated factors of safety against triggering liquefac- tion during Loma Prieta were between 1.2 and 1.5. These “predicted” factors of safety correspond well with the obser- vation of no damage at this dam. COSTS OF MAJOR REPAIRS FOR EARTHQUAKE DAMAGE Although most dams performed reasonably well during the earthquake, some required minor to major repairs. The approximate repair costs for some of the dams discussed in this report are summarized in table 8. CREST ELEVATION 234 feet MODERATELY DENSE CLAYEY AND SlLTY SANDS, SILTS, AND CLAYS. D24 PERFORMANCE OF THE BUILT ENVIRONMENT Table 8.—Approximate repair costs for earthquake damage Darn Approximate Repair Cost Austrian Lexington Guadalupe Chesbro Vasona 1 $ 2,500,000 $ 150,000 $ 100,000 $ 75,000 $ 20,000 1 Another $ 8 million was expended for the design and construction of a new spillway as a result of the earthquake. Table 9.—Peak accelerations measured at earth dams during the Lama Prieta earthquake [T, transverse direction; L, longitudinal direction; V denotes vertical direction] Maximum Peak accelerations (g) Dam height Base Abutment Crest (feet) T L V T L V T L V Lexington 205 .45 .41 .15 .39 .40 .22 .45 .34 .20 San Justo' 135 .26 .16 .40 .49 .33 Leroy Anderson1 235 .26 .25 .17 .07 .08 .05 .39 .26 .19 .23 .18 .16 .43 .32 .16 .38 .32 .23 San Luis 313 .04 .06 .02 .19 .25 .07 .07 .10 .03 .17 .14 .05 O'Neill 70 .08 .10 .05 .11 .11 .06 .14 .10 .06 Martinez 54 .09 .08 .02 .14 .15 .04 Del Valle 222 .04 .06 .03 .08 .08 .07 Contra Loma 88 .07 .05 .03 Note: 1 denotes that other records are available from other instruments at this dam STRONG MOTIONS RECORDED ON EMBANKMENT DAMS the studies by Maley and others (1989), Shakal and others (1989), and Wood and others (1991). Presented in figure 18 is a plot comparing the peak trans- verse accelerations measured at both the base and crest of The Lorna Prieta earthquake provided an excellent oppor- tunity to calibrate dynamic response techniques. As illus— trated in table 9, strong-motion records were recorded at eight embankment dams. This information was summarized from several earth and rockfill dams. These measurements include those made during the Loma Prieta earthquake as well as those made during previous events. As may be observed, the PERFORMANCE OF EARTH DAMS DURING THE LOMA PRIETA EARTHQUAKE D25 0.7 I I I I , / «\O‘fi’ «0 . a \‘Qp/ V 0.6 - 49/ — g 08/ E «0‘27 . E 0.5 _ §Q/ — a «9/ c.) Q/ o I Q/ 2 Q/ ' 0.4 L- 3‘ I _ 5 w/ I.I.I 3/. n: é U ’7 o 3 — :5 - Lu . U) "/ a: L“ l > O . <1) <2: 0.2 — J _ E I-' x / " ' I < O 1 J3“ I 1989 LOMA PRIETA — E o EARTHQUAKE I o PREVIOUS EARTHQUAKES 5'- 0 I I J 0 0.1 0.2 0.3 0.4 0.5 PEAK TRANSVERSE BASE ACCELERATION (9) Figure 18,—Comparison of peak base and crest transverse accel- erations measured at earth dams. points indicate that at low accelerations, the amplification through embankment dams is relatively large. However, as the peak base accelerations become larger, the amount of amplification is relatively low, possibly a result of increased damping or yielding of embankment materials. Also shown in figure 18 is a tentative upper bound curve. This curve should not necessarily be used for design purposes, but it may be useful as a verification tool in the performance of dynamic response analyses. CONCLUSIONS Several earth dams were subjected to relatively strong shak- ing during the Lorna Prieta earthquake and, in general, per- formed satisfactory. One major dam (Austrian) and one mi— nor dam (Soda Lake) developed moderate damage. A small number of other dams developed minor to moderate crack- ing which required repair. A major factor in the good performance of some dams was the fact that the reservoirs were commonly at less than half their maximum height. Consequently, major portions of the embankments were not as saturated as they would have been during full reservoir condition. Thus, the earth— quake was not the most critical test of these structures. Nev— ertheless, the performance and strong-motion data will pro- vide researchers with invaluable opportunities to learn more about how earth and rockfill dams behave during earth— quakes. ACKNOWLEDGMENTS The purpose of this paper was to present an overview of the performance of earth dams during the Loma Prieta earth- quake. To this end, the detailed inspection reports prepared by the California Division of Safety of Dams following the earthquake were invaluable. Providing other details were the generally excellent reports prepared by consulting engineers that are presented in the list of references. The authors also acknowledge Professor Raymond B. Seed and several staff members of the California Division of Safety of Dams for providing many useful pieces of information. The assistance of Thomas Holzer in the publication of this paper is also gratefully acknowledged. REFERENCES Aydin, Atilla, Johnson, Arvid M., and Fleming, Robert W., 1992, Right—lateral-reverse surface rupture along the San Andreas and Sargent faults associated with the October 17, 1989, Loma Prieta, California, Earthquake: Geology, v. 20, p. 1063-1067. Bureau, Gilles, Babbitt, Donald H., Bischoff, John A., Volpe, Rich- ard L., and Tepel, Robert E., 1989, Effects on dams of the Lorna Prieta Earthquake of October 17, 1989: United States Committee on Large Dams, Newsletter no. 90, p. 1-4. Creegan, Patrick, 1990, Newell Dam and the Loma Prieta earth- quake: United States Committee on Large Dams, Newsletter no. 91, p. 9-11. Lee, Kenneth L., and Roth, Wolfgang, 1977, Seismic stability analy- sis of Hawkins hydraulic fill: Journal of the Geotechnical En- gineering Division, American Society of Civil Engineers, V. 103, no. GT6, p. 627-644. Maley, R., Acosta, A., Ellis, F., Etheredge, E., Foote, L., Johnson, D., Porcella, R., Salsman, M., and Switzer, J., 1989, U. S. Geological Survey Strong—Motion Records from the Northern California (Loma Prieta) earthquake of October 17, 1989: U. S. Geological Survey Open-File Report 89-568. Rodda, K. V., Harlan, R. D., and Pardini, R. J ., 1990, Performance of Austrian Dam during the October 17, 1989 Lorna Prieta earthquake: United States Committee on Large Dams, N ews— letter no. 91, p. 7-9. Seed, H. Bolton, Makdisi, Faiz 1., and De Alba, Pedro, 1978, The performance of earth dams during earthquakes: Journal of the Geotechnical Engineering Division, American Society of Civil Engineers, v. 104, no. GT7, p. 967-994. Seed, R. B., Dickenson, S. E., Riemer, M. F., Bray, J. D., Sitar, N., Mitchell, J. K., Idriss, I. M., Kayen, R. E., Kropp, A., Harder, L. F., and Power, M. S., 1990, Preliminary report on the prin— cipal geotechnical aspects of the October 17, 1989 Loma Prieta earthquake: Earthquake Engineering Research Center, Report No. UCB/EERC-90/05, University of California, Berkeley. D26 PERFORMANCE OF THE BUILT ENVIRONMENT Shakal, A., Huang, M., Reichle, M., Ventura, C., Cao, T., Sherburne, R., Savage, M., Darragh, R., and Peterson, C., 1989, CSMIP strong-motion records from the Santa Cruz Mountains (Loma Prieta), California earthquake of 17 October 1989: California Division of Mines and Geology, Report No. OSMS 89-06. Spudich, Paul, 1996, Synopsis, in The Loma Prieta, California, earthquake of October 17, l989—Main-shock Characteristics: U. S. Geological Survey Professional Paper 1550-A. Volpe, R. L. & Associates, 1990a, Investigation of SCVWD Dams affected by the Loma Prieta earthquake of October 17, 1989: Report prepared for the Santa Clara Valley Water District. Volpe, R. L. & Associates, 1990b, Repair of SCVWD dams af- fected by the Lorna Prieta earthquake of October 17, 1989: Report prepared for the Santa Clara Valley Water District. Wahler Associates, 1990, Austrian Dam—Investigation and reme- dial construction following the October 17, 1989 Loma Prieta earthquake: Report prepared for the San Jose Water Company. WahlerAssociates (1992), Seismic Safety Evaluation of Mill Creek Dam: Report prepared for Lockheed Missiles and Space Com- pany. Wood, Chris, Copeland, David, and Viksne, Andy, 1991, Strong— motion data—Loma Prieta Earthquake of October 17, 1989, San Justo dam and dike, San Luis dam, O’Neill dam, Martinez dam: U. S. Bureau of Reclamation. THE LOMA PRIETA, CALIFORNIA, EARTHQUAKE OF OCTOBER 17, 1989: PERFORMANCE OF THE BUILT ENVIRONMENT EARTH STRUCTURES AND ENGINEERING CHARACTERIZATION OF GROUND MOTION ANALYSIS OF SOIL-NAILED EXCAVATIONS STABILITY DURING THE 1989 LOMA PRIETA EARTHQUAKE By Mladen Vucetic, University of California, Los Angeles; Mark R. Tufenkjian, California State University, Los Angeles; Guy Y. Felio, National Research Council, Ottawa, Canada; Pirooz Barar, Pirooz Barar & Associates, Oakland, California; and K. Ronald Chapman, Schnabel Foundation Co., Walnut Creek, Calif. CONTENTS Page Abstract D27 Introduction 27 Soil-nailing practice in California ------------------------------------------ 28 The Lorna Prieta earthquake 29 Characteristics and description of the walls ------------------------------ 30 Mountain View, 2350 El Camino Real (ECR Wall) --------------- 31 Mountain View, Kaiser Permanente Parking Garage (KPG Wall) 31 Santa Cruz, University of California at Santa Cruz (UCSC Wall) 35 San Jose Riverpark Project ------------------------------------------- 35 San Ramon, National Medical Enterprises Community Hospital (NME Wall) 35 San Francisco, Cresta Vista Apartments (CVA Wall) ------------- 36 Walnut Creek, Mini Storage Facility (MSW Wall) ---------------- 37 Richmond, Temporary Shoring Wall (TSW Wall) ---------------- 37 Methods of analysis 37 Possible reasons for the observed behavior ------------------------------- 38 Failure-mechanism assumption -------------------------------------- 39 Role of facing 43 Conclusions 43 Acknowledgments 44 References cited 44 ABSTRACT The performance of nine different grouted soil-nailed ex- cavations in the San Francisco Bay area during the Lorna Prieta earthquake is analyzed on the basis of postearthquake visual inspections, subsequent stability analyses, and dynamic centrifuge model tests. None of the excavations showed any signs of movements or similar distress, even though one of them was located in the vicinity of the earthquake epicenter where there was strong shaking and important seismic- related damage to other structures. The design and construc- tion practices of grouted soil-nailed excavations in Califor- nia are discussed. It is concluded that a combination of con- servative design and construction is the primary reason for excellent seismic stability. It is also confirmed that the method developed and used by Caltrans for calculating the factor of safety is suitable for the stability analysis of the grouted soil- nailed excavations encompassed by the study. This method is based on a bilinear failure surface and the so—called Ger- man mode of failure that considers two sliding blocks. INTRODUCTION Soil nailing is an in-situ technique of mechanically stabi— lizing soil masses which has been used in Europe for more than two decades (Stocker and others, 1979; Chapman and Ludwig, 1993; Federal Highway Administration, 1993). In North America, as well as in Japan (Japan Highway Public Corporation, 1987; Ochiai and others, 1992), soil nailing is steadily gaining popularity because it can be used with con- ventional shoring equipment, it reduces excavation time, it allows construction—related activities to proceed in restricted space, and it can produce significant savings over conven— tional shoring techniques in the proper ground conditions. The main feature of soil nailing is that it is an in—situ method where the existing natural soil is reinforced, as opposed to a backfill reinforcement. As shown in figure 1, the inclusions, commonly called nails, are installed during the excavation using a “top-down” construction procedure, unlike reinforced earth walls which are constructed from the bottom up. This allows soil retention in areas where little space is available for the excavation. The soil-nailing concept is to reinforce the soil with passive inclusions, so that the nailed soil mass behaves as a composite unit, similar to a gravity retaining wall supporting a soil backfill (Juran and Elias, 1991; Mitchell and Villet, 1987). In that sense, soil nailing also differs from the conventional tie-back excavation support since the soil nails are not prestressed; that is, their resistance can be mo- bilized only by the movement of soil mass or the face of the excavation to which the nails are fixed. Figures 2 to 4 show several soil-nailed retaining structures treated in this paper. At present, there are three major concerns about soil-nailed excavations: (l) the adequacy of the analysis or design meth— D27 D28 ods, (2) the long-term behavior, and (3) the performance during seismic loading. Items (1) and (2) have been addressed by many researchers, including recently by Gassler (1992), Juran and others (1990), Plumelle and others (1990), and Stocker and Riedinger (1990). With respect to soil-nailing performance during earthquakes, no full-scale field obser- vations were available until the Loma Prieta earthquake. During the earthquake, nine soil-nailed structures were sub— jected to different levels of shaking, including horizontal ground-surface accelerations probably as high as 0.4 g. In m 1. Excavation ms IIA ///~\ "7"L to m 20 aegrm Np. 2. Drilling of nail holes 3. Installation of nails 2 4. Construction of facing (could be also done before the nails gm installed) ”K— ”RI/l 5. Repeat excavation and steps 2 to 4 Figure l.—Steps in the construction of a grouted soil—nailed excavation. EARTH STRUCTURES AND ENGINEERING CHARACTERIZATION OF GROUND MOTION spite of such relatively high horizontal accelerations, these structures did not show any visible movements or other signs of distress (Felio and others, 1990; Hudson, 1990).A sys- tematic description of these structures and a discussion of possible reasons why they performed so well are the main purposes of this paper. More details about the corresponding investigation can be found inTufenkjian and Vucetic (1993). SOIL NAILIN G PRACTICE IN CALIFORNIA There are three major steps in the construction of a soil— nailed wall, as illustrated in figure 1. They are (l) excava- tion, (2) installation of nails, and (3) construction of facing. The excavation generally proceeds in stages ranging from 1.2 to 1.8 m in depth. One of the major requirements for successful soil-nailed systems is that the excavation be ca- pable of self-support for at least a few hours prior to nailing and construction of facing. For the most economical con- struction, however, the self-support should be able to last 1 to 2 days. As the excavation of each level proceeds, the nails are installed at predetermined locations. These reinforcing elements may be one of several types: driven, grouted, jet- grouted, or even pneumatically propelled into the ground (Myles and Bridle, 1991). However the vast majority of in- stallations are of the open drilled and grouted type (Chapman and Ludwig, 1993; Federal HighwayAdministration, 1993). In California, and NorthAmerica in general, the most popu- lar type of nails are the grouted nails, such as those shown in figures 2, 3 and 4, since in many locations the soil condi- tions allow the excavation to stand open long enough. Grouted nails generally consist of Grade 60 mild steel bars (15 to 45 mm in diameter) placed in boreholes of 100 to 250 mm diameter. Plastic centralizers are often used to ensure proper grout cover of the nail.A cement grout is then placed into the boreholes by gravity flow or low pressure.Typical horizontal and vertical spacings range from 1 to 3 m, de- pending upon the designer’s experience and soil conditions. The nails are generally inclined at 10° to 20° from the hori- zontal. Either before or after the nails are in place, a facing struc- ture is built. The facing is required to control soil erosion at the excavation face and reduce changes in the moisture con- tent of the soil. The most common type of facing is shotcrete layer, 100 to 250 mm thick, which is usually placed by the shotcrete method and which is reinforced with welded wire mesh. A typical detail of the nail connection to such facing is presented in figure 5. If necessary, a blanket of nonwoven geotexile is placed between the natural soil and the shotcrete to control the drainage. The grouted nail is attached to the facing by bolting the steel bar to a square plate usually 300 to 400 mm wide. For additional reinforcement and strength- ening of the facing, horizontal waler bars may be installed to connect the plates. Other methods of attachment are used for driven nails. For permanent walls, the shotcrete facing ANALYSIS OF SOIL-NAILED EXCAVATIONS STABILITY DURING THE 1989 LOMAPRIETA EARTHQUAKE may not provide for the aesthetic requirements of the project. In such cases, either cast-in-place reinforced concrete fac- ing or prefabricated panels can be used. Figure 6 shows pho- tos of large soil-nailed excavation structures recently com- pleted in California. THE LOMA PRIETA EARTHQUAKE The Loma Prieta earthquake (M,=7.1) was one of the most costly single natural disasters in US. history It caused ex- tensive damage, such as landslides in the epicentral region, liquefaction in various areas of the San Francisco Bay re- gion, structural distress to commercial, industrial, and resi- dential buildings, widespread disruption or total destruction of utility systems, and damage to critical transportation sys— tems. The earthquake has been the subject of a wide range of studies, many of them on geotechnical—related failures, as summarized by Seed and others (1991). Figure 7 presents an overview of the regional geology and the recorded peak horizontal ground-surface accelerations during the earthquake. The locations of the nine soil-nailed shotcrete facing | _L I 1.2 to proposed structure _L subgrade I I I I I I I I I I I I ele. 243.23 m : D29 walls considered in this paper are identified on the figure by stars, and the location of the epicenter by a circle.The figure shows that in the epicentral area the measured maximum horizontal ground-surface accelerations, amax, were as high as 0.64 g and the vertical up to 0.60 g. It can be seen that the soil-nailed walls in the northern region (in Richmond, San Francisco, Walnut Creek, and San Ramon) were subjected to seismic forces corresponding to amax of about 0.10 g. In the vicinity of the two walls in Mountain View, an amax of around 0.2 g was measured. In the vicinity of the wall in San Jose, amax was between 0.11 and 0.18 g. The largest amax (0.47 g) recorded near a soil-nailed wall was in Santa Cruz, some 16 km due west of the epicenter Most of these locations were visited and inspected 2 days after the earthquake by a team from the University of Cali- fornia, Los An geles (Felio and others, 1990), and some walls were inspected subsequently by design and construction com- panies. As stated earlier, no signs of distress or correspond- ing deformation were found on the walls, indicating excel- lent performance of such structures during moderate and strong shaking. existing grade soil nails Figure 2.—Cross section of a soil-nailed excavation for a building constructed in Santa Cruz (UCSC wall), Calif. (Felio and others, 1990). D30 EARTH STRUCTURES AND ENGINEERING CHARACTERIZATION OF GROUND MOTION CHARACTERISTICS AND DESCRIPTION OF THE WALLS The main characteristics of the nine walls inspected after the earthquake are summarized in table 1. In figure 8 the dimensions of the walls are presented in a uniform scale. The'variation of the geometry, characteristics of the walls, soil conditions, and estimated ground—surface accelerations are evident. The walls are further characterized in table 2 in terms of the following three dimensionless ratios commonly used as design criteria (Bruce and Jewell, 1987): maximum nail length Length ratio = ———-——-———-—- (1) excavation height hole diameter x nail length Bond ratio = ————————— (2) horizontal spacing x vertical spacing (nail diameter)2 Strength ratio = (3) horizontal spacing x vertical spacing Table 3 further compares the three dimensionless ratios computed for the nine San Francisco walls, with the values computed by Bruce and Jewell (1986, 1987) for soil-nailed structures with drilled and grouted nails constructed all over the world. The bond and strength ratios generally fall within the range of other soil-nailed retaining structures. However, the length ratios for the San Francisco walls are generally much higher than those calculated from other sites, suggest— ing that the San Francisco walls are more conservatively designed. Note from table 2 that the length ratio for the UC Santa Cruz wall is the smallest. This wall is apparently the least conservatively designed of all of the walls, yet it is lo- cated in the vicinity of the highest estimated peak horizontal ground-surface acceleration. In spite of these facts, no ob- V-DITCE / SHOTCRETE WALL #a son. NAIL x 6.1 1:: LONG x g WELDED WIRE MESH E m In a; 190 mm DIA- DRILLED HOLE “In DRAIN a 5 FILLED WITH CONCRETE o. 2 [IL I —H I a: m 0.4 m x 0.3 In DEEP FOOTING J l x \ —|lEll t I 2 #4 0.3 In WIDE X 0.3 m DEEP FOOTING Figure 3.—Cross section of a permanent soil—nailed retaining structure located in San Ramon (NME wall). Calif. (Barar, 1990). ANALYSIS OF SOIL-NAILED EXCAVATIONS STABILITY DURING THE 1989 LOMA PRIETA EARTHQUAKE D31 servable damage was noted on the Santa Cruz structure after the earthquake. The walls are described in greater detail be- low. MOUNTAIN VIEW, 2350 EL CAMINO REAL (ECR WALL) Nearly 280 m2 of soil-nailing construction was used to provide temporary shoring of an excavation for an office building. The concrete wall for the new structure was to be poured in front of the soil-nailed concrete facing. The sub— surface soil consisted of gravelly and clayey sand. The shear- strength parameters used in design were 6 = 9.6 kN/m2 and f = 30°, while the soil unit weight was assumed to be 17.3 kN/ m3. The soil-nailed wall was completed by May of 1989 and the excavation was still open when the earthquake struck. «rt 3 iv 3 n _. M It. 9.8 m =1.1m 15 degrees ”(Xv/- . - 1.5 m " l 300 mm diameter nails Postearthquake observations revealed only a few shallow hairline cracks in the concrete facing, typical of flexural cracking if the facing is considered as a vertical slab with the nails acting as reaction points. Note in table 1 that the facing of this wall was relatively thin (100 mm), while the estimated horizontal acceleration was considerable (0.21 to 0.27 g). MOUNTAIN VIEW, KAISER PERMANENTE PARKING GARAGE (KPG WALL) Approximately 380 m2 of shoring was provided for the construction of a parking garage. Soil nailing was used only on one side of the excavation, while a combination of other! shoring techniques were used on the remaining sides. The soil conditions at the site consisted of stiff sandy to clayey ( gutter reinforced shotcrete 2nd floor H---_- 1st floor J._-__.. 6.1 to 8.5 m long fooling Figure 4.—Cross section of a soil-nailed retaining wall constructed on a steep hill in San Francisco (CVA wall), Calif, to provide adequate space for an apartment complex (Felio and others, 1990). D32 EARTH STRUCTURES AND ENGINEERING CHARACTERIZATION OF GROUND MOTION silt overlying silty to sandy clay. The shear strength param— struction of the shoring was completed just 8 days before the eters used in design were c = 23.9 kN/m2 and f = 14°, while earthquake. The postearthquake observations revealed no the soil unit weight was assumed to be 18.8 kN/m3. The con- visible distress to the soil-nailed wall, while the opposite side Shotcrete ‘ l‘ Second application 50 mm First application 50 mm ‘ 4-#4 horizontal waler bars #8 Rebar (typical) 150 mm X 150 mm X 13 mm plate 100 mm X 100 mm - 8/8 mm wire mesh A Figure 5.——Connection between grouted nail and facing. A, Cross section of a typical connection (from Koemer, 1984). B, Strong reinforcement around the nail tip. ANALYSIS OF SOIL-NAILED EXCAVATION S STABILITY DURING THE 1989 LOMA PRIETA EARTHQUAKE D33 Figure 6.-—Soil—nailed structures recently completed in California. A, Soil-nailed excavation for an underground struc- ture of a building. B. A highway retaining soil-nailed wall showing different stages of the construction of permanent facing. D34 EARTH STRUCTURES AND ENGINEERING CHARACTERIZATION OF GROUND MOTION Walnut Creek / MSW wall \ “LSan Ramon ‘ NME wall .04 A uses 0 CSMIP 2 Mountain \fiew KPG wall Mountain View ECR wall San :lose K RPP wall P \ NA— APPHOX. ZONE \ 0F RUPYUHE Santa Cruz i J ucso wall 7'." SCALE E] DAV uuo D ALLUVIUM D ROCK AND SHALLOW RESIDUALSOIL Figure 7.—Overview of regional geology and recorded peak horizontal ground-surface accelerations during the Lorna Prieta earthquake (from Seed and others, 1991). ANALYSIS OF SOIL-NAILED EXCAVATIONS STABILITY DURING THE 1989 LOMAPRIETA EARTHQUAKE D35 of the excavation, which used cantilevered soldier beams with a concrete facing between the beams, revealed some vertical hairline cracks in the facing. SANTA CRUZ, UNIVERSITY OF CALIFORNIAAT SANTA CRUZ (UCSC WALL) Approximately 350 1112 of shoring was required to con- struct a new science library on the UCSC campus.The soil conditions at the site consisted of sandy silt to sandy clayey silt extending from the ground surface to a depth of approxi- mately 6.4 m. The soil has an average dry unit weight of 13.8 kN/m3 and a moisture content ranging from 26.2 per- cent in the clayey silt near the surface to 13.9 percent in the sandy silt at 6.4 m. Shear strength properties used in design were 0 = 23.9 kN/m2 and f = 25°. The cross section at the highest location of the wall is shown in figure 2. Construc- tion of the wall was completed on September 28, 1989, less than 3 weeks before the earthquake. Since three sides of the excavation were soil nailed, at least one side may have been subjected to the full strength of the earthquake, which pro- I. Mounuin View ECR 2. Mountain View KPG ‘1’ IT 9- SannCnIz UCSC .5 4. San 105: RF? 5. San Jose RPP I T 7.9- 17 6. San Ramon N'ME in 7. San Francisco CVA /// 8- Walnut Creek MSW I 2.7m J— 9. Richmond TSW Figure 8.—Dimensions of the investigated soil-nailed walls. duced in the vicinity of the excavation peak horizontal ground-surface accelerations of about 0.47 g. It should be noted that this wall was located closest to the epicenter and presumably was subjected to the strongest shaking, while at the same time it had the smallest length ratio and thinnest facing among the nine walls examined (see tables 1 and 2). Prior to the earthquake, some wall and column spread foot- ings had been poured (see fig. 2).A postearthquake inspec- tion revealed significant cracking in the concrete of the foot— ings. This cracking was not attributed to shrinkage since foun- dations constructed after October 17 showed fewer cracks. As opposed to that, the inspection of the soil-nailed wall af- ter the earthquake revealed no cracking. A week after the earthquake, nine nails were tested to 150 percent of their design pull—out load. The tests showed no loss in the carry— ing capacity of the nails due to the seismic activity SAN JOSE, RIVERPARK PROJECT (2 RPP WALLS) These two retaining walls were designed and built as per- manent structures along the Guadalupe River in San Jose, approximately 40 km north of the epicenter The subsurface soil consists of silty and sandy clays to a depth of about 4.5 to 6 m. According to the geotechnical report, these clays have an intermediate to high plasticity with an approximate aver- age dry unit weight and moisture content of 14.1 kN/m’ and 22 perecent, respectively, and an undrained shear strength ranging from 72 to 240 kN/mz, as interpolated from static cone penetration tests. The clays are underlain by a 3 m zone of dense, clayey, silty, gravelly sand with an average dry unit weight and moisture content of about 17.3 kN/m” and 15 percent, respectively. The shear strength parameters used in design were c = 23.9 kN/m2 and f = 0°, while the total unit weight was assumed to be 19.6 kN/Irf‘. Since these are per- manent walls, the concrete surface was finished off with ar- chitectural concrete and in some places clad with granite. The postearthquake observations revealed no signs of dis— tress. SAN RAMON, NATIONAL MEDICAL ENTER- PRISES COMMUNITY HOSPITAL (NME WALL) This soil—nailed retaining structure forms a part of a per- manent retaining wall used for the roads and landscape that surround the medical center. The wall cross section is shown in figure 3. According to the geotechnical report, the soil conditions consist mainly of engineered fill up to a maxi- mum depth of 24 m, generated from cut-and-fill operations performed previously. Therefore, the soil-nailed retaining structure was built in fill material. The fill consists of sandy and silty clay of moderate to high plasticity, with an average dry unit weight and moisture content of approximately 17.1 kN/m3 and 18 percent, respectively. The shear strength prop- D36 Table l.—Summary of soil-nailed walls investigated EARTH STRUCTURES AND ENGINEERING CHARACTERIZATION OF GROUND MOTION Project Location Height Nail details Facing General Soil Shear strength Estimated No. (see fig. 7) of wall Spacing Length Diameter Inclination thickness soil unit design parameters horizontal (m) (m) (m) Grout Rebar (degrees) (mm) Type weight c ¢ ground (mm) (mm) (KN/m3 (KN/m1) (degrees) surface ) acceleration (3L Mountain 1.7V €13ch 1 View 6.1 5.2 300 32 20 100 sand 17.3 9.6 30 0.21 to 0.27 ECR 1.5H Mountain 2.0V sandy to 2 View 4.0 3.4 300 25 20 100 clayey silt 18.8 23.9 14 0.21 to 0.27 KPG 1.5H Santa 4.6 1.5V sandy to 3 Cruz 3.4 300 32 20 75 clayey silt 18.1 23.9 25 0.47 UC SC 2-levels 1 .8H San Jose 3.7 1.8V alluvial 4 6.1 180 25 15 200 clay, s1lt 19.6 23.9 0 0.10 to 0.15 RPPl 4—levels 1.811 and sand San Jose 1.8V 311““ 5 2.4 6.1 180 25 15 200 clay. silt 19.6 23.9 0 0.10 to 0.15 RPPZ 1.811 and sand San Ramon st engineered 6 NME 5.5 1 8H 6.1 190 25 15 150 1111 18.8 47.9 0 0.05 to 0.10 fill over San silty clay 7 Francisco 9-3 143V 6-1 1° 300 25 15 200 and highly 19.6 9.6 35 0.10 to 0.15 CVA 2-levels 1.5H 3-5 weathered siltstone Walnut (one fill over 8 Creek 2.7 row) 4.6 190 25 15 200 medium 18.8 14.4 28 0.01 to 0.10 stiff to stiff MSW 181-! Clay Richmond 7_9 1.8V 6.1 to alluvium 9 TSW 1.8H 9.1 190 25 15 200 deposits 18.8 19.1 28 0.05 to 0.15 erties used in design were 6 = 47.9 kN/m2 and f = 0°. Since this is a permanent structure, the facing of the soil—nailed wall was finished with a colored architectural concrete fin- ish. The postearthquake walk-through revealed that the sur- face of the concrete remained smooth and free of cracks. Table 2.—Dimenri0nless ratim~ for San Francisco area soil-nailed walls SAN FRANCISCO, CRESTA VISTA APARTMENTS (CVA WALL) This wall demonstrates the unique concept of using soil nailing on a permanent basis to retain the slope and cut on a Length ratio Bond ratio Strength ratio (103) Project NO- Location max. nail leng th hole dia. x nail 1e ngth (nail dia. ) 2 excav. height H. spacing x V. spacing H. Spacing x V. spacing 1 Mountain View 0.85 0.61 0.40 ECR 2 Mountain View 0.85 0.34 0.21 KPG 3 Santa Cruz 0.74 0.38 0.38 UCSC 4 San Jose 1.6 0.34 0.19 RPP 5 San Jose 25 0.34 0.19 RPP 6 San Ramon 1.1 0.36 0.19 7 San Francisco 1.5 to 1.7 0.68 to 0.94 0.28 CVA 8 Walnut Creek 1.7 not applicable not applicable MSW 9 Richmond 1.2 0.36 to 0.53 0.19 TSW ANALYSIS OF SOIL-NAILED EXCAVATIONS STABILITY DURING THE 1989 LOMA PRIETA EARTHQUAKE D37 Table 3.—Dimensionless ratios for soil-nailed walls Drilled and grouted in Drilled and grouted in granular soils morraine and marl San Francisco (Bruce and Jewell, (Bruce and Jewell, walls 1986, 1987) 1986, 1987) Length ratio 0.5 to 0.8 0.5 to 1.0 0.7 to 0.25 Bond ratio 0.3 to 0.8 0.15 to 0.20 0.34 to 0.94 Strength ratio (103) 0.4 to 0.8 0.1 to 0.25 0.19 to 0.40 steep hill to make room for the development of a housing project. The wall cross section is shown in figure 4.The 9.8- m—high soil—nailed structure was constructed at the toe of a 45.7-m-high slope to allow for the construction of apartment units. The wall is about 90 m long and consists of two levels. Due to the permanent nature of the structure, a 200 mm rein- forced concrete facing and a small footing at the base were used. The soil conditions at the site can be described as col- luvium and residual soil deposits.The design parameters used were cohesion c = 9.6 kN/m2 and an angle of internal fric- tion of f = 35°. The inspection that took place 3 days after the earthquake showed no signs of distress to the wall and no indications of lateral movements or tension cracks in the hill behind the wall. WALNUT CREEK, MINI STORAGE FACILITY (MSW WALL) The project consisted of a three—story building with two levels above grade and one level of basement below The soil-nailed wall was integrated into the final basement wall. The soil at the site consists mainly of fill material up to 3 m depth, including a nonuniform mixture of gravel, sand, and clay. The underlying soil consists of stiif silty clay. The av- erage dry unit weight and moisture content of the fill is 16.5 kN/m3 and 20 percent, respectively. The shear strength pa— rameters assumed in design were 0 = 14.4 kN/m2 and f = 28°, while the total unit weight of the soil was assumed to be 18.8 kN/m3. The postearthquake observations revealed no signs of distress on the surface of the wall or at grade behind the wall. RICHMOND, TEMPORARY SHORING WALL (TSW WALL) Soil nailing was used here to construct a temporary shor- ing wall which has the tallest single-level vertical face of any of the walls examined in this paper A permanent retain- ing wall was eventually built in front of the soil-nailed wall. Unfortunately, soil stratigraphy data is not available for this site. However, the shear strength parameters used in design were c = 19.1 kN/m2 and f = 28°, while the unit weight of the soil was assumed to be 18.8 kN/m3. Awalk-through of the site following the earthquake did not reveal any signs of dis— tress attributable to seismic activity. METHODS OF ANALYSIS Most of the current design methods for soil-nailed retain— ing structures under static loads are derived from classical slope—stability analyses, which incorporate a limit equilib- rium approach. Accordingly, they evaluate global factors of safety along assumed failure surfaces such as those shown in figure 9. They are usually referred to as the German method (Stocker and others, 1979; Gassler and Gudehus, 1981; Lambe and Jayaratne, 1987), Davis method (Shen and oth- ers, 1981; Bang and others, 1992), French method (Schlosser and others, 1983), and Caltrans method (computer program SNAIL: Caltrans, 1993). The differences in the methods re- sult from the definition of the factor of safety assumed fail- ure surface shape, and the assumed contribution of the soil nails to the stability. In that respect, the methods are contra- dictory, and because of the lack of full-scale observations of actual failure mechanisms, different points of View about their applicability have emerged. The German method (fig. 9A) assumes a bilinear failure surface passing through the toe of the excavation.The fail- ing soil mass is broken into two parts.The first part contains most of the nailed soil mass, while the second part forms the active earth pressure wedge behind it—behind the “soil- nailed gravity wall.” The analysis considers the tensile and pull-out resistance of the nails crossing the failure surface and, of course, the forces of interaction between the nailed mass and active wedge behind it. The assumed failure sur- faces are consistent with the concept of soil nailing, that is, the nailed soil mass behaves like a reinforced block. The Davis method incorporates a parabolic failure surface that also passes through the toe, as shown in figure 9B. The sliding surface either passes entirely through the nails or in- tersects the ground surface somewhere beyond the reinforced zone. In the analysis, the tensile and pull-out resistance of the nails crossing the failure surface are considered the gov- erning stabilizing forces. Because of its successful track D38 record and easy implementation, it has been a popular de- sign method in the United States. This has been the case in spite of the fact that the assumption of a parabolic slip sur— face (which does not change slope when crossing from the nonreinforced to the reinforced zone) has not been adequately verified by laboratory or field tests. The French method follows procedures similar to the Davis method, but assumes a circular failure surface passing en- tirely through the nails, as shown in figure 9C. But, unlike the previous two methods, this method considers the shear and bending of the nails, which adds to the complexity of the analysis. The Caltrans method also assumes a bilinear failure sur- face, just like the German method. However, unlike in the German method, the bilinear failure surface may pass en- tirely through the nails (see fig. 9D). More recently, a kinematical limit analysis approach has been proposed for the design of soil-nailed retaining struc- tures (J uran and others, 1990). It differs from the other analy- sis procedures in that it suggests a method for estimating nail forces. In this way, it may provide a check on local sta- bility at each level of nail reinforcement. The method as— sumes that the failure surface is defined by a log-spiral pass- ing partially through the nails and that the failure occurs by rotation of a quasi-rigid body along this surface. All of the San Francisco walls examined in this study were designed using a modified version of the Davis method (Barar, 1990; Felio and others, 1990). Seismic forces were accounted for by using an equivalent static horizontal force H = W x kh, applied at the center of gravity of the potentially unstable soil nailed mass, where W is the weight of the mov- ing soil mass and kh is the horizontal seismic coefficient. It should be mentioned at this point that the Davis method, as well as the German and Caltrans methods, has a certain degree of the inherent conservatism in that the potential stabi- lizing effects of the shear and bending resistances of the nails are ignored. New studies (Jewell and Pedley, 1992; Federal Highway Administration, 1993) show, however, that the ef- fects of bending stiffness are small. Also, the contribution of the steel reinforced facing to the strength of the system is unaccounted for. The lack of full understanding of the role of facing in the global and local stability apparently led to the difference by a factor of 3 (75 mm vs. 200 mm) in the thicknesses of the facing among the nine walls under con- sideration. Some designers and construction companies feel comfortable ‘with thinner facing, while some prefer more conservative thicker facing. Figure 5, for example, illustrates a rather heavily reinforced facing with a sturdy nail contact. The role of the facing in soil reinforcing stability is just be- ginning to be studied as a separate issue (Tatsuoka, 1992), and it should definitely be given more attention in the fu- ture. The factors of safety for the CVA and TSW soil-nailed walls obtained by the Davis method, modified to account for earthquake forces by the pseudostatic technique, are pre- EARTH STRUCTURES AND ENGINEERING CHARACTERIZATION OF GROUND MOTION sented in table 4. The location of the assumed failure sur- faces that yield minimum factors of safety for the TSW wall are shown in figure 10. In general, the factors of safety are relatively low, especially for the range of estimated peak horizontal ground accelerations during the Loma Prieta earth- quake. According to such low factors of safety, and given the fact that some soil-nailed structures were probably sub- jected to much larger horizontal forces, some visible dam- age should have occurred during the earthquake. This should have been expected in particular for the UCSC Wall in Santa Cruz, which had the smallest length ratio and facing thick- ness, and yet is likely to have undergone horizontal seismic forces as large as 0.4 g. The lack of visible damage on any of the walls, except very thin cracks on the ECR wall, suggests that either the design, analysis, or construction, or most likely their combination, may have been more conservative than necessary. The lack of damage also indicates that the assumed failure surface and mechanism of failure of the Davis method may not be fully appropriate for the nine walls treated here. In the following section, the components of the analysis, design, and construction that appear to be on the conserva- tive side, and therefore could be responsible for such excel- lent seismic performance, are discussed. POSSIBLE REASONS FOR THE OBSERVED BEHAVIOR Since soil nailing is a relatively new soil-stabilization tech- nique, with very little practical experience of full-scale static failures and practically no experience of seismic failures, the design and construction are usually quite conservative. The preliminary design of a soil-nailed retaining structure proceeds much like that of retaining walls, by trial and error. Based mainly on the expected excavation height and the soil strength properties, tentative characteristics of nails and fac- ing (length, diameter, horizontal, and vertical spacings of nails, and the thickness and reinforcement of facing, etc.) can be assumed and some sort of stability analysis performed. The assumed values and characteristics depend primarily on the designer’s experience with other satisfactorily constructed soil-nailed walls, which may lead to an overly conservative design, and to a lesser extent on charts and dimensionless parameters derived by others, such as those by Bruce and Jewell (1986, 1987) and Guilloux and Schlosser (1982). Table 3 shows, for example, that the length ratios and bond ratios for the nine walls considered here are on the conservative side in comparison with the values suggested by Bruce and Jewell (1986, 1987). The main components of the conservative design and con- struction for seismic loads include (1) conservative and most probably unrealistic assumption of the failure mechanism, (2) no consideration of the contribution of the facing in the stability analysis, and (3) conservative construction due to the lack of field experience and understanding of the various -"‘ a. ANALYSIS OF SOIL-NAILED EXCAVATIONS STABILITY DURING THE 1989 LOMA PRIETA EARTHQUAKE D39 aspects of soil-nailed excavation seismic response. The first two components are discussed below. FAILURE-MECHAN ISM ASSUMPTION Due to a lack of full-scale observations of failures and corresponding failure mechanisms under both static and seis- mic loads, there is currently no consensus among designers on which failure mode is the most realistic among the four basic modes presented in figure 9. To cast more light on pos- sible modes of failures under dynamic loads, two series of A Bilinear B Parabolic C Circular Bilinear Figure 9.—Assumed failure surfaces used in analyses. A, German Method. B, Davis Method. C, French Method. D, Caltrans Method. dynamic centrifuge tests were conducted, one in 1991 (Tufenkjian and others, 1991; Tufenkjian and Vucetic, 1992; Vucetic and others, 1993) and the other in 1996 (Vucetic and others, 1996). Figures 11 and 12 show the main features of the models tested and results obtained in 1991. The centrifuge tests were performed at the Rensselaer Polytechnic Institute (RPI) Geotechnical Centrifuge Research Center on a 3-m radius Accutronic 665-1 centrifuge (Elgamal and others, 1991). The scale factor was 50in all of the tests. Accordingly, to simulate prototype geostatic stresses, the models had to undergo a centrifugal acceleration of 50 g. For dynamic testing, a servo—hydraulic earthquake simula- tion shaker mounted on the centrifuge platform was used. Four models were tested in 1991. They represented 7.6-m—high soil-nailed excavations with grouted nails, corresponding roughly to an excavation height of a two- to three-story underground garage. The effects of two important characteristics of soil-nailed structures were tested: the length of nails (expressed in terms of the length ratio), and the axial and flexural rigidities of the nails. Three length ratios were tested, 0.33, 0.67, and 1.0, which could be characterized as the ratios corresponding to short, medium, and long nails (see table 3). These ratios cover ap- proximately five out of the nine walls listed in table 2. The four other walls have very large length ratios between 1.5 and 2.5. Two axial and flexural rigidities of the nails were used, one that can be considered regular and the other than can be considered small. By varying the axial and flexural rigidities of the soil nails, their effect on the failure surface geometry and stability could be assessed. As shown in fig- ures 11A and 11C, three displacement transducers (LVDT’s) were used to record the lateral movements of the facing and the vertical soil settlement behind the facing. During dynamic loading, four accelerometers were utilized to measure the accelerations of the model box and in various locatidns within the model box. The soil used in the experiments was fine sand. The sand was partially saturated to generate an appar- ent cohesion, necessary for a rough simulation of in-situ co- hesion and cementation. Other details of the 1991 testing are described by Vucetic and others (1993). Figures 118 and 12 show a typical failure mechanism ob- tained in the tests under horizontal dynamic loads. In all four tests the failure surface never started at the ground surface above the nails. Instead, it started at the ground surface be- hind the ends of the nails. Figure 113 reveals that the failure mechanism involves three soil “zones” and two soil “blocks,” with two failure surfaces, one of which consists of two parts. The primary failure surface extends from behind the nails at the ground surface down to the end of the second row of nails, at which point it changes curvature and continues down to the bottom of the excavation through the toe. The second- ary failure surface develops within the sliding soil mass and divides zones 1 and 3. Such deformation patterns after the tests point to the following failure mechanism. The soil above the second row of nails in zone 1 moves horizontally under D40 EARTH STRUCTURES AND ENGINEERING CHARACTERIZAF ION OF GROUND MOTION Table 4.—Calculated factors of safety using the Davis method (see also Hudson, 1990) Horizontal acceleration Cresta Vista apartments Temporary shoring wall coefficient, k, CVA TSW 0 1.31 1.19 0.1 1.14 1.06 0.2 1.00 0.94 Indicates the factors of safety corresponding to the range of estimated horizontal peak ground—surface accelerations near the site. large inertial forces as a relatively rigid block held together by the nails. Consequently, the soil in zone 2 is pushed out- ward by the horizontal friction along the interface between the upper zone 1 and the lower zone 2.According1y, the fail- ure surface passes through the bottom row of nails. In such a mechanism, the bottom nails obviously act as anchors be— tween the back soil and the facing, while the top nails hold the soil together in the upper part of the excavation.As zones 1 and 2 move horizontally outward during seismic shaking, the lateral stresses in zone 3 are greatly reduced. Conse- quently, zone 3 represents a typical failure wedge behind a retaining wall, the retaining wall being zone 1.This mecha- nism and kinematics of the soil movement resemble the ge- ometry of German method for static stability evaluation, shown in figure 9A (Gassler and Gudehus, 1981), while they contradict the assumption that rotation of one monolith oc- curs along a continuous circular or parabolic failure surface. To examine the factors governing the failure corresponding to the above mechanism, the forces and factors of safety for the TSW Wall in Richmond (see fig. 10 and table 4) are reevaluated. Figure 13 shows the assumed failure surfaces and governing forces, while figure 14 shows the corresponding polygons of forces.To account for the eifects of dynamic horizontal forces the pseudostatic method of analysis is used again, where the dynamic action is represented by the static horizontal forceH = W X kh. Two definitions of the factor of safety, FS, based on the German Design parameters: ‘7 = 18.8 kN/mJ c = 19.1 kN/m2 ¢ = 28° Zone of assumed parabolic failure surfaces passing through the nails. Figure 10.——Assumed failure surfaces passing through the nails for the factor of safety evaluation of theTSW wall in Richmond using Davis method and its modifications. type of failure mechanism are considered below First is the definition for static stability proposed by Stocker, Korber, Gassler and Gudehus (1979), which is adapted here for the dynamic stability by adding to the poligon of forces a horizontal force H = W X kh. The second is the definition proposed and used by Caltrans (1993).Accordingly, the two methods for the calculation of F S are called here the SKGG method and the Caltrans method. According to the SKGG method, the factor of safety is calculated as FS = . (4) where Z, = cumulative axial pull-out force of the nails beyond the failure surface and Z, = mobilized cumulative axial force of the nails beyond the failure surface. There- fore, the entire factor of safety is based on the pullout of the nails. For the TSW wall, F5 for different seismic coeffi- cients, kh, was calculated. The results of this calculation are presented in terms of the FS vs. kh relationship in figure 15. On the same figure the equivalent relationship between F S and k}, obtained by the Caltrans method is presented as well. The bilinear failure surface assumed in the Caltrans method is similar to the failure surfaces assumed in the SKGG method and thus to the deformation patterns and failure sur- faces observed in the centrifuge tests. In fact, as indicated in figure 13, the forces and their positions relative to the free bodies for the TSW wall are the same for the SKGG and Caltrans methods. However, the methods differ fundamen- tally in their definitions of the factor of safety. The Caltrans method applies a unique factor of safety to the soil cohesion, c, soil friction angle, a, and the cumulative nail pullout force, c’ = c/FS = mobilized cohesion, a” = tan‘1 [(tan ¢)/F S] = mobilized friction angle, and Z, = Za/FS = mobilized pullout force. The method then utilizes these “mobilized” parameters in the force equilibrium equations to solve for the interwedge forces, F (see figs. 13 and 14). Since these forces must be equal in magnitude and opposite in direction, assumed value of F S is systematically varied until this condition is fulfilled, which then yields the corresponding FS used in design. The Caltrans method has been coded into a computer program- ANALYSIS OF SOIL-NAILED EXCAVATION S STABILITY DURING THE 1989 LOMA PRIETA EARTHQUAKE D41 ming language and can be run on personal computers. The computer program is called SNAIL (Caltrans, 1993). By in- 152(mm)-—+ FACING 305(mm) f LVDT 3 ' - Q LVDTI 152(mm) - 4 LVDTZ 76(mm) + 457(mm) A - Accelerometer > Lvnr _ _, 206(mm) —T . “JEWEL- 10 ACCELERATION CYCLES B u e a- u MODEL DISPLACEMENT (mm) ‘1 I) C) no I1 III I) TIME (ms) C Figure 11.—Features of the typical centrifuge model test with length ratio of 0.67 (Tufenkjian and others. 1991; Vucetic and oth- ers. 1993). A, Longitudinal cross section of the soil-nailed excava- tion centrifuge model box. B, Failure mechanism obtained in the centrifuge due to strong horizontal shaking. C, Typical records of soil mass movements during shaking with 10 cycles of 0.27 g cy- clic acceleration amplitude. putting the geometry of the slope and details of the soil strength and nail properties, the program can systematically vary the location of the bilinear failure surfaces, until the one producing the lowest factor of safety is found. The pro- gram also has an option to calculate the FS for a specified surface, as well as for considering seismic forces by the pseudostatic technique. Several interesting conclusions can be derived from fig- ure 15. First is that for FS :1 (the conditions of the failure of the wall), kh 20.37 is obtained by both methods. Second, this kh value is much larger than kh z0.1 to 0.20 correspond- ing to FS = 1 calculated according to the Davis method (see table 4). Third, kh 20.37 is in relatively good agreement with the amplitude of the cyclic acceleration of 0.45 g that was required in the centrifuge testing for the failure of the soil-Y nailed excavation model of similar length ratio (see Vucetic and others, 1993). And fourth, the FS versus kh relationship for the SKGG method has a singularity point, while the same relationship for the Caltrans method does not. Based on these observations it can be concluded that the failure mechanism according to the German method seems to be more appropriate than that of the Davis method. However, figure 15 also shows that using the SKGG method to calculate the factor of safety may not be suitable for the calculation of stability involving the horizontal forces (W X kh ), because it is too sensitive to the variation of kh . By varying kh from 0.3 to 0.4, FS varies from -1 to '00 and then from +00 to 0.7—that is, as noted above, the function FS = f (kh) has a singularity point. The reason for such sensitivity of FS with respect to kh can be easily understood from the polygon of forces in figure 148. For example, if the force H1=Wlx kh is increased by only 15 percent, the Z, force will double—that is, change by 100 percent. Consequently, the FS = Z“ /Ze will change dramatically too. Such sensitivity of F S comes from the fact that FS is defined on the basis of forces which are of secondary importance for the stability of the structure. In other words, force 22 is relatively small compared to the other forces in the polygon. More dominant forces are apparently the reaction force Q1 and cohesion force C ‘ mobilized along the failure surface. In the configuration of forces such as shown in the poly- gon in figure 143, corresponding to the German method type of failure mechanism, the role of the nails is predominantly to interact with the soil and form the nailed block. Such a large soil block is evidently seismically very stable, and its stability is governed by the large forces of friction and cohe- sion at the interfaces with the surrounding soil, not by the small forces such as 25. This, of course, would change if 2‘ is relatively large, that is, corresponding to very long nails in- stalled deep beyond the failure surface. In such a case the kinematics of the failure would be different too. Instead of predominantly sliding along the failure surfaces, the facing and thus the soil mass would be forced to rotate around the bottom row of nails which are anchored beyond the failure surface. D42 EARTH STRUCTURES AND ENGINEERING CHARACTERIZATION OF GROUND MOTION MODEL BOX EXCDAVATE ACCELEROGRAM Figure 12.—Excavated model of a test with the length ratio of 0.33 (short nails) shaken by 10 cycles of the uniform acceleration amplitude of 0.10 g. As opposed to the SKGG method of stability evaluation, the Caltrans method does incorporate all significant forces in the definition of F S and consequently yields a more mean— ingful relationship between FS and kh. This, along with the above discussion, leads to the following conclusions: (1) the German-method type of failure mechanism seems to be ap- propriate, (2) the Caltrans method for calculating F S seems to yield appropriate and meaningful results, and (3) the Davis method used in the design of the San Francisco Bay area soil-nailed walls seems to be overly conservative, apparently because it employs an unlikely failure mechanism. To confirm the above conclusions, the factors of safety of the eight soil-nailed walls subjected to the Lorna Prieta earth- quake were calculated by the Caltrans method. The four- level RPP wall located in San Jose could not be accurately reproduced using the SNAIL program. The wall geometries were scaled from figure 8 and the soil design parameters were taken from table 1. The calculated factors of safety are shown in table 5, where those corresponding to the range of estimated horizontal ground-surface accelerations near each soil-nailed wall are indicated by double—headed arrows. Note that the factors of safety are generally much greater than unity for the range of estimated ground-surface accelerations, even for the UCSC wall which was subjected to horizontal accel- erations in the range of 0.4 to 0.5 g. Recall from Table 4, for example, that the Davis method predicted failure (FS = 1.0) for the TSW wall for the acceleration coefficient between Hz=(W,xkh) C3) H,=(w,xk.,) @ . W2 7 Figure 13.——Failure surfaces and forces for the TSW wall employed in both SKGG and CALTRANS methods for the calculation of the factor of safety according to the German type of failure. ANALYSIS OF SOIL-NAILED EXCAVATIONS STABILITY DURING THE 1989 LOMA PRIETA EARTHQUAKE D43 0.10 and 0.20, while the Caltrans method required an ac- celeration coefficient between 0.4 and 0.5. Such large val— ues of F S obtained by Caltrans method are in agreement with the excellent performance of the walls during the earthquake. ROLE OF FACING None of the methods discussed above account explicitly for the contribution of the facing in the evaluation of the global factor of safety, although they do incorporate the evalu- ation of punching shear around the nail connection. In other .ze Q1 Block: ZONES 1 AND 2 Block: ZONE 3 F H1 FB Figure 14.———Polygons of forces identified in figure 13. A, Polygon for block comprising zone 3. B, Polygon for block comprising zones 1 and 2. words, the factor of safety is calculated without considering axial and flexural rigidities of the facing. In that respect, there is no consensus on what the contribution of the facing to the global stability of soil nailed structure really is. However, it is obvious that stronger facing and stronger contact between the facing and the nails will make the nailed soil mass more coherent. The failure mechanism of such a coherent soil mass is likely to be of the German type—that is, behaving as a large seismically stable block. In addition, the inability of the nails (which are firmly fixed to the facing) to move freely decreases the likelihood of local failures, especially in the zones most critically stressed during construction and seis— mic loading. As suggested earlier, the lack of full understand- ing of the role of facing in the global and local stability ap- parently led to the difference by a factor of 3 (75 mm vs. 200 mm) in the thicknesses of the facing between the nine walls considered here. CONCLUSIONS Postearthquake inspections of nine soil-nailed walls fol- lowing the Loma Prieta earthquake indicated superior per- formance and no signs of distress, even though one of the walls was subjected to horizontal accelerations probably as high as 0.4 g. It was shown that the excellent performance may be attributed to a conservative design, generally con- servative stability analysis which is mainly the result of an CALTRANs METHOD - OK I I V l I Figure 15.—Variation of the factors of safety, F S . with the seismic coefficient, kh, for the TSW wall. D44 EARTH STRUCTURES AND ENGINEERING CHARACTERIZAF ION OF GROUND MOTION Table 5.—Calculated factors of safety using the Caltrans method for San Francisco area soil-nailed walls Horizontal Soil—nailed wall acceleration ECR KPG UCSC RPP NME CVA MSW TSW coefficient, k, (single— level) 0.0 2.28 2.53 3.30 5.15 1.88 1.59 3.07 1.64 0.1 2.00 2.26 2.75 4.44 1.52 1.35 2.68 1.51 0.2 1.74 1.96 2.32 2.80 1.27 1.17 2.35 1.39 0.3 1.51 1.71 2.01 1.99 1.10 1.02 1.97 1.19 0.4 1.31 1.43 1.76 1.54 0.97 0.89 1.64 1.03 0.5 1.16 1.18 1.53 1.26 0.86 0.78 1.39 0.91 Indicates the factors of safety corresponding to the range of estimated horizontal peak ground-surface accelerations near the site. unlikely mechanism and geometry of failure, and conserva— tive construction. Because seismic failures of soil-nailed excavations have not occurred in the past and are therefore absent from the literature, dynamic centrifuge testing was performed to pro- vide evidence of the most probable failure mechanism. The centrifuge testing revealed that the most likely failure mecha- nism is the German type of failure mechanism. Furthermore, a simple analysis of the dynamic centrifuge test results and field observations showed that the Caltrans method for cal- culating the factor of safety, which also incorporates the German type of failure mechanism, yields very consistent and logical results. Accordingly, the Caltrans method imple- mented by the computer program SNAIL seems to be an appropriate method for calculating the static and dynamic stability of grouted soil—nailed excavations of the type dis— cussed in this paper. ' ACKNOWLEDGMENTS The investigations described in this paper were supported by the National Science Foundation through the grants BCS- 9001610, BCS-9011819 and BCS-9224479. This support is gratefully acknowledged. We would also like to thank pro— fessors R. Dobry, A.-W. Elgamal, and TE Zimmie, Dr. P. Van Laak, and Dr. L. Liu, all from the RPI Geotechnical Centrifuge Research Center, for their excellent service and continued assistance. Special thanks go to Dr: Macan Doroudian, former student at UCLA, for his help in perform- ing the centrifuge tests. We would also like to recognize Mr Ken J ackura from Caltrans who supplied the computer pro- gram SNAIL and provided valuable suggestions. REFERENCES CITED Bang, 8., Kroetch, PP, and Shen, C.K., 1992,Analysis of soil nail- ing system, in Proceedings of the International Symposium on Earth Reinforcement Practice, Fukuoka, Kyushu, Japan: Balkema, Rotterdam, v. 1, p. 457—462. Barar, P., 1990, The behavior of five soil nailed earth retaining struc- tures during the Loma Prieta earthquake of October 17, 1989: Report prepared for the Department of Civil Engineering, University of California, LosAngeles, 101 p. Bruce, D.A., and Jewell, RA, 1986, Soil nailing—Application and practice, part 1: Ground Engineering, November p. 10-15. 1987, Soil nailing—Application and practice, pan 2: Ground Engineering, January, p. 21-33. Caltrans, 1993, California Department ofTransportation—SNAIL Computer Program, Version 2.06: California Department of Transportation, Office of Geotechnical Engineering, Sacra- mento, California. Chapman, K.R., and Ludwig, C., l993,ADSC participates in FHWA intemational geotechnology tour: Drilling Foundation, Decem- ber/January, p. 17-19. Elgamal, A.-W., Dobry, R., and Van Laak, P., 1991, Design con- struction and operation of 100 g—ton centrifuge at RPI, in Pro- ceedings of the International Conference Centrifuge 1991, Boulder: Balkema, Rotterdam, p. 27-34. Felio, G.Y., Vucetic, M., Hudson, M. Barar; P., and Chapman, R., 1990, Performance of soil nailed walls during the October 17, 1989 Lorna Prieta Earthquake, in Proceedings of the Forty— Third Canadian Geotechnical Conference, Quebec: Canadian Geotechnical Society, p. 165-173. Federal Highway Administration, 1993, FHWA Tour for Geotechnology—Soil N ailing: Federal HighwayAdministra- tion Report No. FHWA-PL-93-020, 84 p. Gassler, G., 1992, Full scale test on a nailed wall in consolidated clay, in Proceedings of the International Symposium on Earth Reinforcement Practice, Fukuoka, Kyushu, Japan: Balkema, Rotterdam, v. 1, p. 475-480. Gassler, G., and Gudehus, G., 1981, Soil nailing—some aspects of a new technique, in Proceedings of the 10th International Con- ference Soil Mechanics and Foundation Engineering: Balkema, Rotterdam, v. 3, p. 665-670. Guilloux, A., and Schlosser, F., 1982, Soil nailing practical appli- cations, in Proceedings of the Symposium on Soil/Rock Im- provement Technique: A.I.T. Hudson, M.B., 1990, The effects of the Loma Prieta earthquake on soil-nailed earth retaining structures in the San Francisco Bay area: Civil Engineering Department, University of California, Los Angeles, M.S. thesis, 108 p. Japan Highway Public Corporation, 1987, Guide for design and construction on reinforced slope with steel bars: September 33 p. ANALYSIS OF SOIL—NAILED EXCAVATIONS STABILITY DURING THE 1989 LOMAPRIETA EARTHQUAKE D45 Jewell, R.A., and Pedley, M.J., 1992,Analysis for soil reinforce— ment with bending stiffness: Journal of Geotechnical Engi— neering, V. 118, no. 10, p. 1505-1528. Juran, 1., Baudrand, G., Farrag, K., and Elias,V., 1990, Kinemati- cal limit analysis for soil nailed structures: Journal of Geotechnical Engineering, v. 116, no. 1, p. 54-72. J uran, 1., and E1ias,V., 1991, Ground anchors and soil nails retain- ing structures, in Foundation Engineering Handbook (2d ed.): Van Nostrand Reinhold, NewYork, p. 868-905. Koemer, R.M., 1984, Construction and geotechnical methods in foundation engineering: McGraw-Hill, Inc. Lambe, RC., and J ayaratne, N.N., 1987, Construction of retaining walls from the top down: Center forTransportation Engineer- ing Studies, Department of Civil Engineering, North Carolina State University at Raleigh, Research Report No. 23241-86- 6, 183 p. Mitchell, J.K., and Villet, W.C.B., 1987, Reinforcement of Earth Slopes and Embankments,Appendix C, Soil Nailing: National Cooperative Highway Research Program, Report 290,Wash— ington, DC, June, p. 258-323. Myles, B., and Bridle, R.J., 1991, Fired soil nails—the machine: Ground Engineering, v. 24, no. 6, p. 38-39. Ochiai, H., Hayashi, S., and Otani, J., eds, 1992, Proceedings of the International Symposium on Earth Reinforcement Prac- tice, IS Kyusha ’92, v. 1 and 2. Fukuoka, Japan: Balkema, Rotterdam. Plumelle, C., Schlosser, F., Delage, P., and Knochenmus, G., 1990, French national research project on soil nailing Clouterre: American Society of Civil Engineers SpecialTechnical Publi- cation No. 25, p. 660-675. Schlosser, F., Jacobsen, H.M., and Juran, I., 1983, Soil reinforce- ment—general report, in Proceedings of the Eighth European Conference on Soil Mechanics and Foundation Engineering: Balkema, Rotterdam, v. 3, p. 1159-1180. Seed, R.B., Dickenson, S.E., and Idriss, I.M., 1991, Principal geotechnical aspects of the 1989 Loma Prieta earthquake: Soils and Foundation, v. 31, no. 1, p. 1-26. Shen, C.K., Herrmann, L.R., Romstad, K.M., Bang, 5., Kim,Y.S., and DeNatale, J.S., 1981,An in situ earth reinforcement lat- eral support system: Department of Civil Engineering, Uni- versity of California, Davis, Report No. 81-03, 188 p. Stocker, M.F., Korber, G.W., Gassler, G., and Gudehus, G., 1979, Soil nailing, in Proceedings of the International Conference on Soil Reinforcement, Paris, v 2, p. 469—474. Stocker, M.F., and Riedinger G., 1990, The bearing behavior of nailed retaining structures: American Society of Civil Engi- neers Special Technical Publication No. 25, p. 612-628. Tatsuoka, F., 1992, Keynote lecture—roles of facing rigidity in soil reinforcing, in Preprints of Special and Keynote Lectures of the International Symposium on Earth Reinforcement Prac— tice Kyusha ’92, Fukuoka, Japan: Institute of Industrial Sci- ence, Tokyo, p. 77-115. Tufenkjian, M.R., and Vucetic, M., 1992, Seismic stability of soil nailed excavations, in Proceedings of the International Sym- posium on Earth Reinforcement Practice, Fukuoka, Kyushu, Japan: Balkema, Rotterdam, p. 573-578. 1993, Seismic stability of soil nailed excavations: Civil and Environmental Engineering Department, University of Cali- fornia, LosAngeles, Research Report ENG-97-169, June, 298 p. Tufenkjian, M.R.,Vucetic, M., and Doroudian, M., 1991, Stability of soil nailed excavations, in Proceedings of the International Workshop on Technology for Hong Kong’s Infrastructure De- velopment: Commercial Press Ltd., Hong Kong, p. 751-762. Vucetic, M., Iskandar, V.E., Doroudian, M., and Luccioni, L., 1996, Dynamic failure of soil-nailed excavations in centrifuge,in Proceedings of the International Symposium on Earth Rein- forcement Practice, Fukuoka, Kyushu, Japan: Balkema, Rotterdam, v. 1, p. 829—834. Vucetic, M.,Tufenkjian, M.R., and Doroudian, M., 1993, Dynamic centrifuge testing of soil nailed excavations: Geotechnical Testing Journal, v. 16, no. 2, p. 172—187. THE LOMA PRIETA, CALIFORNIA, EARTHQUAKE OF OCTOBER 17, 1989: PERFORMANCE OF THE BUILT ENVIRONMENT EARTH STRUCTURES AND ENGINEERING CHARACTERIZATION OF GROUND MOTION EMPIRICAL ANALYSIS OF PEAK HORIZONTAL ACCELERATION, PEAK HORIZONTAL VELOCITY, AND MODIFIED MERCALLI INTENSITY By Kenneth W. Campbell, EQE International, Inc., Evergreen, Colo. CONTENTS Page Abstract D47 Introduction 47 Strong-motion data 48 Ground—motion models 48 Peak ground acceleration ------------------------------------------ 48 Modified Mercalli intensity ——————————————————————————————————————— 55 Peak horizontal velocity ——————————————————————————————————————————— 55 Effect of surficial geology 56 Alluvium 57 Soft and hard rock 57 Bay mud 59 Comparison with previous studies -------------------------------------- 61 Regional attenuation 61 Local site conditions 63 Model verification 64 Acknowledgments 66 References cited 66 ABSTRACT A total of 137 ground—level accelerograms recorded dur- ing the earthquake were used to develop empirical mod- els for predicting the dependence of peak horizontal acceleration (PHA), peak horizontal velocity (PHV), and modified Mercalli intensity (MMI) on distance from the fault, source-to—site azimuth, and surficial geology. These models indicate that (l) PHA recorded on alluvium at distances farther than 50 km from the seismogenic rup- ture zone were significantly higher than those predicted from attenuation relationships available at the time of the earthquake; (2) both the amplitude and rate of at- tenuation of PHA, PHV, and MMI exhibited a strong dependence on azimuth with ground motions recorded at azimuths corresponding to Santa Cruz, San Francisco, and Oakland (west and northwest direction) having sig- nificantly higher amplitudes and lower rates of attenua- tion than those recorded in the east and northeast direc- tion; and (3) PHA recorded at intermediate and far dis- tances at all azimuths exhibited a strong dependence on site geology, with FHA on bay mud (SC-IV, SE), alluvium (SC-III, SD), and soft rock (SC-II, Sc) being on average 2.76, 1.47, and 1.25 times higher than on hard, or pre- dominantly Franciscan, rock (SC—lb, $3). The azimuthal dependence of PHA and PHV is consistent with the ob- served geographic distribution of MMI. The observed azimuthal dependence of PHA, PHV, and MMI is con- sistent with the combined effects of source directivity, shear-wave radiation pattern, and propagation effects (that is, critical reflections from layers within and at the base of the crust). The observed dependence of PHA on surficial geology is consistent with low-strain site am- plification factors recommended for use in building-code and engineering applications. However, these recom- mended factors are generally not consistent with the amplitude-dependent factors predicted from empirical attenuation relationships. A comparison of the model pre— dictions with published data indicates that the models accurately predict the observed dependence of PHA, PHV, and MMI on azimuth and distance throughout central California to distances as far as 250 km from the earth- quake rupture zone. INTRODUCTION The earthquake caused considerable damage to man—made structures in the San Francisco Bay area. The amount of damage observed in San Francisco and Oakland during this earthquake is superseded only by that sustained during the great 1906 San Francisco earthquake (Housner, 1990; US. Geological Survey Staff, 1990). The amount of damage was unusually large compared to other California earthquakes of similar magnitude, especially when one considers the relatively short duration of the ground motions and the relatively large epicentral distance to San Francisco and Oakland. This damage is, however, consistent with the unusually high ground motions that were recorded in San Francisco and Oakland. A compilation of peak accelerations published by Boore and others (1989) and Brady and Shakal (1994) indicates that peak horizontal accelerations as large as 0.20 to 0.25 g were routinely recorded on alluvium and soft soil in the San D47 D48 PERFORMANCE OF THE BUILT ENVIRONMENT Francisco and Oakland areas—the largest being 0.41 g on Holocene bay deposits (Bay mud) in a hanger at the Alameda Naval Air Station approximately 93 km from the epicenter. Campbell (1991) demonstrated that these accelerations are 1.5 to 3 times larger, on average, than those predicted from strong-motion attenuation relationships that were available at the time of the earthquake, but they are consistent with the unusually large amount of damage (Housner, 1990) and unusually high intensities (Stover and others, 1990) observed in these areas. In an attempt to understand this phenomenon, values of peak horizontal acceleration (PHA), peak horizontal velocity (PHV), and Modified Mercalli intensity (MMI) recorded and observed during this earthquake were analyzed to determine their dependence on distance, azimuth, and surficial geology. The dependence on distance, or what is simply referred to as “attenuation” throughout this paper, includes the effects of geometrical spreading, material damping, and scattering. The analysis was intended to be an engineering rather than a seismological study, emphasizing those aspects of ground motion which have the greatest influence on the empirical prediction of ground motion for seismic design. As demonstrated later, these results clearly show why damage in San Francisco and Oakland was more severe than at other locations of comparable distance throughout the region, an observation consistent with the geographic distribution of MMI in the Bay area (Stover and others, 1990). The seismological aspects of the earthquake have been discussed extensively in several special volumes dedicated to this earthquake, including Geophysical Research Letters, v. 17, no. 9, August, 1990; the Bulletin of the Seismological Society of America, v. 81, no. 5, October 1991; and companion volumes of this Professional Paper 1550 on “Earthquake Occurrence” and Professional Paper 1551 on “Strong Ground Motion and Ground Failure”. This information will not be presented in this paper; rather, the reader is referred to these publications for a complete discussion of the seismological aspects of the earthquake STRONG-MOTION DATA The strong-motion database used in this study consisted of 137 ground-level accelerograms compiled by Campbell (1991). These recordings were taken from reports published by Maley and others (1989), Shakal and others (1989), Brady and Mork (1990), and California Strong Motion Instrumentation Program (1991a, b, c). MMI observations were taken from compilations by Stover and others (1990) and C. Stover (written commun., 1991). The source-to—site distance measure used in this study is the closest distance to the zone of seismogenic rupture, a measure first proposed by Campbell (1987) and later adopted by Campbell and Bozorgnia (1994b) and Campbell (1997). It is defined as that portion of the rupture zone responsible for most, if not all, of the seismically radiated energy observed at the ground surface. This zone was interpreted to be approximately 40 km long, extend between depths of 4.5 to 19 km, and dip approximately 70° to the southwest as defined by the location of aftershocks that occurred within several weeks after the earthquake (see cross sections in Plafker and Galloway, 1989) and by seismological, geological, and geodetic studies conducted within several years after the earthquake (for example, Bulletin of the Seismological Society ofAmerica, v. 81, no. 5; Professional Paper 1551). A description of surficial geology for each of the record- ing sites was obtained from Campbell (1989, 1990), Shakal and others (1989), and Fumal (1991). These descriptions were independently verified and supplemented with data derived from 30 geology maps of the region. Based on these de- scriptions, each site was assigned to one of four surficial geologic classifications defined by Campbell (1981): soft soil (artificial fill; Holocene fluvial, bay, and estuarine depos— its; and other soft soils), alluvium (Quaternary deposits >10 m deep), soft rock (primarily Tertiary and Cretaceous sedi- mentary rock), and hard rock (primarily crystalline, meta- morphic, and pre-Cretaceous sedimentary rock). The ap- proximate correspondence between this site-classification criteria and those proposed for building—code and engineer- ing applications by Borcherdt (1994), Borcherdt and Glassmoyer (1994), and the International Conference of Building Officials (1996) is given in table 1. GROUND-MOTION MODELS Boore and others (1989), Plafker and Galloway (1989), Housner (1990), and Borcherdt and Glassmoyer (1992, 1994) found that surficial geology had a profound effect on peak accelerations recorded during the earthquake. In order to minimize this effect on the determination of attenuation, the regression analyses presented below were restricted to the 71 alluvial recordings compiled by Campbell (1991). These data are summarized in tables 2 and 3. Model coefficients were estimated from a nonlin— ear least-squares regression algorithm developed by More and others (1980). PEAK GROUND ACCELERATION A preliminary analysis of PHA using an attenuation rela- tionship whose functional form predicted a monotonically decreasing amplitude with distance resulted in a set of re- siduals that were strongly biased with respect to distance. A similar bias was noted by several other investigators (see fig. 6 of Boore and others, 1989). For distances less than 50 EMPIRICAL ANALYSIS Table l.—Summary of site classification criteria [Vsm average shear-wave velocity in the upper 30 m of the deposit] This study Borcherdt (1994b), Proposed 1997 UBC (Campbell, 1981, 1997) Borcherdt and Glassmoyer (1994) (International Conference of Building Officials) Class Description Class Description Vs 3., Class Description Vs, 3,, (m/s) (m/s) Hard rock Crystalline, metamorphic, and pre- SC-Ib Firm to hard rock 700—1,400 SB Rock 750~1,500 Cretaceous sedimentary rock 808 rock Tertiary and Cretaceous sedimentary SC-ll Gravelly soils 375—700 Sc Very dense soil 360—760 rock, soft volcanics and soft to firm and soft rock rock Alluvium Uneonsolidated Quaternary deposits SC-IlI Stiff clays and 200—375 SD Stiff soil 180—360 >10m deep sandy soils Sofi soil Artificial fill and Holocene fluvial, SC-IV Soft soil 100—200 SE Sofl soil <180 bay. and estuarine deposits (for example, Bay mud') IRefers to Holocene bay mud deposits along the margin of San Francisco Bay. Table 2.—Strong-moti0n data on alluvium ( S C-III, S D )’ used for regression analyses of PHA [PHA, mean peak horizontal acceleration; Rx, closest distance to seismogenic rupture zone; g, acceleration of gravity] Station No. R! PHA Station No. R! PHA (km) (g) (km) (g) 1103 78.3 0.098 57355 18.6 0.100 1116 72.0 0.125 57356 18.9 0.113 1226 51.9 0.005 57357 19.6 0.102 1227 26.4 0.360 57382 15.0 0.320 1230 33.4 0.195 57425 23.2 0.280 1239 75.3 0.110 57458 81.0 0.060 1265 51.5 0.050 57476 11.4 0.265 1439 80.6 0.095 57502 27.6 0.130 1446 74.6 0.095 57504 20.7 0.180 1474 54.1 0.070 57528 64.8 0.040 1479 68.0 0.115 57562 13.4 0.190 1481 51.5 0.165 57563 13.2 0.275 1575 28.1 0.240 58000 77.8 0.100 1601 30.9 0.240 58065 8.2 0.435 1652A 19.3 0.255 58224 71.5 0.228 1652C 19.7 0.205 58233 48.9 0.075 1656 25.4 0.280 58235 8.4 0.295 1675 68.0 0.090 58261 64.4 0.145 1678 79.3 0.180 58393 51.6 0.155 1686 38.1 0.175 58394 60.7 0.115 1687 33.1 0.105 58462 49.5 0.105 1689 57.1 0.085 58483 70.4 0.160 1695 23.2 0.205 58490 60.8 0.130 46173 88.8 0.065 58492 86.6 0.058 47125 15.8 0.505 58496 76.3 0.115 47179 32.8 0.105 58498 53.1 0.160 47288 72.7 0.050 58501 53.2 0.145 47380 11.7 0.350 58503 87.4 0.110 47381 13.4 0.460 58505 87.1 0.120 47459 10.3 0.333 68003 117.0 0.130 47460 81.0 0.080 68150 121.1 0.030 47524 28.6 0.280 68387 150.8 0.060 56012 73.4 0.050 68489 150.5 0.045 57064 38.1 0.120 68491 150.5 0.045 57066 22.9 0.165 69039 154.0 0.040 57191 28.6 0.120 1Definition of site-classification criteria is given in table 1. D49 D50 PERFORMANCE OF THE BUILT ENVIRONMENT Table 3.—Str0ng-m0ti0n stations on alluvium (SC-III, SD)’ used in regression analyses of PHA [PHA, mean peak horizontal acceleraiion. Owner: USGS, US. Geological Survey; CDMG, California Division of Mines and Geology. Location: BSMT, basement; GRND, ground level; TOE, toe of dam] 8:2?“ Owner Station name Location 1103 USGS BERKELEY—2168 SHATTUCK, EAST BSMT 1 116 USGS SAN FRANCISCO—oCSUSF, THORTON HALL GRND 1226 USGS LIVERMORE—VA HOSPITAL, BLDG. 62 BSMT 1227 USGS PALO ALTO—VA HOSPITAL, BLDG. 1 BSMT 1230 USGS MENLO PARK—VA HOSPITAL, BLDG. 37 GRND 1239 USGS SAN FRANCISCO—TRANSAMERICA TOWER BSMT 1265 USGS DEL VALLE DAM TOE 1439 USGS RICHMOND—BULK MAIL CENTER GRND 1446 USGS SAN FRANCISCO—STANDARD 01L BUILDING BSMT 1474 USGS BEAR VALLEY #S—CALLENS RANCH GRND 1479 USGS BEAR VALLEY #IO—WEBB RESIDENCE GRND 1481 USGS BEAR VALLEY #12—W1LLIAMS RANCH GRND 1575 USGS HOLLISTER—CITY HALL ANNEX BSMT 1601 USGS STANFORD UNIVERSITY—SLAC TEST LAB GRND I652A USGS ANDERSON DAM—DOWNSTREAM GRND 1652C USGS ANDERSON DAM—TOE TOE 1656 USGS HOLLISTER—DIFFERENTIAL ARRAY (SMA) GRND 1675 USGS SAN FRANCISCO—1295 SHAFTER GRND 1678 USGS SAN FRANCISCO—GOLDEN GATE BRIDGE GRND 1686 USGS FREMONT—EMERSON COURT GRND 1687 USGS CALAVERAS RESERVOIR!SOUTH GRND 1689 USGS DUBLIN—FIRE STATION GRND 1695 USGS SUNNYVALE4COLTON AVENUE GRND 46173 CDMG BITTERWATER—COALINGA ROAD GRND 47125 CDMG CAPITOLA—FIRE STATION GRND 47179 CDMG SALINAS—JOHN & WORK GRND 47288 CDMG SAN BENITO GRND 47380 CDMG GILROY #2—HIGHWAY 101 MOTEL GRND 47381 CDMG GILROY #3vSEWAGE PLANT GRND 47459 CDMG WATSONVILLE—TELEPHONE BUILDING, N.W. GRND 47460 CDMG GREENFIELD—POLICE STATION GRND 47524 CDMG HOLLISTER—GLORIETTA WAREHOUSE, EF. GRND 56012 CDMG LOS BANOS GRND 57064 CDMG FREMONT~MISSION SAN JOSE GRND 57066 CDMG AGNEWS—AGNEWS STATE HOSPITAL GRND 57191 CDMG HALLS VALLEY—GRANT PARK GRND 57355 CDMG SAN JOSE—GREAT WESTERN BUILDING, SOUTH BSMT 57356 CDMG SAN JOSE—TOWN PARK TOWERS, SOUTH GRND 57357 CDMG SAN JOSE—SANTA CLARA COUNTY BLDG, S.W. BSMT 57382 CDMG GILROY #4——SAN YSIDRO SCHOOL GRND 57425 CDMG GILROY #7—MANTELLI RANCH GRND 57458 CDMG TRACY—SEWAGE PLANT GRND 57476 CDMG GILROY—OLD FIREHOUSE GRND 57502 CDMG MILPITAS—‘Z-STORY BUILDING, EAST GRND 57504 CDMG COYOTE LAKE DAM—DOWNSTREAM GRND 57528 CDMG LIVERMORE—FAGUNDES RANCH GRND 57562 CDMG SAN JOSE—3-STORY BUILDING, EAST GRND 57563 CDMG SAN JOSE—SANTA TERESA HILLS GRND 58000 CDMG BERKELEY—UCB, STADIUM GROUNDS GRND 58065 CDMG SARATOGA—14675 ALOHA GRND 58224 CDMG OAKLAND—TITLE & TRUST BUILDING, N.E. GRND 58233 CDMG LOWER CRYSTAL SPRINGS DAM—DOWNSTREAM GRND 58235 CDMG SARATOGA—SCHOOL GYM, SHEAR WALL, N0. GRND 58261 CDMG SO. SAN FRANCISCO—KAISER MEDICAL CENTER BSMT 58393 CDMG HAYWARD~MUIR SCHOOL (APEEL #ZE) GRND 58394 CDMG SAN BRUNO—US. POSTAL BUILDING, CENTER GRND 58462 CDMG HAYWARD—6—STORY OFFICE BUILDING, CENTER BSMT 58483 CDMG OAKLAND—24-STORY BUILDING, WEST GRND 58490 CDMG SAN BRUNO—6-STORY OFFICE BUILDING GRND 58492 CDMG CONCORD—8-STORY BUILDING, CENTER GRND 58496 CDMG BERKELEY—Z-STORY HOSPITAL BSMT 58498 CDMG HAYWARD—BART STATION, PARKING LOT GRND 58501 CDMG HAYWARD~BART ELEVATED SECTION, BENT 132 GRND 58503 CDMG RICHMOND—3-STORY GOVERNMENT BLDG, E. BSMT 58505 CDMG RICHMOND—CITY HALL, PARKING LOT GRND 68003 CDMG OLEMA—POINT REYES RANGER STATION GRND 68150 CDMG NAPA—NAPA COLLEGE GRND 68387 CDMG SANTA ROSA—S—STORY BUILDING, WEST GRND 68489 CDMG SANTA ROSA—14-STORY BUILDING, CENTER GRND 68491 CDMG SANTA ROSA—HENDLEY & TUPPER GRND 69039 CDMG BODEGA HEAD—DORAN BEACH GRND 1Definition of site-classification criteria is given in table 1. EMPIRICAL ANALYSIS D51 km from the seismogenic rupture zone, this preliminary model tended to overestimate PHA, whereas for distances greater than 80 km the opposite effect was observed. Be- tween these two distances the residuals showed a gradual transition from one trend to the other. Upon further analy- sis it became clear that this bias was caused by the pres- ence of a zone of nearly constant acceleration between distances of about 50 and 80 km from the seismogenic rupture zone. In order to model the observed behavior, a second regres- sion analysis was performed in which the distance limits defining the zone of constant acceleration were included as regression coefficients. The only constraint imposed in this analysis was that PHA remain constant between these two distance limits. This resulted in the following tripartite at- tenuation relationship: In PHA = 1.876 —1.03 In [R5 + 7.79], RSS 50.6 km; =—2.313, 50.6200 133 90—196 1Definition of site-classification criteria is given in table 1. EMPIRICAL ANALYSIS comes an estimate of the standard normal variate. In this case a normalized residual of 0 corresponds to a recording that falls at the 50th percentile (the median) of the distribu- tion of residuals, and normalized residuals of -1 and +1 cor- respond to recordings that fall at the 16th percentile and 84th percentile of this distribution, respectively. The correspon- dence between the site classification criteria given below and those recommended for building-code and engineering ap- plications by Borcherdt (1994), Borcherdt and Glassmoyer (1994), and the International Conference of Building Offi- cials (1996) is given in table 1. ALLUVIUM A plot of NR versus RS for alluvium (SC-III, SD) is dis- played in figure 8. There is no visible trend in this plot. This is expected since these recordings were used to develop the relationship. The diamonds in this figure represent record- ings from embedded sites, primarily from basements of build— ings. A hypothesis test indicated that the embedded record- ings had a mean residual close to zero, no different from the mean residual of the ground-level and free-field recordings at the 90 percent confidence level. This test appears to con- tradict the results of previous studies which have found em- bedded sites to have smaller accelerations than nonembedded sites (see Campbell, 1987, 1988, 1989, 1990); however, the relatively large distances associated with many of these re— cordings has apparently minimized this effect (Campbell, 1987). Figure 9 shows the distribution of residuals with re- spect to RS and azimuth. This plot indicates that distances D57 greater than 80 km between azimuths of 000° and 160° and all distances between azimuths of 160° and 320° are poorly represented by the recordings. A plot of NR versus azimuth is displayed in figure 10. This plot shows that there is a strong azimuthal bias in the residu— als, with NR between azimuths of 320° and 350° biased to- ward positive values, representing an underestimation of PHA by the regression model, and those between azimuths of 350° and 030° biased toward negative values, representing an over- estimation of PHA. The underestimation occurs at azimuths corresponding to San Francisco and Oakland, the overesti- mation at azimuths corresponding to Hayward and San Jose. This bias was accounted for in equation (2). Evidence pre- sented later strongly suggests that this azimuthal bias is due to the combined effects of differences in attenuation, source directivity, and shear-wave radiation pattern. Stover and oth- ers (1990) found a similar azimuthal bias in reported values of MMI. SOFT AND HARD ROCK A plot of NR versus RS for soft rock (SC-II, SC) and hard rock (SC-Ib, SE) is displayed in figure 11. This plot indi- cates that both soft-rock and hard-rock recordings at dis- tances of 20 km and greater tend to have smaller accel- erations than those recorded on alluvial sites. The effect is most significant for hard-rock sites. Hypothesis tests indicated that the mean residuals of both the soft-rock and hard-rock recordings at similar azimuths were sig- nificantly different from the mean residual of the allu- vial recordings at the 90 percent confidence level. A simi— 4 3 .— 2 F 1:1 :1 0 ‘LD 0 '0 >— D g 1 1:530 D D % Do '8 :1 D II 0 DUDE] Cln 0 D D 60% L] B D D U D '3 D D .5 D [EEC] D C] a -1F [:1 D D g 2 _ a 8 11]] U z ' D -3 - _4 1 l 1 1 1 1 1 O 20 4O 60 80 100 120 140 160 Distance to Seismogenic Rupture (km) Figure 8.—Plot of normalized residuals versus closest distance to the seismogenic rupture zone for alluvial sites: squares, ground—level and free-field recordings; diamonds, embedded recordings. D58 PERFORMANCE OF THE BUILT ENVIRONMENT A 160 5 ”D v 140- Q) ‘5 E 120— D U :1 o: .9 100— :: n D “3” 8°- D a ‘3 599 E a B (9 .9 60" [3 ‘5») (21 Q: D g o 4013 D u ... D a) D D as Do U C 8 20 - D [:1 D 0Q; g @ a u: .3 0 I l J l l I J l 0 4o 80 120 160 200 240 280 320 360 Azimuth from Epicenter (deg) Figure 9.-—Plot of closest distance to the seismogenic rupture zone versus epicenter—to-site azimuth measured clockwise from north for alluvial sites: squares, ground—level and free-field recordings; diamonds, embedded recordings. 4 3 _ 2 — E1 _ D g D D D 8 g 1 - [:0 to 5159 1:1 8:) o D 0 Cl: 0 U E10 8 D D C] D C193, CB ,5 D U DO U D D C E -1 - DU D E D D ‘6 —2 - D Q] 2 U -3 —- _4 l a I 1 I I J l 0 40 80 120 160 200 240 280 320 360 Azimuth from Epicenter (deg) Figure 10.—Plot of normalized residuals versus epicenter-to—site azimuth measured clockwise from north for alluvial sites: squares, ground-level and free-field recordings; diamonds, embedded recordings. lar effect was observed during the 1987 Whittier Narrows earthquake (Campbell, 1988). Figure 12 is a plot of the distribution of residuals with respect to RS and azimuth. This plot shows that, although rock recordings are fairly evenly distributed with respect to azimuth, there are few recordings at distances greater than 50 km except at azimuths corresponding to San Fran- cisco and Oakland. A plot of NR versus azimuth is displayed in figure 13. The number of recordings on rock is less than for allu- Vium, and their residuals are biased toward negative val- ues. As a result, it is not obvious from this figure whether the rock recordings show a similar bias with azimuth. The residuals corresponding to azimuths of 320°t0 350° do appear to be larger, on average, than those correspond- ing to azimuths of 350° to 030°, which lends qualitative EMPIRICAL ANALYSIS D59 4 3 _ l. _. 2 Cl ‘3 9K '0 1 - g Cl D 0 9K 3K 3 a W 0 5K .5 -1 .. 9" C] D + C] E 0 @ fil E 9K 0%] DC] a -2 — 0° 2 D -3 — D _4 1 1 1 I I 1 l 0 20 4O 60 80 100 120 140 160 Distance to Seismogenic Rupture (km) Figure ll.—Plot of normalized residuals versus closest distance to the seismogenic rupture zone for rock sites: asterisks, ground—level and free-field recordings on soft rock; crosses, embedded record— ings on soft rock; squares, ground-level and free-field recordings on hard rock; diamonds, embedded recordings on hard rock. A 160 E 35 140 ~ a) Cl ‘5 E 120 - D a: o 100 r 'E *4— 8» so — 2 o W .9 60 _ D ‘1’ ‘5 (D 3K <> 3 40 _ [a 8 D c 20~ c1 cm an * a S a ,2 0 4 jK 1 l l I L D o o 40 80 120 160 200 240 280 320 360 Azimuth from Epicenter (deg) Figure 12.-—Plot of closest distance to the seismogenic rupture zone versus epicenter-to-site azimuth measured clockwise from north for rock sites: asterisks, ground-level and free-field recordings on soft rock; crosses, embedded recordings on soft rock; squares, ground-level and free-field recordings on hard rock; diamonds, embedded recordings on hard rock. support to the suggestion that this bias is a propagation effect and/or a source effect, and not a site effect. BAY MUD A plot of NR versus RS for Bay mud (SC-IV, SE) is dis- played in figure 14. This plot shows that Bay mud sites have substantially larger accelerations than alluvial sites. As indicated in figures 15 and 16, all of these sites lie between azimuths of 320° and 350°. The analysis of the alluvial recordings indicated that these azimuths were associated with higher than average accelerations. There— fore, the large bias associated with these sites is prob- ably caused by a combination of both azimuthal and site effects. A hypothesis test indicated that the mean residual D60 PERFORMANCE OF THE BUILT ENVIRONMENT 9K Normalized Residual 0 $2 944 fiat an 9‘6 DOD _4 I l J l l I l 1 0 40 80 120 160 200 240 280 320 360 Azimuth from Epicenter (deg) Figure 13.—Plot of normalized residuals versus epicenter-to-site azimuth measured clockwise from north for rock sites: asterisks, ground-level and free-field recordings on soft rock; crosses, embedded recordings on soft rock; squares, ground-level and free-field recordings on hard rock; diamonds, em- bedded recordings on hard rock. 4 3.— Normalized Residual -4 l I l l I 1 l O 20 40 60 80 100 120 140 160 Distance to Seismogenic Rupture (km) Figure 14.—Plot of normalized residuals versus closest distance to the seismogenic rupture zone for Bay mud sites: squares, ground-level and free-field recordings; diamonds, embedded recordings. of the Bay mud recordings was significantly different from the mean residual of the alluvial recordings for simi- lar azimuths at the 90 percent confidence level. Thus, the large amplitude bias associated with the Bay mud sites is statistically significant. The effect of surficial geology on peak acceleration is summarized in table 6, where it is compared to similar low—strain site-amplification factors observed or predicted by others, as discussed in a later section. The median site factor from the empirical attenuation relationships is the geometric mean of those calculated for moment magni- tudes (MW) of 6.5 and 7.5. The 68 percent confidence in- terval from the empirical attenuation relationships was derived from the complete set of calculated values for both MW 6.5 and 7.5 assuming a log-normal distribution. According to these results, PHA recorded at intermedi- EMPIRICAL ANALYSIS D61 A 160 E x V 140 " 2 3 o- 120 h I) (I .2 100 ‘ Cl 83 80 0 @ E .22) so — C] U) 3 4O ' I? m D S 20 — 59. .3 o I I l l I I I I 0 40 80 120 160 200 240 280 320 360 Azimuth from Epicenter (deg) Figure 15.—Plot of closest distance to the seismogenic rupture zone versus epicenter-to-site azimuth measured clockwise from north for Bay mud sites: squares, ground-level and free—field recordings; diamonds, embedded recordings. Normalized Residual _4 I I I l I I I 4O 80 120 160 200 240 280 820 360 Azimuth from Epicenter (deg) Figure 16.—Plot of normalized residuals versus epicenter-to-site azimuth measured clockwise from north for Bay mud sites: squares, ground-level and free-field recordings; diamonds, embedded recordings. ate and far distances at all azimuths during the earthquake exhibited a strong dependence on site geology, with FHA on Bay mud (SC-IV, SE), alluvium (SC-III, SD), and soft rock (SC-II, SC) being on average 2.76, 1.47, and 1.25 times, respectively, higher than that on hard, predomi- nantly Franciscan rock (SC-1b, SB). These differences were all found to be significant at the 90 percent confi- dence level. COMPARISON WITH PREVIOUS STUDIES REGIONAL ATTENUATION The zone of nearly constant ground motions observed in figure 1 between distances of about 50 to 80 km from the seismogenic rupture zone when recordings at all azimuths D62 PERFORMANCE OF THE BUILT ENVIRONMENT Table 6.—Summary of relative low-strain site-amplification factors for PHA and short-period response spectral accelera- tion [PHA, mean peak horizontal acceleration; g, acceleration of gravity; MW, moment magnitude; —, no value] Reference Site classification Hard rock Soft rock Alluvium Soft soil (Bay (SC—1b, SB) (SC-II, SC) (SC-III, SD) mud) (SC—IV, SE) 1989 Lorna Prieta earthquake (MW 6.9) This study (median) 0.80 1.00 1.18 2.21 This study (68% confidence interval) 0.53—1.21 0.73—1.37 0.78—1.78 1.62—3.02 Boore and others (1989) 1.002 1.002 1.13 2.81 Other earthquakes 1987 Whittier Narrows (Campbell, 1988), MW 6.1 0.84 1.00 1.47 -—- 1992 Landers (Campbell and Bozorgnia, 1994a), MW 0.81 1.00 1.20 -— 7.3 Empirical attenuation relationships; hard-rock PHA = 0.05 g (MW 6.5, 7.5) Abrahamson and Silva (1997) — 1.00 1.15, 1.15 — Boore and others (1997)3 0.77, 0.77 1.00 1.31, 1.31 — Campbell and Bozorgnia (1994b), Campbell (1997) 0.79, 0.78 1.00 1.21, 1.34 — Idriss (1991, 1993) — 1.00 1.17, 1.17 326,382 Sadigh and others (1997) — 1.00 1.15, 1.34 — MEDIAN 0.78, 0.78 1.00, 1.00 1.20, 1.26 3.26, 3.82 MEDIAN (all MW) 0.78 1.00 1.23 3.53 68% CONFIDENCE INTERVAL (all MW) —— — 1.15—1.31 —— Building—code and engineering applications4 Borcherdt (1994b), Borcherdt and Glassmoyer (1994) 0.77 1.00 1.23 1.54 Proposed 1997 UBC (International Conference of 0.83 1.00 1.33 1.75 Building Officials, 1996) lDefinition of site-classification criteria is given in table 1. 2N0 distinction was made between soft rock and hard rock. 3Calculated for shear-wave velocities of 250, 510, and 1,070 m/s for alluvium, soft rock, and hard rock, respectively. 4Corresponds to hard-rock PHA of 0.075 g (International Conference of Building Officials, 1996) and 0.1 g (Borcherdt, 1994). are combined is not as evident when these ground motions are segregated by azimuth as in figures 2 to 6. It appears that varying the rate of attenuation with azimuth, as was done in equation (2), can explain the geographic distribu- tion of alluvial recordings better than simply varying the rate of attenuation with distance for all azimuths, as was done in equation (1). This observation is statistically con- firmed by the 27 percent reduction in the standard error as- sociated with equation (2). The plot comparing the attenuation of PHA with azimuth (fig. 6) indicates that the highest accelerations and lowest rates of attenuation were located southwest of the epicenter in the direction of Santa Cruz and Capitola. This region is located on the hanging wall of the causative fault and is near a maximum in the shear—wave radiation pattern. As a result, this region was subject to strong amplification due to source effects. The next highest accelerations and lowest rate of at- tenuation was located northwest of the epicenter in the di- rection of San Francisco and Oakland. Seismic waves propa- gating in this direction were subject to amplification due to both source directivity and radiation pattern and were ob- served to have an unusually low rate of attenuation as a re- sult of (1) high Q in the lower crust (Fletcher and Boatwright, 1991) and (2) critical reflections from velocity discontinuities or strong velocity gradients within the crust and from the base of the crust, or Moho (Catchings and Kholer, 1997). The lowest accelerations and highest rates of attenua- tion were located northeast of the epicenter. This region is near a minimum in the shear-wave radiation pattern . and has propagation paths that cross several faults and fault-bounded basins. As a result, seismic waves propagating in this direction were subject to increased attenuation due to scattering and abrupt changes in crustal velocity structure. Seismic waves propagating toward the southeast had acceleration amplitudes in- termediate between those to the northwest and north— east. This region is near a maximum in the shear-wave radiation pattern and was possibly subject to source directivity (assuming bilateral rupture). However, these effects were either weaker than they were toward the northwest, and/or propagation to the southeast was sub- ject to greater attenuation and scattering, possibly as a result of different crustal properties (see Catchings and Kholer, 1997). EMPIRICAL ANALYSIS D63 Seismological observations of test explosions and after- shocks have indicated that critical reflections from within the crust and from the Moho result in a relatively low rate of attenuation along the San Francisco Peninsula beyond dis- tances of a few tens of kilometers that is believed to be the cause of the relatively high ground motions in the San Fran- cisco and Oakland areas during the main shock (see Somerville and Yoshimura, 1990; McGarr and others, 1991; Fletcher and Boatwright, 1991; Borcherdt and Glassmoyer, 1994; Somerville and others, 1994; Stevens and Day, 1994; Catchings and Kholer, 1997). Figures 2 to 6 generally sup— port these observations; however, the sparse strong-motion data set is not sufficient to unequivocally confirm or refute the specific details of these observations. Schneider and others (1993) used a band-limited, white- noise stochastic simulation model to assess the effects of source finiteness, crustal wave propagation, and site response on 23 strong-motion recordings located primarily along the San Francisco Peninsula. They compared a simple spherical spreading propagation model with a seismologically more rigorous model, which included both direct and critically reflected phases, and found that the two methods resulted in nearly the same estimates of uncertainty. They found that simple spherical spreading provided a good fit to the response spectra computed from recordings at fault distances between 1 and 80 km, consistent with the results of this study, and concluded that the effects of the crustal structure could eas— ily be accommodated by other parameters in the model. LOCAL SITE CONDITIONS As demonstrated in table 6, the relative site amplification factors obtained in this study are consistent with those de- rived for the earthquake by Boore and others (1989), Plafker and Galloway (1989), and Borcherdt and Glassmoyer, 1992, 1994); for the 1987 Whittier Narrows earthquake by Campbell (1988); and for the 1992 Landers earthquake by Campbell and Bozorgnia (1994a). They are also consistent with site factors predicted by empirical attenuation relation- ships developed by Campbell and Bozorgnia (1994b) and Boore and others (1997), assuming a hard-rock PHA of 0.05 g. Although not shown in table 6, the relative amplification factors developed in this study are also qualitatively consis- tent with site factors determined from recordings in the San Francisco Bay region of nuclear explosions in Nevada as well as with regional variations in intensity for the 1906 San Francisco earthquake (Borcherdt and Gibbs, 1976). All of these results are biased toward recordings obtained at rela- tively large distances and/or low values of PHA and, there- fore, represent relatively low in-situ strains. Also shown in table 6 for comparison are the relative low— strain site-amplification factors for short—period spectral ac- celerations recommended for building—code and engineer- ing applications by Borcherdt (1994), Borcherdt and Glassmoyer (1994), and the International Conference of Building Officials (1996). These site factors represent hard- rock (SC—Ib, SB) accelerations of 0.1 g for the Borcherdt study and 0.075 g for the International Conference of Building Officials study. The recommended site factors are very similar to those observed and predicted from past earthquakes, indi- cating a reasonable empirical basis for these recommendations. Empirical analyses by Idriss (1991, 1993), Campbell and Bozorgnia (1994b), Abrahamson and Silva (1997), Campbell (1997), and Sadigh and others (1997) have shown that PHA and short-period spectral site-amplification factors become significantly smaller at near-source distances, consistent with the general trend of the amplitude-dependent amplification factors recommended by Borcherdt (1994), Borcherdt and Glassmoyer (1994), and the International Conference of Building Officials (1996). Table 7 compares relative site- amplification factors between alluvium (SC-III, SD) and soft rock (SC-II, SC) predicted from these studies with those rec- ommended for building-code and engineering applications for hard-rock accelerations ranging from 0.05 to 0.4 g. For those studies which do not distinguish between soft and hard rock, rock has been classified as soft rock, consistent with the mean shear-wave velocity of 618 m/s determined for strong-motion recording sites classified as generic rock by Boore and Joyner (1997). The soft-rock PHA corresponding to a given value of hard-rock PHA was estimated from the attenuation relationships of Campbell and Bozorgnia (1994b) and Campbell (1997). The results in table 7 indicate that the amplitude-depen- dent amplification factors recommended for use in scaling short—period response spectral acceleration for building-code and engineering applications have a weaker dependence on cyclical strain than that predicted for PHA and, therefore, short-period spectral acceleration, by empirical attenuation relationships. Borcherdt (1994) described his recommended site factors as being based directly on Loma Prieta strong- motion recordings for low-amplitude rock accelerations and on theoretical dynamic site—response analyses for higher am- plitude rock accelerations. As demonstrated in table 6, the use of Loma Prieta observations to determine the low-strain amplification factors has resulted in factors that are consis- tent with a large number of other empirical observations. However, the results in table 7 indicate that, although the trend with amplitude is similar, the amplification factors rec- ommended for building-code and engineering applications for hard-rock accelerations of 0.1 g and larger are consis— tently higher than those predicted from empirical attenua- tion relationships. This latter observation is independent of the assumption of whether a generic rock site is classified as either soft rock or hard rock. It is possible that the above discrepancy is the result of an intentional built-in conservatism in the recommended am- plification factors, a bias in the theoretical site-response analyses that were used to calculate the high-strain depen- dence of these factors, and/or a bias in the way that the empiri— D64 PERFORMANCE OF THE BUILT ENVIRONMENT Table 7.—Summary 0f amplitude-dependent site-amplification factors for PHA and short-period response spectral acceleration for alluvium (SC-III, SD) relative to sofi rock (SC-II, SC)’ [PHA, mean peak horizontal acceleration; g, acceleration of gravity; MW, moment magnitude; ——, no value] Reference PHA on hard rock (SC-Ib, SB) 0.05 g 0.1 g 0.2 g 0.3 g 0.4 g Empirical attenuation relationships (MW 6.5, 7.5) Abrahamson and Silva (1997) 1.15, 1.15 1.02, 1.02 0.89, 0.89 0.82, 0.82 0.78, 0.78 Boore and others (1997) 1.31, 1.31 — — — -—— Campbell and Bozorgnia (1994b), Campbell (1997) 1.21, 1.34 1.12, 1.23 1.01, 1.11 0.95, 1.03 0.88, 0.96 Idriss (1991, 1993) 1.17, 1.17 1.06, 1.05 0.96, 0.95 0.91, 0.91 0.88, 0.87 Sadigh and others (1997) 1.15, 1.34 1.02, 1.15 0.91, 0.99 0.85, 0.91 0.82, 0.86 MEDIAN 1.20, 1.26 1.05, 1.11 0.94, 0.98 0.98, 0.91 0.84, 0.87 MEDIAN (all MW) 1.23 1.08 0.96 0.90 0.85 68% CONFIDENCE INTERVAL (all MW) 1.15—1.31 1.01—1.16 0.89—1.04 0.83—0.97 0.79—0.91 Building-code and engineering applications Borcherdt (1994), Borcherdt and Glassmoyer (1994) 1.23 1.23 1.17 1.00 0.90 Proposed 1997 UBC (International Conference of 1.332 1.273 1.17 1.09 1.00 Building Officials, 1996) 1Definition of site-classification criteria is given in table 1. 2Corresponds to a hard-rock PHA of 0.075 g. 3Corresponds to average of hard-rock PHA of 0.075 g and 0.15 g. cal attenuation relationships were developed. The possibil— ity of the last consideration is demonstrated by the empiri- cal predictions of Campbell and Bozorgnia (1994b) and Campbell (1997). These relationships were developed spe— cifically to accommodate an independent scaling of site am- plification with distance, thereby minimizing any bias in these factors imposed by arbitrary constraints. The results in table 7 show that the amplification factors predicted by these relationships are greater than those predicted by the other attenuation relationships, but generally consistent with those recommended for building-code and engineering ap- plications. MODEL VERIFICATION The validity of the ground-motion models developed in this study was tested by comparing 37 estimates of PHA, PHV, and MMI derived from equations (2) through (4) with values observed on alluvium at sites located within about 250 km of the seismogenic rupture zone (table 8). These sites were specifically selected to sample a wide range of dis- tances and azimuths. A few sites located on Bay mud were included in the comparison to demonstrate the difference between expected and observed ground motions on soft soils located along the margins of San Francisco Bay. Because the MMI-PGA relationship predicts the MMI contour for a given value of PHA, only the integer part of the calculated MMI should be compared with the observed value of MMI. For example, calculated intensities of 6.3 and 6.7 are com— parable to an observed MMI of VI (6). To make this com- parison easier in table 8, the value of the calculated MMI that should be compared to the observed MMI is given in parentheses after the calculated value. The following adjustments were made to the predicted ground motions for Bay mud sites located in San Francisco and Redwood City to account for the presence of soft soils. Estimates of PHA from equation (2) were adjusted by the relative difference between peak accelerations recorded on soft soils (SC-IV, SE) and those recorded on alluvium (SC— 111, SD) based on charts given by Seed and Idriss (1982) and Idriss (1990). Estimates of PHV from equation (4) were in- creased by a factor of 1.7 based on site—amplification factors developed by Campbell (1992). Estimates of MMI from equa- tion (3) were increased by one intensity unit based on the expected difference in intensity between firm soils (Ground- Shaking Unit L) and soft soils (Ground—Shaking Unit J) rec- ommended by Evemden and Thomson (1988). As discussed below, predictions at several other localities were also ad- justed to reflect the unusually large ground motions and in- tensities observed at these locations during the earthquake. Strong-motion recordings obtained during the earthquake clearly showed that sites located on Bay mud had, on aver- age, significantly higher ground motions than those located on alluvium or rock (see Boore and others, 1989; Campbell, 1991; Borcherdt and Glassmoyer, 1992, 1994). There were, however, some notable exceptions to these observations. For example, accelerograms recorded on Bay mud at Foster City and near the Dumbarton Bridge had peak accelerations simi- lar to those recorded on nearby alluvial sites. EMPIRICAL ANALYSIS D65 Table 8.—C0mparison of predicted and observed values of MMI, PHA, and PH V [MMI, Modified Mercalli intensity; PHA, mean peak horizontal acceleration; PHV, mean peak horizontal velocity; RS, closest distance to seismogenic rupture zone, Pred., predicted; Obs., observed or recorded; —, no value reported] Location Rs MMI PHA (g) PHV (cm/s) (km) Pred.‘ Obs. Pred. Obs. Pred. Obs. Corralitos ----------------------------------- 5 8.4 (8) 8 0.56 0.57 50 51 Los Gatos ----------------------------------- 6 8.2 (8) 8 0.46 — 52 — Watsonville --------------------------------- 11 7.9 (8) 8 0.33 0.33 37 44 Santa Cruz ---------------------------------- 20 8.3 (8) 8 0.51 0.50 57 — Hollister ------------------------------------- 28 8.4 (8) 8 0.26 0.27 36 47 Oakland (Bay mud, Merritt Sand) ------- 72 7.9 (8) 7-8 0.17 0.22 22 29 Oakland (alluvium) ------------------------ 72 6.8 (6) 6-7 0.11 — 13 —— San Francisco (Bay mud) ----------------- 75 7.8 (7) 7-8 0.17 0.15 21 18 San Francisco (alluvium) ----------------- 75 6.8 (6) 6—7 0.1 l 0.1 1 12 14 Gilroy 11 7.9 (7) 7 0.32 0.27 36 34 Morgan Hill --------------------------------- 15 7.6 (7) 7 0.23 — 22 — San Jose ------------------------------------- 19 7.5 (7) 7 0.20 0.16 19 19 Palo Alto ------------------------------------ 31 7.6 (7) 7 0.24 0.29 27 33 Salinas --------------------------------------- 32 7.3 (7) 7 0.17 0.10 19 13 Redwood City (alluvium) ----------------- 35 7.5 (7) 7 0.22 — 25 — Redwood City (Bay mud) ---------------- 35 8.5 (8) —— 0.29 0.25 42 45 Fremont ------------------------------------- 40 7.1 (7) 7 0.14 0.12 15 9 San Mateo ----------------------------------- 51 7.2 (7) 7 0.16 — 18 —— San Bruno ----------------------------------- 61 7.1 (7) 6-7 0.14 0.13 15 15 San Leandro -------------------------------- 61 7.0 (7) 6-7 0.13 ——- 15 — Hayward ------------------------------------- 53 7.0 (7) 6 0.14 0.16 15 13 Livermore ----------------------------------- 58 6.3 (6) 6 0.07 0.07 7 — Los Banos ----------------------------------- 74 6.3 (6) 6 0.07 0.05 8 — Walnut Creek ------------------------------- 79 6.3 (6) 6 0.07 0.07 7 9 Tracy 79 5.9 (5) 6 0.05 0.06 5 — Greenfield ----------------------------------- 81 6.2 (6) 6 0.07 0.08 8 Richmond ----------------------------------- 88 6.6 (6) 6 0.09 0.12 10 16 Napa 125 5.8 (5) 6 0.05 0.03 5 Stockton ------------------------------------- 105 5.4 (5) 5-6 0.04 — 6 ~— Modesto ------------------------------------- 95 5.6 (5) 5 0.04 — 4 - Santa Rosa ---------------------------------- 152 5.6 (5) 5 0.04 0.05 5 —- Sacramento --------------------------------- 160 4.6 (4) 4-5 0.02 —-— 2 — Fresno --------------------------------------- 170 4.8 (4) 3-5 0.02 —— 3 — Coalinga ------------------------------------- 150 5.1 (5) 4 0.03 — 3 — Paso Robles --------------------------------- 172 4.7 (4) 4 0.02 — 3 —— Ukiah 240 4.2 (4) 4 0.02 —— 2 — Santa Maria --------------------------------- 250 3.7 (3) 4 0.01 -— 1 —— Marysville ---------------------------------- 220 3.7 (3) 3 0.01 — 1 —— lValue in parentheses indicates number to be compared with observed value. There were also several sandy sites that exhibited amplifi- cations comparable to those on Bay mud. One such site—a two-story office building in downtown Oakland—is founded on Merritt Sand, a Pleistocene deposit of saturated dune sand that overlies Older bay sediments. These Observations together with those noted above have led some investigators (see Shakal and others, 1990; Hanks and Brady, 1991) to suggest that the response Of the entire soil column above bedrock, not simply the response of Bay mud and other surficial soft sediments, was responsible for the observed amplification at many of these sites. As discussed previously, several investi— gators have suggested that critical reflections from within the crust and from the Moho were at least partly responsible for the high ground motions Observed in San Francisco and Oakland. However, as discussed below, these critical reflec— tions cannot totally explain the unusual pattern of ground motion and damage Observed in parts Of San Francisco and Oakland during the mainshock. D66 PERFORMANCE OF THE BUILT ENVIRONMENT MMI observations by Plafker and Galloway (1989) and E.V. Leyendecker (oral commun., 1992) indicate that dam- age on firm soils on the San Francisco Peninsula north of San Bruno were generally consistent with an MMI of VI, and that areas of saturated fill and Bay mud along the north- ern and eastern margins of San Francisco were consistent with an MMI of VII with isolated pockets of VIII and IX. The areas of higher intensity coincide with areas of rela- tively high ground motions observed during postearthquake seismological investigations (Boatwright and others, 1991a, b; Seekins and Boatwright, 1994). In contrast, Stover and others (1990) assigned a uniform MMI of VII to most of the San Francisco Peninsula. E.V. Leyendecker (oral commun., 1992) indicates that this assessment by Stover and others was somewhat conservative and does not reflect the distri- bution of damage observed from his field reconnaissance. Recorded ground motions in downtown Oakland on Merritt Sand were nearly as large as those recorded at nearby sites on Bay mud at the Oakland Harbor Wharf and at Emeryville. Yet, other nearby sites in east Oakland and Berkeley had ground motions significantly smaller than these. Damage in downtown Oakland and along the eastern margin of San Fran- cisco Bay where saturated artificial fill and Bay mud pre- dominate was found to be generally consistent with an MMI of VII with isolated pockets of VIII and IX (Plafker and Galloway, 1989; Stover and others, 1990; E.V. Leyendecker, oral commun., 1992). These observations are consistent with the high ground motions and intensities observed on similar sites in San Francisco. The lower ground motions recorded in Berkeley appear to be inconsistent with both the MMI of VII generally assigned to this area by Stover and others (1990) and the MMI of VI assigned to the area further north near Richmond, where some— what higher ground motions were recorded. After reviewing his notes, E.V. Leyendecker (oral commun, 1992) suggested that, in his opinion, the area north of downtown Oakland should have been assigned an MMI of VI, more consistent with ground motions and damage observed in east Oakland and Richmond. Based on the above observations, we adjusted the predicted ground motions for sites located on saturated artificial fill and Bay mud, including those sites located along the north- ern and northeastern margins of the San Francisco Penin- sula, and for sites located on Merritt Sand in downtown Oak- land and near Alameda (south of Oakland) to be consistent with the increased ground shaking expected on soft soil. The region around the city of Hollister was another area that exhibited unusually high ground motions during the earthquake. Stover and others (1990) assigned an MMI of VIII to this area, whereas damage at comparable distances in this same general vicinity was found to be more consis- tent with an MMI of VII or less. Steidl and others ( 1991) and Wald and others (1991) were unable to account for these larger than expected ground motions in their seismological models of the main shock and suggested that site effects may have been responsible for the higher ground motions in this area. Because of the strong empirical and theoretical evi- dence that indicated unusually high site amplification in this area, the predicted ground motions for sites located in the Hollister region were adjusted to be consistent with the in- creased ground shaking expected on soft soil. This adjust- ment produces estimates of PHA, PHV, and MMI that are consistent with those observed in the area (table 8). In conclusion, the comparison shown in table 8 indicates that the ground-motion models developed in this study ac- curately predict both the peak ground-motion parameters and the Modified Mercalli intensities observed throughout the central California region during the earthquake over a wide range of azimuths and distances. ACKNOWLEDGMENTS Dave Hampson, formerly with the U.S. Geological Sur- vey and now with EQE International, Inc., assisted in devel- oping the geologic descriptions of the strong-motion record- ing sites used in this study. Glen Reagor of the U.S. Geo— logical Survey provided a digital database of MMI observa- tions collected by the USGS. Klaus Jacob of Lamont-Doherty Geological Observatory, Roger Borcherdt of the U.S. Geo- logical Survey, and several anonymous reviewers provided insightful comments that significantly improved the manu- script. The study was initiated at the U.S. Geological Survey and was completed at Dames & Moore and EQE Interna- tional. The author would like to thank these organizations for their support. REFERENCES CITED Abrahamson, NA, and Silva, W.J., 1997, Empirical response spec- tral attenuation relations for shallow crustal earthquakes: Seis- mological Research Letters, v. 68, no. 1, p. 94-127. Boatwright, J., Fletcher, J.B., and Fumal, T.E., 1991a, A general inversion scheme for source, site, and propagation character- istics using multiply recorded sets of moderate-sized earth— quakes: Bulletin of the Seismological Society of America, V. 81, p. 1754—1782. Boatwright, J ., Seekins, L.C., Fumal, T.E., Liu, H.-P., and Mueller, C.S., 1991b, Ground motion amplification in the Marina Dis- trict: Bulletin of the Seismological Society of America, v. 81, p. 1980-1997. Boore, D.M., and Joyner, W.B., 1997, Site amplification for ge- neric rock sites: Bulletin of the Seismological Society of America, v. 87, p. 327-341. Boore, D.M., Seekins, L., and Joyner, W.B., 1989, Peak accelera- tions from the 17 October 1989 Lorna Prieta earthquake: Seis- mological Research Letters, v. 60, p. 151-166. Boore, D.M., Joyner, W.B., and Fumal, TE, 1997, Equations for estimating horizontal response spectra and peak acceleration from western North American earthquakes—a summary of recent work: Seismological Research Letters, v. 68, no. 1, p. 128—153. EMPIRICAL ANALYSIS D67 Borcherdt, RD, 1994, Estimates of site-dependent response spec- tra for design (methodology and justification): Earthquake Spectra, v. 10, p. 617-653. Borcherdt, RD, and Gibbs, J.F., 1976, Effects of local geological conditions in the San Francisco Bay region on ground mo- tions and the intensities of the 1906 earthquake: Bulletin of the Seismological Society of America, v. 66, p. 467-500. Borcherdt, RD, and Glassmoyer, G., 1992, On the characteristics of local geology and their influence on ground motions gener- ated by the Lorna Prieta earthquake in the San Francisco Bay region, California: Bulletin of the Seismological Society of America, V. 82, no. 2, p. 603-641. 1994, Influences of local geology on strong and weak ground motions recorded in the San Francisco Bay region and their implications for site—specific building-code provisions, in Borcherdt, R.D., ed., The Lorna Prieta, California, earthquake of October 17, 1989—strong ground motion: US. Geological Survey Professional Paper 1551-A, p. 77—108. Brady, AG, and Mork, RN, 1990, The Loma Prieta, California, earthquake, October 18 (GMT), 1989, processed strong-m0— tion records, volume 1: US. Geological Survey Open-File Report 90-247. Brady, AG, and Shakal, AF, 1994, Strong-motion recordings, in Borcherdt, R.D., ed., The Lorna Prieta , California, earthquake of October 17, 1989—strong ground motion: US. Geological Survey Professional Paper 1551—A, p. 9-38. Campbell, K.W., 1981, Near-source attenuation of peak horizontal acceleration: Bulletin of the Seismological Society of America, v. 71, p. 2039-2070. 1987, Predicting strong ground motion in Utah, in Gori, PL, and Hays, W.W., eds., Assessment of regional earthquake hazards and risk along the Wasatch Front, Utah: US. Geo— logical Survey Open-File Report 87-585, V. II, p. L1-L90. 1988, The Whittier Narrows, California, earthquake of Oc— tober 1, 1987—preliminary analysis of peak horizontal accel- eration: Earthquake Spectra, v. 4, p. 115-137. 1989, Empirical prediction of near—source ground motion for the Diablo Canyon Power Plant site, San Luis Obispo County, California: US. Geological Survey Open-File Report 89—484. 1990, Empirical prediction of near-source soil and soft-rock ground motion for the Diablo Canyon Power Plant site, San Luis Obispo County, California: Evergreen, Colorado, report prepared for Lawrence Livermore National Laboratory by Dames & Moore. 1991, An empirical analysis of peak horizontal acceleration for the Loma Prieta, California, earthquake of 18 October 1989: Bulletin of the Seismological Society of America, v. 81, p. 1838—1858. 1992, Recommended models for predicting strong ground mo- tion and MMI in the San Francisco Bay area: Evergreen, Colorado, report prepared for EQE International Inc. By Dames & Moore. 1997, Empirical near-source attenuation relationships for horizontal and vertical components of peak ground accelera- tion, peak ground velocity, and pseudo-absolute acceleration response spectra: Seismological Research Letters, v. 68, no. 1, p. 154-179. Campbell, K.W. and Bozorgnia, Y., 1994a, Empirical analysis of strong ground motion from the 1992 Landers, California, earth- quake: Bulletin of the Seismological Society of America, v. 84, no. 3, p. 573-588. 1994b, Near—source attenuation of peak horizontal accel— eration from worldwide accelerograms recorded from 1957 to 1993, in Fifth US. National Conference on Earthquake Engi- neering, Proceedings, Chicago, 1994: Berkeley, California, Earthquake Engineering Research Institute, v. III, p. 283—292. Catchings, RD, and Kholer, W.M., 1997, Reflected seismic waves and their effect on strong shaking during the 1989 Loma Prieta, California, earthquake: Bulletin of the Seismological Society of America, v. 80, no. 5, p. 1401-1416. California Strong Motion Instrumentation Program, 1991a, Plots to accompany tapes LOMAPRIETA89-IG (phase 1 data) and LOMAPRIETA89-G (phases 2 & 3 data), processed strong— motion data from the Lorna Prieta earthquake of 17 October 1989, ground-response records: Sacramento, California, Di- vision of Mines and Geology, Strong Motion Instrumentation Program Report No. OSMS 91-06. 1991b, Plots to accompany tapes LOMAPRIETA89-IB1, - 1B2, -IB3 (phase 1 data) and LOMAPRIETA89-IIB1, -IIB2, - IIIB (phases 2 & 3 data), processed strong-motion data from the Loma Prieta earthquake of 17 October 1989, building records: Sacramento, California, Division of Mines and Geol- ogy, Strong Motion Instrumentation Program Report No. OSMS 91-07. 1991c, Plots to accompany tapes LOMAPRIETA89—IL (phase 1 data) and LOMAPRIETA89—L (phases 2 & 3 data), processed strong-motion data from the Loma Prieta earthquake of 17 October 1989, lifeline records: Sacramento, California, Division of Mines and Geology, Strong Motion Instrumenta— tion Program, Report No. OSMS 91-08. EQE International, Inc., 1995, The Northridge earthquake of J anu- ary 17, 1994—pre1iminary report of data collection and analy- sis; part A, damage and inventory data: Irvine and Pasadena, California, report prepared for the Govemorr’s Office of Emer- gency Services of the State of California, by EQE Interna- tional, Inc. and the Geographic Information Systems Group of the Governorr’s Office of Emergency Services. Evernden, IF, and Thomson, J .M., 1988, Predictive model for important ground motion parameters associated with large and great earthquakes: US. Geological Survey Bulletin 1838, Fletcher, J.B., and Boatwright, J ., 1991, Source parameters of Lorna Prieta aftershocks and wave propagation characteristics along the San Francisco Peninsula from a joint inversion of digital seismograms: Bulletin of the Seismological Society of America, v. 81, p. 1783—1812. Fumal, TE, 1991, A compilation of the geology and measured and estimated shear-wave velocity profiles at strong-motion sta- tions that recorded the Loma Prieta, California, earthquake: US. Geological Survey Open—File Report 91—31 1. Hanks, TC, and Brady, AG, 1991, The Loma Prieta earthquake, ground motion, and damage in Oakland, Treasure Island, and San Francisco: Bulletin of the Seismological Society of America, v. 81, p. 2019-2047. Housner, G.W., 1990, Competing against time, Thiel Jr., C.C., ed., report to Governor George Deukmejian from The Governor’s Board of Inquiry on the 1989 Loma Prieta Earthquake: Sacra- mento, California, State of California, Office of Planning and Research. International Conference of Building Officials, 1996, Suggested revisions to the 1994 editions of the uniform codes (submit— tals for 1996): Whittier, California. D68 PERFORMANCE OF THE BUILT ENVIRONMENT Idriss, I.M., 1990, Response of soft soil sites in earthquakes, in Duncan, J.M., ed., H. Bolton Seed Memorial Symposium, Berkeley, California, 1990: Berkeley, University of California Press, v. 2, p. 273—289. 1991, written communication to Thomas F. Blake, Newberry Park, California. 1993, Procedures for selecting earthquake ground motions at rock sites: Gaithersburg, Maryland, National Institute of Standards and Technology, Report no. NIST GCR 93-625. Maley, R., Acosta, A., Ellis, F., Etheredge, E., Foote, L., Johnson, D., Porcella, R., Salsman, M., and Switzer, J., 1989, US. Geo- logical Survey strong—motion records from the northern Cali— fornia (Loma Prieta) earthquake of October 17, 1989: US. Geological Survey Open-File Report 89—568. McGarr, A.F., Celebi, M., Sembera, E.D., N oce, TE, and Mueller, CS, 1991, Ground motion at the San Francisco International Airport from the Lorna Prieta earthquake sequence, 1989: Bul- letin of the Seismological Society of America, v. 81, no. 5, p. 1923-1944. Mohraz, B., and FE. Elghadamsi, 1989, Earthquake ground mo- tion and response spectra, in Naeim, F., ed., The seismic de- sign handbook (Structural Engineering Series): New York, Van Nostrand Reinhold, chapter 2, p. 32~80. More, J .J ., Garbow, BS, and Hilstrom, KB, 1980, User guide for MINPACK- 1: Argonne, Illinois, Argonne National Laboratory, Report no. ANL-80—74. Plafl5 Hz) in the near-source region (R<30 km) but reduce to one—half or less at longer periods and far- ther distances. This behavior is also observed at rock sites; however, it is somewhat less pronounced. With a faster attenuation of spectral ordinates at higher frequencies, the shape of the response spectrum is found to change with distance. As expected, the spectral attenuation with distance is generally higher for the vertical spectrum than for the horizontal spectrum. The difference is particu- larly significant at the higher frequency end of the spectrum. Site amplification factors for stiff soil with respect to rock vary between 1.17 and 1.72 for the horizontal spectrum and between 1.01 and 1.81 for the vertical spectrum. Spectral amplifications at four sites underlain by soft soil and artifi- cial fill are also evaluated. This is done by a comparison of the observed spectra with those predicted for rock sites at corresponding distances. As expected, the resulting amplifi— cation factors at soft—soil sites show significant increase rela- tive to those at sites underlain by rock. INTRODUCTION The earthquake was the strongest event to occur on this segment of the San Andreas fault since the great San Fran— cisco earthquake of 1906. It caused loss of life and consider- able damage in the region, including the collapse of the Cy- press section of Interstate Highway 880. The details of source geometry and extent of damage are given in US. Geological Survey (1989a), Earthquake Engineer Research Institute (1990), and Housner (1990). The epicentral region was well instrumented, including over a hundred accelerographs which provided an important suite of high-quality triaxial accelerograms. The behavior of the peak ground accelera- tion (PGA) has previously been studied by Boore and others (1989), Campbell (1991), and Niazi and others (1992). De- spite some differences on the interpretation of the results, these studies suggest that the observed horizontal PGA for this earthquake exceeded previous model predictions. The purpose of this study is to extend the analysis of PGA to the response spectra and to assess the intensity of vertical ground motion in the frequency range of engineering inter- est. Both vertical and horizontal response spectra are ana- lyzed, and the results are given for sites characterized by rock and by stiff soil. D69 D70 EARTH STRUCTURES AND ENGINEERING CHARACTERIZATION OF GROUND MOTION DATA SOURCE The database consists of the pseudovelocity response spec- tra (PSV) calculated at 5 percent of critical damping at 53 sites within 100 km distance of the source as shown in figure 1. Site geology at 22 sites is characterized as rock, 27 as stiff soil (alluvial sites), and 4 as soft soil (Bay mud and artificial fill). The majority of the stations are maintained by the Strong Motion Instrumentation Program (SMIP) project of the Cali- fornia Department of Conservation (Shakal and others, 1989; Huang and others, 1990), and by the US. Geological Sur- vey Strong Motion Program (US. Geological Survey, 1989b). The spectral ordinates used in this study have been calcu- lated and published by these agencies from bandpass filtered accelerograms. The analysis is performed at 16 incremental periods rang- ing between 0.05 and 2.0 s. Vertical PSV’s are directly used in the analysis. For the horizontal PSV’s , the arithmetic mean of the ordinates derived for the two horizontal components are entered. This is consistent with the previous studies by Niazi and Bozorgnia (1992a) and Campbell (1993). In total, nearly 2,500 spectral ordinates are analyzed. Table 1 gives a list of recording stations used in this study and their site ge- ology. The distances are from the vertical projection of the rupture on the Earth’s surface and have been discussed by Boore and others (1989) and Niazi and others (1992). The vertical and horizontal spectral ordinates at periods 0.1 and 0.5 s are also given in table 1. DATA ANALYSIS Data analysis was performed by the application of a multiregression procedure (SAS, 1985) with distance and I 1297-30" \ 121130' —37°4s' ... PACIFIC OCEAN o A Rock A. . Stiff Soil 3 X SOH‘ Soil . Projected Rupture Area . AA 0 -- 36°37’30" Km ... E: o 10 20 Figure 1.—Locations of 53 recording sites considered in this study. The shaded area is the surface projection of the rupture plane (Boore and others, 1989; Niazi and others, 1992). AT TENUAT ION OF VERTICAL AND HORIZONTAL RESPONSE SPECTRA OF THE LOMA PRIETA EARTHQUAKE D71 Table 1.—Rec0rding stations used in this study Recording Distance ROCK (**) Vertical PSV Vertical PSV HOR. PSV HOR. PSV Stations (km) at period at period at period at period 0.1 s 0.5 s 0.15 0.55 (in/s) (in/s) (in/s) (in/s) Corralitos 0.1 1 5.230 13.600 4.325 37.400 Capitola 8.6 0 8.880 6.610 4.675 24.800 Gilroy#1 10.5 1 1.750 6.670 5 .965 26.700 Gilroy G C 10.9 0 2.250 4.770 4.485 18.900 Saratoga 11.7 0 4.720 7.740 4.030 18.500 Gilroy#2 12.1 0 3.070 5.320 3.145 21.850 Gilroy Cm Bldg 12.3 0 2.810 5.270 2.130 18.500 Santa Cruz 12.5 1 8.980 3.880 4.730 12.750 San Js. St.Te. 13.2 0 4.950 6.040 4.545 13.500 Gilroy#3 14.0 0 3 .640 7.000 5 .980 26.800 Gilroy#4 15.8 0 2.250 3.290 2.545 27.250 Gilroy#6 19.0 1 1.370 3.650 1.825 7.585 Anderson D.ds 20.0 0 1 .629 6.276 2.474 16.185 Coyote L ds 2] .7 0 1.490 4.960 1.670 1 1.1 15 Gilroy#7 MR 24.3 0 1.550 1.630 2.510 18.850 Hollister AP 25 .4 0 2.256 4.000 2.042 21 .642 Agnews SH 27.0 0 1.240 5.130 1.510 8.340 Sunnyv. Colt. 27.5 0 1.742 3.407 2.029 13.096 Hollister CH 27.8 0 2.058 4.000 1.492 19.407 Hollister W.H 28.3 0 1 .750 4.640 2.005 29.150 Hollister S&P 28.3 0 2.610 4.920 1.995 28.400 Halls Valley 29.3 0 .843 6.980 .962 9.725 SAGO Vault 29.9 1 .612 1.796 .570 3.352 Salinas 31.4 0 1.520 2.670 .901 6.235 Milpitas 25 31.4 0 1.290 2.980 1.035 9.860 Sago S 34.1 1 .467 3.610 .579 5.005 Stanford SLAC 35.0 1 1.885 6.043 2.099 17.526 Calveras Res. 36.1 0 1.026 1.603 .900 4.352 Men10 Park VA 36.5 0 2.696 2.701 1.719 12.268 Woodside 38.7 0 .388 2.000 .597 6.790 Fremont Msj 42.0 0 1.450 4.460 1.185 7.750 Monterey CH 42.7 1 .434 1.480 .783 3.335 U Crystal #7 46.5 1 .681 3.070 .977 6.615 U Crystal 10 46.6 1 .298 3.060 .700 9.290 Foster City 47.3 * 1.650 4.500 1.895 23.900 Hayward JMS 53.6 0 1.250 1.790 1.565 9.830 (*) Sites specified by an asterisk (*) are on soft soil. (**) Site parameter ROCK is 1 for rock sites and 0 for stiff-soil sites. D72 Table 1 .—Continued EARTH STRUCTURES AND ENGINEERING CHARACTERIZATION OF GROUND MOTION Recording Distance ROCK (**) Vertical PSV Vertical PSV HOR. PSV HOR. PSV Stations (km) at period at period at period at period 0.1 s 0.5 s 0.1 s 0.5 s (in/s) (in/s) (in/s) (in/s) Hayward CSUHFF 56.0 1 .496 2.590 .831 3.740 Ha)IWard BART 57.1 0 .972 3.640 2.240 5.970 SFO 63.2 * .474 3.750 1.940 21.150 S SF Si.Pt. 67.6 1 .274 2.390 .657 3.775 SF Dim Hi ghts 75.9 1 .409 3.200 .903 7.200 Piedmont Jr.H 77 .2 l .236 1.730 .628 5.745 SF Rincon Hill 7 8.5 1 .328 2.590 .825 4.445 Oaklnd O Har W 78.8 * .705 4.360 1.830 18.400 Yerba Buena I 79.5 1 .230 2.090 .425 3.320 SF Pac Hights 80.5 1 .198 2.220 .409 2.980 SF Tel Hill 80.9 1 .234 1.870 .536 4.145 Tresure I 81.7 * .1 13 .574 .937 9.700 Presidio 81.9 1 .462 3.780 1.096 1 1.520 LBL #1 83.2 1 .245 2.980 .547 6.025 SF Cliff House 83.2 1 .385 4.460 .600 6.685 Pt. Bonita 87.5 1 .240 1.700 .539 5.625 Richmond CH 92.0 0 .373 2.370 .915 6.615 (*) Sites specified by an asterisk (*) are on soft soil. (**) site geology as independent variables. Hence, as seen in equa— tion (1), the site parameter “5” would allow an adjustment for site geology of PSV for stiff soil relative to rock. 1n(PSV)=a+d*ln(R+c)+s*ROCK+ e(0,<52) (1) In this equation R is distance in kilometers, PSV is pseudovelocity spectral ordinate in inches/second, ROCK is a site variable which has a value of l for rock and 0 for stiff soil, and a, c, d, and s are regression parameters. The input to the regression is confined only to the information at 49 sites characterized by either rock or stiff soil. The four soft- soil sites are used later for the evaluation of the amplifica- tion due to soft soil material. In equation (1) 8 represents random error with zero mean and (52 variance. A nonlinear regression procedure (SAS, 1985) is applied to the spectral ordinates at each of the 16 frequencies. A sub- jective weighting method is introduced to compensate for the concentration of the observations within specific distance bins. Because the employed methodology and the set up is similar to the one used in the study of PGA behavior by Niazi and others (1992), for further detail the reader is referred to that study. Site parameter ROCK is 1 for rock sites and O for stiff-soil sites. DISTANCE-DEPENDENT RESPONSE SPECTRA The numerical results for the regression parameters de- rived for the horizontal and vertical components are given in tables 2 and 3, respectively, as functions of period. Figure 2 shows the attenuation of the 5 percent damped vertical and horizontal response spectra at a period of 0.1 s for sites with stiff soil geology. The observed spectral ordinates for stiff soil sites are also marked in figure 2. Figures 3 and 4 show the attenuation of spectra at 0.5 and 1.0 s, respectively. It is evident from figures 2 through 4 that the magnitude of the vertical and horizontal spectral ordinates are comparable at the short period (0.1 5), while at the longer periods the horizontal spectra exceed those for vertical motion. This is due to the fact that the vertical ground motion is in general enriched in the higher fre- quency waves relative to the horizontal component. Thus, at the short period end of the spectrum, the vertical mo- tion is pronounced. At longer periods, however, the verti- cal motion does not contain enough energy to strongly excite a long-period oscillator; therefore, as shown in fig— AT TENUAT ION OF VERTICAL AND HORIZONTAL RESPONSE SPECTRA OF THE LOMA PRIETA EARTHQUAKE D73 Table 2.—Computed parameters of regression analysis for horizontal response spectra T(s) a c d s sigma l 0.05 2.882 7.666 —0.924 —0.266 0.372 2 0.075 3.791 9.437 -1.008 -0.264 0.369 3 0.10 4.877 13.094 -1 . 147 —0.268 0.400 4 0.11 5.330 14.990 —l.196 —0.316 0.425 5 0.15 6.118 16.377 -1.257 -0.323 0.433 6 0.20 5.870 12.974 -1.113 -0.378 0.391 7 0.28 5.331 . 6.570 -0.932 -0.203 0.427 8 0.30 5.290 6.324 —0.908 —0.164 0.418 9 0.40 6.020 8.871 —0.997 —0.268 0.422 10 0.50 4.799 3.177 -0.657 -0.380 0.416 1 1 0.60 4.621 1.619 -0.576 -0.477 0.402 12 0.70 4.351 0.780 -0.480 —0.455 0.449 13 0.80 3.847 0.118 -0.337 -0.500 0.451 14 1.00 3.624 0.548 -0.267 -0.379 0.503 15 1.50 4.560 3.882 —0.529 -0.546 0.576 16 2.00 3.575 1.577 —0.308 -0.528 0.543 Table 3.—Computed parameters of regression analysis for vertical response spectra T(s) a c d s sigma 1 0.05 5.289 12.473 -1.577 -0.486 0.479 2 0.075 7.106 18.134 -1.804 -0.596 0.464 3 0.10 6.898 14.182 -1.688 -0.573 0.471 4 0.11 5.938 10.265 - l .461 -0.563 0.453 5 0.15 6.600 12.728 -1.513 —0.579 0.552 6 0.20 5.449 5.624 -1.224 -0.432 0.521 7 0.28 4.479 4.757 —0.954 -0.225 0.356 8 0.30 4.494 6.413 -0.939 -0.138 0.347 9 0.40 4.380 11.148 -0.803 -0.283 0.421 10 0.50 2.884 1.510 -0.460 -0.046 0.336 1 l 0.60 3.078 4.145 —0.451 —0.094 0.470 12 0.70 2.774 1.738 -0.379 -0.013 0.481 13 0.80 2.506 0.544 -0.301 —0.041 0.437 14 1.00 2.520 0.322 -0.222 -0.240 0.436 15 1.50 3.634 9.837 -0.527 —0.200 0.462 16 2.00 3.327 7.806 -0.479 -0.209 0.576 ures 3 and 4, the amplitude of the vibration in the horizontal direction exceeds that of the vertical motion. Figure 5 shows the empirically predicted median horizon- tal response spectra at 10, 20, 40 and 60 km distances from the source, derived from equation (1) by the substitution of the numerical values of the regression, as listed in table 2. Figure 5A shows the attenuation of response spectra for stiff soil and SB is for rock. Figure 6 shows similar plots for ver- tical response spectra based on the coefficients listed in table 3. Figure 6A shows the predicted median response spectra at sites underlain by stiff soil and 6B shows the same for rock. It should be noted that within the distance range of R<30 km, predictions for rock sites are not as tightly constrained as for soil sites, by virtue of fewer recordings (see table 1). Figures 5 and 6 show that the shape of response spectra (for both vertical and horizontal components) is distance dependent; that is, as distance from the source increases, the spectra gradually become depleted from higher frequency waves in a relative sense. This is evident since the medium between the source and recording site is a dynamic system which acts as a low-pass filter. The suggested distance de- pendence of the shape of the response spectrum clearly indi- D74 EARTH STRUCTURES AND ENGINEERING CHARACTERIZATION OF GROUND MOTION lO: _ Q _ _ LLJ (f) — .. \ Z l: j * > i i U) _ D. _ 0.1 llIlI l I lIlII lllll l l I I III! 100 m [MSTANCE (KM) m [NSTANCE (KM) Figure 2.—Attenuation of 5 percent damped vertical (A) and horizontal (B) response spectra at a period of 0.1 s for stiff soil sites. Solid curve is for the median spectrum. One standard deviation band is shown by the dashed curves. The observed spectral ordinates are also marked. mo: : I A 3 B A I 1 Q _ - L|_l (D _ _ \ Z 10—. t V Z : >1: > I I (f) _ _ C1— * 1 Ill|| l l I lllll IIIll I I l Illll 100 10 100 DISTANCE (KM) 10 DISTANCE (KM) Figure 3.——Attenuation 0f 5 percent damped vertical (A) and horizontal (B) spectra at a period of 0.5 s for stiff soil sites. Solid curve is for the median spectrum. One standard deviation band is shown by the dashed curves. The observed spectral ordinates are also marked. cates that scaling of a standard spectral shape with a single parameter, such as PGA, is not justified. This was also noted by Krawinkler and others (1991) regarding the horizontal spectrum. Comparison of figures 5Aand SB shows that the spectral ordinates at stiff-soil sites are generally larger than those at rock sites. This is also true for vertical spectra (fig. 6), ex- cept, by different factors. It is also evident from tables 2 and 3 where the site parameter “s” for soil sites is negative for both horizontal and vertical spectra. The resulting amplifi- cation factor at stiff-soil sites relative to rock varies between 1.17 and 1.72 for the horizontal spectrum and between 1.01 and 1.81 for the vertical spectrum. Application of the regres- sion procedure to equation (1) at discrete periods results in a site parameter “s” as a function of the wave period. The model adopted in this study, however, does not allow evalu- AT TENUAT ION OF VERTICAL AND HORIZONTAL RESPONSE SPECTRA OF THE LOMA PRIETA EARTHQUAKE D75 IOO _ _ Figure 4.—Attenuation of5 Z A I B percent damped vertical (A) I Z and horizontal (B) spectra at A — - x i period 1.0 s for stiff soil 0 - ' F ‘ ‘ x 3; sites Solid curve is for the DJ =o< * ‘ \ .' (f) _ * ’I‘ * * * * ‘ \ * median spectrum. One stan- \ * -- * * dard deviation band is Z 10‘ TI: \ T ‘ \ J: - “>§’~-~__E’_"*_§ ** shownbythedashedcurves. \__/ i * * ‘ r; \ E E “r“‘I— ««««« The observed spectral ordi- Z * * * * — * * nates are also marked. > : _______________ * * * * : (f) :0: "I‘*‘-«.*.___ _ D. _ _ * * _ 1 I I I I I I T I I I I I I I l I | l I | l I I I I I 10 NCE (KM) 100 DIOSTANCE (KM) 100 D I STA STIFF SOIL A 8 Figure 5.—Attenuation ' ‘ ROCK S'TE of 5 percent damped A 10— A 10— spectra: Median hori— B 3 a E zontal response spectra (/3 (f) : at distances R=10, 20, E E - 40, 60 km from the sur— 3 Q ' face projection of the > ‘ > ‘ rupture plane for (A) stiff 8 R=1O (f) soil and (B) rock sites. I i R=2O CL 1'; R=i0 01' R240 01' R=20 E R=60 % R=40 R=60 O‘1001 IMHO.) | 'WI I Y ......10 0"1001 I HmloliI Y 'Wi ‘IHHIO PERIOD (SEC) I PERIOD (SEC) A Figure 6.—Attenuation — STIFF SOIL - ROCK SITE 9 of 5 percent damped spectra: Median vertical 8 10? 8 10—; response spectra at dis- Lfi : a : tances R=10, 20, 40, 60 \ < \ — kmfrom the surface pro— 3 ' é ‘ jection of the rupture plane for (A) stiff soil 5 R: 1 O 5) and (B) rock sites. [1 1F: R=20 0- 1': R=1O D: . [l’ « R=20 g R=4O g : R=60 R=40 4 - R=60 0.1 I IIIIIIII IIII'HI I \fiIIIII 0-1 I I'HII'I I ""“'l I IIHIII 0.01 10 0.01 10 0,1 I PERIOD (SEC) 0.7 PERIOD (SEE) D76 ation of distance dependence of the parameter “s,”unless, as explained below, independent analyses are performed on the subsets of data collected at rock and soil sites, separately. It is of interest to note that for the vertical spectrum the site amplification factor “5” generally tends to decrease with period, whereas for the horizontal spectrum the trend is re— versed, especially for periods longer than about 0.3 s (see tables 2 and 3). UNIFORM GEOLOGY MODELS The regression coefficients of tables 2 and 3, and the re— sulting empirical spectra of figures 5 and 6 are all based on the analyses of the entire data for both stiff soil and rock sites combined. Next, the data files are split into two sub- sets, one containing stiff—soil data and the other rock data. By doing so we search for possible bias in the predictions at stiff-soil sites which may be introduced by the observations at rock sites and Vice versa. As already explained, there are 22 rock sites and 27 sites characterized by stiff soil in the data. Application of the regression procedure given in equa- tion (1), without the soil classification term, to the two data subsets, results in four sets of independent regression coeffi- cients which may somewhat differ from those listed in tables 2 and 3. The resulting empirical models of horizontal and vertical pseudoacceleration spectra (PSA) obtained from the two procedures are compared in figures 7 and 8, respectively. Figure 7A compares the horizontal acceleration spectra for the subset of data observed at stiff-soil sites (dashed curves) 1.60 , _ A 1.40 — p F X0 a ‘1 - d ,I 1.20 — ,1 .y A - f I on , ,1 b V - [I /\ ‘1 1.00 — 0 /\ b/M+SIGMA, R=10 < — ,1 /v 1 ‘ COMBINED SET m j .3 / I *3 CL 080 — ‘1 01‘ i ,0’ W O 0.60 : 1 I J “ R=1O O 40 1 if“: 4 BB R=2O 0.20 15% R=4o - R=60 0.00_ I lllllll' l | 111111] 1 1111111 0.01 0.1 1 RERIOD (SEC) EARTH STRUCTURES AND ENGINEERING CHARACTERIZATION OF GROUND MOTION and for the combined data (solid curves) at 10, 20, 40 and 60 km. Figure 7B shows the same comparison at rock sites. Corresponding plots for the vertical component are shown in figure 8, in which again figure 8A compares the predic- tions of stiff-soil observations with those of the combined data and figure SB compares those of rock observations and the entire data. It is noted that the largest separation between the dashed and solid curves occurs at near-source distances of up to 10 km, where very few observations exist, espe- cially for rock sites. To provide a scale for the comparison, in each example we have also plotted the mean plus sigma (84.1 percent probability) of predictions at 10 km distance for the combined set. In comparison with one sigma, even the largest observed differences between the prediction re- sults are thus relatively small and the models based on the entire data are confirmed. VERTICAL-TO-HORIZONTAL SPECTRAL RATIOS Using the regression parameters of tables 2 and 3, verti- cal-to-horizontal ratios (V/H) of the predicted response spec- tra at four given distances of 10, 20, 40 and 60 km are derived and shown in figure 9. It is clear from figure 9 that V/H spectral ratio is distance—as well as frequency—de- pendent. The highest ratios are observed for the high-fre- quency motion in the near—source region (Niazi and Bozorgnia, 1992b). At more distant sites, the spectral ratios decrease apparently due to a faster rate of attenuation for the 1.50 a B 1.40 3 1.20 ~— /-\ “ 0 ~ In. \_/ ‘ (I ‘0 1.00 — ,1 1‘ $ : ,IM' CL 0'80 _ “ M SIGMA R 10 _ , O‘ + y = Q:- 3 1° / COMBINED SET 0 O 0.60 : q: I a I 1 0.40 — - O _ z< 0.20 — EE 0 _ Q ~ “2% O - m 0 0.00 l I IIIII|[ I I IIIIIII I IIIIIII 0.01 10 0.1 1 RERIOD (SEC) Figure 7.—Comparison of the results of two regression analyses on the spectral ordinates: median horizontal response spectra at distances R=10, 20, 40, 60 km for (A) stiff soil and (B) rock sites. Solid curves are for the application of equation (1) to the combined soil and rock data and assigning proper value for variable “ROCK." Dashed curves are for the analysis of (A) soil data only and (B) rock data only. The 84th percentile spectra for R210 km of the combined analysis are also shown. AT TENUAT ION OF VERTICAL AND HORIZONTAL RESPONSE SPECTRA OF THE LOMA PRIETA EARTHQUAKE D77 hi gh-frequency-rich vertical motion as compared to the low- frequency horizontal component. The observed variation with frequency and distance at stiff- soil sites (fig. 9A) shows similar behavior as those reported previously at stiff-soil sites in widely separated geographic regions of northeastern Taiwan and east—central Iran (Niazi 31.20 ( PSA VERTICAL 1.60 I A 1.40 5 81.20 — 9‘ f1 \_/ _ X, b” 6" <1.00 9 ,1" 1 U7 - ,1 \ / II R O D_ ' 1’ / \ 1 M+SIGMA, =1 0 80 4 1’ V \\'.‘/COMB|NED SET :1:I 4 0’ 1‘1 9 0.60 — E :Z< R=10 g 0.40 5 j: :00 R220’ 0'20 _ M“ R 40 ‘ 2 = ‘ \ - CL R=60- W ' (n 0.00 l llllllll I lIlllllI l Illllll 0.01 01 1 PERIOD (SEC) and Bozorgnia, 1992a). Spectral ratios generally exceed the commonly used value of two-thirds at periods shorter than 0.2 s, particularly at near—source distances of R < 30 km. The ratios, however, fall below one-half at longer periods for all distances; that is, the use of two—thirds factor would likely result in overconservatism at long periods. Spectral 9 m o .0 m 0 pl 1 ’ ‘1/M+SIGMA, R=10 as COMBINED SET .0 4; O 020 MEMAN SPECTRA x H N o R=4O R260 111111111 11 0.1 1 PERIOD (SEC) 000 1111111 1 1111111 O4l_Llelll|ll1|I1lllLlll|lIlllllll .0 Figure 8.—Comparison of the results of two regression analyses on the spectral ordinates: Median vertical response spectra at distances R=10, 20, 40, 60 km for (A) stiff soil and (B) rock sites. Solid curves are for the application of equation (1) to the combined soil and rock data and assigning proper value for variable “ROCK.” Dashed curves are for the analysis of (A) soil data only and (B) rock data only. The 84th percentile spectrum for R=10 km of the combined analysis is also shown. 1.60 A V—PGA/H—PGA 1'40 R=10 0.74 R=20 0.66 R=4O 0.57 1-20 R=10 R=6O 0.52 'o o VER. PSV/HOR. RSV C N c 111lL11|1111111l111|111|111|111 70 H 4; O l 000 I 11111111 1 I OJ 1 PERIOD (SEC) 117111] T |||l|ll O01 1.60 : B : V—PCA/H—PGA >140 _ R=10 0.63 — R=20 0.57 W ~ R=4O 0.50 D_ 1.20 : R260 0.46 . J K L O 1.00 _ I .. \0 80 A > ' T V) : {l 0.60 — It : LI_I 0.40 4 > : 0.20 ~— 000— l Illlllll YTTIIIIII I I llllll 0.01 10 04 1 PERIOD (SEC) Figure 9.—Ratio of predicted vertical to horizontal spectral ordinates at distances R=10, 20, 40, and 60 km from the source at (A) stiff soil and (B) rock sites. Regression was performed on the combined soil and rock data set. The ratios of vertical to horizontal PGA’s are also given (Niazi and others, 1992). D78 EARTH STRUCTURES AND ENGINEERING CHARACTERIZATION OF GROUND MOTION ratios for different distances converge to a common point at a period of about 0.3 s, where the ratio becomes independent of distance at a value of 0.4. The same behavior is also observed at rock sites (fig. 93), however, not as pronounced as for soil. The V/H spectral ratios at stiff—soil sites are larger than those at rock sites for periods shorter than 0.4 s, and slightly less for longer peri— ods. The predicted values of the V/H ratio of PGA for this earthquake are also listed in the insets of figures 9A and 9B for the corresponding site geology. These values are based on the empirical predictions of Niazi and others (1992). We offer the following preliminary explanation for the observed variation of the V/H spectral ratio. Since the verti- cal motion is enriched in higher frequencies, it will be at- tenuated at a faster pace. The behavior of V/H spectral ratio is due to the combination of two factors. The first factor is the differential attenuation of higher frequency P waves (pre- dominantly vertical) relative to shear waves. The second fac- tor is the ray geometry, whereby at short distances the high- frequency P waves have a larger vertical component than shear waves, owing to a near-vertical incidence and align- ment of particle motion with the direction of propagation. The V/H spectral behavior observed in several regions, over a wide range of peak accelerations, therefore appears to be real. The existence of deep sedimentary cover (stiff soil) apparently accentuates this behavior. Whether the nonlinear response of the soil has an influence is not clear. Observa- tions by Darragh and Shakal (1991) of the main shock and several aftershocks at the Treasure Island site clearly dem- onstrate that horizontal response at sites underlain by soft sedimentary material and artificial fill is modified substan- tially by the nonlinear behavior of the soil. However, the same 8.00 SFO OAKLAND WHARF A TREASURE ISLAND FOSTER CITY 39 A \1 'o o DUO-x- .0’ o o 9“ o o P“ o o SOFT/ROCK SPECTR '0 o lIlllllllllllllllllllllllI|lllllllllllll|||lll|lllHltllJlllJllllllllllllllLlll 8'00 * SFO o OAKLAND WHARF B u TREASURE ISLAND (700 A FOSTER CITY SOFT/ROCK SPECTR 0.00 I I llllll I||l|l|[ r RERTOD (SEE) llllllll I .0 o authors failed to detect a similar effect at the Gilroy#2 sta- tion with stiff-soil geology. SPECTRAL AMPLIFICATION OF SOFT SOIL In order to evaluate site amplification at stations under- lain by soft soil, the observed response spectra at four sites with soft-soil geology are compared with the predictions of our empirical results at rock sites, computed at correspond- ing distances. These four sites are denoted by asterisk (*) in table 1. Figure 10 shows the result of this comparison at in- dividual sites for both horizontal and vertical components. The observed site amplifications of horizontal spectra, as seen in figure 10A, show similar features at the four sites considered here. Spectral ratios show a gentle negative trend approximately up to the 0.2—s period. For the Oakland Wharf, Treasure Island, and Foster City sites, amplification peaks occur at 0.6-0.7-s and 1.5-s periods. For the San Francisco Airport (SFO), these peaks occur at shorter periods, imply- ing a thinner soft-soil column. The peak horizontal spectral amplification varies between 4.5 and 7 at these four sites. Darragh and Shakal (1991) also found a peak Fourier spectral ratio of 4 for the strong phase of the horizontal motion at Treasure Island relative to the neighboring rock site at Yerba Buena Island. However, they also noted that the observed site response is systematically higher for smaller magnitude aftershocks. Site amplification for the vertical component (figure 10B) is generally smaller than that of horizontal component. Ex— cept for the 1.5—s peak at the Oakland Wharf with an ampli- P“ o o .0“ o o P4 o o N o o '0 o ET B {Emir |lIlIIIl|lllllllllllLIlllLUlllllllIllllIlllltlIlllllllIl|lllllll|llllll||lllll 0.00 , llllllll I 0.1 l PERIOD (SEC) llTIIIII I .0 0 Figure 10.—Ratio of observed spectra at soft-soil sites over predicted spectra at rock sites for (A) horizontal and (B) vertical ground motion. AT TENUAT ION OF VERTICAL AND HORIZONTAL RESPONSE SPECTRA OF THE LOMA PRIETA EARTHQUAKE D79 fication factor of 3.8, peak amplifications occur predomi- nantly at periods shorter than 0.3 s. As already reported by Niazi and others (1992), the vertical component at the Trea- sure Island site shows deamplification rather than amplifi- cation. DISCUSSION OF RESULTS Previous studies of a more extensive database recorded at SMART— 1 array, Taiwan (Niazi and Bozorgnia, 1989a, 1989b, 1992a), showed, for the first time, a strong dependence of V/H spectral ratios on both distance and wave frequency. Some of the suggestions emerging from that study are as follows: 1.The shape of the response spectra for both vertical and horizontal components of the ground motion, and their ra- tio, is distance—as well as magnitude—dependent. 2.The V/H spectral ratio at the high-frequency end of the spectrum shows a substantial increase relative to the stan- dard two-thirds value. The increase is particularly signifi- cant at near-source distances. 3.At low frequencies and far—field distances, the spectral ratios reduce significantly, such that the two-thirds ratio would become conservative. Most of these suggestions, with the exception of the mag- nitude—scaling effects (inapplicable to a single earthquake) also apply to the Lorna Prieta observations. A study of the magnitude-scaling effect on this ratio requires simultaneous analysis of several earthquakes over a broad range of mag- nitude. Table 3 shows that in general the short—period end of the vertical response spectrum is characterized by higher values of the attenuation parameter d, site amplification parameter s, and to some extent the distance saturation parameter (depth term) c. However, the absolute value of (1 tends to decrease almost monotonically down to a value of 0.22 at 1 Hz. Be- low 1 Hz, the long-period noise may become significant. The correlation of attenuation with frequency is expected as a consequence of intrinsic attenuation and scattering. There may also be a slight possibility of a tradeoff between c and d parameters. For the horizontal motion, a similar trend is ob- served in the variation of the attenuation parameter d and the depth term 6; however, vertical motion attenuates much faster at periods below 0.2 s than the horizontal motion (see tables 2 and 3). The same differential attenuation between vertical and horizontal motions was also observed in the analysis of the SMART-1 data below 0.2 s (Niazi and Bozorgnia, 1992a). However, the SMART-1 data showed much higher attenua- tion at periods above 0.2 s than Lorna Prieta for both vertical and horizontal components. Tables 2 and 3 show a tendency for the site-amplification factor to decrease with period for the vertical spectrum but to increase for the horizontal com- ponent, especially for periods longer than about 0.3 s. As recognized by previous investigators, the effects of the vertical ground motion on the response of structures may not always be neglected. For example, as Newmark and Hall (1982) indicated, columns and walls in compression, beams, and other horizontal elements are particularly vulnerable to the vertical component of ground motion. Other examples include the effects on the seismic response of multistory frames (Anderson and Bertero, 1973), concrete gravity dams (Chakrabarti and Chopra, 1973), and reinforced concrete highway bridges (Saadeghvaziri and Foutch, 1991). In the estimation of near—source vertical ground motion so far, reliance has been placed on the extrapolation of far—field information due to the scarcity of high-quality observations close to the source. Recent studies of SMART-1 data as well as observations of the 1978 Tabas, Iran, and the Loma Prieta earthquakes made it possible to quantify the range of dis- tances and frequencies where the vertical ground motion is pronounced. ACKNOWLEDGMENTS The authors wish to thank the US Geological Survey and California Department of Conservation for providing the accelerograms used in this study. Copyright release provided by John Wiley & Sons, Ltd, is also acknowledged. REFERENCES Anderson, J .W., and Bertero,V.V., 1973, Effects of gravity loads and vertical ground acceleration on the seismic response of multi- story frames, in Proceedings of the Fifth World Conference on Earthquake Engineering, Rome, p. 2914—2923. Boore, D.M., Seekins, L., and Joyner,W.B., 1989, Peak accelerations from the 17 October 1989 Lorna Prieta earthquake: Seismologi- cal Research Letters, v. 60, no. 4, p.151-166. Campbell, K.W., 1991, An empirical analysis of peak horizontal ac— celeration for the Loma Prieta, California earthquake of Octo— ber 18, 1989: Bulletin of the Seismological Society of America, V. 81, no. 5, p. 1838-1858. 1993, Empirical prediction of near-source ground motion from large earthquakes, in International workshop on earthquake haz- ard and large dams in Himalaya. Chakrabarti, P., and Chopra, AK. 1973, Hydrodynamic pressures and response of gravity dams to vertical earthquake component: Earthquake Engineering and Structural Dynamics, V. 1, p. 325-335. Darragh, RB, and Shakal, AF, 1991, The site response of two rock and soil station pairs to strong and weak ground motion: Bulle- tin of the Seismological Society of America, V. 81, no. 5, p. 1885- 1899. Earthquake Engineering Research Institute, 1990, Loma Prieta Earth— quake Reconnaissance Report, Supp. Earthquake Spectra, V. 6. Housner, G.W., 1990, Competing against time; report to Governor G. Deukmejian, Board of Inquiry on the 1989 Lorna Prieta earth- quake: The State of California, Office of Planning and Research, Sacramento. D80 EARTH STRUCTURES AND ENGINEERING CHARACTERIZATION OF GROUND MOTION Huang, M.J., Cao, T.Q.,Vetter, U.R., and Shakal, A.F.,1990, Third interim set of CSMIP processed strong-motion records from the Santa Cruz Mountains (Lorna Prieta), California, earthquake of 17 October 1989: California Strong Motion Instrumentation Program, California Department of Conservation, Division of Mines and Geology, Report No. OSMS 90-05. Krawinkler, H., Nasser, A., and Rahnama, M., 1991, Damage poten— tial of Lorna Prieta ground motions: Bulletin of the Seismologi- cal Society of America, v. 81, no. 5, p.2048-2069. Newmark, N .M., and Hall, W.J., 1982, Earthquake spectra and de- sign: Earthquake Engineering Research Institute. Niazi, M., and Bozorgnia, Y.,l992a, Behavior of near-source vertical and horizontal response spectra over SMART-1 array, Taiwan: Earthquake Engineering and Structural Dynamics, v. 21, p. 37-50. 1992b, Vertical response spectra of Lorna Prieta earthquake as functions of distance and site geology: Seismological Re- search Letters, v. 63, no. 1. 1989a, Behavior of vertical ground motion parameters in the nearfield: Seismological Research Letters, v. 60, no. 1, January- March 1989. 1989b, Empirical modeling of site specific response spectra for Lotung, Taiwan, in Proceedings of International Associa- tion of Seismology and Physics of the Earth’s Interior, Istanbul, August 1989. 15‘! U.S. GOVERNMENT PRINTING OFFICE: 1998 — 773-051 / 20126 Region No. 8 Niazi, M., Mortgat, CR, and Schneider, J .F., 1992, Attenuation of peak ground acceleration in Central California from observa- tions of the October 17, 1989 Lorna Prieta earthquake: Earth- quake Engineering and Structural Dynamics, v. 21, p. 493-507. Saadeghvaziri, M.A., and Foutch, DA, 1991, Dynamic behavior of R/C highway bridges under the combined effect of vertical and horizontal earthquake motions: Earthquake Engineering and Structural Dynamics, v. 20, p. 535-549. SAS, 1985, User’s Guide; Statistics, Version 5: SAS Institute, Cray, N.C. Shakal, A., Huang, M., Reichle, M., Ventura, C., Cao, T., Sherbume, R., Savage, M., Darragh, R., and Petersen, C., 1989, CSMIP strong—motion records from the Santa Cruz Mountains (Lorna Prieta), California, earthquake of 17 October 1989, Report No. OSMS 89-06: California Strong Motion Instrumentation Pro- gram, California Department of Conservation, Division of Mines and Geology. U.S. Geological Survey, 1989a, Lorna Prieta, California earthquake: an anticipated event: Science, v. 247, p. 286-293. 1989b, Strong-motion records from the Northern California (Lorna Prieta) earthquake of October 17, 1989: U.S. Geological Survey Open-File Report 89-568. Selected Series of US. Geological Survey Publications ,1 Books and Other Publications Professional Papers report scientific data and interpretations of lasting scientific interest that cover all facets of USGS inves- tigations and research. Bulletins contain significant data and interpretations that are of lasting scientific interest but are generally more limited in scope or geographic coverage than Professional Papers. Water-Supply Papers are comprehensive reports that present significant interpretive results of hydrologic investigations of wide interest to professional geologists, hydrologists, and engi- neers. The series covers investigations in all phases of hydrol— ogy, including hydrogeology, availability of water, quality of water, and use of water. Circulars are reports of programmatic or scientific information of an ephemera] nature; many present important scientific information of wide popular interest. Circulars are distributed at no cost to the public. Fact Sheets communicate a wide variety of timely information on USGS programs, projects, and research. They commonly address issues of public interest. Fact Sheets generally are two or four pages long and are distributed at no cost to the public. Reports in the Digital Data Series (DDS) distribute large amounts of data through digital media, including compact disc- read-only memory (CD-ROM). They are high-quality, interpre- tive publications designed as self-contained packages for view- ing and interpreting data and typically contain data sets, software to view the data, and explanatory text. Water-Resources Investigations Reports are papers of an interpretive nature made available to the public outside the for- mal USGS publications series. Copies are produced on request (unlike formal USGS publications) and are also available for public inspection at depositories indicated in USGS catalogs. Open-File Reports can consist of basic data, preliminary reports, and a wide range of scientific documents on USGS investigations. Open-File Reports are designed for fast release and are available for public consultation at depositories. Maps Geologic Quadrangle Maps (GQ’s) are multicolor geologic maps on topographic bases in 7.5- or 15—minute quadrangle formats (scales mainly 124,000 or 1:62,500) showing bedrock, surficial, or engineering geology. Maps generally include brief texts; some maps include structure and columnar sections only. Geophysical Investigations Maps (GP’s) are on topographic or planimetric bases at various scales. They show results of geophysical investigations using gravity, magnetic, seismic, or radioactivity surveys, which provide data on subsurface struc- tures that are of economic or geologic significance. Miscellaneous Investigatio s Series Maps or Geologic . Investigations Series (I’s) re on planimetric or topographic bases at various scales; they! present a wide variety of format and subject matter. The series also incudes 75-minute quadran- gle photogeologic maps on planimetric bases and planetary maps. Information Periodicals Metal Industry Ind cator (MII’s) is a free monthly newslet— ter that analyzes and foreca s the economic health of five metal industries wit composite leading and coincident indexes: primary me als, steel, copper, primary and secondary aluminum, and aluminum mill products. Mineral Industry Surveys (MIS’s) are free periodic statistical and economic reports designed to provide timely statistical data on production, distribution, stocks, and consumption of signifi- cant mineral commodities. The surveys are issued monthly, quarterly, annually, or at other regular intervals, depending on the need for current data. The MIS’s are published by commod- ity as well as by State. A series of international MIS’s is also available. Published on an argual basis, Mineral Commodity Summa- ries is the earliest ovemment publication to furnish estimates covering nonfuel mineral industry data. Data sheets contain information on the domestic industry structure, Government programs, tariffs, and 5—year salient statistics for more than 90 individual minerals and materials. The Minerals Yearbook discusses the performance of the worldwide minerals and materials industry during a calendar year, and it provides background information to assist in inter— preting that performance. The Minerals Yearbook consists of three volumes. Volume I, Metals and Minerals, contains chap- ters about virtually all metallic and industrial mineral commod- ities important to the US. economy. Volume 11, Area Reports: Domestic, contains a chapter on the minerals industry of each of the 50 States and Puerto Rico and the Administered Islands. Volume 111, Area Reports: International, is published as four separate reports. These reports collectively contain the latest available mineral data on more than 190 foreign countries and discuss the importance of minerals to the economies of these nations and the United States. Permanent Catalogs “Publications of the US. Geological Survey, 1879-1961” and “Publications of the US. Geological Survey, 1962— 1970” are available in paperback book form and as a set of microfiche. “Publications of the US. Geological Survey, 1971—1981” is available in paperback book form (two volumes, publications listing and index) and as a set of microfiche. Annual supplements for 1982, 1983, 1984, 1985, 1986, and subsequent years are available in paperback book form. 'pa 'JBZIOH _| Fr: 5 g > , , g RETURN EARTH sCIETqEEs LIBRARY _ E... 10> 642-2997 55 LOAN PERIOD 1 2 3 , .E 7 DAYS - 4 5 6 ALL BOOKS MAY BE RECALLED AFTER 7 DAYS Books needed for class reserve are subiecf to immediate recall DUE AS STAMPED BELOW asst ‘u 8380130 :IO axvnomuva 'vmu’d UNIVERSITY OF CALIFORNIA, BERKELEY FORM NO. DD8 BERKELEY, CA 94720 a~zsst Ja‘ded’l9"0!553¥°1d hams leazfioloas 'S'n ISBN 0-507-31373-7 ,_ ® Printed on recycled paper ‘ L ' , l| 8060||7913798