Fig. 7. (A) Geologic map of foundation materials in the Marina District, San Francisco. (B) Seismograms of vertical ground velocity for a M 4.6 aftershock. The top trace (MAS) is from a site underlain by a competent sandstone member of the Franciscan assemblage. The second trace (PUC) is from a site underlain by clean dune sands just onshore of the pre-fill shoreline. At this site, a two-story brick building with a massive turret on its northwest corner, constructed in 1893, rode through the recent earthquake and the great 1906 earthquake without a crack. The third trace (LMS) is

did not reach the surface, the event did leave its signature in the form of the surface fractures discussed above and liquefaction, lateral spread-failures, and warped ground surfaces in sag ponds along the fault trace (23). Thus, the occurrence but not the amount of fault offset of pre-historic earthquakes may be all that is preserved in the geologic record.

Strong Ground Motion

The Loma Prieta earthquake provides a direct test of seismic zonation methods based on detailed classification of geologic units according to their performance in weak ground motion (24). The influence of the local geologic deposits on the amplitudes of ground shaking and extent of damage has been known since the time of the 1906 earthquake. In discussing the 1906 damage, Wood (25) concluded: "... the amount of damage produced by the earthquake

depends chiefly on the geological character of the ground. Where the surface was of solid rock, the shock produced little damage; whereas upon made land great violence was manifested

It has also been known for two decades (26) that weak ground motion on firm alluvium can be amplified by factors of 2 to 4 in the frequency band of a few tenths to several hertz, the band that has the greatest effect on man-made structures. Similarly, amplification factors for the Bay mud and artificial fill can be 5 to 10 or more.

More than 170 records of ground shaking obtained within 200 km of the epicenter (27) form the basis for direct comparisons with expectations. For all practical purposes the records are directly proportional to acceleration; and thus peak accelerations can be obtained and analyzed easily.

Observed accelerations and the predicted values (28) are compared in Fig. 5 for recording sites on various underlying geologic materials. Accelerations from rock sites were in reasonable agreement with the predictions. Accelerations at soil sites were systematically greater than the predictions, and the accelerations at Bay mud sites are much larger than those from most of the alluvium sites. Relative to rock sites, ground motion on young, poorly consolidated, water-saturated alluvium and mud tended to be deficient in high-frequency motion and enriched in longer-period motion; in

0

LMS

from a site one and one half blocks away in the area containing artificial fill. Here, houses were badly deformed by foundation failure, and the north side of the street (North Point) dropped 0.5 m below the south side All three components of ground motion at PUC and LMS are amplified by comparable amounts relative to MAS, but the local damage parterns are significantly different. These data suggest that the problems in the Marina District were fundamentally a result of permanent deformation of the man-made fill rather than local amplification differences.

detail, frequency-dependent amplification is a function of rigidity contrasts and basin geometry. These effects are illustrated by the mainshock accelerograms from five sites in the San Francisco region (Fig. 6).

There is ample evidence, then, in both the historical and instrumental records that the observed patterns of damage and strong ground motion for the Loma Prieta earthquake were both predictable and predicted (24). Whereas the Marina District disaster was fundamentally the result of pervasive failure of the artificial fill, the district also experienced significant local amplification relative to rock sites, as shown by recordings of aftershocks (Fig. 7). Whereas the portion of the Cypress Street section of 1-880 that collapsed long had been recognized as a structure in need of retrofitting for seismic safety, this section was also founded on Bay mud and thus probably sustained higher levels of ground motion than undamaged parts on firmer ground. Both of these areas, as well as many others that sustained significant but less severe damage, had been identified on maps as areas of high potential for damage (Fig. 8).

Landslides

The Loma Prieta earthquake generated landslides throughout a region of approximately 14,000 km2 (Fig. 1). The epicentral region of the earthquake in the steep, rugged, and heavily vegetated Santa Cruz Mountains has historically produced abundant landslides, both during earthquakes and during the region's rainy winters. Even though only about 7 cm of rain had fallen in the preceding 6 months (29), the earthquake generated thousands of landslides throughout the mountains.

The zone of largest and most complex landslides is within and adjacent to the zone of surface fractures (Fig. 1). In this zone, where fissures due to landslide movement are intermixed with other fractures, large, deep-seated blocks of ground moved downslope along a part of the ridge crest. The largest individual landslide block yet identified is more than 0.75 km2 and damaged dozens of residences riding on it.

Away from the zone of surface fractures, the most numerous earthquake-generated landslides in the Santa Cruz Mountains are rock falls, rock slides, and debris slides. Also abundant are deeper

29-577 - 90 - 27

Some of the most devastating damage during the earthquake was caused by liquefaction of loose, saturated sand. Liquefaction (30) occurred in man-made fill around the margins of the San Francisco Bay and in floodplain deposits in the Salinas-Santa Cruz area Within the latter area, most of the catalogued 1906 failures (31) at epicentral distances of up to 40 km were reactivated.

In the San Francisco Bay Area, liquefaction-induced ground failure occurred mainly along the northern shores of the bay in artificial fill. It was most extensive in the hydraulically emplaced sand fills beneath the Oakland International Airport; the Alameda Naval Air Station; the eastern approach and toll plaza of the San Francisco Bay Bridge; the Marina, South of Market, and Mission districts of San Francisco. At these sites, the fill principally consists of hydraulically emplaced sand underlain by water-saturated mud or sand and overlain by broad pavements such as airport runways, streets, and slab foundations. Both the Mission and South of Market Street districts experienced severe liquefaction-induced ground failure and resultant structural damage during the 1906 earthquake.

Significantly, no obvious ground failure occurred in numerous other extensive land fills along the central and southern shores of the bay. Most of these consist of hydraulically emplaced silt and clay, dumped earth and rock, or solid waste covered by compacted earth. There is also no evidence of ground failure along the collapsed

section of 1-880 in Oakland.

From Santa Cruz to Salinas, liquefaction-related ground failure was widespread but restricted to areas underlain by water-saturated, late Holocene alluvial deposits of principally the San Lorenzo, Pajaro, and Salinas rivers, including spits and bars and tidal channels in the Moss Landing area. Lateral spreads and differential settlements disrupted or destroyed flood-control levees; pipelines; approaches, abutments, and piers of bridges; roads, homes, and utilities; and irrigation works, including gradients of irrigated fields.

Implications for Future Earthquakes

In the past, the occurrence of a major earthquake has confronted earthquake scientists and engineers with major surprises. The Loma Prieta earthquake was exceptional, for the likelihood of its coming had been evaluated well in advance of the ear.hquake, and the damage and destruction it wrought occurred in those areas and to those structures known to be at greatest risk. For the San Francisco Bay Area, the lessons of this earthquake differ only in the degree to which we will ultimately measure them from those of every other major historic earthquake in this area.

The strong ground motion generated by the Loma Prieta earthquake was neither particularly long, scarcely a third of the duration of the 1906 earthquake, nor unusually violent. Indeed, areas just north of Oakland and San Francisco that suffered liquefaction. induced ground failure in 1906 did not this time; this distribution indicates that the liquefaction threshold was only slightly exceeded even in heavily damaged parts of San Francisco during the Loma Prieta earthquake (32). When a similar or stronger earthquake strikes closer to the center of the Bay Area, as will happen, the hazard can only be greater. The real question is will its effects be more damaging?

Earthquake intensities previously predicted for the San Francisco Bay region (Fig. 8) for large earthquakes on either the San Andreas or Hayward faults reveal that the potential for damage can vary on a block by block basis depending on the geological character of the ground. The Loma Prieta earthquake served to identify only the most vulnerable structures located at sites underlain by poor soil conditions.

Now that a portion of the 1906 earthquake fault break has reruptured, there can be little doubt that the hazard is real along other active faults in the San Francisco Bay Area. Both the adjacent segment of the San Andreas fault on the San Francisco Peninsula and the Hayward fault, in repose since 1868, are recognized as possible sites for another M 7 earthquake in the coming decades. The aggregate probability of at least one M 7 event from these two faults alone had been judged as 0.5 over the next 30 years (5), and the occurrence of the Loma Prieta earthquake has not lowered that probability (33).

We have known for some time that meaningful reduction of earthquake hazards cannot be achieved through science alone (34). It requires a well-informed and well-prepared public to insist upon mitigation of the hazards before the next earthquake strikes

REFERENCES AND NOTES

1. Loma Prieta, "the dark roiling mountain," highest point in the Santa Cruz mountains at 1157 m, has played a central role in geodetic measurements of crustal deformation in the San Andreas fault system since 1851, including the measure ment of fault shp in the 1906 earthquake and the development of the clastx rebound theory of earthquakes (H. F. Reid, Camegir Inst Washing Publ 87, vol 2 (1910))

2. California Office of Emergency Services estimates as of 9 November 1989

3. A. G Lindh, US Geel. Surv Open-File Rep 83-63 (1983).

4. L. R. Sykes and S. P. Nishenko, J Geophys Res 89, 5905 (1984).

5. Working Group on California Earthquake Probabilities, US Geol. Surv Op-Fil Rep. 88-398 (1988)

6. Geologic effects of the 1906 earthquake were generally well documented (AC Lawson et al, Camegie Inst of Washington Publ 87, vol. 1 and atlas (1908)). The investigation in the southern Santa Cruz Mountains, however, is an exception The investigators apparently veered away from the main trace of the San Andreas fault and may have missed a long section of the main trace between Summit Road ndge and Pajaro Gap. The locations of most of the surface fractures described in the report are poorly documented The report mentions only four sites of hortomal offset in the source region of the 1989 earthquake: Wrights Laurel tunnel, Morl ranch, Pajaro River railroad bridge, and a fence halfway between Chinenden and San Juan Bautista The report from Wrights-Laurel tunnel clearly indicates that there was about 1.5 m of right-lateral offset in the tunnel but not on the surface above. At Morell ranch, left-lateral displacements are described across two frac tures. The report describes damage to the Pajaro River bridge but does not clearly document actual fault displacement. The report describes 1.2 m of horizontal offser

of the fence located between Chittenden and San Juan Bautista but does not give the sense of offset nor characterize its relation to the fault trace. Except for the Wrights-Laurel tunnel, we have little evidence of night-lateral surface faulting in the southern Santa Cruz Mountains during the 1906 earthquake, either because there was none or the investigator did not look in the right places. However, the report does mention extensive fracturing in the Summit Road and Skyland Ridge areas, as we observed in the same areas after the Loma Prieta earthquake. At least one, and probably many, of the 1906 fractures were reactivated in the 1989 earthquake. 7. K. Shimazaki and T. Nakata, Geophys Res Lea 7, 279 (1980). 8. W. Thatcher and M. Lisowski, J. Geophys. Res. 92, 4771 (1987). 9. C. H. Scholz, Geophys. Res. Let 12, 717 (1985).

10. The mainshock epicenter is at 372.19′N and 121°52.98 W The M, -7.1 is based an the average amplitude of 0.05-Ha surface waves at 18 selected stations between 22° and 142" from the mainshock. The hypocentral depth of 17.6 km is deeper than essentially all previously documented events on the San Andreas fault proper and is nusually deep for other faults of the San Andreas system as well.

11. The P-wave first-motion solution for the mainshock defines a fault plane striking N50 ± 8W, dipping 70° 10° to the southwest, with a rake of 130° 15° The southwestern side of the fault, or hanging-wall block, moved to the northwest (right-lateral slip) and upward (reverse ship), relative to the northeast, or footwall block. This solution was determined from 267 P-wave first motions recorded on the U.S. Geological Survey Central California Seismic Network Centroid-momerit-tensor (CMT) inversion of 3.5 to 7-mHz surface waves and 12- to 20-mHz body waves recorded at Pasadena, California; Cambridge, Massachusetts; and Tsukuba, Japan, by H. Kawakatsu of the Geological Survey of Japan (personal communication) gives an almost identical result for the best double couple with a strike of NS4°W, a dip of 72° to the southwest, and a rake of 132 The CMT inversion gave a seismic moment of 2.2 x 10" N-m. Geodetic data collected the day after the earthquake with a Goodelite and meteorological measurements made by aircraft at the time of ranging and with two global positioning systems give a seismic moment of 3.8 x 10" N-m.

12. The slip deficit accumulated in the 83.5 years since the 1906 earthquake equals the time interval times the mean slip rate. Estimates of the slip rate range from a low vahar of 1.2 ± 0.4 cm per year (W. H. Prescott and J. C. Savage, J Geophys Res 86, 10853 (1981)] to 2 cm per year (3) and imply that the slip deficit is between 1.0 m and 1.7 m, in comparison to the geodetically determined right-lateral slip of -1.9 ± 0.2 m.

13. Movement along the San Andreas fault in a restraining fault bend requires the removal of crust if the horizontal motion direction is controlled by the fault orientation outside the bend. If one side of the fault is fixed and the other maintains a horizontal velocity is, then the kinematic solution for a block sliding along a fault plane striking with respect to its constrained motion direction and dipping beneath it gives a horizontal (strike-ship) velocity of cost and a vertical (reverseslip) velocity of sin cos in the plane of the fault. The average difference in fault strike between the Loma Prieta fault plane and adjacent segments is 10° to 15°. For a dip of 70°, this difference implies that the ratio of strike slip to reverse slip on the faul plane is 2:1 to 1.3:1.

14. The high near-term seismic potential of the seismic gap considered by Lindh (3) was first delineated by A. G. Lindh, B. L. Moths, W. L. Ellsworth and Olson (US Geol. Surv. Open-File Rep 82-180 (1981), p. 45] on the basis of the seismically quiet zone along the southeasternmost 45 km of the 1906 rupture, signment of the 1865 earthquake to this fault segment, and (inferred) restoration of the strain released in the 1906 earthquake. Although the 1989 event was larger than anticipated, its intensity pattern closely corresponds the 1865 event. As the magnitude of the 1865 earthquake is based solely on the interpretation of its intensity partern, the two events are probably more similar in size than their 0.6 difference in magnitude would suggest.

15. W. H. Bakaan and A. G. Landh, Science 229, 619 (1985). Analyses of line-length changes on geodetic networks near Parkfield indicate that the 1966 rupture surface has not slipped significantly since 1966 and that the strain released in 1966 will most likely be restored by 1989. These considerations lend independent support for the prediction of the next Parkfield earthquake (P. Segall and R. Harris, Science 233, 1409 (1986)].

16. The State of California issued public advisories following the 27 June 1988 (M 5.0) and 8 August 1989 (M 5.2) Lake Elsman earthquakes. These advisories noted that there was a small but significant chance of a larger earthquake during the next 5 days. All earthquakes have a finite probability of being foreshocks to larger events [P. A. Ressenberg and L. M. Jones, Science 243, 1173 (1989)]. The decision to sue these advisories was driven by their location at the intersection of the Sargent and San Andreas faults, at the northern end of Lindh's (3) postulated earthquake. 17. These results constrict the limit on the seismic moment of possible preseismic deformations by a factor of 10 compared to earlier limits [M. S. J. Johnston, A. T. Linde, M. T. Gladwin, R. D. Borcherds, Tectonophyna 144, 189 (1987)). 18. W. H. Prescott, J. L. Davis, J. L. Svarc, Science 244, 1337 (1989). 19. The unmaal surface breaks occur in an 8-km-long, 1.5-km-wide zone on a ridge top traversed by Summit Road between Highway 17 and Old San Jose Road and on part of the next ridge to the south (Fig. 4). Their continuity along trend or relatively larger displacements or greater length distinguish these from the abun dant small cracks throughout the zone of strong ground shaking that are associated

either with obvious local ground failures or failures in asphalt and concrete pavement. The Summit Road fractures trend northwesterly, and the longest is about 700 m They vary from single fractures to roughly en echelon, anastomosing discontinuous cracks Displacement across these fractures is main extensional, generally with a component of left slip of as much as 75 cm, and locally with a component of dip slip as large as 60 cm Along some fracture sets, the downslope side is consistenth up. Net displacement vectors for the fractures show consistent trends approximately normal to the crest of the ridge (Fig. 4) Repeated measurements across these fractures have detected no post seismic creep

20. A. M. Sarma Wojcicka, E. H. Pampeyan, N. T. Hall, U.S. Geol Surv. Misc Field Stud Map MF-650 (1975).

21. Although fractures of this geometry and sense of displacement can be accounted for with special circumstances for example, if surface materials were stress-free until the deep rupture imposed tractional stresses, or if the shear stress beneath the fracture zone was oriented anticlockwise in plan from the local strike of the fault trace contrary arguments and the lack of corroborative evidence renders such explanations unattractive. Dislocation models suggest that the observed surface displacements are possible south of the San Andreas fault trace, but other explanations are also possible Failure at depth on weak bedding planes, modified by near-surface topography, may also explain the displacements.

22. Similar phenomena have frequently been documented elsewhere in California (C R. Allen, M. Wyss, J. N Brune, A Grantz, R. E. Wallace, US Geol Surv Prof. Pap 787 (1972), p. 87]. In the city of Los Gatos, about 10 km north of the rupture, concrete sidewalks and curbs on northeast-trending streets deformed throughout much of the downtown area. Broken concrete slabs lining Los Gatos Creek indicate about 25 cm of northeast-southwest tectonic shortening. About 3 km northeast of downtown, a 4.5-km-long, cast-southeast-trending zone of broken concrete sidewalks and curbs defines a coseismic thrust or reverse fault. Possibh analogous zones of cracks and concentrated damage also occurred in Los Altos Hills 23. Studies of recurrent liquefaction in the geologic record provide data for dating prehistoric earthquakes and are of particular importance in the central and eastern United States (P Talwani and J. Cox, Scence 229, 379 (1985); S. F. Obermeier et al, ibd 227, 408 (1985)].

24. R. D. Borcherdt, Ed., U.S. Geol Sure. Prof Pap. 941-A (1975). 25. H. O. Wood, in (2), p. 241.

26. R. D. Borcherds, Bull. Srimo! Soc Am 60, 29 (1970).

27. The bulk of the data are from the Strong Motion Instrumentation Program of the California Division of Mines and Geology |Calif Dept. Conserv Rep ÖSMS-89-06 (1989)) and the US. Geological Survey (US Geal Surv Open- File Rep 89.568 (1989)]. Other data are from the University of California Santa Cruz, the California Department of Water Resources, and the US. Bureau of Reclamation. 28. W. B. Joyner and D. M. Boore, Proceeding of the Conference on Recent Advances in Ground Motion Evaluation Earthquaker Engineering and Soil Dynamics 11, Park City, Utah, 27 to 30 June 1988 (Geotech Spec. Publ 20, American Society of Civil Engineer, New York, 1988), p. 43. The peak acceleration, 4, is given by log 0.49 0.23(M6)-log-0.0027r, where r is the slant distance to the nearest point on the rupture at 8 km depth.

29. The area, maximum epicentral distance of effects, and geologic environments of landsliding in this earthquake are generally consistent with carlier predictions based on worldwide data and theoretical considerations (D. K. Keefer et al., Science 238, 921 (1987)).

30. Liquefaction is the process in which sediment composed chiefly of loose, non cohesive, water-saturated sand and silt is temporarily transformed to a slurry of water and sediment by earthquake shaking. It then has scant resistance to flow or to shear forces. So long as excess fluid pore pressure persists, water and sediment may escape to the ground surface and be expressed as sand boils, or the ground may spread laterally downslope, and carry unliquefied overburden, including man-made structures Structures sited above zones of liquefaction also may suffer differential lateral displacement and subsidence if bearing capacity is reduced or if the ground settles beneath the foundation.

31. T. L. Youd and S. N. Hoose, US Geol. Sure Prof Pap 993 (1978) 32. Repeat liquefaction at sites known to have failed in the 1906 carthquake demonstrates that the ground localities such as the Marina District or Treasure Island, which failed by liquefaction in the Loma Prieta earthquake but not the 1906 earthquake, could fail again in future earthquakes. Locations that have experienced liquefaction in earlier earthquakes, but not this time, cannot be assumed to be immune in future earthquakes.

23. The National Earthquake Prediction Evaluation Council has formed a working group charged with the evaluation of the 1988 report (5) in light of new knowledge derived from the Loma Prieta earthquake. Although this work has just begun, it is clear that the stress transferred to the adjacent segments of the San Andreas brings them closer to failure.

34. J. I. Ziony, Ed., US Geol Surv. Prof Pap 1360 (1985);W. W. Hays, Ed., U.S. Geol Surv Open File Rep. 88-11-B (1988).

35. R. D. Borcherdt et al., US Geol Surv Misc Field Stad Map MF-709 (1977). 36. We thank the many individuals and institutions for their assistance in collecting data during the hectic days following the earthquake used in this report In particular, we thank J. B. Berrill, P Cowie, K. Hudnut, M. Jackson, H. Kawakatsu, S. Larsen, Z. Reches, F. Webb, and P. Wood.