Deformation from the 1989 Loma Prieta earthquake near the southwest margin of the Santa Clara Valley, California

In the southwestern part of the Santa Clara Valley (or Silicon Valley, California, USA), a densely populated area between the Santa Cruz Mountains and San Francisco Bay (Fig. 1), the 17 October 1989 Loma Prieta earthquake did not result in major evidence of throughgoing ground rupture, but did produce numerous local breaks in pavement and pipes in addition to widespread damage to structures. The damage to pavement consisted primarily of buckled or thrust concrete curbs (Fig. 2) and sidewalk slabs (Figs. 3 and 4) suggestive of shortening; adjacent asphalt street surfaces typically lacked damage. Such damage principally affected pavement along north-south– and north-northeast–trending streets. Especially near Los Gatos, where such damage was most abundant, the concrete displayed a remarkably regular pattern of deformation consisting of buckling and thrusting in a north-south to northeast-southwest direction (Fig. 3). Damage in Los Gatos during the 1906 earthquake displayed a similar pattern: Lawson et al. (1908, p. 274) reported, “There were about a dozen upheavals of sidewalks, mostly on north and south streets.” Similarly, Scholz (1985), citing Lawson et al. (1908), described a widely distributed “shattered zone” of ground cracks resulting from the 1906 earthquake with various orientations south of Stanford University in the vicinity of currently recognized thrust faults. Eyewitness accounts during the 1989 Loma Prieta earthquake described spectacular results of this deformation, including the violent ejection of a manhole cover and a curb popping up into the air then through the rear window of a parked car. Chimneys generally toppled in north-south directions, and objects on shelves fell most commonly in north-south directions. After the earthquake, evidence of permanent contractional strain included utility-box lids that no longer fit into their casings and distorted storm-drain gratings (Fig. 5).

The earthquake had moment magnitude M_w = 6.9 and surface-wave magnitude M_s = 7.1. Its oblique right-reverse rupture surface extended in depth from 3 to 18 km beneath the Santa Cruz Mountains on a steep (60°–80°) southwest-dipping surface, striking 120°–140°, that sustained ∼1.5 m of right-lateral slip and 1.5 m of reverse slip, as inferred from geodetic data (Lisowski et al., 1990; Marshall et al., 1991; Williams et al., 1993). Modeled in more detail, the slip is heterogeneous in its distribution with a maximum magnitude of 3.6–8 m (Árnadóttir and Segall, 1994; Lisowski et al., 1996; Marshall and Stein, 1996). At the Lexington Reservoir dam, located ∼2.5 km south-southwest of Los Gatos, peak ground acceleration (PGA) was 0.61 g oriented 042° and 0.15 g vertically upward (R.L. Volpe & Associates, 1990; Brady and Shakal, 1994). Analyses of data from the PEER (Pacific Earthquake Engineering Research Center) database (http://peer.berkeley.edu/peer_ground_motion_database) indicate that the highest recorded PGA and peak ground velocity (PGV) proximal to the study area occurred at station LGPC (NGA 779) in the Bear Creek Redwoods Open Space Preserve ∼6 km southwest of Los Gatos. With a fault strike of 130°, the PGA and PGV were 1.08 g oriented 010° and 120.8 cm/s oriented 003°, respectively. Fault-normal PGA and PGV values were generally greater than fault-parallel values. As Somerville et al. (1997) noted, the radiation pattern of shear generated from the earthquake was oriented perpendicular to the rupture strike, resulting in fault-normal forward directivity with higher long-period content. Such shear may have been responsible for damage patterns oriented in a northeast-southwest direction or coseismic triggered slip along nearby thrust faults. Throughout the San Francisco Bay region, claims for earthquake damage to the urban infrastructure and other public property eligible for disaster assistance totaled nearly $600 million (French, 1995).

Although much of the region affected by the earthquake was suburban, the hypocenter occurred under the Santa Cruz Mountains, a sparsely populated region. Here, strong ground motion–induced sackungen-like fractures that formed along preexisting structural discontinuities (Ponti and Wells, 1990) were masked in part by abundant mass wasting; ∼1280 earthquake-induced landslides were mapped over an area of ∼2000 km² (Spittler and Harp, 1990; Keefer, 2000). On the floor of the Santa Clara Valley, however, mass-wasting features were rare, and damage occurred dominantly in buildings and in brittle, human-made structures in contact with the ground, especially pavement and underground utility pipes (Schmidt et al., 1995). Langenheim et al. (1997) documented how distribution of this surficial damage coincides with deeper crustal structure inferred from potential field gravity and aeromagnetic anomalies. It is in this southwestern portion of the Santa Clara Valley that a fold and thrust belt, designated the Foothills thrust belt (FTB) by Bürgmann et al. (1994), surfaces near the front of the Santa Cruz Mountains (Aydin and Page, 1984; Brabb et al., 1998; Wentworth et al., 1999; McLaughlin and Clark, 2004). This belt includes several anastomosing, curvilinear oblique dextral reverse faults, verging northeastward from the San Andreas fault (SAF). Although the future seismic hazard to the urbanized Santa Clara Valley posed by these range-front thrusts in the continuing evolution of the SAF system are poorly understood, interpretation of postseismic global positioning system (GPS) data indicates continued deformation of the FTB in the first years following the Loma Prieta earthquake (Bürgmann et al., 1994; Segall et al., 2000).

The principal aim of this study was to understand the origin of damage in the area so as to better understand hazards in future earthquakes. Although damage to buildings was common, we did not document it because such damage is highly sensitive to design, construction methods, quality of materials, and localized ground shaking. The abundant damage to pavement and pressurized pipelines, which are in direct contact with the ground surface, in our view provided the best mappable indicator of coseismic ground deformation in the absence of mappable ground rupture. The damage to pavement and pipes could conceivably have resulted from shaking, from transient strain, or from permanent strain during the earthquake. All three effects could well have acted in concert or at different places.

To document the damage most indicative of ground deformation, and to provide data for interpretation of causal mechanisms, we first compiled sites of damage to pavement and pipes in the region. In this effort we mapped and compiled the distribution of 1427 sites of coseismic damage to pavement and pipes over a 663 km² area (Schmidt et al., 1995). We then mapped and measured damage to the channel lining of Los Gatos Creek. This 2-km-long concrete lining provided a unique and sensitive concrete strain gauge directly through the zone of greatest damage in the study area. We compiled and conducted surveys to measure ground distortion along lines extending from the Santa Cruz Mountains through the town of Los Gatos, roughly perpendicular to mapped geologic structures. In this paper we briefly describe these efforts and attempt to tie the several lines of evidence into a coherent picture of the origin of damage in the area.

Detectable ground rupture in natural earth materials was rare in the study area depicted in Figures 1 and 6, and so damage to human-made structures and utilities, in particular pavement and pressurized pipelines, provided the major mappable indicator of coseismic ground deformation. Individual breaks typically were not prominent enough to provide measurable displacements, and local alignment of breaks was detectable in only a few places. Consequently, the spatial distribution of this damage provided the principal evidence of its origin (see Phelps et al., 2015). We assembled information on damage by mapping its occurrence in the field and by compiling reports of damage from utility companies and public agencies. The distribution of damage was subdivided into categories by severity of damage, type of materials, sense of deformation, and data source (see Supplemental Files 1¹ and 2² to access digital files; Schmidt et al., 1995). For example, we noted whether pavement breaks occurred in asphalt or concrete and whether the breaks displayed contractional, extensional, or an unspecified sense of deformation. We also distinguished breaks to pressurized pipes, such as water and natural gas lines.

We did not compile all types of earthquake damage in the area, but we were thorough in including information from all likely sources on damage to pavement and pressurized pipelines. The types of damage compiled were limited primarily to those situated at or below the ground surface (depth < 1.4 m). The exception is above-ground breaks of natural gas lines, principally risers, which account for ∼5% of the data set. Subsurface storm drains and sewer pipes were excluded from the search because leaks are not easily detected. Damage to buildings was not documented because it is especially sensitive to design, construction methods, quality of materials, and localized ground shaking in response to peak acceleration or velocity.

The majority of the evidence (almost two-thirds of the data) consisted of freshly broken, buckled, or spalled concrete sidewalks, curbs, gutters, and streets. This type of damage reflected shortening along the length of the concrete. Owing to its rigidity, concrete, where it occurs in long strips such as streets, curbs, and sidewalks, can transmit small amounts of contractive strain across broad zones. Thus, during the earthquake, strain could accumulate over substantial lengths of concrete, to be revealed at a single prominent buckle or break. Similar magnitudes of extensional strain generally were not revealed because concrete pavement typically includes expansion joints or preexisting cracks, which can accommodate small amounts of extension without detectable damage. Likewise, lateral shear across concrete strips was not clearly displayed, perhaps because minor lateral strain could be accommodated along preexisting cracks and expansion joints.

Damage to pavement and water lines was located in the field by driving slowly along streets, then examined on foot from public right of ways. We made a concerted effort to search systematically, without regard to preconceived trends of damage, so as to obtain an objective and complete portrayal of distribution. This field mapping was conducted largely from 21 October 1989, within several days of the earthquake, to December 1989. The principal uncertainties in field observations were distinguishing coseismic damage from other damage and determining the sense of deformation. Damage to concrete was considered coseismic where cracks and spalls showed light colored, unweathered surfaces and the attendant presence of small, readily transportable concrete fragments. We inferred shortening where pavement was damaged in a manner consistent with shortening, typically by thrusting or buckling (e.g., Figs. 2–5). Field checking of address locations from damage listings, and monitoring of select sites of post-earthquake displacements, continued into the autumn of 1990.

To augment field observations, we contacted all recognized governmental and public-service agencies responsible for repairing or cataloging damage within the study area (Fig. 6). More than 30 agencies contributed maps or address listings of damage (see Supplemental Files 1 [see ^{footnote 1}], 2 [see ^{footnote 2}], and 3³; Schmidt et al., 1995, Table 1 therein). Except for a few points identified as pre-earthquake or post-earthquake deformation, all damage compiled from these agencies had been reported to them on the date of the earthquake or shortly thereafter, and so is assumed to have occurred during the main shock. To avoid introducing any bias to the distribution of damage, we did not edit the reports of damage.

The boundary of the study area (Figs. 1 and 6) was chosen to include the bulk of the kind of damage described above and to exclude, as much as possible, ambiguous evidence such as liquefaction-induced damage near San Francisco Bay. Similarly, to exclude any damage resulting from gravitational mass movement, our mapping and compilation focused mainly on the alluvial lowlands, and so our field observations are incomplete in the hills. Figure 6B depicts the relation between the observed damage, mapped geology (Graymer et al., 2006), and the depth of alluvial sediments overlying basement rock (Jachens et al., 2006).

The information on coseismic damage has a number of uncertainties, as the strain markers are also subject to nonseismic damage. Buckling of concrete sidewalks, for example, was probably of mixed origin where upheaval by tree roots appeared to have previously cracked and deflected the concrete. In addition, earthquake-generated strain may have been irregularly reflected in the damage because the strength of the strain markers, and hence their sensitivity to ground deformation, is influenced by variation in quality of the original construction, history of maintenance, and age. Damage in some upland areas may reflect unrecognized slope failures. The times of damage generation are not known in detail, and so some of the damage may actually have occurred in response to foreshocks or aftershocks. The principal limitation is the spatial distribution of concrete and pipes that is largely limited to the semiregular network of straight streets in flatland areas (see Phelps et al., 2015). Therefore, a lack of mapped damage may simply reflect a lack of concrete and pipes rather than absence of ground deformation.

Los Gatos Creek is lined to some extent by concrete for ∼2 km, but only the northernmost portion of the lined channel, in which the displacements occurred, has been diverted to an artificial channel continuously lined by concrete slabs on the sides and bottom (Figs. 7 and 8). Slabs are separated by seams that serve as expansion joints (Fig. 9). Individual side slabs, which are inclined at an angle of 32°, are ∼18 m long (along the channel) by 9 m wide (from the base, at the foot slab in Fig. 9, to the top of bank), and 8 cm or more thick. Undersides of slabs are irregular and rough, indicating that they were poured in place. No steel reinforcement, in the form of wire mesh or reinforcing bar, was observed, even at fresh breaks, but the regular spacing of aging cracks suggests the possibility of reinforcement. Some of the seams between slabs still contain the original fibrous packing compound. Apparently the lining was constructed using a variety of seam widths; in areas of channel lining that appear undisturbed, seams range in width from <1 cm to 2.5 cm and have an average width of ∼1.3 cm.

Within the side slabs are systematic cracks, which we call aging cracks, that have a fairly uniform pattern from slab to slab (Fig. 9). These cracks extend both horizontally and vertically through the slabs; vertical cracks are spaced ∼2.5–3 m from each other or a seam (Fig. 9). These cracks typically appear as irregular hairlines and seldom exceed 2 mm in width. At some locations along the channel, arrays of inclined parallel concrete bars, called energy dissipaters, are attached over the side slabs. These bars cross seams as well as aging cracks (Fig. 8).

During our mapping, the slabs that formed the bottom of the channel were covered by water as much as 0.5 m deep and were also variably covered by sediment, moss, and other aquatic plants, including thick stands of cattails. Thus, the side slabs offered the best exposures, and they were continuously exposed along the mapped part of the channel except for the areas covered by vegetation.

The entire length of channel lining (Fig. 7) was examined for fresh breaks and displacements. Examination was conducted on foot from the channel bottom and from accessible places along the top. Areas of particular interest on the channel sides were examined directly with rope support. The possibility of displacements across the trend of the channel was investigated visually by sighting along the edges of the concrete lining and along persistent slab seams. Locations of displacements were mapped at 1:1200 scale using tape, rangefinder, and a detailed base map provided by the California State Department of Transportation (Caltrans). Generally, one or two displacements within a group were located with the rangefinder by averaging several distance measurements to a feature clearly identifiable on the base map. Nearby displacements were located with respect to these master points by tape measure. Calibration indicated that measurements with the rangefinder were accurate to within 2% for distances as great as 100 m.

Measurements of displacement in the concrete lining, along the trend of the channel, were made using standard tape measures and rulers. Measurement was complicated by the crude form of the displaced concrete and by variations in displacement along breaks. Such variations were particularly troublesome in the places where displacements extended only partially through the side lining, suggesting rotation of the slabs. Because of these complexities, the measurements reported should be considered best estimates based on several measurements obtained over the length of a given feature.

The most accurate pre-earthquake surveys for horizontal control were selected through discussions with experienced local land surveyors and with appropriate officials of Los Gatos, Santa Clara County, Santa Clara Valley Water District, and Caltrans. These discussions indicated that before the use of electronic distance meters (EDMs), beginning in the 1960s, horizontal distances were measured by chain, resulting in typical accuracies of only within two or three tenths of a foot (∼6–9 cm) under good conditions. Surveys were known to be uneven in quality, so that a given accuracy could not be assured. The advent of EDMs increased accuracy, but their adoption by land surveyors was gradual, and recorded surveys did not indicate the equipment used. Thus, land surveys in Los Gatos, which we had hoped might provide tight local control on deformation in specific zones on damage, were of uneven and questionable accuracy. As a consequence, we searched for major pre-earthquake surveys that employed EDMs for distance measurements.

Two separate lines were selected for horizontal control (Fig. 10). The first was a single 4.7 km shot, oriented 011°, between bronze monuments at Vasona Dam and Saint Joseph’s Hill, part of a 1972–1973 Caltrans second-order survey network (Table 1; Fig. 10). Inspection of these monuments after the earthquake showed no visible evidence of disturbance or local ground movement, although displacement had been monitored at Vasona Dam (described herein). The shot between these two monuments extends across as many as six fault strands of the FTB (McLaughlin et al., 2002) (Fig. 10). This leg was originally surveyed using AGA (Svenska Aktiebolaget Gasaccumulator Co.) model 6 and model 8 geodimeters that resulted in a standard error of 15 mm + 1 ppm (ppm) of the distance; we used 15 mm + 2 ppm as a more conservative estimate. The data were reduced to sea level arc (geodetic distance) based on the UTM NAD27 datum (Universal Transverse Mercator North American Datum of 1927). This leg was resurveyed on 2 October 1990 by a U.S. Geological Survey (USGS) crew using an EDM and procedures that resulted in a standard estimate of error of 3 mm + 2 ppm of the distance. These data were likewise reduced to geodetic distance based on the NAD27 datum. The interseismic velocity, or secular deformation, field between these survey monuments was assessed by (1) applying the empirical model for time-dependent positioning of horizontal crustal motion (TDP-H91; presented in Snay et al., 1991) and (2) the three-dimensional block model (BAVU; presented in d’Alessio et al., 2005) with data provided by R. Bürgmann (2007, personal commun.). Horizontal control of Snay et al. (1991) was derived from very long baseline interferometry (VLBI), EDM, and triangulation-trilateration data and is representative of the time period between 1 January 1950 and 16 October 1989, including accumulated secular motion as well as any coseismic displacements associated with the 1979 Coyote Lake and 1984 Morgan Hill earthquakes. The surface-velocity control of d’Alessio et al. (2005) was derived from GPS data for the time period of 1993–2003 and may include some postseismic deformation from the Loma Prieta earthquake. Our resulting pre–Loma Prieta (1950–1989) and post–Loma Prieta earthquake (1993–2003) estimates of interseismic strain, projected into the Vasona–Saint Joseph’s Hill line, are ∼1.8 and 1.4 mm/yr, respectively. Here we use the pre–Loma Prieta rate inferred from Snay et al. (1991) to correct for the total coseismic change from EDM surveys listed in Table 1.

The leg from Vasona to Saint Joseph’s Hill was also independently resurveyed during March and April of 1990 by Snay et al. (1991) using GPS (Table 1). In this work, the differences detected between the 1972–1973 survey (by geodimeter) and the 1990 resurvey (by GPS) were already adjusted to reflect only coseismic deformation by subtracting the interseismic strain estimated for the period between 1972 and the earthquake in 1989 (Snay et al., 1991). Snay et al. (1991) reported the standard error for individual horizontal components of coseismic displacement vectors as 73 mm, and acknowledged that their plausibly large standard error estimation was the product of their conservative approach. To provide comparison with the EDM survey, we projected the components of horizontal displacement at Vasona and Saint Joseph’s Hill (Snay et al., 1991, Table 2 therein) into the line of the Vasona–Saint Joseph’s Hill leg.

The second line selected for horizontal control was a 1966 Caltrans survey line of unspecified order along State Highway 17. This survey, called the LG line, extended from steep mountainous terrain in the vicinity of Lexington Dam (Figs. 7 and 10) across the flatlands of Los Gatos to the vicinity of Campbell along Highway 17, a heavily traveled four-lane highway. We obtained the data for this survey from the Caltrans Surveys Office; estimates of the range of error had not been recorded, and so we obtained information of the equipment used and estimates of the range of error typical for procedures used in the 1966 survey from party chiefs of the Caltrans survey crews that had conducted the survey. Wherever there was uncertainty, we adopted the widest reasonable range in error so that our estimates of error are conservative. We calculated the combined effect of the several sources of error using standard methods of error propagation (Davis et al., 1981) assuming independence of the several sources of error.

Our resurvey on 17 November 1990, by a USGS crew (including us) covered the central 12 legs (4.5 km) of the original 18 leg Caltrans survey, between stations LG-5 and LG-17 (Fig. 10). A Wild T2002 total-station theodolite (http://www.wild-heerbrugg.com) was used with an estimated standard deviation of error of 1 mm + 1 ppm of distance, <5 cm of arc for horizontal angles, and 25.4 cm of arc for vertical angles. Shortening measured in the steep sloping ground of legs 5–6 and 6–7 was discarded from our analysis because calculations showed that changes in both horizontal angle and leg length were consistent with downslope movement of ∼10 cm at stations LG-5 and LG-6. Vertical control was insufficient to detect such movement, but downslope movement is reasonable at these stations, which are on highway fills founded in steep hillslopes underlain by unstable materials. Further, the concrete lining along Los Gatos Creek, which parallels Highway 17 in this area, was not broken along these legs, suggesting that deformation was confined to the hillslope above. Our use of the LG survey line is thus constrained to the flatlands between stations LG-7 and LG-17 (Fig. 10).

Monuments in flatland along the LG line consist of bronze disks on overpasses and 2.54 cm diameter steel pipes with plug and tack in the median strip between northbound and southbound lanes. The steel pipes had been paved over between 1966 and 1989, then exposed by Caltrans survey personnel in 1989. The monuments on overpasses showed no sign of movement with respect to the overpass, but expansion joints as wide as a few centimeters at each end of each overpass suggested that the overpasses were somewhat free to move, principally in a direction about normal to the direction of survey.

The Vasona Dam is an earthfill dam on Los Gatos Creek that retains the Vasona Reservoir (Figs. 10 and 11). The horizontal and vertical positions of a series of monuments along the dam crest (Fig. 11) have been monitored since 1964 by the Santa Clara Valley Water District (previously the Santa Clara County Flood Control and Water District). Monitoring has been accomplished by surveys from presumably stable stations near the dam; intervals between surveys have ranged from about six months to several years. The monitoring shows gradual settlement and downstream lateral movement of the dam crest, punctuated by abrupt movement at the time of the Loma Prieta earthquake. The monitored stations, spaced at ∼30 m intervals along the dam crest, bracket station Vasona and also help constrain movement at leveling stations G386 and G875 (Fig. 11). The displacement of station Vasona during the interval between 1972 and 1990 revealed lateral displacement in a downstream direction of 1.5 cm and vertical downward settlement of 3.7 cm. Similarly, stations G386 and G875 record settlement of 2.4 and 2.1 cm, respectively, during the period between 1967 and 1990 (Table 2). The displacement of station Vasona (Fig. 11) between 1964 and the dates of subsequent surveys is shown in Table 2. Estimates of error are not reported, but the history of measurements over decades of little significant movement indicates that standard deviation of readings is <3 mm.

Changes in elevation detected by precise leveling surveys shed light on the nature of ground deformation because they reveal any vertical changes that accompanied the abundant signs of horizontal deformation. Leveling surveys have been conducted in the area by the National Geodetic Survey (NGS) periodically since 1934 (T.D. Gilmore, 1990, 1991, written commun.). Many of these surveys are first order, the remainder are second order, and all other sources of vertical control at the time of the earthquake were poor in comparison. Benchmarks along NGS leveling line 3, the focus of our analysis, extend >15 km through Los Gatos about parallel to Highway 17 (Figs. 10 and 12B). These were releveled by the NGS during February and March of 1990 to assess earthquake-related changes in elevation.

Except for the pronounced progressive subsidence related to groundwater withdrawal in the northern part of the profile (Poland and Ireland, 1988), little significant change in elevation was evident between 1934 and 1967. Between 1967 and 1990, however, significant elevation changes took place on the level line in the vicinity of Los Gatos (Fig. 12B). These changes contrast so strongly with the uneventful previous three decades that we assume they are coseismic, as did Marshall et al. (1991) and Árnadóttir and Segall (1994). The leveling surveys of 1967 and 1990 were first order, with a field tolerance of 4.0 mm (Marshall et al., 1991). The 1967 survey was double-run (leveled in both directions); the 1990 survey was single-run, but spur stations were run two or more times (Richard Snay, 1990, written commun.).

Because only a dozen or so stations were available to define elevation changes across the Los Gatos area, and because elevation changes of only a few centimeters at a given station could significantly influence the profile, we inspected most monuments for evidence of disturbance or local ground movement, and in hillside areas, for general stability. All hillside monuments except R878 were inspected and found to occupy what appeared to be stable ground, although numerous deep-seated landslides are present in the area. Other stations inspected were G386 and G875, at Vasona Dam, and the spurs at F180 and A887R; all had no signs of disturbance or local ground movement. Grading for a parking lot had removed station C177 in downtown Los Gatos since the time of the 1990 resurvey, so it could not be examined. The remaining stations in flat land were assumed stable because of their positions along an old railroad grade and the long-term stability of nearby stations.

The form of the 1990 elevation-change profile can be simplified considerably (Figs. 12B, 12C) by correcting for settlement at Vasona Dam and by trimming spurs from the profile. Spurs are stations that are out of the line of profile, to the sides of the main traverse, and so their elevations include effects of deformation across the line of profile. Stations R878, F180, and A887R constitute the spurs in the 1990 traverse (Figs. 10, 12B, and 12C). In the practice of geodetic leveling, accurate measurement of spurs is not required for, or checked by, closure of the main traverse, which is a critical measure of leveling accuracy (see Marshall et al., 1991, p. 1662–1666). Also, confidence in interpretation of leveling data rests more in the trend of a number of these delicate observations than in the value of any individual observation; because each spur of this survey gives an elevation change at just one station, the spurs do not provide trends. The spurs also created profile spikes, defined by adjacent sections that have large tilts of opposite sign (Fig. 12B). For this analysis, we trim the spurs because the resulting profile (Fig. 12C) provides a smooth form that can be compared to regional geodetic models of coseismic deformation.

Of the 1427 locations of damage to pavement and pipes (see Fig. 6), concrete breaks constitute 64% of the mapped damage sites, asphalt breaks constitute ∼3%, pipe breaks constitute ∼31%, and other damage categories constitute the remaining 2%–3% of the sites. Damage at 30% of the sites was indicative of shortening. Spatially, the sites are clustered; 54% are within the Los Gatos USGS 7.5 min quadrangle (Phelps et al., 2015).

Much of the damage was consistent with shortening normal to the mountain front. Breaks and buckles in concrete streets, sidewalks, curbs, and gutters occurred dominantly along north-south– or north-northeast–trending streets (Fig. 7), and so suggest strain with a large component of shortening in those directions. Most of the pavement damage was limited to concrete; adjacent asphalt was typically devoid of cracks.

At three sites of concentrated deformation, however, cracks in asphalt were 30 m long or more (Fig. 13). These sites displayed complex cracking, the details of which appeared controlled largely by human-made structural features of the pavement. Each of these breaks occurred near the center of a damage zone defined largely by breaks in concrete, and each trended northwest, about parallel to the trend of surrounding damage. These breaks marked the hearts of the damage zones and were the closest features to actual ground rupture that we observed in the study area. Their appearance at the hearts of the zones makes sense because breakage of ductile asphalt appears to have required greater or more rapid deformation than breakage of more brittle concrete or buried pipes.

Viewed at broad scale, the swath of damage trends ∼N50W and extends from the Santa Teresa Hills north to Stanford University. This swath contains four discrete zones of concentrated damage that trend from N30W to N65W (Figs. 1 and 6; Plate 1). Three of these zones are near the southwestern valley margin, and one is farther north well away from the valley margin. The three zones of damage near the mountain front are here called the Los Gatos, Blossom Hill, and Los Altos–Cupertino zones; the zone north of the valley margin, which consists mainly of water-line breaks, is designated the Santa Clara zone (Figs. 1 and 6). These zones are described separately in the following.

The Los Gatos zone is a concentration of damage in and near downtown Los Gatos. It trends northwest for a distance of ∼2.5 km, with a width of ∼1 km (Fig. 7). It is located mainly on alluvium that occupies a topographic reentrant carved by Los Gatos Creek into the bedrock uplands (Fig. 6B). This zone contains the most severe and concentrated damage of the study area, as well as the principal measured coseismic displacements.

The length of the zone of damage is limited by the extent of concrete pavement; the few roads that cross likely extensions of this zone lack concrete components. Damage occurred mainly where long strips of concrete that trend north-northeast were thrust or buckled, suggesting a large component of shortening in this direction. Similarly, the storm-drain grate shown in Figure 5 displayed deformation consistent with this sense of shortening. Most of the damage that we recorded in this zone consisted of contraction in concrete, but damage of other kinds, including damage to buildings, was abundant.

The belt of most severe damage within this zone trended about N35W from breaks in the concrete median and asphalt surface of State Highway 17, through the lining of Los Gatos Creek, where we measured shortening of ∼80 mm (described herein), toward cracks in an asphalt parking lot in downtown Los Gatos (Fig. 7). On trend between the creek lining and the parking lot were several contractional breaks in concrete, including at least one with measurable shortening (Fig. 3). The parking lot, located just east of Santa Cruz Avenue between Royce Street and Grays Lane, showed more than 30 m of complex cracks in the asphalt pavement, as well as crushed brick and cracked concrete in walkways and curbs. Northwest from the parking lot, this belt includes damaged buildings, numerous breaks in concrete pavement, and sites of post-earthquake deformation.

All sites of documented post-earthquake shortening occurred along the trend of concentrated damage within the Los Gatos zone. One site was in the sidewalk along the northwest side of Massol Avenue, southwest of the intersection with Almendra Avenue. Sidewalk slabs that buckled during the earthquake had been partly removed and repaired with an asphalt patch, which, by 5 December 1989, had shortened by ∼24 mm along a discrete thrust fracture. At a second site, in the parking lot mentioned above, three measurements were made by steel tape across a 27-m-wide zone that included the cracking in asphalt. The measurements, taken over the period from 7 December 1989 to 13 July 1990, showed progressive shortening that totaled >10 mm. Although the measurements were crude and the shortening slight, we consider the measurements to reflect real ground deformation because (1) the measured shortening is far greater than the potential thermal contraction of the measuring tape, and (2) measurements of similar length at another site during the same period remained constant within ±1.0 mm. At two other locations along this trend of concentrated damage, concrete pavement deformed noticeably during the period between a day and a month after the earthquake, according to a U.S. Postal Service employee who had walked the sidewalk daily to deliver mail (A. Morrisette, 1989, oral commun.; Fig. 14).

The Los Gatos zone includes the only known report of pre-earthquake ground deformation in the study area. The sidewalk in front of a business on East Main Street near Maple Lane developed a fresh spall that was patched with asphalt during the two months before the earthquake. The concrete sidewalk and the old building at this site were then damaged during the earthquake. Old concrete patches in the sidewalk here, apparently to accommodate buckled sidewalk slabs, attest to a prolonged history of pre-earthquake deformation.

The Berrocal fault passes through the Los Gatos area as several subparallel strands that in detail show considerable complexity, including reverse faults that reveal northeast-southwest shortening and related fault-parallel folds (Bailey and Everhart, 1964; McLaughlin, 1974; Sorg and McLaughlin, 1975; McLaughlin et al., 1991; Wentworth et al., 1999; McLaughlin and Clark, 2004). The fault strands typically strike about N50W. One strand near Main Street in downtown Los Gatos passes through the site of premonitory strain. Strands show both normal and reverse motion that locally offsets the folded gravels and finer grained materials of the Pliocene and Pleistocene Santa Clara Formation (QTs in Fig. 6B). McLaughlin and Clark (2004) documented an exposure where Jurassic and Cretaceous rocks of the Franciscan Complex (fsr in Fig. 6B) are thrust over the Pliocene and Pleistocene Santa Clara Formation. Concentrated damage within the Los Gatos zone, including all documented sites of post-earthquake deformation, is northeast of the mountain front fault, on or between traces of the Berrocal fault mapped by Bailey and Everhart (1964) and McLaughlin et al. (1991) within Pleistocene or Holocene alluvium (Qpa and Qha, respectively, in Fig. 6B).

In the foothills east of downtown Los Gatos, coseismic movement along a strand of the Berrocal fault mapped by Bailey and Everhart (1964) and McLaughlin et al. (1991) demonstrated the relationship between fault-juxtaposed bedrock lithologies and damage to adjacent concrete pavement. A roadcut along Top of the Hill Road near Kennedy Road exposes this fault contact between sedimentary rocks and serpentinite. Fresh cracking of a bench in the roadcut, as well as fresh breaks in bedrock of the shear zone, indicated coseismic movement on this fault (Fig. 15). Several meters along strike to the east, northeast-trending concrete curbs on both sides of Top of the Hill Road were deformed. Buckling of the easternmost curb indicated 40–60 mm of shortening along the trend of the curb. The asphalt road surface between the curbs showed minimal deformation.

The Blossom Hill zone is an alignment of three concentrations of damage within a larger band of less concentrated damage. The core of the zone has a width of 400–800 m and extends from near the Vasona Dam southeast for ∼5 km along the foot of Blossom Hill (Figs. 6 and 10). Marking the core of this zone was damage at the Village Square shopping center, which included one of the complex breaks in asphalt (Fig. 13) surrounded by abundant breaks in concrete (Fig. 2B). This site coincided remarkably with a short isolated segment of faulted Quaternary alluvium mapped by Bailey and Everhart (1964), part of a branch of the Shannon fault that is largely concealed by alluvium (Bailey and Everhart, 1964; McLaughlin et al., 1991).

The Blossom Hill zone extends for ∼13 km parallel to Blossom Hill, which is a fault-parallel syncline uplifted between strands of the Shannon fault (McLaughlin et al., 1991). Evidence of damage in this zone diminished progressively toward the southeast, possibly because of the waning occurrence of concrete. Northwest of monument F180 (Fig. 10), just outside the dense concentration of the Blossom Hill zone, is a sparse concentration of damage coincident with the intersection of mapped faults. Toward the northwest, in an area of abundant concrete, scattered damage continued for only a few kilometers beyond Vasona Dam, but 5 km further along the same trend was another concentration of damage, here called the Los Altos–Cupertino zone of damage.

The Los Altos–Cupertino zone of damage extends from the intersection of Page Mill Road and Highway 280 in Palo Alto southeastward to the city of Saratoga. This zone could be considered the northwestern extension of the Blossom Hill zone, but it is distinguished here for descriptive purposes (Fig. 6A). The zone contains two main clusters of damage that are somewhat isolated, probably by the paucity of concrete strain markers in the Rancho San Antonio County Park and the Gate of Heaven Cemetery. The northern cluster, which is located along Highway 280 near the town of Los Altos Hills, showed a broad distribution of damage near traces of the Monte Vista fault (Brabb et al., 1998), including numerous coseismic fractures in asphalt and abundant water-line ruptures. Damage to concrete was abundant near the intersection of Eastbrook Avenue and Mora Drive. Pampeyan and Loney (1995) discussed coseismic damage in part of this area, including damage to buildings. This northern cluster also correlates with the Black Mountain shattered zone of ground deformation generated by the 1906 earthquake (Scholz, 1985).

The other main focus of damage within the zone was in the vicinity of Cupertino, where a narrow band ∼5 km long displayed a cluster of damage near the intersection of Regnart Creek and Bubb Road. Here a concentration of concrete breaks suggestive of north-south shortening was punctuated by a crack in asphalt more than 30 m long trending N40W. This deformation correlates with vegetation lineaments near the Regnart scarp, a 6-m-high northeast-facing escarpment (Hitchcock et al., 1994).

The Stanford Linear Accelerator provided a sensitive indicator of deformation near the northwest end of this zone. Laser measurement along the 3.2-km-long accelerator housing recorded two sites of coseismic fill settlement, each several millimeters in magnitude, and a site of abrupt permanent coseismic displacement totaling more than 10 mm of vertical offset (Fischer, 1989; Riordan, 1989).

The Santa Clara zone is a diffuse zone of damage situated in the city of Santa Clara, 8 km northeast from the Santa Cruz Mountains and the three other damage zones (Fig. 6). It consists of water-line ruptures interspersed with less abundant concrete breaks. The zone is located on Quaternary alluvium 8 km south of the margin of San Francisco Bay, and trends roughly parallel to the zones of damage that are along the foot of the Santa Cruz Mountains. The origin of damage in this zone is elusive; it may have resulted from coseismic subsidence of the thick alluvium, but the zone is beyond the area of major subsidence related to ground-water withdrawal (Poland and Ireland, 1988). Areas of land subsidence related to aquifer compaction may not directly relate to coseismic shaking induced subsidence. Unlike the other zones, the Santa Clara zone is not along mapped faults or breaks in slope. It is, however, in the general vicinity of the San Jose fault, a potential concealed fault inferred from aeromagnetic anomalies in the bedrock floor of the alluvium-filled basin (Brabb and Hanna, 1981; Carter and Jachens, 1993). Furthermore, a gentle gradient in the gravity field coincides roughly with the damage zone (Langenheim et al., 1997), and the map of thickness of alluvium in the Santa Clara Valley (Bishop and Williams, 1974), as well as the modeled basement bedrock depth contours of the Cupertino basin (Jachens et al., 2006) (Fig. 6B). The Cupertino basin formed during transtension associated with the development of the San Andreas fault associated with the passage of the Mendocino Triple Junction (Langenheim et al., 2015). The northeast margin of the Cupertino basin exhibits depth contours aligned generally parallel to the damage concentration.

The concrete lining of Los Gatos Creek served as a sensitive strain gauge that spanned the entire Los Gatos zone of damage (Fig. 16). We measured 12 contractional displacements, expressed as thrusts, in two distinct belts. The principal displacements in the lining occurred at preexisting seams or aging cracks, and thus only incidentally caused fresh breaks in concrete of the channel lining. Fresh breaks were commonly of two types: (1) broken and missing segments of the energy dissipaters where they cross seams or aging cracks (Fig. 8), and (2) spalls along edges of seams and aging cracks (Figs. 9, 17, and 18). Spalls ranged in size from a few centimeters to 30 cm or more across, were generally 1–2 cm deep, and were conchoidal in form, having their largest surface subparallel to the plane of the concrete slab. Spalls exposed white or light gray concrete in their fresh surfaces, which contrasted with the darker gray of the weathered faces, seams, and aging cracks of the lining (Fig. 17B). Using color as an indicator of freshness, it appeared that almost all of the fresh breaks occurred along preexisting seams and aging cracks.

Transient displacements were evidenced by fresh breaks in concrete that were not accompanied by permanent displacement. These occurred most notably where energy dissipaters crossed slab seams (Fig. 8) and, less frequently, as small, isolated spalls along edges of aging cracks. These isolated spalls, though few in number, were observed throughout the channel length, and indicate that adjacent pieces of concrete moved independently, perpendicular to the ground surface, during the earthquake. Such transient displacements provide a potential source of error in measuring permanent displacement. It is possible that some displacements that we identified as transient are actually permanent contractional displacements of less than a few millimeters.

Contractional displacements consisted principally of thrusts, and these generally produced abundant spalls along seams and aging cracks (Figs. 16–19). At a few locations, slabs were thrust as much as 12 cm over adjacent slabs (Figs. 17B, 18B, and 19). At eight of the 11 locations where slabs were thrust, southwestern slabs were thrust over adjacent northeastern slabs. This sense of displacement on reverse faults implies structures with southwestward dips. Other contractional displacements were more subtle, such as seams that had closed tightly to produce a slight (1–2 mm) overlap of spalled concrete edges. A few instances of buckling, or tenting, of small parts of adjacent slabs occurred. Vertical displacements at thrusts were equal to the thickness of the thrust slab, and thus did not reflect the magnitude of any vertical ground displacement.

The principal problem in measuring permanent displacement was that the pre-earthquake widths of seams and aging cracks could not be established with certainty. In measuring contractional strain, we assumed that the seams along which displacement had occurred had a pre-earthquake as-built width of 1.3 cm, the average for apparently undeformed seams. Aging cracks were assumed to have had no significant pre-earthquake gap.

Extensional displacements were more difficult to identify and measure. By the nature of extensional movement, fresh spalls along slab edges would not be expected and were not observed. Only where energy dissipaters crossed seams or aging cracks that had extended did breakage occur. In two places, extension was revealed where moss colonies had spanned pre-earthquake aging cracks. The moss colonies had been torn apart by extensional movement on these cracks, but their halves remained connected by delicate filaments of moss. Otherwise, interpretation of extension relied on the presence of a gap that appeared freshly opened and was not filled with plant roots, plant detritus, or soil. In the case of open aging cracks, a gap of 3 mm or more was generally considered to have resulted from extension during the earthquake. Seam gaps of 2 cm or less were generally assumed not to be extensional because they were within the apparent range of as-built spacing. If the original seam packing was still in reasonable condition and a significant gap existed, however, this gap was assumed to have resulted from coseismic extension.

We measured 12 contractional displacements in two distinct belts, identified as the northern and southern belts (Figs. 7 and 16–19). The northern belt is ∼70 m wide, as measured along the creek channel. The southern belt is 125 m wide, although most displacement occurred within a belt 40 m wide. The two belts are separated by 330 m of undeformed lining, and undeformed lining extends more than 150 m both north and south of these belts.

Displacements were measured principally within the side linings (Table 3). Measured contraction at individual sites ranged from 0.3 to 12.5 cm. Extensional displacements were small in comparison; their observed magnitudes ranged from 0.3 to 1.6 cm (Table 3). This smaller magnitude, combined with the more subtle features generated, made extensional displacements more difficult to recognize, and thereby contributed considerable uncertainty to measurement of total displacement. All eight measured extensional displacements are located in the southern belt, where they occur on both sides of the channel (Fig. 16). No extensional features were identified with certainty in the northern belt, and any extensional features in the bottom slab of the channel lining were concealed.

Both side linings demonstrated measurable net shortening along the trend of the channel in two discrete belts (Table 3). In the northern belt, the average net shortening (rounded to the nearest 0.5 cm) is ∼9.5 cm in the northwest bank and ∼6.5 cm in the southeast bank (Fig. 16). In the southern belt, average shortening measured ∼7.0 cm in the northwest bank and ∼9.0 cm in the southeast bank. We consider this degree of agreement between the two creek banks within each belt surprisingly good considering the uncertainties discussed here. No significant lateral displacements (normal to the channel trend) were detected.

We totaled the average net shortening of the two belts to obtain an overall measured shortening along the trend of the channel of ∼16.5 cm in the northwest bank and ∼16.0 cm in the southeast bank. A maximum total shortening calculated from the ranges of measurements reported in Table 3 is 17.5–19.0 cm, whereas a minimum shortening using these ranges is 14.0–14.5 cm. This total is smaller than the 20–25 cm of channel lining shortening reported previously (Plafker and Galloway, 1989; Haugerud and Ellen, 1990a, 1990b), which was based on reconnaissance observations.

The southern belt of deformation in the lining of Los Gatos Creek aligns with a concentration of damage in downtown Los Gatos (northwest of the channel) and distortion in the pavement of Highway 17 (southeast of the channel) (Fig. 16). This belt is crudely parallel to young faults in the vicinity and ∼300 m north of the concealed location of a young fault shown in Figure 16. The northern belt of deformation in the channel lining is not related to such concentrated damage, but it is ∼100 m south of a concealed fault (Fig. 16). Faults depicted in Figure 16 represent structural boundaries between mapped units or structural discontinuities within mapped units that underwent relative offset, but fault identification is obscured by overlying geologic units. Map locations are inferred by extrapolating known locations or follow topographic lineaments such that positional accuracies are estimated to be within ∼120 m on the ground.

The two measurements of horizontal deformation along the Vasona–Saint Joseph’s Hill line (Fig. 10) are generally consistent in indicating significant shortening across the Los Gatos area (Table 1). Both the 1990 EDM and 1990 GPS measurements were shorter than the 1972–1973 measurement. This is despite the monitored downstream movement during that period of station Vasona (on Vasona Dam), which would tend to lengthen the horizontal distance between stations (Tables 1 and 2). Thus, the downstream (022°) horizontal movement at station Vasona between 1972 and 1990 and the small lateral effect of the measured dam settlement during that period on the inclined EDM shot, oriented 011°, were added to the measured shortening between Saint Joseph’s Hill and Vasona (Table 1). To determine the coseismic strain represented by the EDM data, we removed the estimated interseismic strain for the period between 1972 and the earthquake. The GPS-derived measurement reported by Snay et al. (1991) included the correction for estimated interseismic strain. The results are a shortening of 4.5 cm by EDM and 12.2 cm by GPS. The lower shortening estimate from the EDM data is probably conservative as the interseismic strain correction inferred from Snay et al. (1991) exceeds that representative of the period preceding and immediately following the Loma Prieta earthquake derived from d’Alessio et al. (2005).

The effects of any possible change in elevation of Saint Joseph’s Hill on the inclined EDM shot were not incorporated. Settlement of the hill by ∼1 m, however, would be required to account for the observed shortening. Settlement of this magnitude is unlikely; measured coseismic elevation changes in the region were much smaller (Marshall and Stein, 1996), and the level line discussed in the following shows the part of the profile near Saint Joseph’s Hill to be uplifted with respect to the area near Vasona Dam.

For the Highway 17 survey line, shortening measured in the steep ground within the Santa Cruz Mountains was discarded from our analysis because calculations showed that changes in both horizontal angle and leg length were consistent with downslope movement of ∼10 cm. Downslope movement is entirely plausible at these stations, which are on highway fills founded in steep hillslopes underlain by unstable materials. Further, the concrete lining along Los Gatos Creek, which parallels Highway 17 in this area, was not broken along these legs, suggesting that deformation was confined to the hillslope above. Our use of the Highway 17 survey line is thus constrained to the flatlands between stations LG-7 and LG-17 (Table 4; Fig. 10). For these legs, comparison of the 1966 and 1990 surveys reveals an overall shortening of 13.4 cm along the survey line (Tables 4 and 5). Two legs of the Highway 17 survey show significant shortening, one shows significant lengthening, and most of the other legs show shortening smaller than the standard deviation of error. The association between damage and measured shortening along the survey line is variable; significant shortening accompanies the abundant damage in the heart of Los Gatos, but greater measured shortening to the north lacks significant damage, and no measured shortening accompanies the abundant damage of the Blossom Hill zone near Vasona Dam.

Measurements of horizontal deformation along the Vasona–Saint Joseph’s Hill and the Highway 17 lines are consistent in indicating significant shortening (Table 5). The surveyed shortening in response to the Loma Prieta earthquake ranges from 4.5 cm to 13.4 cm over different spans of ∼5 km that include the abundant damage of the Los Gatos zone. Measurement of offsets in the concrete lining of Los Gatos Creek provides a larger value of 16–17 cm over a shorter distance that includes the entire zone of damage. We express the general magnitude of shortening indicated by these several measurements as ∼10 cm, oriented roughly orthogonal to strikes of FTB faults. This shortening accounts for a maximum of 20% of the total fault-normal shortening produced by the earthquake (e.g., Lisowski et al., 1990).

The trimmed and corrected profile of coseismic elevation changes is compared to profiles of regional models of coseismic geodetic changes in Figure 12C. The models rely on rectangular, uniform-slip dislocations in a homogeneous elastic half-space to represent the coseismic deformation. These elastic dislocation models reflect deformation due to the elastic strain derived from buried rupture sources. Lisowski et al. (1990, their Table 2, model 3) calculated changes in horizontal position measured by EDM, GPS, and VLBI in a dislocation model that included buried oblique slip along the San Andreas fault. Marshall et al. (1991, their Table 5, two-rake planar model) calculated changes in vertical position from leveling data. Árnadóttir and Segall (1994, their Table 1, distributed slip model) modeled EDM, GPS, VLBI, and leveling data to produce a similar profile (not shown here). Although these regional models differ in some respects (Marshall et al., 1991, p. 1681), they are similar in form and magnitude near Los Gatos, showing uplift in the mountains south of Los Gatos relative to a downwarp centered in Los Gatos.

The profiles of the models in Figure 12C were obtained by determining the values provided by each model at the stations of the 1990 leveling profile (NGS line 3). The three profiles plotted are accurate in horizontal and vertical dimensions, but the vertical positions have been shifted to facilitate comparison. Clearly, most of the form and magnitude of the trimmed and corrected elevation-change profile is explained by the models, which show the form of broad-scale warping expected from an earthquake-generating fault at depth.

Three main features distinguish the surveyed profile from the model profiles. The first is a bulge in the left center of the profile between stations G386 and M878, in which the profile of the surveyed elevation changes diverges upward from the smooth arcs of the models. As shown in Figure 12C, the base of this bulge coincides with (1) the axis of the modeled downwarp, (2) the Los Gatos zone of damage, and (3) the zone of greatest shortening as measured along Highway 17 and in the lining of Los Gatos Creek, which in the profile is within the Los Gatos zone. Note that the bulge extends above modeled values in Figure 12C because the profiles were shifted vertically to superimpose their forms. Leveling near Los Gatos actually showed values well below the regional leveling model, resulting in large local negative residuals (Marshall et al., 1991, their Fig. 4C). Regardless of vertical position, the bulge is a departure from the arcuate form of the regional models.

The second main feature is the ∼20 mm spike within the zone of significant groundwater-driven subsidence at survey distance 15–16 km (Fig. 12C). This two-station divergence (E875 and U176R) occurs outside of all generally recognized strands within the FTB and does not correlate spatially with damage (Fig. 10). It is located 2.5 km northeast of the Cascade or New Cascade fault, inferred from gravity, seismic refraction, and geomorphic data (Hitchcock and Kelson, 1999; Boatwright et al., 2004; Hanson et al., 2004, Catchings et al., 2007). It coincides within a few hundred meters with a concealed fault mapped by Bortugno et al. (1991). If this leveling divergence is related to ground deformation, it may result from folding above a blind thrust fault.

The divergence of values at spurs F180 and A887R from the profile of the main traverse is ∼40–50 mm, well above estimates of error (Fig. 12B). This divergence suggests abrupt lateral discontinuities in coseismic elevation change near the main level line. Thus, the smooth form of the trimmed profile may be an accident of the position of leveling stations in a complex deformation field. The divergence at the spurs weakens the significance of the bulge, which is similar in magnitude, although defined by a sequence of stations. Thus, the two-station spike and superposed bulge of Figure 12C appear worthy of considerably less confidence than the more prominent crude downwarp evident even in the raw profile (Fig. 12B).

The downwarp represented by the elastic strain field of the regional geodetic models probably accounts for some of the measured horizontal shortening. The elastic half-space model by Lisowski et al. (1990) results in a horizontal shortening of 64 ± 46 mm between station D177 and station G386 at Vasona Dam (Michael Lisowski, 1991, written commun.), a value comparable to the range of measured shortening, 45–122 mm, over the slightly longer distance of the Vasona–Saint Joseph’s Hill shot (Table 5; Fig. 10). Of the modeled horizontal shortening between D177 and G386, the geometric effect of warping of the ground surface within the elastic half-space is consistent with the chord of the shot being shorter in response to ground deformation. Thus, it appears that a broad warp is sufficient to explain much of the measured horizontal shortening and does not require significant shallow slip on strands of the FTB, but it is also permissible that several centimeters of permanent localized fault deformation may have occurred.

Several explanations for the damage are plausible: shaking alone, possibly amplified by subsurface conditions; transient broad-scale strain; permanent broad-scale strain, such as the downwarp near Los Gatos; sympathetic movement on preexisting faults or folds, released either by shaking during the main shock, by fault movements accompanying the main shock, or by broad-scale strains, such as the downwarp near Los Gatos; and development of new faults or folds.

Ground shaking levels at a site depend on the distance from the earthquake rupture, the local rock and soil conditions, and variations in the propagation of seismic waves due to complexities in crustal structure (e.g., Rubinstein and Beroza, 2004). Ground shaking is typically amplified at soft soil (low velocity) sites. For larger magnitude earthquakes such as Loma Prieta, high levels of ground shaking, associated with greater damage to human-made structures, likely correlate better with peak velocity than peak acceleration (Wald et al., 1999; Wu et al., 2004). As Boore and Asten (2008) determined from invasive and noninvasive methods, shear-wave slowness (inverse of velocity) increases with distance from the edge of the Santa Clara Valley as fine-grained sediments dominate. For the 10 stations within 15 km of the rupture surface extracted from the PEER database, we determined that the PGV orientation was generally north-northeast with a mean and standard deviation of 58.3 ± 31.6 cm/s. Shaking alone seems a plausible source for much of the damage to pavement and pipes, especially when one considers the strength and orientation of ground shaking. This is especially true in Los Gatos, where the systematic broad zone of northeast-southwest orientation of concrete breaks was broadly consistent in orientation with the strongest measured acceleration and velocity.

Three-dimensional modeling of seismic waves in Santa Clara Valley revealed that duration of shaking is strongly influenced by position in the basin (Frankel and Vidale, 1992, 1994). Even though most of the observed damage occurred near the edge of the basin (Fig. 6B), where duration of shaking was shorter, the long-period content of the seismic waves may still be responsible for damaging pavement and pipes. That is, shorter duration shaking does not preclude long-period content enhancement, especially when forward-directivity effects are present with fault-normal dominance at longer periods. Although sites near the edge of the basin are predicted to undergo shorter duration of shaking than sites in the middle, basin edge effects and amplification in sedimentary basins filled by shallow sediments with relatively low shear-wave velocities may be responsible for heightened ground motion. For example, Harmsen et al. (2008), using three-dimensional velocity structure models of the Santa Clara Valley to predict ground motions from scenario earthquakes, determined that the Cupertino basin (Fig. 6B; Langenheim et al., 2015) tends to be highly excited by earthquake scenarios for the San Andreas and Monte Vista faults. The damage pattern presented here, for example, correlates well with peak horizontal velocity simulated for an earthquake occurring at the northwest end of the Monte Vista fault.

Other studies suggest that abrupt changes in thickness or in seismic wave propagation characteristics of alluvium, which tend to occur near valley margins, can create the potential for local seismic amplification (Borcherdt, 1975). Although shaking duration at basin margins may be shorter, edge effects can lead to more intense ground motions through constructive interference and impedance between basin sediment and rock. As Graves (1993) detailed, site response in the Marina District of San Francisco during the Loma Prieta earthquake was amplified with long shaking durations and complex waveforms generated at basin margins. Similarly, Graves et al. (1998, for the 1994 Northridge earthquake) and Kawase (1996, for the Kobe earthquake) concluded that shallow basin-edge structure (∼1 km deep) amplified ground motion response responsible for heavy damage. Such studies suggest that shaking alone, or shaking-induced triggered slip on preexisting faults, could have caused the observed damage near the valley margin.

Another possibility is that concentrations of damage near Los Gatos and Los Altos may have arisen from intensified ground shaking caused by reflection of seismic waves from velocity gradients of shallow to mid-crustal reflectors (∼8–12 km deep). To test this hypothesis, Catchings and Kohler (1996) conducted controlled-source seismic experiments. Their results indicated that seismic reflections from subhorizontal crustal velocity structure intensify ground shaking at discrete distances from the hypocenter, which, from the Loma Prieta source, correspond to the Los Gatos, Los Altos, and Oakland areas. Their experiments, however, indicate that ground shaking intensified by this mechanism should generally occur in bands concentric to the hypocenter. This orientation is roughly normal to the zones of damage shown in Figures 1 and 6 (Phelps et al., 2015), suggesting that this mechanism was not the dominant control on the damage zones.

Another possible explanation is that transient strain might have contributed to the damage. Studies from GPS and leveling measurements of postseismic strain by Bürgmann et al. (1997b) documented unanticipated contraction normal to the strike of the SAF system. The 5 yr of observed postseismic observations documented aseismic reverse slip averaging 2.4 cm/yr along the FTB, approaching the 2.9 cm/yr measured average along the SAF. Bürgmann et al. (1997b) observed that contractions with the steepest gradient in postseismic elevation changes are associated with the hanging wall of the Berrocal fault (similar to Fig. 12C), with negative elevation changes continuing outboard of the range front to the Shannon fault zone. Therefore, the localized contraction and differential uplift identified during the postseismic period across the FTB faults likely involves buried slip on one or more faults and is generally consistent with the possibility that some fault slip may have initiated during the Loma Prieta earthquake.

Ground distortion and resulting damage may have been expressed as movement along preexisting faults (and possibly folds), regardless of whether the distortion resulted from shaking, from broad-scale strain such as the downwarping, or from other tectonic adjustments along these planes of weakness. Permanent coseismic downwarping and associated contraction of the ground surface near Los Gatos, for example, may account for some of the damage in the Los Gatos zone and for the relatively great breadth of this zone (Figs. 12C and 20A). However, damage near the valley margin, including in the lining of Los Gatos Creek, is consistent in several respects with deformation expected from faults of the FTB (Fig. 20B). First, distribution of damage shows linearity parallel to the trend of faults, both regionally and to lesser degree locally within each zone (Figs. 6, 7, and 10; Phelps et al., 2015). Second, damage at many sites occurred on or near faults or tight high-amplitude folds in Pleistocene alluvium, as mapped by Bailey and Everhart (1964), McLaughlin (1974), Sorg and McLaughlin (1975), McLaughlin et al. (1991), Wentworth et al. (1999), and McLaughlin and Clark (2004). Phelps et al. (2015) conclude that the observed damage decreases with distance from the lineaments, with the greatest amount occurring within ∼200 m of the nearest fault or lineament. The damage is not only proximal to the mapped faults and lineaments, but also roughly logarithmically decays with distance, and exhibits the most contractional strain close to mapped tectonic structures. Third, the orientations of contractional damage in concrete where orientations were well defined by regularity of concrete strips do not radiate from the earthquake epicenter, as might be expected if the damage were caused by shaking, and at one locale seem unrelated to even the extent of the main rupture surface of the earthquake as modeled by Lisowski et al. (1990). Rather, the damage orientations appear to be related to local sources, such as nearby faults or folds with the dominant offset mode observed within the Los Gatos Creek concrete lining expressing northeast-vergent thrusts.

Damage to the lining of Los Gatos Creek provides a local example of the fault-like character of coseismic deformation. The most striking characteristic is that the displacements occur in two belts bounded by lengthy stretches of essentially undeformed channel lining (Fig. 16). The magnitude of net horizontal displacement recorded in opposite sides of the channel is similar (Table 3; Fig. 16), consistent with displacement along geologic structures, such as faults, that cross the channel. The third characteristic is the planform northwest bearing in each belt. Although these bearings could be measured only approximately, they ranged from N10W to N34W and N5E to N25W in the northern and southern belts, respectively, similar to the northwest-trending regional geologic structure. The southwest slabs were generally thrust over adjacent northeast slabs consistent with FTB kinematic sense (Fig. 20B). Contractional displacements did not cut cleanly across the channel, but rather were dispersed among the rectilinear planes of weakness in the concrete, over belts several tens of meters wide. Such dispersion might result in part from a veneer of unconsolidated, dominantly coarse-grained alluvium (Bailey and Everhart, 1964; McLaughlin et al., 1991) that probably overlies bedrock beneath at least some of the channel. Alluvium could be expected to transform discrete fault displacements in buried bedrock into diffuse belts of small intergranular displacements near the ground surface (Bray et al., 1994a), with differential movement decreasing as the shear propagates to the ground surface (Bray et al., 1994b).

Alternative explanations for the breaks in the channel lining deserve consideration. Because the shortening is located on the gently sloping valley floor, well away from the nearest mountain slopes, it is unlikely that large-scale landsliding caused the displacements; landsliding of such scale and extent into alluvial flatlands has not been recognized in the area. Local failure of the channel banks cannot account for the cross-channel continuity of the displacement belts. Any effects of seismic shaking, such as settlement, fail to explain the significant permanent shortening; settlement would be expected to extend such a straight piece of pavement. The most plausible alternative explanation is that the measured shortening results from coseismic shifting of the channel slabs such that numerous small extensions, undetected by our examination, were opened in compensation for the measured thrusts. This possibility is countered by surveying measurements that document shortening of similar magnitude across the area (Table 5).

Other lines of evidence suggest that damage resulted from permanent ground deformation rather than shaking. The zone through Los Gatos showed the highest concentration of damage and included evidence for pre-earthquake and post-earthquake deformation. Less direct evidence is an active geologic and tectonic context of young folds and reverse faults that indicates progressive northeast-southwest shortening and recent seismicity in the FTB. Activity of some of these faults is suggested by geomorphic evidence, such as sag ponds, fault troughs, and stream offsets (Rogers and Williams, 1974; Hitchcock et al., 1994; Hitchcock and Kelson, 1999). Using gravity model inversions, R. Jachens (2006, personal commun.) modelled the tip of the primary fault in the FTB surfacing through the Vasona Reservoir area, corroborating mapped inferred faults. Tectonic activity is indicated by a cluster of small earthquake epicenters near Vasona Dam between 1969 and 1974 (Sherburne et al., 1974; Brabb and Hanna, 1981), by other seismicity that shows first-motion solutions consistent with mountain-front thrusting along southwest-dipping planes (Kovach and Beroza, 1993), and by focal mechanisms of aftershocks from Loma Prieta that reveal reverse focal mechanisms northeast of the SAF (Bürgmann et al., 1997a). These various features and events appear to accompany the long-term uplift of the Santa Cruz Mountains, in which the downwarp shown in Figure 12C provides a foot-of-the-hill counterpart to the upwarp suspected (Cotton, 1990; Ponti and Wells, 1990) and surveyed and modeled (Lisowski et al., 1990; Marshall et al., 1991; Bürgmann et al., 1997a, 1997b) near the crest of the Santa Cruz Mountains.

The presence of contractional features along subsidiary faults parallel to the SAF zone is consistent with interpretations of regional tectonic strain partitioning along the transpressional plate margin of central California. The present boundary between the Pacific and North American plates is a complicated zone that exhibits both contractional and transverse deformation (McLaughlin, 1974; Aydin and Page, 1984; Zoback et al., 1987) in response to oblique convergence across the plate boundary (Harbert and Cox, 1989). As a result of the apparently negligible shear strength of the SAF system, the oblique strain in the lower crust or lithospheric mantle is partitioned upsection into nearly pure tangential and normal components of strain in the upper seismogenic crust (Zoback et al., 1987). It appears that the oblique slip at depth during the Loma Prieta earthquake produced distributed coseismic strain at the ground surface. The area surrounding the trace of the SAF near the crest of the Santa Cruz Mountains demonstrated primarily lateral and extensional offsets in response to the Loma Prieta earthquake (Johnson and Fleming, 1993). In contrast, the region along the northeastern front of the Santa Cruz Mountains exhibited the damage documented here, which is consistent with reverse slip on planes striking subparallel to the SAF (Kovach and Beroza, 1993; Bürgmann et al., 1997a) (Fig. 20B). These two areas thus appear to represent the main surface manifestations of coseismic strain partitioning in the upper crust that evolved with the SAF system.

Although the SAF zone has been the topic of most seismic hazard evaluations, subsidiary faults in the region, such as those reflected in the damage described here, may be more active than previously recognized. Our results support the interpretation that reverse displacements occur sympathetically, or undergo aseismic triggered slip, along the foot of the Santa Cruz Mountains in conjunction with large-magnitude events along the SAF. Phelps et al. (2015) conclude that approximately two-thirds of the damage could arise from triggered slip. Hitchcock and Kelson (1999) asserted that broad anticlinal folding and distributed hanging-wall deformation is occurring above buried faults beneath the valley, and that triggered slip, as discussed here, and postseismic creep (Bürgmann et al., 1997a) serve to distribute strain over an area much larger than the SAF zone. Unresolved issues include the near-surface locations of buried or blind thrusts, the timing of past thrust earthquake events, and their recurrence in relation to earthquakes on the SAF. The location of potentially seismogenic reverse faults may be refined by the damage patterns depicted here (and the data available in Supplemental Files 1–3 [see ^{footnotes 1–3}] and Plate 1).

The timing of either aseismic triggered slip or primary rupture along the FTB remains elusive. McLaughlin and Clark (2004) concluded from multiple lines of evidence that 4 km of reverse slip has occurred since the Miocene and that slip has averaged ∼0.3 mm/yr since 10 Ma. Hitchcock and Kelson (1999) concluded that the long-term average deformation rate along the FTB related to aseismic processes is 0.25–0.4 mm/yr, based upon the projected average return period for Loma Prieta–style earthquakes and measurements of triggered and postseismic slip. It is clear that the style of damage described here is not unique to the Loma Prieta earthquake. The description by Lawson et al. (1908) of pavement damage in Los Gatos from the 1906 earthquake is remarkably similar to the 1989 damage described here north of Blossom Hill and in downtown Los Gatos, where damage to concrete was concentrated on north-south–oriented streets. The lack of concrete in 1906 probably limited the development of recognizable deformation features. Similarly, southwest of Palo Alto, near the intersection of the Monte Vista fault with Page Mill Road, damage occurred during both the 1906 (McLaughlin, 1974) and 1989 earthquakes, though the Loma Prieta damage was located to the north of the mapped fault trace, at the intersection of Page Mill Road and Highway 280. Thus, historical evidence points to repeated triggered slip on the range-front thrust system in response to large SAF ruptures.

The distribution of damage to pavement and pipes near the southwestern margin of the Santa Clara Valley resulting from the Loma Prieta earthquake may help to refine the location of potentially active faults concealed by alluvial sediments. Damage in the area was concentrated in zones suggestive of faulting, both on broad and local scale. The mapped distribution of breaks in pavement and utility pipes shows northwest-trending, discontinuous concentrations of damage arranged in four zones that roughly parallel the front of the Santa Cruz Mountains. Much of this damage appears at least crudely associated with mapped northwest-striking oblique dextral reverse faults of the FTB, and the northeast-southwest shortening reflected in much of the damage is consistent with their reverse or thrust nature. Specific sites where damage coincided with mapped fault traces suggest aseismic slip on these faults triggered by the main shock.

Within the Los Gatos zone of damage, the concrete channel lining of Los Gatos Creek was permanently displaced during the earthquake in two belts bounded by long stretches of relatively undeformed channel lining. Cumulative net shortening of ∼7–10 cm occurred on both sides of the creek in each belt, for a total measured net shortening of ∼16–17 cm. The discrete nature of these belts, the similar magnitudes of displacement on both sides of the creek, the similar trends of the displacements across the creek, the subparallelism of these trends with regional geologic structure that includes young faults, the alignment of the southern belt with nearby concentrated damage, and the proximity of these belts to concealed locations of mapped young faults all suggest that these displacements in the channel lining were produced by displacements along faults. The dominance in the channel lining of overthrusts from the southwest suggests that the southwest side of the parent faults moved upward, and this sense of movement on reverse faults implies southwestward dips, compatible with Quaternary reverse faults mapped nearby (Fig. 20B).

Geodetic measurements in the Los Gatos area revealed permanent coseismic shortening of ∼10 cm in a northeast-southwest direction; this shortening accounts for a maximum of 20% of the total fault-normal shortening produced by the earthquake. The sense of this deformation is consistent with the systematic northeast-southwest sense of coseismic contractional failures in pavement that was produced in the Los Gatos zone of damage and less dramatically in the Blossom Hill and Los Altos–Cupertino zones of damage. The main element of the leveling profile is a broad downwarp similar in form and magnitude to three independent regional geodetic models of coseismic deformation. The axis of the warp coincides with the Los Gatos zone of pavement damage and with measured shortening in the Los Gatos Creek lining, and the magnitude of the warp is sufficient to explain most or all of the horizontal shortening measured by the Vasona–Saint Joseph’s Hill surveys.

Any contractional strain resulting from the downwarp may have been distributed across the zone of bending, concentrated in zones of weakness, or expressed in both ways; damage in Los Gatos suggests both forms of expression. Despite the absence of surface rupture, the three zones of damage near the valley margin suggest movement along local faults, principally because of the abrupt concentrations of damage, the linearity and broad-scale continuity of some of the zones, and the crude spatial association with mapped faults of likely Holocene activity. The northeast-southwest sense of shortening indicated by much of the damage is consistent with movement expected along recognized faults, but in much of the area near Los Gatos is also consistent with the direction of peak ground velocities during shaking.

We appreciate the cooperation from public and private agencies that provided information for the damage compilation (see Schmidt et al., 1995, for details). Major contributors to compiling and conducting the surveys were Ralph Haugerud, U.S. Geological Survey (USGS), who initiated some of the work and ideas presented here, Ed Bagnani, California State Department of Transportation (Caltrans) District 4 Surveys, who provided data from pre-earthquake horizontal surveys; Thomas D. Gilmore, who provided historical leveling data; John C. Hamilton, USGS, who conducted the 1990 Highway 17 survey; Michael Lisowski, USGS, who provided the 1990 Vasona–Saint Joseph’s Hill electronic distance meter survey; Grant A. Marshall, USGS, who provided leveling data; Chris Fischer, Frans Lind, Gene Campbell, Scott Baker, and staff at the Engineering Department of the Town of Los Gatos; Gary Faler, Tom Collabro, and staff at the Surveying Department of the Santa Clara Valley Water District; Bob Teple and Trish Gomes of the Water Facilities Branch of the Santa Clara Valley Water District; Bill Ferry, Billy Rae Bovee, and Bob Miller at Caltrans; Don Marcott of Santa Clara County, and local surveyors Roger Dodge, Herb Killmeyer, Tom Riley, and Bob Raznatovich, for information on past local surveying practices; and Richard Snay of the National Geodetic Survey. Our field observations were supplemented by information from other USGS personnel, including Allyn Foss, Edwin L. Harp, Ralph A. Haugerud, Randall W. Jibson, Robert J. McLaughlin, Elise Pendall, Dennis H. Sorg, and John C. Tinsley, as well as USGS volunteers Julian Bommer (Imperial College, London, England), Mauro Cardinale (Istituto di Ricerca per la Protezione Idrogeologica del Consiglio Nazionale delle Ricerche, Perugia, Italy), K.H. Chang, and Sam Donaldson. Donna L. Knifong and Geoffrey A. Phelps provided initial digital preparation of the maps and assisted in the field. Comments from Roland Bürgmann, Tinsley, Carl M. Wentworth, David R. Montgomery, Jonathon Stock, and an anonymous reviewer improved the text.