Abstract

Many fault zones trend through developed urban areas where their geomorphic expression is unclear, making it difficult to study fault zone details and assess seismic hazard. One example is the Holocene-active Rose Canyon fault zone, a strike-slip fault with potential to produce a M6.9 earthquake, which traverses the city of San Diego, California (USA). Several strands trend through densely populated areas, including downtown. Much of the developed environment in San Diego predates aerial imagery, making assessment of the natural landscape difficult. To comply with regulations on development in a seismically active area, geotechnical firms have conducted many private, small-scale fault studies in downtown San Diego since the 1980s. However, each report is site specific with minimal integration between neighboring sites, and there exists no resource where all data can be viewed simultaneously on a regional scale. Here, geotechnical data were mined from 268 individual reports and synthesized into an interactive geodatabase to elucidate fault geometry through downtown San Diego. In the geodatabase, fault segments were assigned a hazard classification, and their strike and dip characterized. Results show an active zone of discontinuous fault segments trending north-south in eastern downtown, including active faults outside the mapped regulatory Earthquake Fault Zone. Analysis of fault geometry shows high variability along strike that may be associated with a stepover into San Diego Bay. This type of geodatabase offers a method for compiling and analyzing a high volume of small-scale fault investigations for a more comprehensive understanding of fault zones located in developed regions.

INTRODUCTION

The city of San Diego is the 8th most populous city in the United States and is located on the southernmost coast of California (USA), within the border zone between the Pacific and North American plates. At the latitude of San Diego, the plate boundary includes a wide zone of faulting from the San Andreas fault in the east to the coastal and offshore faults of the California Borderlands (Fig. 1). The Rose Canyon fault zone (RCFZ) is a coastal fault zone, characterized by right-lateral motion and a long-term slip rate of ∼1–2 mm/yr, which is capable of producing a M6.9 earthquake (Anderson et al., 1989; Lindvall and Rockwell, 1995; Rockwell and Murbach, 1996; Rockwell, 2010). The RCFZ poses a major seismic hazard for the San Diego region, as it strikes through densely populated areas, including downtown.

Downtown San Diego has been well developed since the early 1900s, prior to aerial imagery or high-resolution topographic maps, making geomorphological recognition of faulting difficult. Furthermore, the dense development of the downtown area precludes traditional fault zone studies, in which fault exposure is required. This is similar to several other major cities in California (e.g., Los Angeles, San Francisco Bay area) and worldwide (e.g., Izmit, Turkey; Wellington, New Zealand; Kumamoto, Japan), where fault zones are obscured by development. Additionally, the RCFZ is a complex fault zone, and downtown San Diego sits at the edge of a major releasing stepover where the RCFZ steps offshore across San Diego Bay, a pull-apart basin (Fig. 2). The total step from the RCFZ to the offshore Descanso fault is >10 km, so a throughgoing rupture is not predicted (Wesnousky, 2006). Nevertheless, rupture models show that the presence of smaller faults within the stepover can have a complicated effect on rupture propagation across a step, with some scenarios suggesting that rupture could propagate onto intermediate faults within the step (Lozos et al., 2015). Stepover geometry also evolves over time and could result in temporally complex rupture patterns with changes to the amount of slip accommodated on various fault segments (e.g., Wakabayashi et al., 2004; Wu et al., 2009). Therefore, a detailed understanding of fault geometry near and across the San Diego Bay stepover is important for accurate hazard assessments for the region and for improving our understanding of stepover evolution.

This project represents the first attempt to synthesize geotechnical data from downtown San Diego, gathered by the geotechnical community, to understand better the geology and seismic hazard of the area. The RCFZ’s location through the populated city and classification as Holocene active (Lindvall and Rockwell, 1995) place restrictions on development. These restrictions are in place through the Alquist-Priolo Earthquake Fault Zoning Act, which prohibits the location of most structures for human occupancy across the traces of active faults, and through the City of San Diego Downtown Special Fault Zone, which requires a fault evaluation for any new or additional development near active fault traces. These regulations define a zone around known active fault traces where the restrictions are in effect, herein referred to as the Alquist-Priolo zone (AP zone) (Fig. 3). Within the AP zone, fault investigations are routinely conducted for proposed development projects by various geotechnical firms. The fault investigations can include trenching, sediment borings, cone penetration tests (CPTs), and geophysical subsurface imaging, which are then used to define the stratigraphy and fault geometry across the proposed development site, typically about the size of a city block. These types of data are included in site reports submitted to the firm’s clients and city officials, but the data from the reports have never been compiled before into a single geodatabase for a more complete view of the RCFZ through the entire downtown area.

For this study, data were pulled from 268 geotechnical investigations performed by various consulting firms between 1979 and 2016, and were compiled into a comprehensive GIS fault and seismic hazard map of downtown San Diego. The resulting geodatabase could contribute to an active fault database for use in updating the city’s seismic safety element, and aid the science community by helping to establish fault characteristics and complexities along strike, map subsurface stratigraphy beneath downtown for use in ground acceleration and liquefaction models, and potentially illuminate recurrence intervals, patterns of multi-segment ruptures, and evidence for long-term slip rate. Advances in these areas would greatly improve our understanding of faulting and earthquakes and improve our ability to assess seismic hazard to populated regions. As a case study, the compiled geotechnical data were used to assess RCFZ geometry as it relates to evolution of the San Diego Bay stepover. Specifically, the orientations of fault segments were compared with the timing of the most recent activity on those segments to assess whether the evolution of the pull-apart basin may have resulted in a change in active fault orientation through time.

GEOLOGIC BACKGROUND

Rose Canyon Fault Zone

The RCFZ is the southern continuation of the Newport-Inglewood fault zone that strikes south from Los Angeles, continues along the continental shelf edge, and then trends onshore just north of Mount Soledad in La Jolla, California (Fig. 1) (Fisher and Mills, 1991; Rockwell, 2010; Sahakian et al., 2017). From La Jolla, the RCFZ extends south along the Interstate Highway 5 corridor and then diverges near Old Town San Diego (Kern and Rockwell, 1992; Singleton et al., 2019). One strand trends toward the San Diego International Airport and the other toward downtown San Diego (Fig. 2). Both faults splay offshore into San Diego Bay, where they continue as the Spanish Bight fault (from the airport) and the Coronado and Silver Strand faults (from downtown). The offshore faults accommodate transtension across San Diego Bay created by a releasing stepover between the onshore Rose Canyon fault and offshore Descanso fault (Fig. 2) (Moore and Kennedy, 1975; Legg, 1985; Rockwell, 2010; Maloney, 2013). The Descanso fault continues south along and adjacent to the coast of Baja California, Mexico.

The stepover across San Diego Bay has resulted in localized transtension and subsidence and the evolution of multiple north-trending en echelon faults within the bay (Moore and Kennedy, 1975; Rockwell, 2010; Maloney, 2013). To the east, the north-south–trending La Nacion fault zone has been interpreted as the eastern boundary of the pull-apart basin (Anderson et al., 1989) (Fig. 2). The La Nacion fault zone is composed of west-dipping, anastomosing normal faults with >60 m of vertical offset observed in the Pliocene San Diego Formation (Hart, 1974).

Downtown San Diego contains several Holocene-active splays of the RCFZ. Since the late 1970s, geotechnical investigations have identified two zones of faulting that comprise the AP zones downtown (Fig. 3). The western zone encompasses the San Diego fault, which trends N5°–6°W and dips 60°–80°E with measured vertical offsets, east-side down, of between 3 and 10 m and a minimum lateral offset of ∼3 m (Treiman, 1993; 2002). The eastern zone extends south into San Diego Bay and includes the “downtown graben,” which is a transtensional basin inferred via three north-south–trending faults (Treiman, 1993, 2002). The western graben fault is described as striking from north-south to N15°W with eastward dip angles that range from 63° to 86°. The normal sense of displacement is east-side-down, and evidence for lateral displacement has been identified. The central graben fault includes multiple changes in strike, but dip angles are relatively consistent, ranging from vertical to 68°E. Both vertical and horizontal displacement have been observed on the fault. The central graben fault appears to continue offshore across San Diego Bay as the Silver Strand fault. The eastern graben fault trends N10°–20°W and shows evidence for both vertical and lateral offset. Dip angles vary from vertical to westward dipping at ∼60° with apparent displacement down to the west. The surface expression of faulting in the “downtown graben” was also observed and illustrated in an 1876 bird’s-eye-view drawing of downtown San Diego, prior to most development (Glover, 1876). The fault is drawn laterally offsetting topographic features.

Historic seismicity has been observed in the downtown San Diego and San Diego Bay vicinity. Astiz and Shearer (2000) located several earthquakes occurring between A.D. 1981 and 1997 of <M4 between Coronado Bank and Point Loma, which could be attributed to motion along either the offshore Coronado Bank fault or the Descanso fault (the possible western boundary of the San Diego Bay stepover) (Fig. 1). Additional seismicity includes several small magnitude earthquake clusters recorded beneath San Diego Bay from A.D. 1985 to 1987 (Magistrale, 1993). Paleoseismic investigation at a site east of Mount Soledad revealed six Holocene events on the RCFZ yielding a recurrence interval of ∼1800 yr (Lindvall and Rockwell, 1995; Rockwell, 2010), but recent trenching near Old Town suggests a recurrence interval of 700–800 yr during the late Holocene (Singleton et al., 2019). Data from several geotechnical studies were used to further constrain the RCFZ most recent event to an age of A.D. 1650 ± 120 yr (Rockwell, 2010).

Stratigraphy

San Diego is a coastal city that has been experiencing regional uplift (0.13–0.14 m/k.y.) with localized transpression and transtension during the eustatic sea-level cycles of the Quaternary, which resulted in stratigraphy that was deposited in both marine and nonmarine environments (Kern and Rockwell, 1992; Haaker et al., 2016). Formations found in the stratigraphy of the downtown San Diego area include the San Diego Formation (3–1.5 Ma), the Lindavista Formation (1.5–0.7 Ma), and the Bay Point Formation (0.13–0.08 Ma) (Kennedy, 1975; Kennedy and Peterson, 1975; Kennedy and Tan, 1977; Tan and Kennedy, 1996; Abbott, 1999). There is a regional unconformity between the Lindavista and Bay Point Formations corresponding to a gap in deposition of ∼0.57 m.y. Both “Lindavista Formation” and “Bay Point Formation” are older terms that have been recently discontinued, and the associated deposits now are commonly referred to as paralic deposits. However, because the terms are used in many of the earlier reports included in the project to describe subsurface stratigraphy, the terms will be used here. Also present in most sites downtown are paleosols and Holocene overburden deposits such as alluvium, colluvium, stream or river terrace deposits, and undocumented fill placed by humans for development. Identification of these deposits in the subsurface in geotechnical studies allows for relative dating of fault offset to determine the recency of fault activity.

METHODS

Data for this project came from geotechnical reports that were completed by several different private geotechnical consulting firms (Table 1). Within the reports, data were primarily found in the form of trench logs, boring logs, CPT logs, and geologic cross sections, with lesser contributions from seismic lines and test pits. The information contained in the logs was originally obtained on site by a licensed engineering geologist. Most of the reports collected for this project were bound paper reports, which were unbound and scanned into PDF format, and then cataloged (Table 1).

Geodatabase Creation

For geospatial analysis, the gathered data were imported into an ESRI GIS ArcMap project. The ESRI World Imagery layer, ESRI World Street Map layer, and a parcel map of San Diego from the San Diego Association of Governments (SANDAG) were all downloaded from ESRI’s online database and used as georeferencing base maps. Within each geotechnical report was a site map showing locations of collected data. The site map from each report was georeferenced to the base layer individually using the ArcMap georeferencing toolbar.

None of the site maps contained markings for geographic coordinates, so all georeferencing was done using other mapped features tied from the site map to the base map layers. Most site maps showed parcel boundaries or property lines for the study area. For those maps, control points were placed along the parcel boundaries or property lines and tied to the SANDAG parcel map. For site maps where these boundaries were unavailable, control points were placed on street corners and around buildings and tied to the ESRI World Imagery and World Street Map base layers. There was a small number of site maps that showed neither parcel boundaries nor building locations. For these sites, other landmarks, such as vegetation, parking lots, and driveways, were tied to the ESRI World Imagery and World Street Map base layers. Once the site maps were georeferenced, they were moved into group layers in the project table of contents, organized by the decade in which they were created (Fig. 4).

To estimate the total uncertainty in the process of georeferencing site maps, 10 georeferenced maps were randomly selected from the downtown area. For each site, the ESRI World Imagery accuracy was determined with the ArcMap Identify tool. All sites reported an accuracy of 4.06 m. The other two base maps do not include reported accuracy. Therefore, the center of an intersection adjacent to the site was used to measure offset distance between the World Imagery and World Street Map layers. Additionally, the offset between the World Street Map and parcel map layers was determined by measuring the distance between an identifiable corner of a building, lot, or park on the World Street Map layer and the same corner on the parcel map. Finally, the total root mean square (RMS) error reported in the ArcMap georeferencing control point table was recorded for each site map. All of these measurements are reported in Table 2. The average uncertainty from each source of error was calculated, and then the total uncertainty (U) was calculated to be the square root of the sum of squares for all sources of error (World Imagery [wu], World Street Map [su], parcel map [pu], and georeferencing control points [gu]), calculated as: 
graphic

This resulted in an average estimated total uncertainty of 5.20 m for the 10 sites. While this is an average estimate, it offers a method of quantifying uncertainty that can be applied across the entire project. Where knowing site uncertainty is crucial, we suggest calculating total uncertainty by this method for individual sites of interest. As this project is based on the work of others, it is difficult to quantify other sources of uncertainty introduced during field investigations and generation of original site maps.

The georeferenced site maps were used to digitize locations of the various data recorded on each map. Using the drawing toolbar in ArcMap, data locations were traced using point symbols for boring, CPT, hand auger, and test pit locations, and lines or polygons for trench, cross section, and seismic profile locations. The drawn symbols were then converted to features and added to the ArcMap project. A feature was created for each individual report and then grouped by data type and organized by year within each group (e.g., a group containing all boring data from all reports was broken into sublayers, organized by year, of boring locations from each site) (Fig. 4). This allows the user to turn on and off all of one type of data at a time.

Once data locations were digitized and divided into sublayers, additional information for each sublayer was added to its attribute table (Fig. 5). Columns were created that included: Name, the name of the feature (e.g., B-2, Trench Log, etc.); ReportName, the name of the associated report; Company, the geotechnical firm that did the work; Date, the report date; and ProjectNo, a reference number used by the geotechnical firm for record keeping (Fig. 5). Additionally, the columns Latitude and Longitude were created to give coordinate data to boring, CPT, hand auger, and test pit sublayers.

Next, the image(s) of the data log(s) from the original report were linked to each corresponding feature in ArcMap. This was accomplished by creating a folder within the project catalog where all images of data logs were stored as individual files. To link the image of a specific data log with the corresponding feature on the map, a column was created in every data attribute table called Image. Here, the image file name for the given data log was placed within the row for that feature, and the Hyperlink property of the sublayer was enabled. The user could then left-click on a feature using the hyperlink tool to display the corresponding data log.

Hazard Classification

In ArcMap, faults identified and recorded in geotechnical reports were traced using the georeferenced site map to create line features for each fault classification. The faults were classified and grouped as active, potentially active, or less potentially active. These definitions are modified from the previous California Geological Survey (CGS) fault classifications, which were adopted by the geotechnical community during the fault investigations synthesized in the geodatabase. Herein, an active fault is defined as one that displays evidence of movement within the Holocene (the last 11,500 yr). By the CGS definition, a potentially active fault displays evidence for movement prior to 11,500 yr ago, but within the Quaternary period (the last 1.8 m.y.) (Bryant, 2010). This project further splits the CGS-defined potentially active faults into two groups; potentially active and less potentially active. This was done to accommodate the moving of the Pliocene-Pleistocene boundary from 1.8 Ma to 2.6 Ma in 2009. Herein, a fault defined as less potentially active shows evidence of movement within the San Diego Formation and/or the Lindavista Formation (3.0–0.7 Ma). A fault defined as potentially active shows evidence of movement within the Bay Point Formation (0.13–0.08 Ma). Due to the gap in deposition within the region, activity between 0.7 and 0.13 Ma could not be identified. Fault classification in this project is directly based on the classifications made by the engineering geologists that performed the original work. To distinguish potentially active from less potentially active, the trench logs and geologic cross sections were used to view and interpret the extent of faulting through the stratigraphic sections. The CGS updated their fault classifications in 2018 to simply “Holocene-active” if younger than 11,700 yr, or “pre-Holocene” if older than 11,700 yr. Nevertheless, the reports used in this study were based on the older definitions, and so those definitions were adopted for hazard classification in the geodatabase.

Finally, an earthquake hazard group layer was made based on the mapped faults and their classification. For every site included in this project, its boundary was outlined atop the base map in ArcMap and grouped to represent the level of seismic hazard risk at that site. Sites where no faults were encountered were grouped as hazard level 0 sites. Sites where faults classified as less potentially active were encountered were grouped as hazard level 1 sites. Sites where potentially active faults were encountered were grouped as hazard level 2 sites. And, sites with active faulting were grouped as hazard level 3 sites. These hazard levels, too, were given a group layer with sublayers based on classification. It should be emphasized that a hazard level 0 does not indicate that a site would suffer no impact from an earthquake on the RCFZ, but rather that the site is not located directly above any known fault traces.

Fault Geometry

Strike orientation data were taken from all faults plotted in the project from geotechnical reports. This was done by downloading and installing the EasyCalculate add-on for ArcMap (https://www.ian-ko.com/free/EC10/EC10_main.htm), which measures an azimuth direction for line features (e.g., fault traces) within ArcMap. Because some fault traces change direction, the mapped faults were first split at all vertices along the polyline using the Split Line at Vertices ArcMap tool. Then strike was measured from the middle of each fault segment. Measured strike orientations for each fault classification were placed in that layer’s attribute table. Dip angles and directions for each fault, taken from associated geotechnical reports, were also recorded in attribute tables when available. A histogram of dip angles for each group of faults was also generated. Strike orientation data were exported to MATLAB to generate rose diagrams that plot strike orientations and mean direction for each fault classification group. For the rose diagrams, and subsequent statistical tests, the strike data were normalized by fault segment length. Each 10 m length of a segment was counted as one data point (e.g., a 10 m segment produced one data point of the same strike while a 100 m segment produced 10 data points of the same strike).

Three basic statistical analyses were run in MATLAB, following Trauth (2007), on the strike orientation data for active faults, potentially active faults, and less potentially active faults. The tests were performed on a 95% confidence level and include a Pearson chi-squared test (Pearson, 1900) for randomness of directional data, a Rayleigh test (Mardia, 1972) for the significance of a mean direction, and an F-test (Snedecor and Cochran, 1989) for the difference between two sets of directions. The chi-squared test compares the empirical frequency distribution of fault directional data (strike orientations) with a uniform distribution to determine randomness. The Rayleigh test uses the mean resultant length, a measurement based on the computed sine and cosine of each strike direction, which increases proportionally to the significance of the mean direction. If the measured mean resultant length is greater than the critical mean resultant length (taken from table 10.1 in Trauth [2007] from Mardia [1972], using significance level of 0.05), then the null hypothesis is rejected and the data are claimed to have a preferred direction. The F-test compares resultant lengths of each of two data sets with the combined resultant length of both data sets to give an F-value. If the measured F-value is greater than the critical F-value (taken from standard F-value tables), then the null hypothesis is rejected and it is concluded that the data sets are not from the same population.

RESULTS

Geodatabase

Within ArcMap, the finished project contains 23 group layers (Fig. 4), which include group layers for the three fault classifications (active, potentially active, and less potentially active), four hazard classifications (hazard levels 0–3), seven types of data logs (borings, CPTs, augers, test pits, trenches, cross sections, and seismic lines), and five layers for site maps grouped by decade (1970s, 1980s, 1990s, 2000s, 2010s). The total size of the ArcMap project is ∼5.2 gigabytes, which includes digitized images of site maps and data logs from geotechnical reports. Also included are the background layers for the San Diego parcel map (from SANDAG), the U.S. Geological Survey and California Geological Survey (USGS-CGS) Quaternary fault map (USGS-CGS, 2006), and the ESRI layers that make up the base of the map (World Imagery, World Street Map). To create the geodatabase, >400 reports were collected and assessed from various geotechnical firms. Many of the reports were decades old, and some reports were found to be inadequate for inclusion due to lack of data, uninterpretable data (e.g., faded pages, illegible hand-drawn logs), missing pages, or incomplete data logs. From the reports collected, 268 were included in the project. Still, 42 of these reports have missing data. These are reports that are only missing one or two data logs but still contain enough data to be useful. The 268 reports included in the study yielded a total of 2020 georeferenced data points. These data points are made up of 922 boring locations, 290 CPT locations, six hand auger locations, 54 test pit locations, 554 trench locations, 189 cross section locations, and five seismic lines (Figs. 6 and 7). The geodatabase was made publicly available through Weidman et al. (2019).

Hazard Classification

Reports were assessed to assign a hazard classification to each site and plot faults that were encountered by the engineering geologist during field work. A total of 93 faults were mapped and classified (Fig. 8). Nine were classified as less potentially active faults, 35 as potentially active faults, and 49 as active faults. Faults were traced from each original report’s site map. Fault traces on site maps are generally confined only to the study area, which resulted in many short and discontinuous fault segments (each about a block in length). These segments were not connected within the ArcMap project to best represent the original data, but mapped faults may represent strands or segments of a larger, more continuous fault zone, which could be interpreted by the user.

Hazard levels for each site were assessed based on the presence (or absence) of faults within the different classifications. In total, 223 sites within the downtown area were given a seismic hazard classification (Fig. 8). Of those sites, 172 sites were classified as hazard level 0 and outlined in green; eight sites were classified as hazard level 1 and outlined in yellow; 22 sites were classified as hazard level 2 and outlined in orange; and, 21 sites were classified as hazard level 3 and outlined in red.

There are three sites in the study that do not have a hazard classification and are outlined in black on the map (Fig. 8). Each site contains some geologic data (e.g., boring data, CPT data, etc.), but lacks any information on faulting because a fault evaluation was not performed. Therefore, they could not be given a hazard risk classification. There are also three sites that have a hazard classification, but no fault classification. In all three cases, the reports indicated the existence of faults classified by this study as less potentially active, but did not plot the faults on the site map. Because this study is meant to show results based on geotechnical data, the sites were classified as hazard level 1. However, there was too little information to accurately plot faults on the map at these sites.

The faults plotted from geotechnical reports generally follow the pattern of faults from the USGS-CGS fault database (Fig. 9). However, many faults are mapped differently for at least part of their length. Eight of the nine less potentially active faults, all 35 potentially active faults, and 40 of the 49 active faults differ from the USGS-CGS faults by >5.2 m (estimated uncertainty) for at least part of their extent. The biggest differences were observed in the potentially active faults, where many of the faults do not appear to align with any part of a USGS-CGS fault (Fig. 9). Most active faults track more closely with the USGS-CGS faults, but mismatch for part of their extent. The active faults also fall within the AP zones, with the exception of three faults classified as active that were identified one block east of the current western AP zone boundary and were not included in the USGS-CGS fault map (Figs. 9 and 10A). The site map associated with the faults shows two fault zones, fault zone A and fault zone B. Fault zone A comprises two faults that trend to the northwest at approximately N17°W to N20°W and dip to the west at 65° (Geocon, Inc., 2013). Displacement on these faults was not resolved, but a west-side-down sense of motion is suggested (Geocon, Inc., 2013). Minor faulting also discovered during trenching operations shows vertical displacement of 0.3–0.6 m (Geocon, Inc., 2013). Fault zone B comprises one fault that trends northwest from the southeastern corner of the site, parallel with faults in fault zone A, but turns toward the northeast. Fault zone B dips to the east and shows an east-side-down sense of displacement, with estimates of vertical offset per event to be on the order of 0.3–0.6 m (Geocon, Inc., 2013). Additionally, at a bend in a fault segment located south of the “downtown graben,” the maximum distance between an active fault mapped in geotechnical investigations and that mapped in the USGS-CGS database is ∼28.3 m (Fig. 10A). The total georeferencing uncertainty for this site was calculated to be 7.31 m by the same calculations outlined in the Methods section, which is smaller than the measured difference. Furthermore, the fault was identified by the investigating geotechnical firm in a trench that extended across the entire southern length of the site, including across the location of the USGS-CGS mapped fault (Geocon, Inc., 2004) (Fig. 10A). The investigation did not show evidence of a fault at the location of the USGS-CGS mapped fault, justifying a mismatch in the two databases at this location.

There are also sites in the downtown area, located across faults in the USGS-CGS database, where reports from geotechnical investigations uncovered no evidence for active or potentially active faults. Because this study directly reflects the geotechnical data, these sites were assigned to hazard level 0. In some locations, the USGS-CGS faults just barely cross the corner or edge of the site, and therefore the inconsistency could be a result of location uncertainty or lack of data at the site edges. However, there were also some larger inconsistencies observed. For example, there are two sites located in the Little Italy section of northern downtown where trenches were dug directly across a mapped segment of the RCFZ but a fault zone was not encountered (Fig. 10B). This northern San Diego segment is not included in the AP zones, but is mapped in the USGS-CGS database as a segment of the RCFZ extending south from Old Town. At the northern site, where the investigation was conducted in 2008 by Geocon, Inc., the trench was dug to 2.4 m depth where unfaulted Bay Point Formation (Pleistocene age) was encountered (Geocon, Inc., 2008). At the southern site, a 1.5-m-deep trench was excavated in 2016 by Construction Testing & Engineering, Inc. (CTE), and unfaulted Pleistocene-age soils exhibiting very well-developed argillic horizons were encountered (CTE, 2016). The San Diego segment also crosses a third site in Little Italy that has been classified as hazard level 0. However, at this site, located between the other two trenching sites, only soil borings were collected, and the boring transect did not cross the trace of the USGS-CGS fault location (Fig. 10B).

Fault Zone Geometry

To identify possible trends in fault activity that may be related to fault orientation or dip, the strikes and dips for all faults mapped in the study (if available) were analyzed (Table 3). All faults classified as less potentially active exhibit strike orientations that vary from approximately N40°W to N30°E, with a mean strike direction of N9°W (Fig. 11A). The chi-squared test results indicated that strike directions are not random. The mean direction was also tested using the Rayleigh test, which found the data to have a preferred direction. Based on data from included reports, dip angles of the less potentially active faults vary from 50° to near vertical (Fig. 12; Table 3). Dip directions are both to the east and west and exhibit normal separation in most cases, with beds separated from as little as 5 cm up to 2.7 m.

Faults classified as potentially active exhibit strike orientations that vary from approximately N26°W to N50°E, with a mean strike direction of N16°E (Fig. 11B). The chi-squared test results showed that the data are not random, and the Rayleigh test indicated that there is a preferred direction. According to the geotechnical reports, most dip angles exhibit minor deviations from vertical to the east and west. Only three of the 35 potentially active faults have a dip angle shallower than 70° (Fig. 12; Table 3). Offset beds along the potentially active faults exhibit oblique offset in most cases that varies from centimeters to meters. Spatial analysis of these trends shows that potentially active faults in the western half of the study area, near San Diego Bay, are dominantly north to NW trending, while the eastern half, near Interstate Highway 5, contains potentially active faults that dominantly trend NE.

Strike orientations of the active faults vary from approximately N78°W to N48°E with a mean direction of N9°W (Fig. 11C). The chi-squared test results indicated that the data are not random, and the Rayleigh test showed that the data have a preferred direction. In map view, there is an obvious change in active fault strike moving from south to north (Fig. 9). To the south, active faults trend to the north coming out of San Diego Bay, whereas further north, they take a more northeast trend and then turn back toward the northwest near the “downtown graben” (Fig. 9). Approximately half of the active faults have near-vertical to vertical dip angles. The other half exhibit dips that vary from east to west and have dip angles ranging from 50° to 79° (Fig. 12; Table 3). Active faults that have shallower dips are concentrated around the “downtown graben,” with faults on the west side of the graben dipping to the east and faults on the east side of the graben dipping to the west.

The strike data from each group of faults were tested against each other using the F-test to determine if the difference between the data sets is statistically significant. When comparing the active faults with the potentially active faults, the F-test concluded that the two groups have statistically significant directional differences. The same conclusion resulted between the potentially active faults and the less potentially active faults. However, the F-test concluded that a statistical difference cannot be determined between the strikes of the active faults and those of the less potentially active faults.

DISCUSSION

Geodatabase

The compilation of individual geotechnical investigations provides a centralized database for study of the RCFZ and associated hazards. Prior to creation of this database, the community would have needed to sift through hundreds of ungeoreferenced paper reports to extract data. The ArcMap project allows the user to view the location of various types of data and directly open images of the data logs through hyperlinks in the project. The subsurface data provide detailed information on stratigraphy and faulting that would otherwise be difficult to obtain in the urban environment of downtown San Diego. These data include geologic descriptions of sedimentary units from borings and trenches, geotechnical properties of sedimentary units from CPT logs, and descriptions of structural observations including offset or folded strata, bioturbation, fissures, and soft-sediment deformation structures. Future work based on this geodatabase could include mapping of three-dimensional stratigraphic architecture and modeling of ground acceleration and liquefaction for different earthquake scenarios. Additional layers could also be added for future seismic hazard investigations, such as building zonation or population density. The geodatabase was made publicly available through Weidman et al. (2019).

The work of compiling, scanning, and reviewing reports, the building of the geodatabase, and the initial fault geometry analysis were all conducted by a single student for a Master’s thesis project (Weidman, 2017). The onerous task of compiling reports was made easier by working directly with several consulting firms that provided copies of reports. This had the added benefit of helping to build relationships between the geotechnical and academic communities. Similar geodatabase projects could be useful in other areas of the world where major fault zones trend through densely populated and developed regions. One such area is the San Francisco, California, region where multiple strike-slip faults trend through cities surrounding San Francisco Bay, including the San Andreas and Hayward faults. Fault zones can be highly complex along strike, and the detailed fault investigations that are possible through GIS can improve understanding of fault zone geometry, mechanics, and evolution, which in turn can improve seismic hazard assessments.

Hazard Classification

The geodatabase was used for initial assessment of seismic hazard for the downtown San Diego area and for comparison with existing fault maps and regulatory zones. Two active fault zones, fault zones A and B, were discovered outside the current AP zone in eastern downtown (Figs. 9 and 10A). These sites should be considered when the AP zone is updated, which may result in moving the eastern AP zone boundary to include the active fault traces. The faults are close to the “downtown graben”, and therefore could be considered part of the deformation associated with localized transtension at the graben. Alternatively, the northeasterly trend of fault zone B along with its sense of slip suggest that it could represent the southern terminus of the Florida Canyon fault. If fault zone B is in fact the southern terminus of the Florida Canyon fault, then it implies a series of en echelon graben structures between the Rose Canyon and Florida Canyon fault zones, and that the region of transtension and seismic hazard associated with the San Diego Bay stepover affects a much larger area of San Diego.

The data also revealed areas with a contradiction in fault zone location between the geotechnical report data and the USGS-CGS fault map (Figs. 9 and 10). Some of these contradictions may be related to uncertainty between fault maps, but many differences were greater than our total estimated uncertainty and therefore should be considered when these fault databases are used for hazard assessment and zoning decisions, especially where differences are large, as is the case south of the “downtown graben” (Fig. 10A). The USGS-CGS San Diego fault segment through Old Town may also require a more significant adjustment, as two separate trenches across its trend did not reveal evidence for faulting (Fig. 10B). However, it is also possible that the fault zone is correctly located, but that the age of the most recent event is older than stratigraphy exposed in both trenches (Pleistocene age).

The geodatabase also revealed faulting considered potentially active that is not included in the USGS-CGS database, or the AP zones. For example, there is a roughly north-south–trending series of potentially active fault segments mapped near the waterfront, south of Little Italy, and some other segments are mapped trending southwest between the “downtown graben” and San Diego Bay (Fig. 9). Based on our classification, these faults have been active in the late Quaternary, but not during the Holocene. Although the earthquake risk for these faults is considered lower due to lack of Holocene activity, the faults may still be important for models of hazard or fault zone evolution. Furthermore, recent work has demonstrated that earthquakes may propagate across both active and inactive structures in the same event (Vallage et al., 2016). The strike of the potentially active fault along the waterfront suggests that it may be a northern continuation of the offshore Coronado fault zone, which is included in AP zone regulations (Fig. 9).

Fault Zone Geometry

The analysis of fault strike and dip illustrates that there is much variability in fault geometry along strike in the downtown San Diego section of the RCFZ. This may locally be associated with the San Diego Bay stepover, but highly variable geometry has also been observed elsewhere along the RCFZ, including the offshore extension north of La Jolla (Sahakian et al., 2017). Although the strike of active faults was found to be statistically different than that of potentially active faults, it was found not to be statistically different from less potentially active faults (Fig. 11). If we consider that the less potentially active faults are oldest (longest time since rupture), followed by potentially active faults, and then active faults (most recently ruptured), this pattern could indicate a change in stress that first made rupture on approximately NW-trending faults more likely, followed by a period where more NE-trending faults were active, and then a return to activity on more NW-trending faults. In this scenario, it appears that the less potentially active faults were not reactivated with the active faults following a similar trend, but this may be related to the lower dip angles of the less potentially active faults compared to the more vertical active faults (Fig. 12). It has also been shown that in areas of slip partitioning, a temporal change in the stress field is not needed to explain differences in slip between adjacent faults (Wesnousky and Jones, 1994). Alternatively, the observed fault geometry could be related to spatial patterns associated with regional and local stress or relative location within the San Diego Bay pull-apart basin.

The maximum horizontal stress direction in this region is N20°E–N40°E (Yang and Hauksson, 2013), consistent with strike-slip faulting along a principal displacement zone with roughly similar trend to the active and less potentially active faults, and Riedel shears with roughly similar trend to the potentially active faults (Cloos, 1928; Riedel, 1929; Sylvester, 1988). However, stress fields are known to be highly variable on regional and local scales (Zoback, 1992; Yale, 2003; Montone et al., 2012), and the subsampling of the fault zone shown here is limited to a very small section of the RCFZ, making it difficult to fully assess relationships between fault zone geometry and stress field. Changes in fault activity with fault strike in this area could also be related to the evolution of the San Diego Bay pull-apart basin, and could depend more on variability in fault dip and sense of motion, which can be highly complex at fault stepovers and bends (e.g., McClay and Dooley, 1995; Wakabayashi et al., 2004; Wu et al., 2009). A preliminary assessment of fault dip indicates that a higher percentage of active and potentially active faults are steep (>85°) compared to less potentially active faults, but there is a high degree of variability in the data with few patterns that emerge (Fig. 12). It could also be interpreted that all faults are part of the same active zone with the less potentially active faults and potentially active faults representing evolutionary phases of the system that shut off as the main faults were established. Other recent examinations of fault zone geometry used high-resolution surface mapping technology (e.g., lidar) across greater fault zone extents, and included more information on fault dip, sense of slip, and fault zone width (e.g., Barth et al., 2012; Teran et al., 2015; Vallage et al., 2016; Scott et al., 2018). These studies were also conducted in undeveloped areas, which again highlights the relative difficulty of assessing fault zones through urban areas where much of this information is not available. Expanding the geodatabase beyond downtown San Diego to include information from both geotechnical and scientific investigations would improve interpretations of larger-scale fault zone geometry.

The detailed fault mapping and assessment of fault dip compiled by this study illustrate that the RCFZ exhibits localized normal deformation in downtown San Diego as it approaches the releasing stepover at San Diego Bay, and that the southernmost faults of the RCFZ appear to extend south across the downtown area and into the stepover. Both observations suggest a highly complex geometry of faulting that could influence earthquake rupture patterns and should be considered for seismic hazard assessments. Further investigation into the pull-apart basin evolution is warranted, especially with consideration of the fault segments that step offshore into San Diego Bay.

CONCLUSIONS

  • The compilation of geotechnical data from hundreds of individual reports into a geodatabase allows for more comprehensive study of a highly complex urban fault zone.

  • Geotechnical data from the subsurface beneath urban areas is an effective way to map stratigraphic architecture and fault zone geometry that are otherwise obscured by development at the surface.

  • The Rose Canyon fault zone in San Diego, California, exhibits transtensional deformation as the main fault strands approach a releasing stepover in the downtown area.

ACKNOWLEDGMENTS

We are indebted to the geotechnical firms that contributed reports to this project, especially AECOM, Geocon, Kleinfelder, Leighton and Associates, and Construction Testing & Engineering. This research was funded by the California Geological Survey through student support. We are also grateful for thoughtful reviews by Nicolas Barth and two anonymous reviewers, which have greatly improved this manuscript.

Science Editor: Shanaka de Silva
Associate Editor: Jose M. Hurtado
Gold Open Access: This paper is published under the terms of the CC-BY-NC license.