California earthquake history

This paper presents an overview of the advancement in our knowledge of California’s earthquake history since ~ 1800, and especially during the last 30 years. We first review the basic statewide research on earthquake occurrences that was published from 1928 through 2002, to show how the current catalogs and their levels of completeness have evolved with time. Then we review some of the significant new results in specific regions of California, and some of what remains to be done. Since 1850, 167 potentially damaging earthquakes of M ~ 6 or larger have been identified in California and its border regions, indicating an average rate of 1.1 such events per year. Table I lists the earthquakes of M ~ 6 to 6.5 that were also destructive since 1812 in California and its border regions, indicating an average rate of one such event every ~ 5 years. Many of these occurred before 1932 when epicenters and magnitudes started to be determined routinely using seismographs in California. The number of these early earthquakes is probably incomplete in sparsely populated remote parts of California before ~ 1870. For example, 6 of the 7 pre-1873 events in table I are of M ≥ 7, suggesting that other earthquakes of M 6.5 to 6.9 occurred but were not properly identified, or were not destructive. The epicenters and magnitudes (M) of the pre-instrumental earthquakes were determined from isoseismal maps that were based on the Modified Mercalli Intensity of shaking (MMI) at the communities that reported feeling the earthquakes. The epicenters were estimated to be in the regions of most intense shaking, and values of M were estimated from the extent of the areas shaken at various MMI levels. MMI VII or greater shaking is the threshold of damage to weak buildings. Certain areas in the regions of Los Angeles, San Francisco, and Eureka were each shaken repeatedly at MMI VII or greater at least six times since ~ 1812, as depicted by Toppozada and Branum (2002, fig. 19). Mailing address: Dr. Tousson Toppozada, California Geological Survey, 801 K Street, Sacramento, CA 95814, U.S.A.; e-mail: ttoppoza@consrv.ca.gov

of shaking intensities, and ultimately earthquake epicenters and magnitudes (table I).
The routine seismographic determinations of earthquake epicenters and magnitudes (M) started in southern California in 1932. By comparing the shaking intensities of the modern instrumentally determined earthquakes to those of pre-1932 events, the epicenters and magni-tudes of the pre-instrumental earthquakes could be estimated with some consistency.
The short earthquake history has shown that most regions in California are susceptible at various levels to earthquake damage. This information has been applied to earthquake hazard assessment and mitigation at critical structures such as nuclear power  .  Stover and Coffman (1993). ( * ) The date is given as year, month, day (Greenwich). ( ** ) Proposed epicenter on or near San Andreas Fault .
plants, dams, schools and hospitals. The historical earthquake information is also an important component of the probabilistic seismic hazards assessments that are part of the modern building codes.

Before 1970
The most comprehensive catalog of historical California earthquakes available before 1970 was that of Townley and Allen (1939). This provided mostly narrative descriptions of historically reported felt earthquakes, from the 1769 Spanish exploration of California to 1928. It did not generally provide epicenter coordinates or magnitudes for the earthquakes. The sources of information were mostly Catholic Mission documents before the 1849 Gold Rush, and Newspapers and varied correspondence thereafter.
The Bulletin of the Seismological Society of America (BSSA) started publication in 1910, and has included many articles on various California earthquakes. The BSSA section titled «Seismological Notes» describes the shaking and felt effects of contemporary earthquakes, with some seismographic information generally for post-1932 events.
Since 1928, the annual publication United States Earthquakes (USE) has provided descriptions of felt effects for earthquakes in the United States, and available seismographic information, such as epicenter coordinates and magnitude. Since 1950 USE has included rough isoseismal maps of some of the most widely felt earthquakes.
Special publications were also available for numerous significant events, including the major San Andreas Fault earthquakes of 1857 (Wood, 1955) and 1906 (Lawson, 1908).

The 1970s
Hileman et al. (1973) published the California Institute of Technology (CIT) seismographic catalog of earthquakes from 1932 to 1972 for Southern California, roughly south of the 36th parallel. This catalog lists instrumentally determined epicenter coordinates and magnitudes, and provides annual and multi-year epicenter maps of earthquakes of various magnitudes, generally of M > 3. Online updates of the CIT catalog are available at: http://www.scec dc.scec.org/catalogsearch.html Bolt and Miller (1975) published the University of California, Berkeley (UCB) catalog of earthquakes from 1910 to 1972, for northern California, roughly north of the 36th parallel. Their estimates of pre-1942 earthquake location and size were approximate, and generally not determined from seismographs. These estimates were largely based on Townley and Allen's (1939) descriptive catalog, except for a few seismographic values of epicenter and magnitude (M) quoted from Gutenberg and Richter (1949) and CIT. Starting in 1942, UCB seismographic records were used to determine epicenters and M for generally M > 3 earthquakes. Online updates of the UCB catalog are available at: http://quake.geo.berkeley.edu/nced c/catalog-search.html Toppozada (1975) analyzed the size of the areas shaken by earthquakes in California and Western Nevada at various Modified Mercalli Intensity levels (MMI), and related them to Richter local magnitude. These relations were subsequently used in the 1970s and 1980s to estimate the magnitudes of preinstrumental earthquakes. Toppozada et al. (1978) interpreted the felt effects and estimated epicenters and magnitudes for the most strongly felt California earthquakes occurring from 1900 through 1931. They based their estimates on the felt effects and size of areas shaken at various levels of intensity as reported by Townley and Allen (1939) and USE. Supplemental information came from felt shaking effects described in pre-1932 newspapers, BSSA, and available seismographic information including the UCB Bulletin of the Seismographic Stations. They selected 517 events of probable M~4 or larger based on Toppozada's 1975 relations, having Modified Mercalli Intensity (MMI) of V or larger or felt area ~7000 square km or larger. They determined that at least 229 earthquakes were of M ≥ 4, but the available MMI and felt area data were not sufficient to allow robust estimates of magnitude for the remaining events, that were probably of M~4. Real et al. (1978) combined this pre-instrumental 1900-1931 catalog with a 1932-1975 catalog that they created by merging the CIT and UCB catalogs, and the USGS catalog for Central Coastal California since 1969. Their resulting 1900-1975 catalog was available on computer tape, and on an epicenter wall map of 3600 earthquakes of M~4 or larger, or MMI V or greater if M was unknown. The catalog lists the event dates, epicenter coordinates, and M, and includes 452 events of M ≥ 5. This catalog is essentially pre-instrumental before 1932 in Southern California, and before 1942 in Northern California.

The 1980s
Toppozada et al. (1981) analyzed in detail the felt effects of the most strongly felt California earthquakes between 1769 and 1900. They selected earthquakes from Townley and Allen (1939)  Pre-1932 earthquakes were best known in the areas that had Catholic Missions before the 1840s, and newspaper coverage after 1849. The State epicenter map ( fig. 3) employs three symbol shapes, Xs, circles and triangles, to identify  the three approximate thirds of the 200-year record. It shows that before 1869, earthquakes indicated by Xs were relatively well known in the central and southern coast and in the gold and silver fields northeast of Sacramento, but not in the northwestern or southeastern regions of California. After 1868 when the newspaper coverage expanded to these remote regions their high seismicity became apparent, as indicated in fig. 3 by the circles and triangles in these areas. The San Andreas Fault region between the 36th and 37th parallels was much more active before than after 1932, as illustrated by the Xs and circles greatly outnumbering the triangles in this region ( fig. 3). This will be discussed below under the «Parkfield region». Most of Toppozada et al. (1981)'s data resulted from searching about 12 000 newspaper issues for towns that might have felt the earthquakes, for several issues after each event. About a quarter of the issues searched provided some earthquake descriptions, although many descriptions were duplicated in various newspapers. A bibliography of the newspapers searched for each earthquake was appended to the 1981 report. Also, a summary of the reported felt effects for each earthquake and the corresponding MMI value at each reporting town was appended to the 1981 report. Toppozada and Parke (1982) completed a similar study of the areas damaged by California earthquakes that caused MMI VI to VII or greater effects in California, from 1900 to 1949. This study provided isoseismal maps and summaries of reported effects, which were not available in USE (1928USE ( -1949 or in Toppozada et al. (1978)'s study of pre-1932 events. A newspaper bibliography and a summary of the reported effects were also appended to the 1982 report.

1990 to 2003
Ellsworth (1990) reviewed the history of major California and Nevada earthquakes from 1769 to 1989, and compiled a comprehensive catalog of M ≥ 6 earthquakes. This included the known values of the various magnitude types for each earthquake, derived from MMI, short and long period seismic records, seismic moment, and a summary or preferred M. Stover and Coffman (1993) summarized the information from USE, Toppozada et al. (1981), and other available sources, and provided a catalog and felt descriptions for historical earthquakes of M ≥ 4.5. They also provided isoseismal maps for some of the most significant California events, generally of M ≥ 6.5. Toppozada et al. (2000) compiled a wall map of epicenters of M ≥ 5 earthquakes from 1800 to 1999. This map lists the date, co-ordinates, and magnitude of the 383 events of M ≥ 5.5. An inset map shows the areas damaged at MMI VII or greater by these earthquakes, which is the threshold of damage to weak buildings, and the number of times the various areas have been damaged.
Certain areas in the regions of Los Angeles, San Francisco, and Eureka were shaken repeatedly at MMI VII or greater at least six times each since ~1800. A generalized subset of this wall map is available on the web at: http://www.consrv.ca.gov/cgs/rghm/quakes/MS49.htm  summarized the information on the M ≥ 5.5 historical earthquakes, and the areas that they damaged. For each earthquake they listed all known values of M, and identified a preferred M. They obtained instrumental magnitudes generally for post-1932 events from the earthquake catalogs of CIT and UCB, Ellsworth (1990) for M ≥ 6.0 events, Hutton and Jones (1993), Stein and Hanks (1998), and Bakun (1999Bakun ( , 2000. The preferred magnitude was moment M when available. Otherwise, the order of preference was generally Mg-r (Gutenberg and Richter, from 20-sec surface waves); M s (from surface waves, before 1935 generally from un-damped Milne seismographs according to Abe, 1988); ML (local M of Richter, 1935); Ma (based on the size of the areas shaken at or above MMI V, VI, VII); Mi (from Bakun 1999, 2000, using Toppozada et al., 1981. They also showed the spatial and temporal evolution of the seismicity and the areas damaged at MMI VII or greater from 1800 to 2000.

Updated relations between areas
shaken and M  developed new relations between the moment magnitude of modern California earthquakes of M 5.5 to 7.3 and the areas in square km that they shook at or greater than MMI V, VI, or VII ( fig. 4)   and  used these relations to derive estimates of magnitude of pre-instrumental earthquakes from the areas shaken at these intensities or greater. For example the area shaken at MMI V or greater by the 12 April 1885 earthquake is that enclosed by the innermost two contours in fig. 5. That is the whole area from Visalia to Santa Cruz (E to W) and from Oakdale to Santa Maria (N to S). The values of MMI are determined at the points that described the shaking effects, and the contours are drawn to distinguish between areas having different values of MMI.
The new equations result in magnitudes that are approximately 0.3 to 0.4 units higher than those resulting from Toppozada (1975)'s relations, which were based on local magnitude values and partly on Nevada earthquakes. Ellsworth (1990) independently found that Toppozada (1975)

Selected regions
The earthquake history of selected regions of California has also been studied. For example, Agnew (1991) studied the completeness of the historical earthquake record in Southern California, in light of the distribution of newspapers and other available sources of information. Bakun (1999Bakun ( , 2000 studied the seismicity of the San Francisco Bay region, and of California's North Coast, using the method of Bakun and Wentworth (1997) to determine epicenters and magnitudes for pre-instrumental earthquakes.
Uhrhammer (2003) is currently studying the early seismographic record  in the San Francisco Bay region, using the UCB seismic records and notebooks to revise the early epicenter and M determinations. This should distinctly improve the earthquake record before 1942 when UCB started to routinely determine epicenters and magnitudes using seismic records. Fig. 4. Moment magnitude as a function of the areas, in km 2 , shaken at or greater than MMI V, VI, VII (modified from .

Parkfield region
The Parkfield Earthquake prediction by Bakun and Lindh (1985) followed, and was based largely on the historical earthquake catalog work available at that time. The prediction was based on the observation that M~6 events had occurred quasi-regularly near Parkfield in 1857 (identified by Sieh, 1978), 1881 (identified by Toppozada et al., 1981), 1901, 1922, 1934, and 1966(Bakun and McEvilly, 1979, 1984. This simple model predicted that the next M~6 earthquake would occur before 1993. The predicted earthquake has not yet occurred as of this writing (September 2003).  undertook a detailed study of the earthquake history of the Parkfield region and adjoining segment of the San Andreas fault, based on exhaustive research of newspapers and local diaries. This study showed that the pre-1932 seismicity of Parkfield and the adjoining segment of the San Andreas fault to the north was much greater than previously thought, and has been decreasing steadily from 1857 to the present time. The simple 1985 model of Bakun and Lindh does not include the high pre-1900 San Andreas Fault seismicity within 70 km NW of Parkfield. This seismicity brackets the northern end of the major 1857 San Andreas fault rupture, which is at the largest 1857 epicenter symbol (X) in fig. 6a.
Epicenters of the strong 1885 earthquake of M~6.5 ( fig. 5) and other pre-1932 M~6 earthquakes in this zone are shown in fig. 6a. The epicenters were estimated at the centers of strongest MMI. Relative locations of neighboring earthquakes were determined by comparing their MMI values at the same towns (e.g., Parkfield, Salinas, etc.). The earthquakes in fig. 6a have isoseismal maps or descriptions of MMI in . Their rate of occurrence indicates a decaying earthquake sequence ( fig. 6b). This suggests post-earthquake stress decay at the northern end of the 1857 fault rupture, and helps explain the delay of the predicted earthquake. Figures 6a and 6b show that M ≥ 5.5 earthquakes have occurred in the San Andreas fault zone within 70 km northwest of Parkfield ten times from 1857 to 1932, but only twice after 1932. Ben-Zion et al. (1993) modeled a decreasing stress at the terminus of the 1857 rupture that is generally compatible with that indicated by our observed decrease in earthquake rate. They concluded that the decreasing stress could delay the predicted Parkfield earthquake from 1988 to ~1995.  re-evaluated the major earthquakes that occurred on 8 and 21 December 1812, damaging the Missions from San Juan Capistrano to Purisima Concepcion ( fig. 7), in Orange, Los Angeles, Ventura and Santa Barbara counties. They found that the 1812 San Andreas faulting indicated by Jacoby et al. (1988), extending ~170 km from near Cajon Pass to near Fort Tejon, could have ruptured in two events and generated both earthquakes ( fig. 7). Tree ring studies by Jacoby et al. (1988) placed the first event that damaged mostly San Juan Capistrano and San Gabriel, on the Wrightwood segment that includes Cajon Pass and Pallett Creek.  concluded that the second faulting event 13 days later could have started from the west end of the first, with rupture directed to the west toward Fort Tejon and Purisima Concepcion. It damaged mostly Santa Barbara, Santa Inez and Purisima. Missions San Fernando and San Buenaventura were damaged by both the 8 and 21 December 1812 earthquakes ( fig. 7).

December 1812 Southern California earthquakes
The damage at Purisima Concepcion was stronger than that at Santa Inez. This was apparently due to Mission Purisima's location on sloping ground at the edge of a marsh (Toppozada et al., 1981). Also, Rancho de San Antonio, 17 km north of Purisima, was not damaged (fig.  7). The adverse site effects at Mission Purisima were recognized by the padres and noted in their 1813 written communications. They consequently rebuilt Mission Purisima on flat dry ground, 5 km north of the 1812 ruins.
If the 1812 faulting extends westward to Fort Tejon, or beyond to Mil Potrero as suggested by Fig. 6a,b. a) Epicenters of M ≥ 5.5 earthquakes in the Parkfield-Bitterwater region and surrounding areas. Symbol size is proportional to M and shape indicates time period. The dashed region is that considered in fig. 6b (Modified from . b) Cumulative seismic moment released in the Parkfield-Bitterwater zone of the San Andreas fault from 1857 to 2001 (Modified from .  Davis (1983), it cannot all have ruptured on 8 December. Faulting that extends that far westward would damage Santa Barbara, which was not damaged on December 8th but on the 21st. Thus the total 1812 faulting suggested by Jacoby et al. (1988) and depicted in fig. 7 would suggest that the western segment ruptured on the 21st. Definition of the western extent of the total 1812 faulting is complicated by the major 1857 earthquake rupture that extended from near Parkfield to Cajon Pass, and re-ruptured the 1812 faulting after only 44 years. The conventional Santa Barbara Channel epicenter of the 21 December earthquake was based largely on a tsunami that resulted from the event. However, it is becoming clear that tsunamis are often generated by submarine landslides. Lander et al. (1993) interpreted the description of the 1812 sea wave as «a submarine landslide-generated wave and not one generated by a (submarine) fault movement source». Such landslides are mapped frequently in the Santa Barbara Channel (Greene and Kennedy, 1989), and could have been triggered by the 1812 earthquake shaking (Greene et al., 2000).

1830s San Francisco Bay area earthquakes
The destructive 1989 Loma Prieta earthquake provided the impetus for Toppozada et al. (1996) to study the San Francisco bay area earthquake history in detail. This lead to a bet-ter understanding of the major 1838 San Andreas fault event and to a radical re-evaluation of the 1836 «Hayward Fault earthquake» by Toppozada and Borchardt (1998).
They concluded that the 1838 earthquake was of M~7.4, significantly larger than previously thought. The 1906 San Francisco earthquake of M 7.8 occurred 68 years later as a result of major San Andreas faulting that overlapped and reruptured the 1838 faulting. This pair of events is similar to the 1857 Central San Andreas Fault earthquake of M 7.9 whose major faulting overlapped and reruptured the faulting associated with the1812 earthquakes of M~7.3 and ~7.1.
Thus the 1857 and 1906 M 7.8 and 7.9 San Andreas faulting events have each overlapped a M~7.4 faulting event that occurred 44 to 68 years earlier. Earthquake ruptures that occur 44 to 68 years apart are difficult to differentiate in fault trenches because of the limitations in stratigraphic resolution and paleoseismic dating. This suggests that overlapping ruptures occurring a few decades apart can be more common than paleoseismology would indicate. Toppozada and Borchardt (1998) also found that the 1836 event was not a M~7 event on the Hayward fault as was previously believed, but rather a M~6 1 /2 event ~100 km to the south, in the general vicinity of Monterey Bay. The idea of an 1836 Hayward Fault earthquake had been based on successive false assumptions and misinterpretations of information, which Toppozada and Borchardt summarized in their table 2 : 'Evolution of the Myth concerning the 1836 «Hayward Fault earthquake»'.

Current work
We are building a database of the felt effects of M > 5 historical earthquakes. For each earthquake there are various towns that reported felt effects. The database will have a separate entry for each earthquake and each town that reported the event. We have digitized the data previously analyzed by Toppozada et al. (1981) and Toppozada and Parke (1982). We will supplement this with unpublished data that we have analyzed, and also add information for post-1950 events from USE. The database will be searchable for event dates, towns, geographic regions, and levels of MMI shaking.
This will allow us to determine how many times specific towns, as well as regional areas, have felt different levels of MMI shaking. The database will be linked to digital isoseismal maps of the corresponding earthquakes. These maps and data will be published on a website to allow comparing the felt effects of current earthquakes to those of neighboring historical events. This will permit a rapid assessment of the potential MMI shaking and damage from a current strong earthquake from the shaking of neighboring historical events of similar magnitude. It will also allow recalibrating the epicenters and magnitudes of preinstrumental earthquakes by comparing their intensities of shaking to those of nearby well-instrumented modern earthquakes.