March/April 2006

A Look Back At 1906—Perspectives On Great Earthquakes and Post-Earthquake Investigations

At 5:12 am on April 18, 1906, a fracture formed some 10 km below the Earth’s surface offshore of the Golden Gate. It propagated upward and outward, reaching the surface and extending north and south. To the south it broke along the San Francisco peninsula, climbed through the Santa Cruz Mountains, and stopped abruptly at San Juan Bautista. To the north it hugged the coast in Marin, Sonoma, and Mendocino counties, passed into the Pacific Ocean north of Pt. Arena, and came to halt on land in Humboldt County. Approximately one and one-half minutes after it began, the rupture stopped. Fourhundred- seventy km of the Pacific— North American plate boundary had shifted an average of 4 ½ m along the San Andreas Fault, and the M 7.9 San Francisco earthquake had occurred.

This year, 2006, is the centennial of the 1906 San Francisco earthquake. This anniversary has prompted a broad range of earthquake-centered scientific, historical, cultural, and political activities. Scientifically it has refocused attention on the ground motions and the regional distribution of shaking intensity that the 1906 earthquake generated, the potential effects of its repeat on the San Francisco Bay Area and northern California, and its place in the development of earthquake science. In July 2005, I was in northern Mongolia for an international field trip celebrating the 100th anniversary of another great earthquake, the 1905 rupture of the Bulnay Fault. I admit to a fascination for these great on-land surface-faulting earthquakes, having had the good fortune to work not only on the 1906 and 1857 San Andreas ruptures but also on other great fault ruptures in some of the most remote and beautiful parts of the planet. An on-land strike-slip earthquake the size of 1906 is not an everyday occurrence, and diverse messages are to be taken away from it. Many of these already have appeared in a host of books, articles, and television documentaries, and more will be presented and discussed at the 100th Anniversary Earthquake Conference convened jointly by the Seismological Society of America, Earthquake Engineering Research Institute, and Disaster Resistant California (California Office of Emergency Services). For me, one of the most compelling messages is the incalculable importance of post-earthquake investigations— both for what they provide at the moment and the legacy they leave for the future.

The 1906 Rupture and Its Place in the Firmament of Faults

Faults come in a broad range of lengths, dips, and widths, and they generate a similarly broad range of earthquake magnitudes. Large, great, and mega-magnitude earthquakes are no strangers to subduction zones, as evidenced in Chile (1960), Alaska (1964), and most recently and tragically in Sumatra (2004). Moderate and large magnitude earthquakes occur routinely on faults around the globe—take a daily look at the Internet or subscribe to one of the global earthquake pager services. However, great earthquakes on faults in continental crust on land have been relatively infrequent, yet these events are extremely important for what they reveal about earth fundamentals. They are the major products of the cycle of strain accumulation and release. Their occurrence provides insight into how plates move and continents deform, the physics of the earthquake generation process, controls of strong ground motion, and a wide range of hazards such as surface faulting, landsliding, liquefaction, and lateral spreading.

To put 1906 in perspective, there have been only six other on-land strike-slip earthquakes in continental crust since 1900 with similar moment magnitudes (M 7.9±) and rupture lengths (> 250 km). These occurred on the Bulnay Fault, Mongolia, 1905; Haiyuan Fault, China, 1920; North Anatolia Fault, Turkey, 1939; Bogd Fault (Gobi-Altay earthquake), Mongolia, 1957; Kunlun Fault, Tibet, 2001; and most recently on the Denali Fault, Alaska, 2002. (The 1989 Macquarie Ridge earthquake qualifies for this group but was underwater.) There have been other large strike-slip earthquakes during this period. Ruptures of the Fu Yun Fault, China, 1931; Fagnano Fault, Tierra del Fuego, 1949; Fairweather Fault, Alaska, 1958; Motagua Fault, Guatemala, 1976; and North Anatolia Fault, Turkey, 1999, are some examples. These have been important and in many cases severely damaging events, but they do not have the magnitude and dimensions of the “on-land seven.”

Of these seven, 1906 was the longest, 1905 the largest magnitude, and 1957 the most complex in terms of surface rupture distribution. At an estimated 470 km, 1906 was slightly longer than its next closest competitor, the 2001 Kunlun event. Its length qualifies it as the longest historical on-land strike-slip surface rupture, and its width of about 10 km makes it oddly skinny for its length. On the morning of July 23, 1905, the surface of the northern Mongolian steppes ruptured for 350 km along the Bulnay Fault in what was ultimately to become the largest magnitude on-land strike-slip earthquake of the 20th century; it averaged 8 m of left-slip with maximum offsets of 11 m, and it set off seismographs around the world. For sheer complexity, the December 4, 1957, Gobi-Altay earthquake rupture is the winner, with 250 km of strike-slip rupture of the Bogd Fault accompanied by rupture of the Dalan Turu, Gurvan-Bulag, and Toromhon thrust faults.

The Report of the State Earthquake Investigation Commission— The California Earthquake of 1906

Investigations of these great surface-faulting earthquakes, particularly observations recorded at the time of occurrence or shortly after, often have had an immediate scientific and social impact. But even when not fully appreciated at the moment, these post-earthquake observations can age like a fine wine, waiting to be poured at some later date in conjunction with new ideas, theories, and analytical techniques. Following the 1906 earthquake, a major effort was undertaken under the direction of Berkeley’s Professor Andrew C. Lawson to collect information on essentially everything that occurred during the earthquake. These observations were published in two volumes and an atlas by the Carnegie Institution of Washington as the Report of the State Earthquake Investigation Commission—The California Earthquake of 1906 (hereafter referred to as the 1906 report or, simply, the Report).

Volume I appeared in 1908 along with the atlas. It thoroughly documents the location and surface extent of faulting; provides geodetic measurements (pre- and post-earthquake triangulation surveys); and contains an amazingly thorough description of damage and intensity effects throughout California (almost half the volume). Volume 2, authored by Henry Fielding Reid, was published in 1910. In a section titled “On Mass Movements in Tectonic Earthquakes,” it presents Reid’s elastic rebound theory, which has become a physical cornerstone for earthquake recurrence and is the basis for present- day concepts of time-dependent earthquake behavior. Part II of this volume also contains a description of the instrumental records of the earthquake obtained from more than 90 seismographs at different locations worldwide. The accompanying atlas of maps and seismograms is both a scientific and artistic gem. In addition to photographic and line-drawing reproductions of the seismograms, most printed at their original size, the atlas contains a fascinating selection of maps. You’ll find, among others, a geological map of San Francisco; a street map highlighting the “burnt” area of the city; a detailed map of Santa Rosa, California (the second most-heavily damaged city in 1906), showing the distribution of buildings destroyed by the earthquake and those burned in the subsequent fire; a map of the Coast Range region with the distribution of earth movement in 1906 based on the 1906—1907 triangulation resurvey; and a map of the distribution of apparent intensity in California and Nevada. For active fault aficionados, the maps are also a wonderful snapshot of what was known about fault locations early in the 20th century. There is a series of topographic quadrangle maps with the location of the San Andreas rift zone and the active fault trace extending south of the 1906 rupture into southern California as far as San Gorgonio Pass. And the geomorphic map of California and Nevada (with portions of Oregon and Idaho, Figure 1 of the atlas) captures the then-known distribution of faults across this region. In the Bay Area, the Hayward, northern Calaveras, southern Green Valley, and part of the San Gregorio faults had been located, while the central Calaveras, Rodgers Creek, Greenville, and Concord faults had yet to be identified. In southern California, the San Jacinto, Elsinore, Sierra Madre, and Garlock faults were known, but the southern San Andreas Fault between San Gorgonio Pass and the Salton Sea had yet to be placed on the map.

The Search for Coseismic Slip

In 1969 the Carnegie Institution reprinted the Report, making it widely available for the first time. The 1989 Loma Prieta earthquake, the investigations that followed, and the resulting focus on earthquake hazards in northern California renewed broad interest in the Report and its contents. I certainly had perused and been impressed by the reprint, but the occurrence of Loma Prieta sent me and others back to the Report to carefully compare what occurred in 1989—faulting, liquefaction, damage to structures—with what was observed and reported in 1906. It was obvious from our fieldwork in the Santa Cruz Mountains that the main San Andreas Fault had not been involved in the 1989 earthquake. As we delved into the Report to see what had happened there in 1906, it became clear that G. A. Waring, a graduate student who was responsible for mapping this section of the rupture under the supervision of Stanford professor and Investigation Commission member J. C. Branner, had somehow wandered off the main fault trace and followed other features, mainly shaking-related slope failures and fractures, south through the Santa Cruz Mountains. With the descriptions and photographs in the Report, USGS geologist Carol Prentice and I were able to trace Waring’s route and show that some 40 km of the San Andreas Fault had actually not been mapped in 1906.

One of the most important observations from large surface ruptures is the distribution of coseismic slip. We are learning that locations of changes in the amount of slip along strike may indicate the presence of asperities that affect the generation of high-frequency ground motion, as observed in the 2002 Denali Fault earthquake. Variations in coseismic slip distribution, coupled with information on fault slip rate, provide insights into how a fault may fail in the future. And knowledge of the amount of slip in a historical event can be compared to paleooffsets at the same location to evaluate similarities or differences in past fault rupture behavior. For a geologist, however, it is a sad admission that information on coseismic surface offset during 1906 is not one of the strengths of the Report. Of the total rupture length of 470 km, approximately 150 km occurred north of Pt. Arena, with most of this offshore, and another 95 km is hidden beneath the waters of the Golden Gate, Tomales and Bodega bays, and the Pacific Ocean between Bodega Bay and Fort Ross. Of the 225 km of on-land faulting south of Pt. Arena, there is the 40-km gap in the Santa Cruz Mountains. In effect, only about 40 percent of the total 1906 rupture length was actually traversed after the earthquake.

A review of the Report shows there are 36 locations, primarily fences and roads, where the amount of 1906 surface offset is reported or can be estimated; of these, 20 are reliable measurements. In contrast, 132 piercing points were measured along 290 km of the 2002 Denali- Totschunda strike-slip rupture to define its surface slip distribution, more than 200 offset measurements define slip along 450 km of the 2001 Kunlun Fault rupture, and 281 piercing points were measured on the 108 km-long 1999 Izmit earthquake rupture on the North Anatolia Fault. Appreciating that the 1906 coseismic slip distribution could never be adequately obtained from the surface geology, USGS geophysicists Wayne Thatcher and Michael Lisowski went back to the pre- and post- 1906 triangulation data contained in the Report. In 1987 they inverted geodetic data to model 1906 slip south of the Golden Gate, showing general agreement with surface offsets on the northern San Francisco peninsula but distinctly larger geodetic values (2.5 m) on the southern peninsula and in the Santa Cruz Mountains than were reported in 1906 (1.5 m at Wright’s tunnel in the northern Santa Cruz Mountains). Following Loma Prieta, and again using 1906—1907 and more recent data, they modeled the entire 1906 slip distribution. Interestingly, this analysis shows that the largest offsets in 1906 actually occurred along the submarine section of the fault north of Pt. Arena and in the remote sections of Humboldt County.

The amount of coseismic surface slip reported in 1906 and the recent modeling of it from geodetic data also played a role in the evolution of California earthquake forecasts. Following Loma Prieta, the Working Group on California Earthquake Probabilities, which had released the first consensus 30-year earthquake probability estimates (earthquake forecasts) for the San Andreas fault system only one year earlier, was convened again to reevaluate Bay Area probabilities in light of the Loma Prieta event. At that time, these time-dependent probabilities were calculated using the time-predictable model, which required knowledge of the amount of slip in the prior earthquake, in this case 1906, as a basis for estimating time to the next earthquake. A contentious debate quickly arose over which slip values to use for the San Francisco peninsula—the 1906 surface geologic offsets in the Report or the geodetic slip values. The choice significantly affected the probabilities. I was a member of the 1990 Working Group and remember that the disagreement about this was considerable, though it eventually led to a positive outcome. It resulted in alternative slip and segmentation models being proposed for the Peninsula segment of the San Andreas Fault. This was the debut of the logic tree approach, which was already being employed by the seismic hazard industry and was to later provide the calculation framework for the Bay Area’s 2003 Earthquake Probability Working Group report and official California earthquake probability estimates.

Post-earthquake Investigations— Yesterday, Today, and Tomorrow

In the introduction to the 1969 reprint, geophysicist William W. Rubey provides an excellent discussion of why the Report was reprinted and places it in context with reports on other major earthquakes that relied on extensive field observation. These include John Mitchell’s 1761 report on the 1755 Lisbon earthquake, Robert Mallet’s study of the devastating 1857 Neapolitan earthquake published in 1862, C. Dutton’s 1889 report of the 1886 Charleston earthquake, and R. D. Oldham’s 1899 report of the 1897 Assam, India, earthquake. Each of these post-earthquake investigations helped to advance earthquake science and contained primary observations for future investigators. One important pre-1906 earthquake not mentioned by Rubey, likely because no single report was prepared, is the disastrous 1891 Nobi event in central Japan. Rupture of the causative strike-slip Neodani Fault was identified at the time, mapped, and subsequently reported by the great Japanese seismologist J. Omori in 1894. Investigation of the Nobi earthquake marks an early modern recognition that earthquakes are produced by slip on faults rather than the view that faulting is a secondary effect of shaking (for more on this topic, see the article by Max Suter on the 1887 Sonora, Mexico, surface rupture and earthquake in this issue of SRL). Interestingly, Omori contributed a short note to the Report in which he briefly described the 1891 event and compared aspects of its fault slip to the 1906 San Andreas rupture. The Nobi earthquake had the additional impact of leading to the formation of the Imperial Earthquake Investigation Committee in 1892, which guided earthquake research in Japan until the Earthquake Research Institute was established in 1923.

Mallet’s study of the 1857 Neapolitan earthquake, subtitled “The First Principles of Observational Seismology,” is worth additional mention. Beautifully reprinted in 1987 by the Istituto Nazionale di Geofisica, Rome, Mallet’s report is widely available. Mallet spent several months in the southern Apennine Mountains documenting structural damage and ground deformation effects of the earthquake with drawings, maps, and photographs that included stereo pairs. (This was an early use of this photographic technique, and the reprint includes an old-fashioned stereoscope for viewing; this technique would later be used to extensively record 1906 effects.) Based on the distribution of damage, Mallet produced an isoseismal map for the earthquake and used the directions in which objects fell to calculate a focal depth (5 to 8 km) and an approximate epicenter. Interestingly, the source for the 1857 earthquake remained unidentified well into the 1990s. In 1997, Lucilla Benedetti, a geologist working on her doctorate in Paris, used Mallet’s map to geographically focus the location of a SPOT satellite imagery analysis and identified what appeared to be a very recent normal fault scarp in the Val di Agri. With additional fieldwork she concluded that this fault was the likely source of the 1857 earthquake. In 1999, I visited this fault scarp; it still exhibits a retreating free-face, which testifies to its recent formation. Had it not been for Mallet’s work, it is quite probable that the source of this major earthquake would still be unrecognized.

As we push into the 21st century, GPS, synthetic aperture radar interferometry (InSAR), improved inversions of teleseismic waves, analysis of LIDAR and high resolution digital photographs taken of faults before and after large earthquakes, other remote approaches yet to be developed, and always-increasing computer capabilities will play a growing role in post-earthquake investigations. This is to the good, and these methods have the further advantage of being able to be performed from the comfort and security of an office or laboratory. Yet the importance of researchers venturing into the field to collect detailed observations cannot be overstated. Fieldwork provides the record of much that is perishable, as well as the fundamental data required to constrain and test interpretations—offset, rupture length—based solely on remotely recorded information.

But field investigations of large earthquakes are not always easy, particularly when the events occur in remote and physically challenging areas. From the comfort and warmth of a large Mongolian tent in 2005, watching PowerPoint presentations from generator-run laptops and computer projectors, it was impossible not to marvel at the effort of A.V. Vosnesenskii. In the fall of 1905, this Russian seismologist, regarded as the first scientist to observe an active fault rupture in the Earth’s crust in Central Asia, rode on horseback from Irkutsk, Siberia, essentially alone, to map and photograph the surface effects of the Bulnay earthquake. If not for this, the length of this great surface rupture, and the distribution of liquefaction features along it, would be undocumented and a guess at best.

The response to the 1957 Gobi-Altay earthquake was a joint Russian-Mongolian expedition that produced a classic post-earthquake report. Edited by N. A. Florensov and V. P. Solonenko, the expedition’s chief scientists, it documents the mapping of the complex 1957 surface faulting during the Mongolian winter between December 1957 and February 1958. A December aerial survey carried out in the most severe winter weather over completely unknown air routes and with no ground service classified the earthquake as a “world earthquake,” which led the Soviet Academy of Sciences to fund a follow- up expedition between June and November 1958. But even this later work was difficult in a roadless region with incessant wind and abruptly changing weather. The report notes that in September 1958, the expedition’s base camp on the southwest shore of Lake Orog-Nuur (a site where I camped in 1995 and 1996), was twice destroyed by storms. Florensov and Solonenko go out of their way to cite the “dedication and courage of the expedition members while examining the earthquake damages.”

Difficult post-earthquake field conditions have not been relegated to the past. The mapping of the 2001 Kunlun rupture by Chinese and French scientists posed a similarly severe physical test, with travel possible only when the ground was frozen (and the temperature commensurately low) and altitude sickness affecting many of the scientists working along the fault at elevations over 5000 m. Immediately after the 2002 Denali earthquake, geologists had a five-day window in early November to helicopter in subzero temperatures during limited daylight hours to begin initial mapping of the surface faulting. This allowed critical measurements to be made of offsets through glacial ice, which covered 30 percent of the rupture. The amount of dip-slip on the Susitna Glacier thrust could actually be measured along the planar underside of the hanging wall of the fault, which hung frozen in time and space above the glacier’s surface, and crevasses offset 4 and more meters gave slip data where the Denali Fault traversed the Black Rapids and Canwell glaciers. When I traveled to Alaska in July 2003 to complete the slip distribution mapping, many of the glacial offsets looked like melting ice in a summer drink and could not be measured, and by 2004 they were gone.

Post-earthquake investigations require resources. I don’t know how much the 1906 investigation cost in 2006 dollars (it would be interesting to calculate this), but I am concerned about funding to properly conduct similar studies in the future. The Denali Fault earthquake is a case in point. The response should have been a no-brainer: It was the largest on-land earthquake in the United States since 1906 and a proxy for what the southern San Andreas Fault will do during its next great California earthquake. Furthermore, the surface rupture passed beneath one of the United States’s most critical facilities—the Trans Alaska Pipeline, which carries 17 percent of the country’s crude oil. Yet raising funds to study this earthquake has been difficult. Perhaps it’s because it wasn’t urban, or no one died, or the pipeline did not break. Without the support of the Alaska Pipeline Service Company (Alyeska), which wanted fault information to reset the pipeline for possible future fault slip; or Lloyd Cluff, head of geosciences at the Pacific Gas and Electric Company, who had an engineering interest in pipeline response to shaking for gas lines in California as well as a personal interest (he led the 1973 fault study that developed the design inputs for the pipeline’s active fault crossings), we would have been left with the initial reconnaissance and a stack of aerial photographs. While we have been able to map the coseismic slip distribution and begin paleoseismic studies, the extensive landsliding and the liquefaction and lateral spreading the earthquake caused have not received the attention they require; they will most assuredly remain poorly studied with information forever lost as features are absorbed by the tundra and dismembered by the advancing glaciers.

Great continental earthquakes such as 1906 and the others noted here may be infrequent, but their post-earthquake investigations, as the relatively brief history of our science has shown, have been invaluable. As a community we must maintain an explicit commitment to conduct post-earthquake studies as thoroughly as possible and to invest the resources to do so, because of the knowledge and understanding gained about fundamental earth processes, the application of this knowledge to mitigating the effects of future earthquakes everywhere, and the data they provide for the next generation of earthquake scientists. While I have emphasized surface faulting and deformation in this discussion (I am, after all, a geologist), these comments apply equally to the full spectrum of post-earthquake investigations— aftershocks, GPS, strong motion, geotechnical, damage to structures, and social and economic impacts.

Andrew Lawson said it well in the preface to the 1906 report: “The effort to condense the record as far as possible has been tempered by the desire to omit no significant fact, so that the record may be as complete as possible for purposes of comparison with similar events which may occur in years to come.

David P. Schwartz
Menlo Park, CA

To send a letter to the editor regarding this opinion or to write your own opinion, contact the SRL editor by e-mail.




Posted: 23 March 2006