For this scenario, we selected a magnitude 6.7 earthquake on the southern Hayward fault. This event should fall between the "occasional" and "rare" events used by Mary Comerio and her co-authors in The Economic Benefits of a Disaster Resistant University.

An earthquake of this size would cause damage throughout the San Francisco Bay Area. For comparison, the 1989 Loma Prieta earthquake was an M6.9. According to the new earthquake probabilities, there is 70% probability of one or more earthquakes of M6.7 or greater from 2000 to 2030 in the San Francisco Bay Area.

This segment of the Hayward fault last ruptured in the 1868 earthquake. The scenario event initiates near Fremont and ruptures toward the UC Berkeley. However, this scenario is not a repeat of the 1868 event. Instead, we used a fault length of 38 km with ~1 m of slip to produce a M6.7 event. 1 m of slip is approximately the amount which has acculmulated on this segment of the fault since the 1868 earthquake (132 years x 9 mm/year = 1.2 m).

Although this event focuses strong ground shaking toward UC Berkeley (since it is rupturing toward the campus), the fault break does not go through the campus and thus does not provide the hazard of surface rupture through Cal Memorial Stadium and other fault-crossing structures.

The aftershock probability was calculated by Dr. Lind Gee using the software of Dr. Paul Reasenberg of the USGS. This is the same software used to generate reports after earthquakes in northern California, and is based on the statistics of California earthquakes.

The ground motion maps were produced by Dr. Asya Kaverina and Prof. Doug Dreger. This methodology predicts ground motions, based on information about the earthquake rupture. Using the model of the finite rupture above, they generated synthetic seismograms on a fine grid of points assuming a simple model of Bay Area structure. The synthetic seismograms are then processed to estimate peak ground acceleration (PGA) and velocity (PGV) at each grid point and the results are contoured. This methodology can be combined with tools for estimating fault parameters automatically to produce maps of strong-ground shaking emergency response. It was applied to the 1999 Hector Mine earthquake.

In addition, Prof. Dreger ran a finite-difference simulation of this earthquake using his model of 3-dimensional structure in the San Francisco Bay Area. This computation accounts for the complexities of focusing and defocusing due to variations in geology (such as in San Pablo bay). His computations show ground motions higher than the motions estimated from the simplified model used above. Click here to see a movie of the ground motions generated by this earthquake.

If this had been a real event, the UC Berkeley/USGS joint earthquake notification system would provide information about the earthquake's location, magnitude, and extent of strong-ground shaking. At UC Berkeley, earthquake processing is handled by the Rapid Earthquake Data Integration (REDI) system, which provides rapid access to earthquake information. Typically, a preliminary location is available within 30 seconds after an event. This is followed by a revised location within 2-4 minutes with a preliminary magnitude. This is followed by a revised magnitude within a minute. This figure illustrates the typical flow of information following an earthquake.

Given this scenario earthquake and the location of UC Berkeley monitoring equipment, would it have been possible to warn of the impending event? "Warning" or "alerting" is that situation where sensors monitoring a fault detect an earthquake and recognize that it will be damaging. Since seismic waves travel more slowly than the electromagnetic waves, it may be possible to transmit a message warning of the impending ground shaking in some cases. Mexico City has an early warning system for earthquakes in the Guerrero gap, over 250 km away. For these events, the inhabitants of Mexico City may have up to 70 seconds to prepare for strong shaking.

Here in the San Francisco Bay Area, we live right on top of the faults, which significantly complicates the problem. In addition, the density of instrumentation is not sufficient at the present time to realistically solve this problem. Consider the case of this scenario event, where the nearest UCB-operated seismic site is 24 km away. In order to produce a warning, the P waves must first propagate from the earthquake source to WENL (~4 s). That leaves 11 sec before the S-wave will arrive at UC Berkeley. Allowing for a few seconds to transmit and process the data, ~7 sec might be available for warning. This amount of time is too short for a human response, but MIGHT be long enough for automated processes, such as stopping elevators and opening fire house doors.

The best case for this scenario would be if seismic instrumentation were located at MONB, currently a continuous GPS site. This station is located 5 km away from the epicenter and might provide ~10 sec of warning. In the case of this scenario, it would have been possible for us to install temporary equipment at MONB and use radio to transmit the signals to UC Berkeley. However, this is the best case. If a large aftershock initiates closer to the campus (say, at the termination of the rupture), there will be only a second or two between the strong shaking starts.