Earthquake early warning in northern California

Richard M. Allen and Gilead Wurman
University of California Berkeley

SCEC Annual Meeting, September 2006

The ElarmS algorithms for earthquake early warning have been developed and implemented for offline testing in northern California. Implementation in northern California presents specific challenges that require improvements to the ElarmS methodology used in southern California. Broadband velocity and strong-motion accelerometer data must be integrated from two networks, the Berkeley Digital Seismic Network and the Northern California Seismic Network. Improvements to the magnitude estimation algorithms include utilization of the predominant period information as well as the peak velocity and/or displacement amplitude of P-wave arrivals. Using both these parameters significantly improves the magnitude estimates.

Since February 2006, ElarmS has been running automatically in a non-interactive fashion 10 minutes after all M > 3 earthquakes. The 65 events that have been processed to date range in magnitude from 2.9 to 4.7. The first magnitude estimate for each event is available when 1 sec of the P-wave at the closest station is available. The uncertainty in the estimate is large at this point in time, the standard deviation of the magnitude estimate errors is 0.8 magnitude units, but it also provides the greatest warning time. We define the "alarm time" as when 4 sec of P-wave data are available from 4 stations. This is a somewhat arbitrary choice but is intended to represent when it might be appropriate to issue a warning. At the alarm time the standard deviation of the magnitude estimate errors is 0.5 magnitude units.

The two largest events to occur during the period of automatic processing were the M 4.7 Gilroy event (June 15, 2006) located near the southern Calaveras fault south of the Bay Area, and the M 4.7 Santa Rosa event (August 2, 2006) near the Rodgers-Creek fault north of the Bay Area. The location of these events is similar to likely hazardous earthquake scenarios; they therefore provide insight to the future performance of ElarmS in damaging events. We will present the warning output from these events.

© Richard M Allen