The Mendocino Triple Junction is the most seismically active region in Northern California, with a variety of unusual seismic events in addition to regular earthquakes. It also represents the southernmost part of the Cascadia subduction zone (CSZ) where potential damaging thrust earthquakes can occur. The current real-time earthquake monitoring is a cascade-type procedure following a joint effort of the USGS Menlo Park and the Berkeley Seismological Laboratory at UC Berkeley. For offshore earthquakes occurring outside of the seismic network, as is the case in the Mendocino region, such a procedure can generate errors in the event detections and locations resulting in incorrect source determinations. With the goal of more efficiently monitoring the offshore region of Northern California, particularly for slow/low-stress-drop and large, possibly tsunamigenic earthquakes, we develop an automatic scanning of continuous long-period ( 20 sec) broadband seismic records following the method proposed by Kawakatsu (1998) and currently in use in Japan (Tsuruoka et al., 2009). In addition, we are proposing an improved algorithm for great earthquakes occurring along the CSZ, that, if implemented in real-time with a continuous scanning algorithm, will provide information that could be utilized for near-source tsunami early warning.
We are implementing a continuous seismic scanning algorithm following the method proposed by Kawakatsu (1998) that makes use of continuous long-period filtered data. By using this approach, it is possible to calculate moment tensors every two seconds for each point of a spatial grid. We generated a grid of nearly 5,000 virtual sources, with each point being separated by 0.2 degrees in latitude (between 40N and 43N) and longitude (between -123E and -128E) and a 3 km interval in depth (5 to 38 km). We selected four Berkeley broadband stations (HUMO, ORV, WDC, and YBH) for which we computed a catalog of Green's functions using a 1D velocity model of Northern California. In order to detect all earthquakes of small and large magnitudes within the region, we are implementing two parallel-running systems. The first one is an inversion of 380 seconds of data filtered between 20 and 50 sec to detect earthquakes with magnitudes between 3.5 and 7. The second system considers longer records (480 sec) filtered at longer periods (100 and 200 sec) to account for M8+ earthquakes.
The CSZ that marks the subduction of the Pacific Plate beneath the North American Plate can produce very large and tsunamigenic earthquakes - magnitude 9.0 or greater - if rupture occurs over the entire area. The last known great earthquake in the region was in January of 1700, and geological evidence indicates that great earthquakes may have a return time of 300 to 600 years. Because there is no available seismic data for a magnitude 8+ earthquake in the study region, we performed a series of synthetic tests for such large events, defined with uniform and variable slip models along the subduction zone. Figure 2.40 illustrates the need to look for longer period data for a M8+ earthquake. Indeed we find that inversions using the 20-50 sec passband fail to recover the seismic moment and location of the tested event. The moment is underestimated, yielding only a M6.7 for a M8.2 synthetic earthquake, and our best solution is located onshore (Figure 2.41a). The narrow band processing and the point-source synthetic only fit a small portion of the record, and the inversion is not sensitive to the total moment of the event.
However, for the 100-200 sec passband, the inversion yields a point-source location near the fault centroid (Figure 2.41), indicating that this passband works well in identifying the earthquake magnitude, mechanism and location. Furthermore, for the heterogeneous slip models, the detections show better variance reductions (VR) due to the concentration of slip that is better represented by a point-source assumption (Figure 2.41b).
For great earthquakes, the proximity of the seismic stations to the rupture segment leads to the problem of a near-field component, for which the single point-source assumption considered in the previously described method could fail. To account for the finiteness of the near-field rupture, we are testing the algorithm with multi-point sources obtained after summing together single point sources accounting for the respective distances and azimuths between source and receiver for each of the sources. By doing this, it is possible to study any type of rupture and directivity: north to south, south to north, and bilateral, depending on the grid points considered.
Figure 2.42 demonstrates that the multi-point source assumption (here a summation of three points of the grid) can deliver a better source solution of a large earthquake located off the coast of Northern California than a single point source. Indeed the VR increases from 66 (Figure 2.42b) to 75 (Figure 2.42d).
Such scanning will provide complete information on the events in real-time using a single stage of processing, and for this reason it will be faster than the current procedures. The method that we are implementing makes use of regional seismic recording stations with continuous real-time telemetry that will enable autonomous detection, location, estimation of scalar seismic moment, and determination of the mode of faulting (i.e. dip-slip versus strike-slip) within approximately 8 minutes of the earthquake occurrence and before the damaging tsunami waves reach the coastline. In the future, our efforts toward a real-time source parameter reporting system for great earthquakes may aid in the development of a tsunami early warning system in Northern California.
Kawakatsu, H., On the realtime monitoring of the long-period seismic wavefield, Bull. Earth. Res. Inst., 73, 267-274, 1998.
Tsuruoka, H. H. Kawakatsu, and T. Urabe, GRID MT (Grid-based Realtime Determination of Moment Tensors) monitoring the long-period seismic wavefield, Phys. Earth Planet. Int., Special issue: Earthquakes in subduction zones: A multidisciplinary approach, 175, 8-16, 2009.
Berkeley Seismological Laboratory
215 McCone Hall, UC Berkeley, Berkeley, CA 94720-4760
Questions or comments? Send e-mail: email@example.com
© 2007, The Regents of the University of California