Site Map/Alternate Navigation Skip to Main Content

125 Years of Sound Science, serving society

Rapidly Determining the size and type of Faulting for Potentially Tsunamogenic Earthquakes

Few Californians know that the region just off Cape Mendocino is the most seismically active region of California. Called the Mendocino Triple Junction, this is a place where three tectonic plates meet. North of Cape Mendocino lies the Cascadia Subduction Zone, capable of generating megaquakes far larger than the magnitude 8's possible on the San Andreas Fault. Subduction zone quakes from Alaska, Japan and, potentially, Cascadia, make the coast of Northern California the most tsunami-prone area of the continental US.

Map showing plate boundaries surrounding the Mendocino Triple Junction offshore of Northern California, dots for earthquake locations, beachball-like mechanism diagrams, and an evenly-spaced grid of plus signs. Four seismic stations are labeled.
For large earthquakes that may generate tsunamis, it is important to determine the size and type of faulting of the event rapidly. We have developed a method to continuously scan an offshore region for earthquakes, and quickly determine their size and mechanism. The method is currently being applied to the region of the Mendocino Triple Junction, using data from four Berkeley Digital Seismic Network (BDSN) stations (HUMO, ORV, WDC, and YBH). In the figure above, the plusses are the grid locations used in the search, while black dots show the region's seismicity since 1900. Mechanisms for the events studied are given from this analysis, color-coded by variance reduction, and compared with solutions from the Berkeley moment tensor catalog (in brown). Click to view larger.

Initial early warnings for tsunamis are, more often than not, issued before the mechanism - thrust, strike-slip, or normal - of the potentially tsunamogenic earthquake is known. Since earthquake-induced tsunamis arise most often from the ocean bottom's response to a thrust, or subduction, event, rapid determination of the mechanism is an important improvement to make to the standard processing systems.

Drs. Guilhem and Dreger of the Berkeley Seismo Lab extended and improved a rapid mechanism determination technique pioneered in Japan using catalogued Mendocino area quakes as test cases. The method utilizes computer algorithms for calculating moment tensors, mathematical representations of earthquake forces that can be used to determine mechanisms.

An advanced tsunami warning algorithm could look for subduction events by computing moment tensors over and over for a grid of potential earthquake locations and depths, using streaming, real-time seismic data. A real, moderate-to-large earthquake would result in a moment tensor with a high "variance reduction," a measure of how well the real data fits synthetic data generated from our knowledge of the way seismic waves travel through the Earth's crust in this region, while wiggles caused by far-away quakes (teleseisms), tiny quakes, or noise would not. In other words, this "variance reduction" is a measure of the "goodness" of the moment tensor solution, and only real earthquakes can result in a good solution computed by the algorithm at one of the grid points. This approach can provide the earthquake location, magnitude, and mechinism within a few minutes of the origin time.

Guilhem and Dreger tested this technique on a score of events offshore of Northern California whose moment tensor solutions had already been computed by human analysts during the course of regular Berkeley Seismo Lab operations. The solutions generated and flagged as "good" by the automated grid-search algorithm compared favorably with the human-reviewed solutions, providing evidence that this method is suitable for analyzing events offshore of Northern California.

The largest tested earthquake was of magnitude 7.1. Magnitude 8+ earthquakes pose special problems for early mechanism detection. But Guilhem and Dreger's tests on two synthetic M 8+ events* showed that including information about the different directions the fault could rupture as well as the duration of the actual rupture during the earthquake in the models used for the moment tensor algorithm could allow these major events to be detected more rapidly than at present, with a more rapid determination of their mechanisms. The method was also tested using Japanese data from the magnitude 9.0 Tohoku earthquake of March 2011. In contrast to Japan's current system, which calculated the mechanism 20 minutes after the quake, Guilhem and Dreger's approach could have produced the mechanism within only 8 minutes. In all, by providing complete earthquake information, including the mechanism, within 6 to 8 minutes of an offshore quake, this technique could allow for up to tens of minutes of onshore warning in the area less than 60 miles from the epicenter of the quake.

* Real seismic data was not available, as the last known megathrust Cascadia Subduction Zone earthquake was in 1700.


Read More About BSL Research on this Topic:

Read the Geophysical Research Letters paper on this topic

BSL research reports:

Rapid Detection of Large Earthquakes Using Quasi-Finite-Source Green's Functions in Moment Tensor Analysis (PDF)
Towards a Real-time Earthquake Source Determination and Tsunami Early Warning in Northern California (html)
Detecting the 2011 M9.0 tohoku Earthquake with Moment Tensors (PDF)