151st ASA Meeting, Providence, RI


Short-term Non-Poissonian Temporal Clustering of Magnitude 4+ Earthquakes in California and Western Nevada

John Ebel 1- ebel@bc.edu
Daniel W. Chambers 2, Alan L. Kafka 1, and Jenny A. Baglivo 2

1
Weston Observatory
Department of Geology and Geophysics
Boston College
381 Concord Rd.
Weston, MA 02493

and
2Department of Mathematics
Boston College
Chestnut Hill, MA 02467

Popular version of paper 2pSP1
Presented Tuesday afternoon, June 6, 2006
151st ASA Meeting, Providence, RI

One of the major challenges for forecasting earthquakes is to identify patterns in the occurrences of the routinely recorded earthquakes of a region that might indicate that a stronger earthquake is imminent. While intuition might suggest that there must be patterns in the earthquake activity that would be diagnostic of coming large earthquakes, so far scientific research has failed to find any repeatable patterns of earthquake activity that make possible short-term forecasting of coming seismic events. However, recent research by seismologists at Boston Colleges Weston Observatory teaming with mathematicians from the Department of Mathematics at Boston College finds statistical evidence suggesting that there might be some patterns in the seismicity that would indicate an enhanced chance of a stronger earthquake.

Our analysis of earthquakes above magnitude (M) 4.0 since 1932 in California and western Nevada shows that there are more short-term clusters (earthquakes within a few to several days of each other) than expected from random, uncorrelated earthquake activity. We found that when an M≥4.0 earthquake takes place somewhere in California or western Nevada, during the next 5-10 days there is a slightly enhanced chance that another one or more M≥4.0 seismic events will strike somewhere else in that region.

This statistical analysis is not able to determine the cause of these short-term earthquake clusters. Possible causes include one earthquake triggering another or some type of regional strain event that triggers all the earthquakes in a cluster. We find only a few weak spatial correlations of earthquakes in different parts of California and western Nevada. For example, M≥4.0 earthquakes east of the Sierra Nevada in eastern California and western Nevada tend to occur within a few days of earthquakes on the San Andreas Fault and other nearby faults in central California.

In order to better pinpoint when future short-term earthquake clusters might occur, we are developing an earthquake forecasting method for M≥4.0 earthquakes in California and western Nevada based on a Hidden Markov Model (HMM). The method we have developed uses an analysis of the past earthquake activity of an area to estimate, after an earthquake, the probability that another M≥4.0 earthquake will occur in the next several days. At the present time, the model only tries to forecast the time of the next earthquake, without regard to its location or magnitude. In future versions of the method we will attempt to incorporate forecasts of magnitudes and locations along with the times of M≥4.0 earthquakes.


[ Lay Language Paper Index | Press Room ]