The 2009 Samoan and Sumatran Earthquakes: A Coincidence with Precedents (Invited)
Abstract
The 29 September 2009 M8.1 earthquake near Samoa was followed in less than a day by an M7.5 earthquake near Sumatra. This raises the question of whether two such events can be considered a coincidence or are evidence for long-range triggering of large earthquakes. I use coincidence as the null hypothesis and examine global earthquake catalogs to determine if that null hypothesis can be rejected with 95% significance. First, I define a large earthquake pair as two or more M≥7.5 events, which are not part of an aftershock sequence, within 24 hours. Aftershocks are removed using the Gardner and Knopoff (BSSA, 1974) method modified to use distances based on the Wells and Coppersmith (BSSA, 1994) fault lengths so that the method works on a global catalog. The catalogs used are the Centennial Catalog (Engdahl and Villaseñor, 2002), the USGS Preliminary Determination of Epicenters (PDE) catalog from 1973 to October 2009, and a merged catalog by using the Centennial catalog through 1972 and the PDE catalog thereafter. After declustering, the Centennial catalog has distant pairs of large earthquakes on single days in 1901, 1906, 1934, and 1993 while the PDE catalog has such a pair in 2002. Thus, large earthquake pairs have precedents. The 1993 pair is not found in the PDE catalog due to magnitude differences. Interestingly, the 2009 pair does not appear because the Sumatra event was determined to be an aftershock of the 2007 M8.5 earthquake there. The statistical analysis is done by first computing the Poisson probablity (Pp) of 2 or more events occuring within 24 hours based on the long-term rate of M≥7.5 earthquakes after declustering (R events/year). While these Poisson probabilities are very small, one must consider the probability of observing these high activity days during the length of the catalogs (Nyears). To do this, I simulate Poisson sequences of earthquakes given the observed rate and, by repeating these simulations many times, compute the probability of obtaining at least the number of observed pairs (Npairs) during the length of the catalog. If this simulation probability (Ps) is less than 0.05 then the null hypothesis is rejected. To include the 2009 pair in the analysis, I did a second declustering where aftershock sequences are limited to 18 months duration (* in the table below). Including the 2009 pair in the test of the hypothesis is questionable because it was used to develop the hypothesis. In all cases Ps is much greater than 0.05 and thus the data do not reject the null hypothesis of coincidence. My interpretation is that these pairs are coincidental and do not provide evidence for long-range triggering between large earthquakes.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2009
- Bibcode:
- 2009AGUFM.U22B..01M
- Keywords:
-
- 7223 SEISMOLOGY / Earthquake interaction;
- forecasting;
- and prediction;
- 7230 SEISMOLOGY / Seismicity and tectonics