Comment on “Tidal synchronicity of the 26 December 2004 Sumatran earthquake and its aftershocks” by R. G. M. Crockett et al.

Abstract

[1] Crockett et al. [2006] report a striking coincidence of some of the larger recent earthquakes around the Sunda/ Andaman/Java trench regions with the times of the full and new moon. However, it may be worth noting that the results contradict nearly all previous studies and theoretical expectations, and the statistics invoked have some problems. [2] Two flavors of earthquake-tide correlations have been sought, with either the diurnal stress variations or the biweekly modulations in the amplitude of the stress envelope. A correlation with both tidal periods has occasionally been noted in the case of seismicity just below the ocean floor at mid-ocean ridges, but in each reported case the correlation with biweekly tides was much weaker than the correlation with diurnal tides [Wilcock, 2001; Tolstoy et al., 2002]. Extensive data on global tectonic earthquakes have sometimes shown evidence of a correlation with diurnal tides [Tanaka et al., 2002, Tsuruoka et al., 1995], especially in the case of the strongest tides [Cochran et al., 2004]. However, none of the studies report seeing a correlation with the biweekly tide. Even regional studies of far larger sets of earthquakes in the San Francisco Bay Area [McNutt and Heaton, 1981; Kennedy et al., 2004] and the Pacific Northwest [Kennedy et al., 2004] find no measurable correlation. As stated by Hartzell and Heaton [1989], the difference in the amplitude of the biweekly tidal stress envelope is much smaller than the diurnal peak-to-peak stress variation, with diurnal stress variation being over 5 times larger than the biweekly variation. Along the same vein, theoretical studies predict a correlation with the diurnal but not biweekly tides due to the small overall amplitude variation of biweekly tides [Dieterich, 1987]. [3] Crockett et al. [2006] base their conclusions on a comparison with the lunar phase only; however, it has been well documented, for example in a review by Emter [1997], that the tidal time series must be computed in relation to the focal mechanism to derive a useful correlation. In addition, Crockett et al. [2006] do not include the ocean tide component of the tidal stress in their correlation even though the ocean tide component is often an order of magnitude larger than the solid earth tide in subduction zone (coastal) environments. Depth matters as well; for example, Tsuruoka et al. [1995] show that tides calculated at the surface can be opposite in phase to tidal time series at hypocentral depths when ocean loading is taken into account. [4] Perhaps most importantly, the statistics employed by Crockett et al. [2006] are somewhat problematic. Schuster’s test is only valid for independent datasets in which events occur at random. This test can produce a false positive result if the data set is not random and so should not be used for statistical tests on catalogs that include aftershocks. Therefore, the statistics given by Crockett et al. [2006] for the raw (clustered) catalog are invalid. Specifically, in the clustered catalog, the times of aftershocks of the M9.3 megathrust dominate the sample, and they are not uncorrelated in time. Rather they tend to occur just after the mainshock, as aftershocks are prone to do. [5] The most compelling statistic of Crockett et al. [2006] is the one derived from 13/14 declustered events happening near the quadrants of either the full or new moons. An only 1-in-a-100 chance of finding the result at random is cited, but the paper documentation suggests the degrees of free-