No clear evidence for localized tidal triggering of earthquakes in the Japan region

Abstract:

Because cyclical stressing rates from the Earth tides are typically much higher than the long-term buildup rate of tectonic stresses, many studies have searched for a possible correlation between earthquake occurrence and tidal phase. Although most of these studies find no clear relation between earthquakes and tides, evidence for at least some correlation has been found in a number of regions. Here we examine possible localized tidal triggering of earthquakes near Japan, including time periods before and after the 2010 Mw 9.1 Tohoku-Oki earthquake. Ideally studies of possible tidal triggering of earthquakes should compute the tidal stresses on the earthquake faults to establish directly when earthquakes should be encouraged and discouraged. However, because event catalogs typically only include focal mechanisms for relatively large earthquakes, this prevents analysis of the much more numerous smaller events. Here we adopt the simpler approach of searching for non-random distributions of event times with respect to tidal phase, without establishing any explicit relationship to stress or the most likely tidal phase to promote triggering. Our method could fail to detect tidal triggering if a variety of focal mechanisms are present in a region, which trigger at different times during the tidal cycle. However, because nearby events typically have similar focal mechanisms, it is likely that sufficiently localized regions will show clear peaks in occurrence time phase if tidal triggering is present. We examine 74610 earthquakes of M ≥ 3 from the JMA catalog from Jan. 2000 to Apr. 2013. We experiment by dividing the data into cells with a range of spatial (0.2, 1.0, 1.5) and temporal dimensions (100, 200, 400 days). We compute p-values using Schuster's test applied to the distribution of events times with respect to both semidiurnal and semimonthly tidal phase. Because these p-values may be biased by temporal earthquake clustering (e.g., aftershock sequences), we apply two different declustering methods, including that of Reasenberg [1985] and an approach that retains only one event per tidal cycle per phase increment. Our results show a wide range of p-values for the localized earthquake bins, but the number of bins with very small p-values (e.g., p ≤ 0.05) is no more than might be expected due to random chance.