Application of Dynamic Strains to Earthquake Source Characterization and Earthquake Early Warning

Abstract:

Borehole strainmeters can provide data that are useful for rapid earthquake source characterization, which is necessary for earthquake early warning. Our analysis of high-frequency (1-Hz) strains from 180 earthquakes which occurred between 2004 and 2012, recorded by 68 Plate Boundary Observatory (PBO) borehole strainmeter (BSM) stations, with moment magnitudes M ranging from 4.6 to 7.2, depths ranging from 12 km to 33 km, and hypocentral distances ranging from 13 km to 500 km; reveals that peak dynamic strains can be predicted, with high statistical confidence, from the magnitude of the earthquake and its hypocentral distance. Our regression model also holds for high-rate GPS derived strains during the 2011 M9 Tohoku-oki earthquake, from GEONET subnetworks located as close as 140 km, indicating that the model does not saturate for large earthquakes. Moreover, using linear mixed-effects regression, we show that the largest source of bias in the residual mean squared error in the magnitude-distance regression arises from effects associated with the source and/or propagation path, rather than with the station. For instance, earthquakes on the Blanco fracture zone produce dynamic strains with lower amplitudes on average than earthquakes around the Sierra microplate, indicating that dynamic strains are also affected by earthquake source parameters other than the magnitude; these source and path effects are not large enough to degrade the relationship between dynamic strain and magnitude and distance though. We also show that by including PBO stations into current source characterization efforts would enhance station density significantly in critical regions. That is, 32 (41%) of the BSMs are located within 280 km from the trench axis of the Cascadia subduction zone where, for example, the probability of a M9 earthquake within 50 years might be as high as 15% and the probability of a smaller but still damaging M8 event might be as high as 40% (Goldfinger, et al., 2012). Finally, we demonstrate the applicability of BSMs to earthquake early warning and discuss challenges associated with real-time implementation.