Item

Abstract

We investigate characteristics of spatial biases in the Back-projection method and the effectiveness of the Slowness Enhanced Back-projection (SEBP). The spatial bias refers to the location error in Back-projection caused by travel time errors due to 3D Earth structure. SEBP reduces such bias by calibrating the gradient of travel time (slowness) in the source. We first analyze 22 M4-M6 earthquakes in the Commander Island and find the amplitudes and directions of spatial biases follow distinct regional patterns. In light of this, we propose a regional SEBP approach that introduces spatially variable slowness correction. The regional SEBP in the Commander Island resulted in a ~ 50% reduction in the average length of spatial bias from ~20 km to ~10 km, which is more effective than the uniform SEBP that gives a 25% error reduction. We then analyze 109 M4-M7 earthquakes in the Tohoku region and also find a 50% error reduction by regional SEBP. This indicates that half of the spatial biases are aleatory uncertainties that are caused by regional structural complexity and can be calibrated with SEBP, while the residue errors are epistemic uncertainties that are random and caused by local velocity heterogeneities. With regional SEBP applied to the 2011 Tohoku earthquake, we find that high-frequency radiators did not reach beyond the down-dip limit of interplate seismicity, indicating that the coseismic slip unlikely penetrated into the brittle-ductile transition zone. Such observations suggest that the enhanced dynamic weakening mechanism due to thermal pressurization effects may not be activated during this event.