The investigation of back projection location errors in Commander Island

The uncertainty of Back-projection inferred (BPI) location of earthquakes is caused by the complexity of earthquake sources, BP data processing and the 3-D heterogeneous Earth structure. Here we investigate BPI location errors due to the performance of waveform alignment and BP’s parameterization that are included in the BP data processing. Using the Command Island region, Alaska, as a test region, we analyzed 22 M4-7 earthquakes between 2013 and 2017, that are evenly distributed over the entire region, and classified them into two groups (Mw 5.5) based on the degree of the finite-source effect relevant to the earthquake magnitude. By comparing BP results of two examples (a Mw 6.3 and a Mw 5.5 earthquake) of events with hypocenter timing corrections from different alignment ways, we confirmed the significance of the fine alignment on initial P-wave arrivals for the acquisition of accurate traveltime corrections, especially for the earthquakes with magnitudes larger than Mw 5.5 due to their finite-source effects. Subsequently, based on accurate hypocenter timing corrections from fine alignment on first-arriving P waves, we explored the effect of BP’s parameterization on BPI location errors of the two groups of earthquakes. Our experimental results suggest that for an earthquake without finite-source effect, BP’s parameterization will not affect the location accuracy of BP results with the kernel center of BP images invariably located at its hypocenter. Improper BP’s parameterization would bias the kernel center of BP images from the hypocenter of earthquakes affected by the finite-source effect, reducing the accuracy of BPI location. To further investigate the BPI location error affected by the 3-D velocity structure in this region, we took 9 different earthquakes as the reference events and measured the deviation distance (De) of BPI location of the other events using hypocenter timing corrections from each reference event. An obvious increasing trend of BPI location error can be found with the increase of the separation distance between the reference event and the registration event, implying that the 3-D velocity heterogeneity in Commander Island area would strongly limit the effectiveness of timing corrections for different events. To improve the BPI location accuracy, we subsequently tested the effectiveness of the aftershock calibration method on mitigating the BPI location uncertainty by comparing the De of BP results before and after the aftershock calibration. The effectiveness of the aftershock calibration, especially with the zonation strategy (zoning this region into three subregions and separately performed aftershock calibration within each subregion to further refine traveltime corrections of P-waves) significantly decreases of BPI location errors compared to the errors before the calibration, which implies that the BPI uncertainty caused by the heterogeneous 3-D velocity structure can be effectively mitigated by the aftershock calibration method with the consideration of the regional tectonic characteristics.