A Forecast of Peak Ground Motion Due to Aftershocks Based on the Extreme-Value Analysis of Seismograms

Date: 4/24/2019

Time: 02:45 PM

Room: Pike

Aftershock forecasting is one of the most required information after a large earthquake, especially at an early stage. However, earthquake detection becomes very difficult at this stage because many wave packets from aftershocks overlap in seismograms, which unable automatic identification of P- or S-phases at many stations. For this reason, both quality and quantity of earthquake catalog deteriorate, and eventually aftershock forecasting becomes unavailable at the early stage.

We propose a method that will overcome this situation. Instead of using an earthquake catalog, we directly analyze continuous seismograms to forecast frequency of occurrence of large peak ground motion by aftershocks. we calculate maximum amplitude within each unit time interval (e.g., 1 minute) for a continuous seismogram. We propose a nonstationary extreme-value distribution for the maximum amplitudes by considering the generalized Pareto distribution and the Omori-Utsu type temporal decay. By optimizing three parameters that control this distribution, we evaluate current aftershock activity and forecast peak ground motion in near future at each station. Unlike earthquake catalog, this method is not suffered from deterioration of the data at the early stage. In addition, this method can provide forecast of peak ground motion rather than magnitude, which should be more desired information for public use.

We apply the proposed method to the 2016 Kumamoto earthquake sequence in Japan (M_W 7.3 for the mainshock). We find only three hours-long records can provide acceptable forecasting performance at most stations that experienced strong ground motion by the mainshock. Of note, our forecast performs well not only at stations near the mainshock source but also at relatively distant area, where many large events were triggered after the mainshock.

Presenting Author: Kaoru Sawazaki

Authors