Item

Abstract

We developed a real-time forecast method for long-period (4-10 s) ground motions that appear in the distant basins when a large earthquake occurs. The Green’s functions (GFs) from the vicinity of the source to the forecast points are obtained from ambient noise interferometry, recently used by e.g. Viens et al. (2017) and Denolle et al. (2018). Since the absolute amplitude of the GFs obtained by interferometry is unknown, it is calibrated by comparing it to the earthquake waveform according to Viens and Denolle (2019). In this study, one station in the northern Kanto, Japan was selected as the input station, and three stations in the Kanto and Nobi basins were selected as the forecast target stations. The Z-component ambient noise waveforms of the input station and the target stations were deconvolved and stacked for 1-year period to retrieve GFs. The amplitude calibration factor is calculated by comparing Fourier spectra of the forecasted and observed waveform of a Mw7 earthquake in the 4-10s period band. By convolving the obtained GFs with the input waveform of any earthquake, the long-period waveform at the target sites can be forecasted. The method has been tested against a Fukushima-Oki earthquake (Mw7.4). The shape of forecasted waveforms and their Fourier spectra were compared to those observed at each target site and found to be in good agreement. For the practical application, application to horizontal motion of a larger earthquake, such as Tohoku-Oki Earthquake (Mw9.0), is required.