IAG-IASPEI 2017

講演情報

Oral

Joint Symposia » J04. Geohazard early warning systems

[J04-3] Geohazard early warning systems III

2017年8月3日(木) 13:30 〜 15:00 Intl Conf Room (301) (Kobe International Conference Center 3F, Room 301)

Chairs: Mitsuyuki Hoshiba (Meteorological Research Institute, JMA) , Hiroaki Tsushima (Meteorological Research Institute, Japan Meteorological Agency)

14:00 〜 14:15

[J04-3-03] Real-Time Ground Motion Prediction based on Radiative Energy Transfer using Front-Site Waveform Information and Data Assimilation for the Application to Regional Earthquake Early Warning

Mike Lindner1, 2, Masato Motosaka2 (1.Karlsruhe Institute of Technology, Karlsruhe, Germany, 2.Tohoku University, IRIDeS, Sendai, Japan)

In-time warning of strong ground motion for citizens in affected areas is one of the primary objectives of Earthquake Early Warning. To satisfy this requirement, continuous front-site observation and quick analyzing algorithm to ensure enough warning time are indispensable. Existing network systems depend on fast source parameter estimation while on-site methods try to predict ground motion from an observation at a target site. Latter systems are currently the fastest approach for highly vulnerable sites, but they lack on quality in prediction. This study uses the radiative transfer theory based on ray scattering, to model a real-time numerical ground motion map for the region of Miyagi. The authors use a two-layered model to represent processes in geological structures by simulating the energy transfer in the bedrock as a function of scattering. The processes on the surface layer are then modeled as a function of local amplification and the bedrock information. Prediction is done for a circular area with an observation site as the central point. The predictive model uses band-passed filtered acceleration information from surface stations, projected onto the bedrock layer, by using a scalar transfer function. An energy impulse, which is a function of the PGA for a time length $dt$ is then used to simulate the distribution of energy within the area. Surface intensity is estimated by multiplying the bedrock model with the surface amplification. This process is repeated in steps of $dt$ second and can be done almost instantaneously. Furthermore, by using other observation sites within the prediction radius, evaluation and correction of the predictive field are achieved using data assimilation. The developed system allows it to model an interpolated wave-field behind, and a predicted wave-field in front of the wave-front, thus providing ground motion information around recording sites. The authors currently work on the extension of the proposed method into a network.