Aim of the Project
In Japan, more than one thousand seismic observatories have been continuously acquiring digital data of high-precision. In addition to this, a "seismic big database" that collects a large amount of vibration data measured by sensors installed in buildings, public utilities and smartphones will be available in the near future. In order to make the best use of this database for earthquake disaster prevention, mitigation, and clarification of earthquake phenomena, we should start immediately to develop analysis methods that efficiently extract information on various types of earthquakes from the super-large-amount seismic data continuously obtained. In this project, we develop algorithms for "intelligent seismic data processing" based on the cutting-edge Bayesian statistics towards the coming big-data era, through a collaboration of experts in seismology at Earthquake Research Institute, The University of Tokyo, those in statistics at Graduate School of Information Science and Technology, The University of Tokyo, and those in hydrodynamics at Graduate School of Engineering, Tohoku University. We believe this project will drastically deepen our understanding of earthquake phenomena and contribute to earthquake disaster prevention and mitigation.
What's New in the Project
The seismic big database contains data acquired by tens of millions of various types of sensors with different signal-to-noise ratios and noise characteristics. In order to extract, from such the hetero-quality data, the information related to oscillations originated from the solid Earth having extremely wide distribution in space and time, we implement an autofocus function, in addition to conventional seismic data processing methods, that automatically and data-adaptively seeks appropriate space and time ranges to be focused on. This research, which develops statistical analysis methods to extract essential information from such a large amount of data, perfectly matches the strategic goals of the research area "Intelligent Measurement Analysis" in JST CREST. "Intelligent seismic data processing", which comprehensively analyzes various oscillating phenomena, is challenging in seismology, and the utilization of high-dimensional state-space models that makes the best of spatial information obtained by the recent high-density observatories is also pioneering in the world.
This project, which develops the fundamental methods by a collaborating group of experts in seismology and statistics will bring big progresses in understanding the mechanism of earthquake occurrences, and impact Japanese society that has high expectations and demands for the earthquake researches. The methods for "Intelligent Seismic Data Processing" will be implemented on real-time monitoring systems of earthquakes and volcanoes that run continuously 24 hours a day, every day, essential in the disaster prediction, prevention and mitigation in Japan in the future.