Data mining method in seismology by applying cellular automaton equivalence of ground vibration fluctuations recorded near the epicenter of the 2011 Mw 9 East Japan earthquake

We propose a data mining method to extract physical phenomena from the ground vibration velocity fluctuation (GVF) which is a daily observed weak signal and has no specific waveform. Our aim is to detect any traces of the return of ground motion to its normal state after earthquakes, and the evolution of ground motion towards earthquakes in the GVF. We focus on the GVF recorded near the epicenter of the magnitude 9 Great East Japan Earthquake (GEJE). Although the GVF can be discussed in the framework of the master equation which represents the thermodynamics of a stochastically fluctuating non-equilibrium system, the relation between the thermodynamic parameters and seismology-related phenomena is hardly known. So, we introduce cellular automata (CA) to analogically interpret the physical phenomena of interest as thermodynamic parameters, and show the thermodynamic equivalence between GVF and CA. These CA-derived thermodynamic parameters are then expected to deliver physical phenomena similar to those of the target when evaluated on the GVF data. In the demonstrated example of data mining, the stress relaxation signal immediately after GEJE is discovered from the recorded GVF data by deriving the thermodynamic parameters representing the stress relaxation of CA and evaluating these thermodynamic parameters for the recorded GVF data. Furthermore, the elastic rigidity and viscosity of the underground structures are estimated from the stress relaxation signal. The novelty of this study lies in drawing attention to the GVF, and combining standard techniques of CA with GVF through the use of the master equation.