Scaling of Seismic Memory with Earthquake Size

It has been observed that the earthquake events possess short-term memory, i.e. that events occurring in a particular location are dependent on the short history of that location. We conduct an analysis to see whether real-time earthquake data also possess long-term memory and, if so, whether such autocorrelations depend on the size of earthquakes within close spatiotemporal proximity. We analyze the seismic waveform database recorded by 64 stations in Japan, including the 2011 “Great East Japan Earthquake”, one of the five most powerful earthquakes ever recorded which resulted in a tsunami and devastating nuclear accidents. We explore the question of seismic memory through use of mean conditional intervals and detrended fluctuation analysis (DFA). We find that the waveform sign series show long-range power-law anticorrelations while the interval series show long-range power-law correlations. We find size-dependence in earthquake auto-correlations—as earthquake size increases, both of these correlation behaviors strengthen. We also find that the DFA scaling exponent $\alpha$ has no dependence on earthquake hypocenter depth or epicentral distance.

💡 Research Summary

The paper investigates whether seismic waveforms exhibit long‑range memory and how such memory depends on earthquake size. Using continuous broadband recordings from 64 stations of Japan’s F‑net (46 stations selected for data quality), the authors focus on the vertical component (U) sampled at 1 Hz, covering the year 2003 and the March 11, 2011 “Great East Japan” event (including two additional M ≈ 7.3–7.6 shocks on the same day).

For each earthquake the raw acceleration trace w(t) is normalized to zero mean and unit variance, then a sub‑segment w′(t) is extracted from the time of maximum amplitude to the end of the normalized trace. Two derived binary time series are constructed: (i) the sign series s_t = sign