[Paper Review] Dynamics of the Markov Time Scale of Seismic Activity May Provide a Short-Term Alert for Earthquakes

Reza Mansouri, Muhammad Sahimi+10 more|ArXiv.org|Oct 6, 2005

Complex Systems and Time Series Analysis3 references3 citations

TL;DR

This paper proposes a short-term earthquake prediction method based on the dynamics of the Markov time scale ($t_M$) and extended self-similarity ($T_1$) in seismic data. When both $t_M$ and $T_1$ exceed critical thresholds simultaneously, a reliable alarm is triggered, with the method achieving near-zero failure rates in testing across Iranian earthquakes.

ABSTRACT

We propose a novel method for analyzing precursory seismic data before an earthquake that treats them as a Markov process and distinguishes the background noise from real fluctuations due to an earthquake. A short time (on the order of several hours) before an earthquake the Markov time scale $t_M$ increases sharply, hence providing an alarm for an impending earthquake. To distinguish a false alarm from a reliable one, we compute a second quantity, $T_1$, based on the concept of extended self-similarity of the data. $T_1$ also changes strongly before an earthquake occurs. An alarm is accepted if {\it both} $t_M$ and $T_1$ indicate it {\it simultaneously}. Calibrating the method with the data for one region provides a tool for predicting an impending earthquake within that region. Our analysis of the data for a large number of earthquakes indicate an essentially zero rate of failure for the method.

Motivation & Objective

• To develop a method for short-term earthquake prediction based on statistical dynamics of seismic time series.
• To distinguish real precursory signals from background noise using Markov process analysis.
• To reduce false alarms by requiring simultaneous signals from two independent statistical indicators: $t_M$ and $T_1$.
• To provide a quantifiable alert time dependent on earthquake magnitude, enabling actionable early warnings.
• To calibrate the method regionally using historical seismic data for practical deployment in real-time monitoring.

Proposed method

• The method computes the Markov time scale ($t_M$) as the time over which seismic data can be modeled as a Markov process, determined by testing the Chapman-Kolmogorov equation for conditional probabilities.
• It evaluates the validity of the Markov assumption by comparing directly measured transition probabilities with those computed via the Chapman-Kolmogorov equation over varying time lags.
• A second indicator, $T_1$, is derived from the concept of extended self-similarity to detect anomalous scaling behavior in seismic fluctuations before an earthquake.
• Critical thresholds $t_{cM}$ and $T_{c1}$ are calibrated using data from previously occurred earthquakes in a specific region.
• An alarm is triggered only when both $t_M > t_{cM}$ and $T_1 > T_{c1}$, minimizing false positives.
• The alert time $t_a$ is estimated via a logarithmic relation: $ ext{log } t_a = -1.35 + 2.4 ext{ log } M$, with $t_a$ in hours and $M$ the earthquake magnitude.

Experimental results

Research questions

• RQ1 Can the Markov time scale ($t_M$) of seismic activity serve as a reliable short-term precursor to large earthquakes?
• RQ2 Can the extended self-similarity indicator $T_1$ enhance the reliability of $t_M$-based alarms by reducing false positives?
• RQ3 Is there a quantitative relationship between the magnitude of an impending earthquake and the alert time provided by the method?
• RQ4 Can the method be calibrated regionally to provide accurate, real-time short-term earthquake warnings?
• RQ5 What is the failure rate of the method when both $t_M$ and $T_1$ are required to trigger an alarm simultaneously?

Key findings

• The method achieved an essentially zero rate of failure when both $t_M$ and $T_1$ provided simultaneous alarms, with only two failures observed over two years of analysis.
• For an earthquake of magnitude $M = 4.5$, the predicted alert time is approximately 2 hours, based on the relation $ ext{log } t_a = -1.35 + 2.4 ext{ log } M$.
• The alert time $t_a$ increases with magnitude, indicating longer warning times for larger quakes, with $t_a$ reaching up to 10 hours for $M > 5.7$.
• The method successfully provided a seven-hour alarm for the Baladeh earthquake, despite no foreshocks in the region, demonstrating its sensitivity to precursory signals.
• Stations located at depths greater than 50 m and oriented perpendicular to active faults yield more correlated data and more accurate alarms.
• The method was validated using 173 Iranian earthquakes with magnitudes between 3.2 and 6.3, confirming its robustness across a wide magnitude range.