A Markov chain model for earthquake occurrence analysis in Megathrust 4 (M4), Sumatra, Indonesia

Earthquakes are natural events with complicated spatio-temporal dynamics that can cause serious threats to human lives, infrastructure, and the environment. Understanding and effectively predicting earthquakes is critical for disaster preparation, mitigation, and response actions. This paper demonstrates modeling mainshock earthquakes in Megathrust 4 (M4), Sumatra, Indonesia, using a Markov chain framework with a K-Means cluster to form the state, which takes advantage of its probabilistic nature to reflect the stochastic aspects of seismic activity. The results of declustering with the Gardner-Knopoff process fulfill the assumption of the Poisson process, so that only the mainshock data are processed. A stationary distribution for the region and the mean recurrence time of the earthquake of each cluster were also determined after conducting a training and test on the model. According to the study findings, cluster 3 formed by K-Means clustering in M4 was found to have the shortest mean recurrence time, indicating more frequent seismic activity in that cluster.