Earthquake Forecasting Using Hidden Markov Models

Abstract

This paper develops a novel method, based on hidden Markov models, to forecast earthquakes and applies the method to mainshock seismic activity in southern California and western Nevada. The forecasts are of the probability of a mainshock within one, five, and ten days in the entire study region or in specific subregions and are based on the observations available at the forecast time, namely the inter event times and locations of the previous mainshocks and the elapsed time since the most recent one. Hidden Markov models have been applied to many problems, including earthquake classification; this is the first application to earthquake forecasting

Prediction of earthquake hazard by hidden Markov model (around Bilecik, NW Turkey)

“…The use of HMM to predict earthquake hazard is carried out by Granat and Donnellan [37], Chambers et al [38], Ebel et al [39], Orfanogiannaki et al [3,40,41], Wu [42] and Chambers et al [4]. All previous studies about earthquake data modeling by HMM except Orfanogiannaki et al [3,40,41] are based on the continuous earthquake data and interevent earthquake times.…”

“…However, HMM has not as widely implemented as it should be in earthquake modeling. There have been few applications of HMM on earthquake problems [37][38][39] Orfanogiannaki et al [3,4,[40][41][42]). Granat and Donnellan [37] used HMM as an unsupervised learning method for clustering of earthquakes and determining the classes of similar earthquakes in southern California region.…”

“…Wu [42] proposed a simple approach for HMM to decluster earthquakes in central and western regions of Japan. Chambers et al [4] extended HMM to forecast the location of the earthquakes as well as interevent earthquake times in southern California and western Nevada. Let T = {0 1 2 T } a discrete-time horizon.…”

“…HMM first introduced by Baum and Petrie [2] allows for overdispersion together with the dependence between the data [3,4]. HMM assumes that the state of the underlying system at any time is unobservable.…”

Identification of temporal patterns in the seismicity of Sumatra using Poisson Hidden Markov models

“…The time-dependent interaction of seismic events that is extracted by the transition probability matrix of HMMs identifies relationships between earthquakes due to stress changes within a fault system. The outline of this methodology was used by Chambers et al 6 to produce 1, 2 and 10 days forecasts in the southern California and western Nevada regions. Orfanogiannaki et al 7 used HMMs as a tool to identify through the estimated sequence of hidden states the seismic cycle of strong earthquakes in the area of Killini, Western Greece.…”