THE PROBABILITY MODEL OF EARTHQUAKE FREQUENCY IN SUMATRA MEGATHRUST ZONE BASED ON IDENTIFICATION OF UNOBSERVABLE PROCESSES FROM PLATE TECTONIC DYNAMICS

• Main Author: Rizal, Jose
• Format: Dissertations
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/64422
• Institution: Institut Teknologi Bandung

The Sumatran megathrust zone contains five segments of major earthquake sources, namely Aceh-Andaman (AA), Nias-Simeuleu (NS), Mentawai-Siberut (MS), Mentawai-Pagai (MP), and Enggano (EO). The level of earthquake hazard in this zone can be mapped through the Probabilistic Seismic Hazard Analysis model. In the formation of that model, the frequency of earthquake occurrences is assumed to follow the Poisson process which does not depend on the time variable. However, it is reasonable to suppose that the frequency of earthquake occurrences may have a dependence on the time variable and also dependence on spatial. In addition, from the physical phenomenon of earthquake occurrences point of view, the frequency of earthquake occurrences can be affected by the tectonic plate dynamics in each segment. Based on these three research problems, two hypotheses are formulated, namely the frequency of earthquake occurrences in a segment of the Sumatran megathrust zone depends on time, and the frequency of earthquake occurrences in each of two adjacent segments of the Sumatran megathrust zone has spatial and time-dependent properties. The purpose of this study is to determine the validity of the two hypotheses and to produce the frequency probability model of earthquake occurrence in line with the research hypothesis. The research methods include scenario design of modeling, preparation of seismic data, model diagnosis, selection of probability models with parameters estimation, and their interpretation. The seismic data used are the frequency of major earthquakes from 1971 to 2018 with magnitude Mw 4; 6 and earthquake depth 70 km. The seismic data are diagnosed using the Ljung-Box, Kendall’s , and empirical Copula serial tests. The model diagnosis results are used to determine the validity of the research hypotheses and as a guide in the construction of Probability Models I and II. The probability models applied to the Probability Model I are the Poisson distribution, the Mixture-Poisson, and the Poisson-Hidden Markov models (Poisson-HMMs), while for the Probability Model II are the bivariate Copula, Mixture-Copula, and Copula-Hidden Markov models (Copula-HMMs). Based on the results of model diagnosis toward seismic data in the Sumatra megathrust zone, the statement of the two research hypotheses is true. According to this, the Probability Model I for the five major earthquake source segments in the Sumatra megathrust zone produce the same model, namely 3-state Poisson-HMM, but with different model parameters for each segment. Meanwhile, the Probability Model II for four cases of two adjacent segments, namely AA-NS, NS-MS, MS-MP, and MP-EO, is 2-state Gumbel-HMM, 2-state Clayton-HMM, Copula Frank, and 2-state Clayton-HMM, respectively.