Recurrence Time Modeling for Earthquake Prediction

Abstract

Abstract In this study, statistical approach is used for predicting the probability of occurrence of earthquake using recurrence and elapse time. The data are from 1907 to 2008, in the area between 00 - 50 N and 900 – 1100 E, with magnitude ≥ 6.0 Richter scale, which include tsunami earthquake that occurred on Dec 26, 2004. Four reccurrence models (i.e. Gamma, Exponential, Log-normal, and Weibull) are examined for determining the best one that represents earthquake data. The maximum likelihood method is better parameter estimation than least squares and moment, because it has a lower Mean Squared Error. Based on Akaike and Bayesian Information Criterion, Weibull is better fitting than the others using Kolmogorov-Smirnov, Anderson-Darling, and Chi-Square test statistic. Cumulative distribution function of Weibull is used to predict earthquakes that will be occurred. The methods of maximum likelihood and moments are similar, least squares has a lower prediction. In general, for zero elapse time, the occurrence probability are higher than non-zero elapse time