Time series modelling of maximum tsunami wave heights recorded in Sibolga (Sumatra)

Abstract

This study reports univariate modeling methodologies applied to the maximum tsunami wave height over Sibolga, Sumatra. The univariate time series models fitted are autoregressive model (AR), autoregressive integrated moving average (ARIMA) and autoregressive neural network (AR-NN). Goodness of fit of the models to the time series of maximum tsunami wave height has been assessed using percentage of prediction error, Pearson correlation coefficient, and Willmott's indices. After rigorous skill assessment using the above three models, the AR-NN model with seven previous values as predictor has been identified as the best predictive model for the time series under study.