|
Abstract:
In this paper, the author scrutinized the time series of a number of earthquakes and energy released by them in each month for the whole of Taiwan. The data obtained from January 1995 to April 2019, that is, totally 292 months were used in this paper. Both labeled and unlabeled earthquakes recorded in the Central Weather Bureau (CWB) were analyzed to realize the characteristics of these time series. The time series of the number of earthquakes of each month, either total (labeled plus unlabeled) or labeled, were stationary by the check of Augmented Dickey-Fuller (ADF) unit root test. The time series of energy released by the earthquakes in each month, either total or labeled, were also stationary. The ARIMA (Autoregressive Integrated Moving Average) model were used to identify the time series patterns of either monthly earthquakes or energy. The author found that ARIMA (3,0,0) and ARIMA (1,0,0) were suitable to the time series of the total (labeled plus unlabeled) and labeled number of earthquakes per month, respectively. Whereas, ARIMA (1,0,0) was suitable for energy released by both total and labeled earthquakes in each month. The energy released by earthquakes in each month was huge, hence it was normalized by the energy generated by the atomic bomb dropped at Hiroshima, Japan. The ARIMA (1,0,0) model was used to forecast the energy released in each month in the year 2019. In each month, there will 2.8 atomic bombs equivalent energy be released for the total earthquakes and 2.7 for the labeled ones. The relationship between number of earthquakes and energy released was obtained by the regression equation. The adjusted coefficient of determination can be as high as 25.43%, which explained the relationship between the energy released and number of the labeled earthquakes. Since the stationarity of each time series had been checked, the regression equation was not spurious, which usually occurred when two nonstationary time series were used in regression.
|
