Arima model for forecasting Poisson data: Application to long-term earthquake predictions.

摘要

Earthquakes that occurred worldwide during the period of 1896 to 2009 with magnitude greater than or equal to 8.0 on the Richter scale are assumed to follow a Poisson process. Autoregressive Integrated Moving Average models are presented to fit the empirical recurrence rates, and to predict future large earthquakes. We show valuable modeling and computational techniques for the point processes and time series data. Specifically, for the proposed methodology, we address the following areas: data management and graphic presentation, model fitting and selection, model validation, model and data sensitivity analysis, and forecasting.