iPlex Lunch spring-2017

The spatial distribution of aftershocks in the Epidemic Type Aftershock Sequences (ETAS) model has been estimated under various parametric forms. Typical formulations include isotropic, anisotropic, and magnitude dependent. Nonparametric modeling procedures have been developed, such as in Marsan and Lengliné 2008, 2010, and more recently in Fox et al. 2016. In this study, we examine nonparametric extensions for the previously proposed parametric forms of the spatial response function. Simulation studies are conducted to assess the methods and they are demonstrated on an earthquake catalog for California from 1985 to 2005. Using the proposed extensions, multiple time-independent forecasts are developed, which provide five-year forecasts for magnitudes m ? 4.95 in the California CSEP testing region. Model performance is evaluated by the comparison of our models to the long-term forecast of Helmstetter 2007 using both Deviance and Voronoi Residuals. We show improved performance compared to Helmstetter 2007 in various regions while using a full nonparametric estimation and forecasting approach.