Improving Earthquake Forecasting for Europe Using Elliptic ETAS Aftershock Distribution Kernels

Abstract In the standard Epidemic‐Type Aftershock Sequence (ETAS) model, widely used for earthquake forecasting, aftershocks are distributed isotropically in space—in circular symmetry around the triggering event. This simplification neglects fault geometry and observed anisotropy in aftershock distributions. In this study, we introduce directional information into the ETAS model by replacing the circular kernel with an elliptic one during parameter inversion procedure based on Expectation–Maximization. Using real catalogs from the European region, we investigate the impact of this modification on forecasting capability in retrospective and pseudo‐prospective evaluations. Our findings suggest that incorporating directional anisotropy by estimating ellipses using aftershock spatial distribution can significantly improve model performance, as well as reduce bias in the estimated ETAS parameters, specifically those controlling the productivity component. We also explore the limitations of estimating anisotropy in real time, proposing strategies to overcome the delay in obtaining reliable directional information.