A1461
Title: Short-term forecast of repeating earthquakes using a spatiotemporal nonstationary renewal process model
Authors: Shunichi Nomura - Waseda University (Japan) [presenting]
Naoki Uchida - The University of Tokyo (Japan)
Yosihiko Ogata - Institute of Statistical Mathematics (Japan)
Abstract: Small repeating earthquakes observed along plate boundaries occur more frequently than large repeating earthquakes and have much shorter recurrence intervals. Hence, short-term forecasts for small repeating earthquakes yield relatively high occurrence probabilities, allowing for validation of forecasting methods by comparing predictions with observations over short time spans. On the other hand, forecasting small repeating earthquakes is challenging because their recurrence intervals can vary due to external influences such as afterslip following nearby large earthquakes and slow slip events. Under such conditions, stationary models assuming constant recurrence intervals can easily fail. To address this issue, a forecasting method for small repeating earthquakes based on a nonstationary renewal process model is proposed. In the proposed model, spatiotemporal variations in interplate slip rates in the surrounding region are estimated from the activity of small repeating earthquake sequences distributed along the plate boundary. These spatiotemporal variations are represented using tensor-product spline functions and estimated by maximizing a penalized likelihood with smoothing constraints. Moreover, by extrapolating the slip rates beyond the observation period, the occurrence probabilities of small repeating earthquakes over a short-term future horizon can be evaluated. The proposed model is applied to a catalog of small repeating earthquakes in northeastern Japan.
