Exciting news! We're transitioning to the Statewide California Earthquake Center. Our new website is under construction, but we'll continue using this website for SCEC business in the meantime. We're also archiving the Southern Center site to preserve its rich history. A new and improved platform is coming soon!

The earthquake rates they are a-changin’: Improving forecasts during earthquake swarms

Andrea L. Llenos, Andrew J. Michael, Morgan T. Page, Nicholas J. van der Elst, & Sara K. McBride

Published July 30, 2018, SCEC Contribution #8226, 2018 SCEC Annual Meeting Poster #051

Earthquake swarms present challenges for operational earthquake forecasting (OEF), because they are often modeled as time-varying changes in background seismicity that are driven by external processes such as fluid flow or aseismic creep, in addition to inter-earthquake triggering based on aftershock statistics. While the time decay of aftershock sequences can be modeled with the modified Omori law, it is difficult to forecast how long a swarm is likely to last or how seismicity rates may vary without being able to forecast the behavior of the external processes. The 2016 Bombay Beach, CA swarm highlighted the need to improve OEF for swarms. Because of its proximity to the southern San Andreas fault, the swarm caused concern that a larger earthquake could be triggered on that fault. However, computing that likelihood was not a trivial task, because current forecast methodologies typically do not account for changes in background rate.

Here we summarize some ways to improve OEF during swarms that we have developed and tested on the 2016 Bombay Beach and 2015 San Ramon, CA swarms. We make forecast models using the Epidemic-Type Aftershock Sequence (ETAS) model (Ogata, JASA, 1988), which can account for background rate changes during a swarm. Additional improvements include: 1) developing a regional swarm model based on fitting previous swarms to capture higher background rates; 2) creating a duration model to estimate how long a swarm is likely to last based on actuarial statistics of previous swarms; 3) updating background rates periodically during a swarm using a number of different lookback windows over which to estimate the background rate; and 4) constructing ensemble forecasts, which enables us to combine different models weighted according to their performance and avoid making arbitrary decisions at the outset of a swarm as to which single model will perform the best. We also briefly explore how forecasts for the 2016 Bombay Beach swarm were communicated and lessons learned from that experience to inform future communication about swarm forecasts.

Llenos, A. L., Michael, A. J., Page, M. T., van der Elst, N. J., & McBride, S. K. (2018, 07). The earthquake rates they are a-changin’: Improving forecasts during earthquake swarms. Poster Presentation at 2018 SCEC Annual Meeting.

Related Projects & Working Groups
Earthquake Forecasting and Predictability (EFP)