Prospective evaluation of multiplicative hybrid earthquake forecasting models in California

José A. Bayona, William H. Savran, David A. Rhoades, & Maximilian J. Werner

Published January 18, 2022, SCEC Contribution #11011

The Regional Earthquake Likelihood Models (RELM) experiment, conducted within the Collaboratory for the Study of Earthquake Predictability (CSEP), showed that the smoothed seismicity (HKJ) model by Helmstetter et al. (2007) was the most informative time-independent earthquake model in California during a five-year evaluation period. The diversity of competing forecast hypotheses in RELM was suitable for combining multiple models that could provide more informative earthquake forecasts than HKJ. Thus, Rhoades et al. (2014) created multiplicative hybrid models that involve the HKJ model as a baseline and one or more conjugate models. In retrospective analyses, some hybrid models showed significant information 1 gains over the HKJ forecast. Here, we prospectively assess the predictive skills of 16 hybrids and 6 original RELM forecasts, using a suite of traditional tests implemented in CSEP. Traditional CSEP tests are based on a likelihood function that approximates earthquakes as independent and Poisson distributed. However, the Poisson distribution insufficiently captures the observed spatiotemporal variability of seismicity. Therefore, we additionally introduce new tests based on a binary probability function that are less sensitive than CSEP evaluations to clustering. The evaluation dataset contains 40 M 4.95 events recorded within the CSEP California testing region from 1 January 2011 to 31 December 2020, including the 2016 MW 5.6, 5.6, and 5.5 Hawthorne earthquake swarm in southwestern Nevada, and the MW 6.4 foreshock and MW 7.1 mainshock of the 2019 Ridgecrest sequence. In particular, we evaluate the consistency between the observed and the expected number, spatial, and likelihood distributions of earthquakes, and compare the performance of each forecast to that of HKJ. Prospective test results show that none of the hybrid models is significantly more informative than the HKJ benchmark forecast, mainly due to temporal instabilities in the method of forming hybrids. These results give indication that smoothing high-resolution, small earthquake data is a robust method to describe seismicity patterns in California.

Key Words
earthquake forecasting; hybrid modeling;

Citation
Bayona, J. A., Savran, W. H., Rhoades, D. A., & Werner, M. J. (2022). Prospective evaluation of multiplicative hybrid earthquake forecasting models in California. Geophysical Journal International, 229(3), 1736-1753. doi: 10.1093/gji/ggac018.

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