Pseudo-prospective Evaluation of UCERF3-ETAS Forecasts During the 2019 Ridgecrest Sequence

William H. Savran, Maximilian J. Werner, Warner Marzocchi, David A. Rhoades, David D. Jackson, Kevin R. Milner, Edward H. Field, & Andrew J. Michael

Under Review April 29, 2020, SCEC Contribution #10082

The Ridgecrest sequence began with several small earthquakes preceding a Mw 6.4 event on July 4, 2019, that was followed 36 hours later by the Mw 7.1 mainshock. This sequence included the largest earthquake in California since the 1999 Mw 7.1 Hector Mine event. Forecasts from the Uniform California Earthquake Rupture Forecast Version 3 – Epidemic Type Aftershock Sequence (UCERF3-ETAS) were started within an hour of the initial Mw 6.4 event. In contrast to traditional ETAS models, UCERF3-ETAS provides probabilities of triggering earthquakes on specified nearby faults with spatially varying magnitude distributions. The Ridgecrest sequence provides the first opportunity to evaluate the model in a pseudo-prospective sense. As a benchmark, we include a version of the model without explicit faults, more closely mimicking standard ETAS models. We evaluate the forecasts with new metrics developed within the Collaboratory for the Study of Earthquake Predictability (CSEP). The metrics consider the synthetic catalogs simulated by the models rather than synoptic probability maps, thereby relaxing a simplifying assumption in previous CSEP tests, namely that of Poissonian fluctuations. Our approach uses Turing-style tests that compare metrics from the synthetic catalogs directly against observations, including their distributions, allowing a flexible approach that can accommodate arbitrary non-parametric distributions and dependencies encoded in models and their aleatory and epistemic uncertainty. We find that UCERF3-ETAS overpredicts the rates of earthquakes in the vicinity of the fault zone. The observed magnitude and spatial distributions of observed seismicity fall within the variability of the forecasts. The spatial forecast tends to be more diffuse spatially than the observed catalog, while the observed interevent time distribution is largely consistent with the simulations. During the Ridgecrest Sequence, the performance the evaluations indicate no significant differences between forecasts produced by UCERF3-ETAS and a more traditional ETAS model that does not consider faults.

Savran, W. H., Werner, M. J., Marzocchi, W., Rhoades, D. A., Jackson, D. D., Milner, K. R., Field, E. H., & Michael, A. J. (2020). Pseudo-prospective Evaluation of UCERF3-ETAS Forecasts During the 2019 Ridgecrest Sequence. Bulletin of the Seismological Society of America, (under review).

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