The Collaboratory for the Study of Earthquake Predictability in China: Experiment Design and Preliminary Results of CSEP2.0

Shengfeng Zhang, Yongxian Zhang, Maximilian J. Werner, Kenny Graham, David A. Rhoades, & José A. Bayona

Submitted September 7, 2025, SCEC Contribution #14222, 2025 SCEC Annual Meeting Poster #TBD

In earlier CSEP1.0 experiments in China, the PI model was used for alarm-based, intermediate-to-long-term forecasts of earthquakes with magnitude MS ≥ 6.0 in the Yunnan–Sichuan region (Zhang et al., 2016). That study adopted a magnitude cutoff of 3.0, using a background window (1999–2008), an anomaly window (2009–2013), and a forecast window (2014–2018), with 0.2° grid resolution. Forecast evaluation was conducted based on performance during the forecast window (Zhang et al., 2021).

In the current CSEP2.0 study, we refined the region and adopted a higher-resolution grid (0.1°) to align with international standards. Both PI and RI models were modified to produce rate-based forecasts for MS ≥ 6.0 events using the same parameters. The PI model predicted higher per-cell earthquake rates than RI.

During the 2014–2018 forecast window, four MS ≥ 6.0 earthquakes occurred: Yingjiang (MS 6.1), Ludian (MS 6.5), Jinggu (MS 6.6), and Kangding (MS 6.3). Forecast evaluation using the unified catalog from CENC and applying N- and S-tests shows that both models passed the S-test (spatial consistency), but only the PI model passed the N-test (event count consistency).

Citation
Zhang, S., Zhang, Y., Werner, M. J., Graham, K., Rhoades, D. A., & Bayona, J. A. (2025, 09). The Collaboratory for the Study of Earthquake Predictability in China: Experiment Design and Preliminary Results of CSEP2.0. Poster Presentation at 2025 SCEC Annual Meeting.

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