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!

Evaluating the use of declustering for induced seismicity hazard assessment

Andrea L. Llenos, & Andrew J. Michael

Published July 29, 2016, SCEC Contribution #6403, 2016 SCEC Annual Meeting Poster #309

The recent dramatic seismicity rate increase in the central and eastern US (CEUS) has motivated the development of seismic hazard assessments for induced seismicity (e.g., Petersen et al., 2016). Standard probabilistic seismic hazard assessment (PSHA) relies fundamentally on the assumption that seismicity is Poissonian (Cornell, BSSA, 1968); therefore, the earthquake catalogs used in PSHA are typically declustered (e.g., Petersen et al., 2014) even though this may remove earthquakes that may cause damage or concern (Petersen et al., 2015; 2016). In some induced earthquake sequences in the CEUS, the standard declustering can remove up to 90% of the sequence, reducing the estimated seismicity rate by a factor of 10 compared to estimates from the complete catalog. In tectonic regions the reduction is often only about a factor of 2. We investigate how three declustering methods treat induced seismicity: the window-based Gardner-Knopoff (GK) algorithm, often used for PSHA (Gardner and Knopoff, BSSA, 1974); the link-based Reasenberg algorithm (Reasenberg, JGR,1985); and a stochastic declustering method based on a space-time Epidemic-Type Aftershock Sequence model (Ogata, JASA, 1988; Zhuang et al., JASA, 2002). We apply these methods to three catalogs that likely contain some induced seismicity. For the Guy-Greenbrier, AR earthquake swarm from 2010-2013, declustering reduces the seismicity rate by factors of 6-14, depending on the algorithm. In northern Oklahoma and southern Kansas from 2010-2015, the reduction varies from factors of 1.5-20. In the Salton Trough of southern California from 1975-2013, the rate is reduced by factors of 3-20. Stochastic declustering tends to remove the most events, followed by the GK method, while the Reasenberg method removes the fewest. Given that declustering and choice of algorithm have such a large impact on the resulting seismicity rate estimates, we suggest that more accurate hazard assessments may be found using the complete catalog.

Llenos, A. L., & Michael, A. J. (2016, 07). Evaluating the use of declustering for induced seismicity hazard assessment. Poster Presentation at 2016 SCEC Annual Meeting.

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