Two years stress drop estimates for induced earthquakes in Oklahoma

Nana Yoshimitsu, William L. Ellsworth, Gregory C. Beroza, & Martin Schoenball

Submitted August 15, 2017, SCEC Contribution #7729, 2017 SCEC Annual Meeting Poster #047

Stress drop is important factor for a better understanding of ground motion intensity, which is commonly calculated from the seismic moment and corner frequency by comparing theoretical models to data in the frequency domain. In this study, we focus on assessing the fit to the data across the measurable frequency range to understand the trade-offs among the parameters of the theoretical model.

We gathered co-located events within 1 km from large (Mw >= 4) event in Oklahoma to apply eGf spectral ratio approach. There are 64 clusters obtained from relocated highly accurate event list from May 2013 to November 2016 (Schoenball and Ellsworth, 2017). We formed spectral ratios of 100 sps waveforms between a large and eGfs event pair by stacking all components of four USGS/OGS stations. To suppress the influence of directivity and radiation pattern, we analyzed 5.12 seconds of S-coda wave later than 2.5 times the S-wave travel time. The best model was estimated using least square grid search by fitting Brune spectral model to the data over a frequency range of 1 to 40 Hz. We evaluated the trade-off among three unknown parameters, two corner frequencies and the moment ratio, by comparing the misfit at each grid point with the misfit of the best model. We discarded event pairs with a large trade-off between any of two parameters. In addition, we assessed the residual variation by dividing the residual into five parts with equal log frequency intervals. If the residuals in any frequency band show a large trend (slope) we also discard the event pair.

After assessing the suitability of the eGf with above procedures, the variation of the estimated stress drop in our data set decreased. For example, the stress drop in one cluster varied between 4 to 511 MPa using all available eGf events, but reduced to 35 to 64 MPa after the eGf event selection. Considering these results, we conclude that event selection using trade-off and residual slope increases stress drop estimation accuracy.

Key Words
Stress drop, induced earthquakes, Oklahoma

Citation
Yoshimitsu, N., Ellsworth, W. L., Beroza, G. C., & Schoenball, M. (2017, 08). Two years stress drop estimates for induced earthquakes in Oklahoma. Poster Presentation at 2017 SCEC Annual Meeting.


Related Projects & Working Groups
Seismology