Towards improved stress drop accuracy
Nana Yoshimitsu, Stanford University
Wednesday, March 29, 2017 at 10:30 AM
- Building 3, Rambo Auditorium
- Annemarie Baltay
In this talk, we introduce the trial to improve the estimation quality of the stress drop using the residual evaluation of the corner frequency estimation process, and show the estimated stress drop of potentially induced earthquakes in Oklahoma. In the last several years, disposal of wastewater increases the seismic activity in central United States, especially in Oklahoma. The similarity or difference of the source characteristics between induced and natural earthquakes would contribute to hazard assessment. While stress drop is one of the most important source parameter, accurate estimation was very difficult. The process in the stress drop estimation need to obtain the corner frequency by comparing the spectral ratio of the co-located event pair and the theoretical model. The accuracy of the corner frequency strongly affects the stress drop estimation since stress drop has proportional relationship to the cube of the corner frequency.
We formed clusters of co-located events which occurred from 2016 March 1st to August 31. The spectral ratios between the smaller events (Mw < 4.0) and the larger event (Mw > 4.0) in each cluster were calculated to remove path effects. Spectral ratios of the all components of 4 stations with 100 Hz sampling frequency were stacked. We analyzed 5.12 seconds from 2.5 times after the S wave arrival time after applied the band-pass filter of 0.1 to 40 Hz. Corner frequencies and moment ratio of each event pair were searched by the least square fitting with Brune model.
To improve the accuracy of the results, we evaluate the quality of the fitting process by using the residuals between the best fitted curve and data. The residual on each data point at each grid search trial demonstrated the large uncertainty of higher corner frequency estimation (search range of a larger event: 0.1 to 40 Hz, smaller event: 0.1 to 60 Hz). We selected the events with the size and shape of the small residual region around the best fitted point. The event pairs which has small residual region show a rounded compact shape around the best fitted point were defined as a good quality, and utilized them for the stress drop calculation. Comparing to the results without the selection, stress drops of selected data in the same cluster showed small variation in three clusters; thus, the quality control with residual would be effective. Estimated stress drops were 1-10 MPa for the smaller eGf events and 6-63 MPa for the larger events. This result is consistent with the stress drop estimated in other regions of central United States as well as natural earthquakes.