S51A-2636
Using the Gauss-Newton Method to Invert for Brune Model Moment, Corner Frequency, and Kappa Parameters: Results from the Canterbury, New Zealand Earthquake Sequence

Friday, 18 December 2015
Poster Hall (Moscone South)
Corrie Neighbors1, Elizabeth S Cochran2, Kenny J Ryan1 and Anna E Kaiser3, (1)University of California Riverside, Riverside, CA, United States, (2)Organization Not Listed, Washington, DC, United States, (3)GNS Science-Institute of Geological and Nuclear Sciences Ltd, Lower Hutt, New Zealand
Abstract:
The seismic spectrum can be modeled by assuming a Brune spectrum and estimating the parameters of seismic moment (M0), corner frequency (fc), and the high frequency site attenuation (κ). Traditionally studies either hold fixed or use a predefined set of trial values for one of the parameters (e.g., fc) and then solve for the remaining parameters. Here, we use the Gauss-Newton nonlinear least-squares method to simultaneously determine the M0, fc, and high-frequency κ for each event-station pair.

We use data collected during the Canterbury, New Zealand earthquake sequence. The seismic stations include the permanent GeoNet accelerometer network as well as a dense network of nearly 200 Quake-Catcher Network (QCN) MEMs accelerometers installed following the 3 September 2010 M 7.1 Darfield earthquake. We examine over 180 aftershocks ≥ Mw3.5 that occurred from 9 September 2010 to 31 July 2011 and are captured by both networks.

We use Fourier-transformed S-wave windows that include 80% of the S-wave energy and fit the acceleration spectra between 0.5 and 20 Hz. We apply a path and site correction to the data as described in Oth and Kaiser (2014). Then, the records are smoothed using a Konno and Omachi (1998) filter and uniformly resampled in log space. Initial “best guesses” for M0 and fc are determined from GNS catalog magnitudes and by assuming a 100 bar (10 MPa) stress drop and an initial κ is determined from an automated high-frequency fit method. We use a parametric inversion technique that requires a single M0 and fc for each event, while κ is allowed to vary by station to reflect varying site conditions. Final solutions for M0, fc, and κ are iteratively calculated by minimizing the residual function. After Brune (1970, 1971), the stress drop is determined from the best-fit fc.

Moment magnitudes determined agree well with the GNS catalog, with a median difference of 0.12 Mw and 0.20 Mw for GeoNet and QCN inversions, respectively. Stress drop results are within the range reported by Oth and Kaiser (2014) who use a non-parametric inversion of GeoNet data.

<< Previous Abstract | Next Abstract >>