Electronic Supplement to
A New Strategy to Compare Inverted Rupture Models Exploiting the Eigenstructure of the Inverse Problem

by F. Gallovič and J.-P. Ampuero

Table S1. Inverted Models of the SIV2a Benchmark with Descriptions of Applied Inversion Techniques.

Model name; Method Description Data processing
Gallovic0.01; Gallovič et al. (2015) Linear multitime window inversion approach with long duration of slip rate functions (equal to the assumed duration of the rupture process). Constraints: (1) smoothing by means of a prior k-2 covariance functions, and (2) positivity of the slip rate function. The smoothing weight is relatively small (perhaps not applicable in real-data application). Butterworth bandpass filter in range of 0.05–0.5 Hz (four poles, causal)
Gallovic0.1; Gallovič et al. (2015) Same as Gallovic0.01 but with more severe smoothing (similar to that used in real-data applications). Same as Gallovic0.01
Hoby; Razafindrakoto and Mai (2014) Parametric (single-time window) inversion assuming triangular slip rate function. Parameters: rupture times, rise times, and peak slip rates. Metropolis algorithm is used to optimize the parameters considering L2 norm. Butterworth bandpass in range of 0.01–1 Hz
CedricT3; Twardzik et al. (2012) Simplified source model considering two elliptical subfault patches, triangular slip rate function, and constant rupture velocities along the patches. Parameters: location and size of the ellipses, rupture velocities, and onset times of the subfault patches. Neighborhood algorithm is used to find the best fitting parameters considering L2 norm. Butterworth bandpass filter in range of 0.1–1.0 Hz (four poles, two passes—acausal)
Asano; Sekiguchi et al. (2000) Multitime window linear inversion with spatiotemporal smoothing constraint. Weight of smoothing is determined by minimizing Akaike Bayesian information criterion. Bandpass filter in range of 0.05–1.0 Hz

[ Back ]