 |
| Titel |
More power to kinematic earthquake source inversions: With new tools from mismodelling to uncertainties |
| VerfasserIn |
Sebastian Heimann, Henriette Sudhaus, Rongjiang Wang, Simone Cesca, Torsten Dahm |
| Konferenz |
EGU General Assembly 2014
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 16 (2014) |
| Datensatznummer |
250095329
|
| Publikation (Nr.) |
 EGU/EGU2014-10777.pdf |
|
|
|
|
|
| Zusammenfassung |
| The notorious discrepancies among finite fault slip inversion results have attracted much
attention over the last years. In consequence, much effort has been put into methods to
improve the robustness of such inversions and to quantify uncertainties on results. The
techniques exploited include controlling the smoothness of the inferred slip distribution,
reducing dimensionality of parameter-space, propagation of observational errors through
Bayesian inference, Monte-Carlo modelling and bootstrapping.
The difficulties in earthquake finite source parameter estimation arise from three distinct
origins: (1) observational errors, (2) the (in)ability of the earthquake source model to
represent nature, and (3) mismodelling of synthetic seismograms. While observational errors
can often be formally included in the source parameter estimation process, the latter two are
much harder to to handle. Appropriateness of the source model (2) is hard to achieve
because more realistic models require more model parameters and quickly lead to
underdetermined systems. Mismodelling of synthetic seismograms (3) has not been
investigated much, probably because the technical effort to deal with it is usually
high (because forward modelling may have to be repeated for many earth model
variations).
In this presentation, we will show that freely available precomputed Green’s functions for
ensembles of different earth models will make such investigations feasible for routine
practice. We will illustrate this with a synthetic test case of a regional kinematic source
parameter optimization.
The presented work is closely related with the development of a new open source Python
toolbox for the handling of precomputed Green’s functions and for synthetic seismogram
generation (http://emolch.github.io/pyrocko/gf). Ultimately, we would like to launch a
community driven open access Green’s function sharing platform and web services for
synthetic seismogram and test scenario generation (http://kinherd.org/). |
|
|
|
|
|
|