![Hier klicken, um den Treffer aus der Auswahl zu entfernen](images/unchecked.gif) |
Titel |
Fréchet kernel sensitivity |
VerfasserIn |
Tarje Nissen-Meyer, Alexandre Fournier |
Konferenz |
EGU General Assembly 2011
|
Medientyp |
Artikel
|
Sprache |
Englisch
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 13 (2011) |
Datensatznummer |
250057030
|
|
|
|
Zusammenfassung |
The mapping from data to model ("sensitivity kernels") forms the basis for seismic inverse
problems, and has received much attention in recent years as developments in theoretical
foundations, computational resources, and high-quality broadband data converge to tackle
geophysical scales of interest. Although much progress has been achieved, it is clear that the
generally non-unique inversion of seismic data still poses a formidable validity challenge on
all ends.
We quantify the variability and scope of subjective choices one takes upon
a certain data and model selection as illuminated by the dependence of Fréchet
sensitivity kernels on a multi-dimensional parameter space. We apply this to the
global scale by factorizing the raw waveform sensitivity upon spherically symmetric
reference models from all specifications such as source time functions and radiation
patterns, receiver components, azimuth, epicentral distance and source depth, model
parameterization, time and frequency windows, and a range of misfit functions. Having such
spatio-temporal/spectral sensitivity kernels at hand, these choices are computationally trivial
to undertake, being mere operators on the waveform kernel. Specifically, we depict the
imprint of data/synthetic processing onto the sensitivity kernel as illuminated by
traveltime picking, cross-correlations, complete waveforms, and phase-/envelope
misfits.
It is worth assessing these dependencies for two reasons on either end of the tomographic
workflow: 1) To choose appropriate a priori sub-sets of the data that significantly illuminate a
given region of interest; 2) To understand the scale of influence on the mapping as induced by
these choices with respect to the a posteriori tomographic image. Point 1) can be seen as a
semi-automated data selection process, which, in times of the availability of vast
amounts of digital data, may streamline subjective and time-intensive previous
efforts. Point 2) touches upon the non-uniqueness and illposed-ness of seismic
inverse problems: Subjective choices on the above-mentioned parameter space
have an inevitable and quantifiable effect on the inversion and "the" final model.
Thus, we argue that multiple inversions for many such choices may help in better
constraining the nature and reliability as well as uncertainty of tomographic images. |
|
|
|
|
|