![Hier klicken, um den Treffer aus der Auswahl zu entfernen](images/unchecked.gif) |
Titel |
How to select a prior subsurface covariance model from indirect geophysical data observations |
VerfasserIn |
T. M. Hansen, M. C. Looms, L Nielsen |
Konferenz |
EGU General Assembly 2009
|
Medientyp |
Artikel
|
Sprache |
Englisch
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 11 (2009) |
Datensatznummer |
250026565
|
|
|
|
Zusammenfassung |
Least squares type inversion is applied to infer information about the subsurface for
many types of near surface geophysical data (e.g. electric, electromagnetic, gravity
data). Such inverse problems are typically under-determined inverse problems, and
therefore additional information needs to be provided. For least squares type inversion
problems, such information is given in form of a covariance model (CM) describing the
prior assumptions of the variability between sets of points in the model parameter
space. In recent years, stochastic least squares inversion methods have been used for
more wide spread applications. Here the solution to the inverse problem is not just
one smooth model, but a set of realisations consistent with data observations and
the prior CM. For such methods the choice of prior CM plays a very important
role. Sometimes one may have knowledge about the subsurface covariance model
from for example reflection georadar/seismic profiles, and well logs, from which
one can try to infer an appropriate prior CM. In other cases such additional data
are not available, and it can be difficult to choose a prior CM. We have recently
proposed a method to identify the prior CM most consistent with data observations
for least squares types of problems. We present the methodology and an analysis
of the method for inferring properties of the subsurface covariance from indirect
geophysical measurements. We illustrate that for a synthetic cross-borehole georadar
tomography example we are able to infer the properties of the subsurface with an accuracy
comparable to a traditional semivariogram analysis of the actual subsurface model
(which in a real case is unknown). We also illustrate the effect such a choice of
prior CM has on both stochastic realisations and the least squares mean estimate. |
|
|
|
|
|