HR: 1340h
AN: H13C-1342    [Abstracts]
TI: Cross-Gradients Joint Inversion of Disparate Geophysical Data for Characterization of Heterogeneous Environments
AU: * Gallardo, L A
EM: lgallard@cicese.mx
AF: Earth Science Division, CICESE, km 107 Carretera Tijuana-Ensenada, Ensenada, 22860 Mexico
AB: Reliable characterization or discrimination of different materials in complex heterogeneous environments such as fractured or contaminated sites is only possible when complementary geological, geochemical or geophysical information is combined to produce an integral model of the subsurface. A common practice in geophysics is to estimate independent geophysical images for each data type and use these images to characterize the different materials in the subsurface. The success of this practice is limited, because the individual images are non-unique and some features of the particular images selected as optimal for each data type may not be compatible and, therefore, not appropriate for an integral model. In a joint inversion approach, the geophysical data are simultaneously inverted to produce only compatible images that can reduce significantly the misleading features in the final integrated model. A key issue in joint inversion is how to define objectively the compatibility of the images. For instance, in relatively homogeneous geological environments there are some petrophysical relationships that may gauge the compatibility of two geophysical parameters such as electrical resistivity and seismic velocity. In heterogeneous environments, however, such relationships are either complex or inexistent. I follow a different philosophy and posit that the geometrical distribution of the underlying materials can be used as a common factor for joint inversion. I developed algorithms for the joint inversion of disparate geophysical data that seek for two-dimensional images of the subsurface that are in geometrical accord. In the objective function, the geophysical data fit is defined in a least squares sense whereas the geometrical resemblance is defined by constraining the coupled cross product of the gradients of their physical properties to be zero. The objective function also includes smoothness and ridge regression-type constraints. The developed algorithms account for the joint inversion of DC-resistivity, magnetotelluric, compressional and shear wave travel time data and are tested on diverse synthetic and field data. The results show that the joint inversion using cross-gradients finds more accurate images of the subsurface with largely improved geometrical resemblance where geophysical relationships can be better distinguished.
DE: 0902 Computational methods: seismic
DE: 0925 Magnetic and electrical methods (5109)
DE: 1835 Hydrogeophysics
DE: 3260 Inverse theory
DE: 7270 Tomography (6982, 8180)
SC: Hydrology [H]
MN: Fall Meeting 2005