22.11.2012 Views

Airborne Gravity 2010 - Geoscience Australia

Airborne Gravity 2010 - Geoscience Australia

Airborne Gravity 2010 - Geoscience Australia

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

<strong>Airborne</strong> <strong>Gravity</strong> <strong>2010</strong><br />

where is the vector of model parameters (densities, ) of length ; is a vector of observed<br />

data of length ; and is a rectangular matrix of size formed by the corresponding<br />

kernels. The integral form of the gravity fields and their gradients are convolutions of the<br />

corresponding Green’s functions and density inside each cell. As the background density is assumed<br />

to be homogeneous and the Earth model is discretized into a regular grid, then becomes a Toeplitz<br />

or block-Toeplitz matrix, whose number of different entries is much less than the number of elements.<br />

This makes storage of more economical. Moreover, matrix multiplications involving are of an<br />

order of rather than . This results in considerable speed up for iterative inversion<br />

based on repetitive matrix multiplications (Zhdanov, 2002).<br />

Regularized Focusing Inversion<br />

<strong>Gravity</strong> field and gradient data is reduced to the solution of the matrix equation (21). This inverse<br />

problem is ill-posed; i.e., the solution can be non-unique and unstable. Therefore, we have to use<br />

methods of regularization in order to recover unique and stable density models (Tikhonov and Arsenin,<br />

1977; Zhdanov, 2002). This is achieved by minimizing the Tikhonov parametric functional, :<br />

where is a misfit functional, is a stabilizing functional, and is a regularization<br />

parameter that balances (or biases) the misfit and stabilizing functional. The stabilizing functional<br />

incorporates information about the class of models used in the inversion. Equation (22) can be<br />

minimized any number of ways. The inverse problem is a linear one. We choose to use the reweighted<br />

regularized conjugate gradient method (Zhdanov, 2002), since it reduces the iterative scheme to a<br />

series of matrix-vector multiplications that can be evaluated very quickly with FFT-based matrix-vector<br />

multiplications.<br />

The choice of stabilizing functional should be based on the user’s geological knowledge and prejudice.<br />

Traditionally, the stabilizers have been based on minimum norm or first or second derivatives of the<br />

density distribution, which recover models with smooth density distributions (e.g., Li and Oldenburg,<br />

1998). Smooth models bear little resemblance to economic geology. Moreover, smooth stabilizers can<br />

result in spurious oscillations and artifacts when the density distribution is discontinuous. As such, it is<br />

useful to search for unique and stable solutions within those models that have sharp density contrasts.<br />

This can be accomplished by introducing the so-called focusing functionals (Portniaguine and<br />

Zhdanov, 1999, 2002; Zhdanov, 2002, 2009b). First, we present the minimum support (MS) stabilizer:<br />

where is a focusing parameter introduced to avoid singularity when . The minimum<br />

support stabilizer minimizes the volume with non-zero departures from the a priori model. Thus a<br />

smooth distribution of all model parameters with a small deviation from the a priori model is penalized.<br />

A focused distribution of the model parameters is penalized less. Similarly, we present the minimum<br />

gradient support (MGS) stabilizer:<br />

which minimizes the volume of model parameters with non-zero gradient.<br />

Case study – Nordkapp Basin<br />

The Nordkapp Basin is located in the Norwegian sector of the Barents Sea, and is an intra-continental<br />

salt basin containing over 30 salt structures. The salt is of an Early Permian age, and was mobilized<br />

by Early Triassic sedimentation. Tertiary uplift and erosion removed nearly 1400 m of Cretaceous and<br />

younger sediments (Neilsen et al., 1995). The petroleum plays are mainly salt-related traps. Only two<br />

199<br />

(22)<br />

(23)<br />

(24)

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!