Airborne Gravity 2010 - Geoscience Australia

Airborne Gravity 2010
Inversion for density and susceptibility is performed using a Minimum Mean Square Error (MMSE)
method. A statistical approach is used with the MMSE inversion method to obtain the density or
susceptibility models (Sæther, 1997; Haase, 2008). Every density or susceptibility value is treated as
an ’expected value’ with its uncertainty as the ’variance’. In an MMSE inversion, the most probable
densities and variances are derived such that the differences between observed and measured data
are minimised. The advantage of this approach is that the inversion is well-behaved and has a unique
solution. For future application, we are working on the implementation of evolutionary algorithms for
three-dimensional inversion of potential fields and their derivatives (gradients). Evolutionary algorithms
would be used to optimize objective functions for three-dimensional modelling of susceptibility, density,
and conductivity distributions.
Model construction
The process of developing a 3D model with complicated topological structures is a laborious task and
does not easily fit into modern work flows where users demand quick answers and decisions. IGMAS+
provides some special features to minimise this problem:
� The geometry of a model is defined using parallel vertical planes. This configuration allows
easy automatic triangulation as well as visualization in the form of cross sections.
� Semi- or fully automated import of predefined models or information. An import option for
predefined seismic horizons has recently been developed.
� Use of standardized data exchange formats that allow integration of work carried out with
IGMAS+ into complex workflow environments that require the use of several specialized
computer software packages.
Interpreted seismic horizons are generally defined using a large number of irregularly spaced points.
In comparison, the modelling resolution of potential fields is quite modest. If not down sampled, these
picked points will not only inflate the storage size of the model structure, but will place unnecessary
demands on computational resources to perform potential field modelling calculations.
Down sampling may be achieved through a variety of procedures. In IGMAS+, the “Binned Average
Values” algorithm is used. The model space is subdivided into cells. An average value is calculated for
all of the points within each gridded cell. The user can visually control the degree of generalization by
adjusting the cell size of the filter.
Since IGMAS+ needs a polygonal geometry on each of the cross sections of the 3D structure, each
horizon surface has to be cut with vertical sections. The cutting algorithm is very fast and may be
applied to a Delaunay-triangulated surface directly or after an intermediate step of gridding. This
produces a more generalized and smoothed initial model geometry.
Conclusions
IGMAS+ is a software tool which can be used for integrated interpretation in today’s industrial oil and
gas exploration projects. IGMAS+ allows interactive gravity and magnetic modelling to play an
important role in the depth imaging workflows of these complex oil and gas projects (Figure 4).
Integration of workflow and tools is important to meet the needs of today’s more interactive and
interpretative depth imaging techniques.
Current research activities focus on the use of natural language processing (NLP) techniques to
extract semantic constraints from various texts such as reports, publications and books. The aim is to
close the gap between data modelling and “intellectual summaries/descriptions” that are used in free
text documents.
94