Spatially Constrained Inversion for Quasi 3-d Modelling of Aem Data

Airborne electromagnetic (AEM) surveys carried out around the globe produce thousands of km of data every year. Because of the enormous computational costs involved in a full 3D inversion, these data are usually inverted using a 1D forward model. The 1D model assumption is legitimate in quasi layered sedimentary areas. In some cases the output models are simply stitched together (Auken et al., 2003; Huang and Fraser, 2003), resulting in abrupt variations in neighboring models, because of noisy data and equivalent models. This is a non-optimal result for sedimentary environments where the lateral variations are expected to be relatively smooth. Models with smooth lateral variations can be achieved either working in the data domain, or in the model domain. The first approach, widely used both with frequency and time domain airborne EM data, entails smoothing the raw data before inversion. In the second approach, constraints are applied between adjacent output models during the inversions. There is not equivalence between working in the data or the model domain and it is always should always be preferred to work in the model domain. Examples of constrained inversions methodologies are the laterally constrained inversion (LCI) of galvanic, (Auken and Christiansen, 2004), and EM data (Santos, 2004), and the simultaneous inversion method of galvanic data (Gyulai and Ormos, 1999). Each of these processing or inversion techniques is profile-oriented, in the sense that it aims at producing a continuum along a line. However, they don't create any connection between neighbouring lines. Features that are perpendicular to lines benefit only partially from in-line constraints or averaging, as there no information is passed between adjacent lines. This means that profile oriented techniques favour structures following the flight direction. Producing area maps based on such methodologies often results in some lineation following the fight paths. In this abstract we expand the concept of along-profile, laterally constrained inversion (LCI), to the Spatially Constrained Inversion (SCI), which operates both along and across profiles.