Magnetic Interpretation By Scanning Multiple Models For Best Least Squares Fit

Magnetic interpretation is generally accomplished by comparing discrete anomalies in the<br>residual field with theoretical profiles of reasonable models. Simplified ‘rules of thumb”<br>based upon sets of model curves are also employed. Direct analytical inversion of field<br>data is impractical due to non-uniqueness and computational difficulties. The problem is<br>that even the simplest models may be adjusted in depth, horizontal position, and<br>orientation. More complicated models may also be varied in two- or three-dimensional<br>shape, and may involve multiple bodies. It is therefore very difficult to optimize model<br>parameters and measure the degree of fit or agreement with field data.<br>We have developed a procedure called MAGFIT for automatically comparing actual and<br>model profiles, and calculating a best least squares fit. With our algorithm, it is possible to<br>rapidly scan a simple set of models (such as dipoles) over many locations and orientations,<br>and determine the best fit as a function of position (and/or other model variables). “Best<br>fit” plots generally show closed contours around a best tit of 60% to 90%, identifying the<br>likely locations of magnetic bodies.