A multimodel method for depth estimation from magnetic data

The local wavenumber and a multimodel wavenumber are complex attributes derived from a complex analytic signal. These quantities have been used to interpret anomalies arising from contacts, thin sheets, and horizontal cylinders. A new multimodel wavenumber can be used for computing depths of 2-D thick dikes and 2-D sloping steps. These two multimodel wavenumbers have been incorporated into a depth-estimation algorithm based on automatic curve matching. This algorithm works on profile data and has three appealing features: (1) the most appropriate of these five models is selected automatically; (2) the automatic curve matching uses a least-squares technique to reject responses that do not conform to the model assumptions; and (3) interference from distant sources can be accounted for as a base-level shift of the multimodel wavenumber curves. Applying the automatic technique to survey data from the Western Canada sedimentary basin yields four thick dikes between 3400 and 4300 m below sensor. These depths are equivalent to 2.2 and 3.1 km below sea level, which is consistent with the basement depths derived from drillhole information. Using these solutions as a starting point in an iterative forward modeling exercise, the measured data were explained with a geologically reasonable model.