Abstract
The results of the inversion of the COPROD2 magnetotelluric data set are presented using two implementations of the AIM inverse method. The two algorithms invert magnetotelluric data over 2D conductivity structures. Both algorithms use an approximate inverse mapping based on approximate sensitivities that arise from the 1D conductivity profile beneath each station; this avoids the large computations normally required to approximate the exact inverse mapping, for example, by using the full 2D Jacobian. The primary difference between the algorithms is that the first algorithm minimizes the l1 norm of the data misfit with regularization provided by minimizing the l1 norm of the conductivity. This problem is solved with standard linear programming techniques. The second algorithm uses the l2 norm in place of the l1 norm and is solved using a subspace approach., Both algorithms produce models with predicted data in satisfactory agreement with the COPROD2 data. The difference between the final models is evidence of the well known nonuniqueness in geophysical inverse problems; the similarities of the final models give a suggestion of that which might be common to all models which fit the data.
