Imaging of transient electromagnetic soundings using a scaling approximate fréchet derivative

Abstract

The imaging algorithm is fast, robust, fully automated, and requires no initial model. The computation time, including the transformation from dH/dt to H-field, is approximately 0.5 sec/sounding/Mflop, which is 2–3 orders of magnitude faster than conventional least squares iterative inversion. The imaging produces models which fit the original data typically within 5–15%.

The imaged models may be used as very good input models to an iterative least squares inversion program, can be implemented as an online interpretation in TEM instruments, and the imaging procedure lends itself readily to AIM strategies (Oldenburg 1991). Contoured model sections based on imaged models from soundings along profile lines give a very fast insight in the subsurface conductivity distribution.

The principles of the imaging algorithm based on the Fréchet kernel can be extended to the 2D and 3D case of transient data. In fact, the idea of an instantaneous scaling of the Fréchet kernel according to some “average value” of the investigated parameter is applicable in other areas, as long as an “average parameter” can be defined and if the Fréchet kernel depends on the investigated parameter.