Inversion of IP-affected TEM responses of a two-layer earth

Abstract

A numerical experiment has been applied to explore the potentialities and the limitations of the inversion of IP-affected TEM responses of a two-layer earth with a polarizable layer above (Model 1) or below (Model 2) a nonpolarizable layer. The IP effects have been incorporated into the models via a Cole–Cole complex frequency-dependent conductivity. One of us computed synthetic in-loop and coincident-loop transient responses with added Gaussian noise, and the other performed single and joint inversion of the two sets of pseudo-experimental data. Model 1 turns out to be advantageous over Model 2 in TEM applications and gives a good fit of the Cole–Cole parameters (chargeability, IP time constant, and exponent) even in the absence of a priori information. In the case of Model 2, the lack of a priori information causes problems as to recognize which layer is polarizable, and the fit of the Cole–Cole parameters is generally worse. The layer thicknesses and resistivities are rather accurate in both groups of models, irrespective of whether a priori information is available. As the upper layer increases in thickness ( H 1), the fit of its parameters ever improves in both models while the parameters of the lower layer, on the contrary, contain a greater error. Joint inversion of in-loop and coincident-loop transients improves the fit in most cases. Relative rms error (σ rel) does not depend on the upper layer thickness for Model 1 but decreases as H 1 increases in the case of Model 2. The error in joint inversion is times that in single inversion, which means that additional criteria other than σ rel may be useful to estimate the inversion quality.