Joint inversion of IP-affected TEM data

Abstract

A numerical experiment was applied to explore the potentialities and limitations of joint inversion of IP-affected transients measured with different loop configurations above a uniform half-space with a Cole-Cole complex conductivity. One of us calculated 200 m × 50 m and 50 m × 50 m loop responses of a uniform polarizable conductor with varied Cole-Cole parameters and imposed synthetic Gaussian noise that simulated measurement errors. Then the generated pseudo-experimental data passed to the other co-author who performed single and joint 1D inversion twice: first being unaware of the “true” underlying models and then after being told that they all were represented by a uniform polarizable earth. More than a half of the fitted models provided a good idea of the true models though misfit was quite large in some cases. The fit was better in single inversion with a priori information available, and improved further through joint inversion of central-loop and coincident-loop responses. Joint inversion with a priori information known was of good quality even at a chargeability as low as 0.02. The standard error in joint inversion was times the measurement error and depended mainly on fitting errors for smaller-loop data. The reason is that the smaller-loop transients included a non-monotonous interval where the signal changed rapidly under the effect of fast-decaying induced polarization.