A Comparison of Electromagnetic Physical Scale and Numerical Modeling for Geophysical Exploration

Abstract

Numerical modeling has, in the twenty-first century, become the dominant form of modeling for electromagnetic exploration geophysics, but few studies have been undertaken that compare numerical models with physical scale models to determine the constraints on numerical algorithms. In this thesis, a physical scale model system was constructed and twenty-six profiles run to analyze the strengths and limitations of four algorithms in PetRos EiKon's EMIGMA V8.6 modeling software: Free Space EikPlate (FS Plate), VH Plate, Inductive Localized Non-Linear (ILN), and EMSPHERE. A low-power vertical-array system with a ten-turn square transmitter loop and receiver coils in the Bx, By, and Bz directions with ten turns, ten turns, and two turns, respectively, was designed and constructed for this thesis. Profiles were taken in either a lab setting, which provided more space, or a tank setting, which allowed for lower noise and modeling of a conductive host. A total of twenty-six profiles taken with ten targets are presented here together with their geometric configurations. Through comparison of the measured and the simulated data, the following conclusions are made: (1) The VH Plate and ILN algorithms produce less accurate simulations for small targets; this may be redressed by increasing the scale of the targets. (2) Every algorithm designed to account for galvanic responses, when otherwise operating within its constraints, does so effectively and corroborates the measured data. (3) FS Plate simulations reasonably approximate the measured responses of long sheet targets when the sheet center is distal to the receiver by more than one third the sheet length, but do not approximate the measured data as well when the center of the target is proximal to the receiver. (4) EMSPHERE, which is only supported for dipole sources, does not approximate a target with a loop source well when the target is directly beneath the transmitter, but it does approximate it well when the target is displaced from the transmitter; a sphere may be approximated for the ILN algorithm by a cube that fits flush within the sphere. (5) FS Plate peak responses tend to be smaller than the measured responses by a factor of two, but the side skirts match well; FS plate also appears to diverge from the Kramers-Kronig relations for induction numbers smaller than 2e-4. (6) VH Plate peak responses tend to be larger than the measured responses by a factor of roughly two, and the side skirts do not match well. (7) ILN breaks down for high conductivity contrasts, such as a graphite cube in air; this issue may be avoided by using a more resistive host for the model so long as inductive effects still dominate, which may be determined using LN.