2D inversion modeling of synthetic data is used to evaluate the performance of five electrical resistivity arrays, with the primary criteria being the reproduction of sharp model boundaries. 2D synthetic noise free data were calculated simulating a modern fixed spacing multi-electrode cable. Twelve 2D synthetic models, resembling a number of different geologic situations, were used to investigate performance of the dipole-dipole, pole-dipole, pole-pole, Wenner and Schlumberger arrays

Although the synthetic, noise-free data were well matched for all inversions, many of the inversion results exhibit substantial mismatches from the true model. The greatest resistivity mismatches are near model discontinuities. Resistivity mismatches become worse with depth and the geometry of geologic boundaries in the deep portion of the models are not well reproduced by any of the arrays. Field surveys must be designed so that the geologic target is in the middle of the data constrained region. Different arrays performed best for different models and a practical table is presented allowing the practitioner to choose the optimal array for the particular geologic situation under investigation. Although the dipole-dipole and pole-dipole arrays may not be the optimal array for a given geology, they rarely fail for any model, and thus are recommended for reconnaissance or preliminary investigations in regions of unknown geology.

Contrary to traditional advice found in textbooks, based on 1D profiling and sounding, and data plot comparison, this study, using 2D data and 2D inversion, finds the Wenner and Schlumberger arrays, thought to perform poorly for vertical boundaries, performed well for a vertical boundary and a thin vertical resistor. Similarly, the dipole-dipole and pole-dipole arrays, thought to perform poorly for horizontal and dipping boundaries, performed well for several models containing those geometries. Another interesting finding is that changing the polarity of geologic units from resistors to conductors changed relative array performance in most models.