Laboratory study of the cross-hole resistivity tomography: The Model Stacking (MOST) Technique

Model experiments in the laboratory are used to find the optimal measuring and interpretation parameters that affect the quality of the results for the improved application of resistivity tomography between boreholes or between the surface and a tunnel. The experiments have shown that the detection ability of each crosshole electrode array is different and depends on the form of the sensitivity pattern. The bipole-bipole array, with current and potential pairs in different boreholes, has quite low signal and very good resolution of confined bodies in homogeneous half-spaces, but the resolution decreases as the half-space becomes more complex. The bipole-bipole array with the electrodes of each bipole in different boreholes produces a stronger signal, causing the models to be greatly influenced, even by the presence of small targets. However, the resolution quality is poor, mainly in the middle of the area between the boreholes. Pole-bipole array shows good resolution of the targets detection, but it is less accurate when conditions become noisier. The pole-tripole array has the best resolution, even when the environment becomes more complex or noisier, but the produced models have also many artifacts. The combined arrays' data inversion yields the greatest influence of the targets on resistivity models, usually with very good positioning or shape resolution, but with many more intense artifacts, since this method inevitably combines the advantages and disadvantages of each array. A new approach to improve the quality of resistivity models has been developed, based on a stacking technique, through the processing of different arrays' inverted models ("MOST" technique). The improvement of the MOST models' quality has been measured, comparing each final resistivity model with the corresponding real one, highlighting this way the efficiency of this technique, in contrary to the combined arrays' data inversion. (C) 2012 Elsevier B.V. All rights reserved.