Comparison of seismic and electromagnetic inversion at the Idaho Springs tunnel detection test facility, Idaho Springs, Colorado

Abstract

Eight second sweeps were used with frequency varying from 80 Hz to 420 Hz. The recorded signals Crosshole variable frequency electromagnetic and were then filtered, cross-correlated with a source shear wave seismic data sets were collected across a reference signal, deconvolved and five-fold stacked known mine tunnel at the Idaho Springs Tunnel to create sections from which shear wave travel times Detection Test Facility in Idaho Springs, Colorado. were picked. A total of 250 usable travel times ware The data sets were inverted using simultaneous obtained out of 318 stacked traces. iterative reconstruction techniques (SIRT) and damped least-squares inversion. Resulting models for SEISMIC INVERSION RESULTS seismic velocity did not detect the tunnel due to the attenuation of the high frequencies but did exhibit The travel times were inverted using two methods: linear trends in the geology. The electromagnetic a straight-ray simultaneous iterative reconstruction results showed both the tunnel anomaly, arising from technique (SIRT) using a code obtained from the U. S. conducting material in the tunnel, and linear trends Bureau of Mines (Tweeton, 1988); and a damped leastsimilar to those seen in the seismic data. squares method (Levenberg, 1944; Marquardt, 1963) Comparison of the inversion results with geology that also employs a straight-ray approximation. indicates that the linear trends match mineralized Parameterization of the two-dimensional velocity fracture zones passing through the tunnel. Using model varies between the two methods. The tomography multiple geophysical methods not only provides used 1782 constant velocity 1 m square cells with greater confidence in the interpretation, but can smoothing applied so that no cells were unaffected by also give successful results if one of the methods passing rays, even if no rays passed directly through fails. the cell. The least-squares method, on the other hand, used 209 defined node points on a 3 m by 3 m INTRODUCTION grid with velocities interpolated between them. Overall good resolution, except near the top and In the summer of 1989, a suite of crosshole bottom of the model, was indicated by resolution variable frequency electromagnetic (EM) and shear matrices obtained from the least-squares inversion. wave seismic data was collected surrounding a known Velocity models derived from each of the mine tunnel at the Idaho Springs Tunnel Detection inversion methods are shown in Figures 2 and 3. Both Test Facility operated by the Colorado School of methods show similar gross structural features. Mines. The experiment was designed to test the Unfortunately, the tunnel does not appear evident. capabilities of the two methods and to determine how Considering the good resolution of the data set, we well they would complement each other in the believe this is due to the attenuation of the high interpretation of tunnel detection data sets. frequencies resulting in a dominant frequency of The site is located in a terrain of folded, about 150 Hz for the crosscorrelated wavelet. The interlayered gneisses with varying amounts of tunnel, therefore, has a diameter of only about l/4 biotite, pegmatites, and schists. Fault and shear wavelength and cannot be resolved with these zones give rise to localized highly-fractured inversion techniques. regions. These regions commonly contain metallic There are similar features that show up in both sulfide mineralizations. The tunnel itself is from 3 velocity models. These include a low velocity (.75 m to 4 m in diameter at a depth of approximately 46 m to 1.25 km/s) region down to 30 m depth, a high . below the surface. The tunnel has a flat floor and velocity area between 32 m and 42 m out to 10 m