Abstract
Total attenuation (Q(t)(-1)) in ground-penetrating radar (GPR) data is a composite of intrinsic and scattering attenuations (Q(in)(-1) and Q(sc)(-1)). For nonmagnetic materials, Q(in)(-1) is a combination of the effects of real conductivity and dielectric relaxation. The attenuation for real conductivity >1.0 mS/m in the GPR frequency band is a function of frequency while the dielectric relaxation is frequency-independent. These frequency behaviors allow separation of the attenuation types by attributing and fitting the Q(t)(-1) decay shape with frequency to the conductivity, and by attributing the magnitude of Q(t)(-1) to the sum of conductivity and dielectric relaxation attenuations at each frequency. Total attenuation is calculated from GPR data using spectral ratios, and Q(in)(-1) is obtained by fitting a smooth lower bound to Q(t)(-1); the difference between Q(t)(-1) and Q(in)(-1) estimates the scattering contribution Q(sc)(-1). Scatterer size spectra are evaluated using KA = 1 for 2D, and KA = 1.5 for 3D, propagation (where K is wavenumber and A is the scatterer size). We illustrate with 2D synthetic data and three field 2D crosshole profiles from an outcrop of an Ellenburger collapsed paleocave environment in central Texas. Between the three pairs of holes, we estimate the breccia sizes from the. scattering spectra Q(sc)(-1). To image the anisotropic electrical conductivity distributions, we use simultaneous iterative reconstruction tomography. There is a correlation between the low wavenumber features of the results Of the. current conductivity tomography and those in previous velocity tomography, and with surface data results that are predicted and calculated from GPR data attributes. Low- and high-conductivity zones tend to follow either the GPR facies distributions, lithological boundaries, or the larger of the fractures. Correlations are not visible Where the breccias are finer because these tend to be more randomly oriented, and/or below the resolution of the GPR data.