Analyses of traveltimes and amplitudes of crosshole georadar data
provide estimates of the electromagnetic velocity and attenuation
of the probed media. In contrast to inversions of traveltimes, which
are well established and robust, ray-based inversions of amplitudes
depend critically on a priori assumptions about the directive properties
of the antennas. In particular, the influence of electric material
property fluctuations on antenna performance may lead to serious
distortions of the radiation pattern. Such distortions cannot be
accounted for by currently employed ray-based amplitude inversion
algorithms. To explore the problem of antenna coupling to local heterogeneities,
we generate synthetic crosshole georadar data for a suite of stochastic
models using a finite-difference time-domain (FDTD) solution of Maxwell's
equations in cylindrical coordinates. Analyses of the radiation patterns
extracted from the synthetic data indicate that distortions of the
radiation patterns are primarily due to propagation effects and not
to dipole-coupling effects. We do, however, find that the quality
of the amplitude tomograms diminishes rapidly with increasing heterogeneity,
probably because of inherent inadequacies in ray-based inversion
methods. In contrast, the quality of ray-based traveltime tomograms
is surprisingly high, even in the presence of strong heterogeneity.