Large-scale, full-wave modeling of multistatic target imaging in a rough ground environment is described. The emulation methodology employs a parallelized three-dimensional “near-field” finite-difference time-domain algorithm in characterizing the electromagnetic scattering from the ground surface and buried and on-surface targets in the form of landmines and unexploded ordnances; subsequent focusing of the scattered fields into an image is obtained by exploiting the time-reversal technique applied over a spectral band consistent with one used by a step-frequency system under development. Imaging performance is investigated with numerical experiments for both single- and multi-aperture sensing geometries. The emphasis of this study is on examining the responses of discrete ground targets in the presence of distributed variable ground clutter as relevant to performance prediction for ultra-wideband forward-looking radar applications.