[1] We study the applicability of deterministic strong ground motion simulations at very near fault distances for a subvertical strike-slip fault model corresponding to the 2004 M6 Parkfield, California, earthquake in the frequency range up to 1 Hz. Theoretical modeling under the assumptions of a planar rupture and 1-D medium shows that as a consequence of the S wave radiation pattern, the particle motion for such close stations should be almost linear in the fault-normal (FN) direction, having fault-parallel (FP) and vertical (V) components of almost zero. However, as shown on the Parkfield earthquake recordings, observed particle motions are rather circular with peak velocities at FP and V components comparable to those at FN components. We investigate several realistic features that could explain this controversy, namely, nonplanar fault, realistic three-dimensional (3-D) medium, and the topography of the area. We test and quantify these hypotheses using discrete wave number and discontinuous Galerkin modeling methods applying 1-D and 3-D velocity structures, respectively, and two nonplanar rupture models. We compare the synthetic and observed particle motions and peak velocity ratios and conclude that deviations from a planar rupture geometry in reasonable bounds for the Parkfield fault and the influence of topography only partially explain the behavior of the observed seismograms. On the contrary, the heterogeneous 3-D velocity structure significantly reduces the synthetic peak ratios to values closer to 1 and provides rather circular particle motions. Therefore, the 3-D velocity model is crucial to obtain realistic estimates of ground motions at near-fault distances and is more important than the detailed fault geometry or topography in the Parkfield area.Citation: Gallovič, F., M. Käser, J. Burjánek, and C. Papaioannou (2010), Three-dimensional modeling of near-fault ground motions with nonplanar rupture models and topography: