A high-resolution nested global-Japan coastal FVCOM system was used to simulate the March 11, 2011 Tōhoku M9 earthquake-induced tsunami waves and coastal inundation along the northeastern coast of Honshu Island in the western Pacific Ocean. Experiments were made with initial fields provided by five seismic rupture models under realistic conditions with inclusion of the Kuroshio, tides and wind forcing. Results show that the model-computed intensities and distributions of tsunami waves and subsequent coastal inundation could be significantly influenced by initial conditions, even though all five cases were capable of reproducing key features of the tsunami waves. Modeled tsunami waves featured a low dispersive, weakly nonlinear long wave controlled by hydrostatic dynamics. Non-hydrostatic effects only became significant when tsunami waves reached the inner shelf and the amplitude of the leading tsunami wave grew within O(1) of the local water depth (10m or shallower). In both hydrostatic and non-hydrostatic cases, significant mixing occurred when the ratio of wave amplitude to local water depth grew to about 0.25 or greater. Model-predicted run-up was in good agreement with 2-D N-wave analytical solutions on the northern coast around South Iwate where inundation was small, but not in the central Sendai coastal region where inundation was large and 3-D wave dynamics became significant. The experiments suggest that once local bathymetry is accurately configured and the intensity and shape of the initial bottom movement can be predicted, this nested FVCOM system is capable of making accurate predictions of tsunami waves and coastal inundation.