The grain size distribution of a tsunami deposit may have a correlation with the tsunami inundation process, and further with the tsunami source characteristics. We test this hypothesis using thickness and grain size distribution data of the 2011 Tohoku tsunami deposit in Numanohama coast, Iwate Prefecture, Japan. Here, we build and validate a tsunami sediment transport model that can simulate deposit thickness and grain size distribution. Our numerical model has three layers: parent (bed), deposit (bed), and suspended load layers. The two bed-layers contain information about the grain size distribution. This numerical model can handle a wide range of grain sizes from 0.063 (4ϕ) to 5.657 mm (−2.5ϕ). The grain size distributions at 12 sample points along a 900 m transect from the beach, through a marsh, and up to the inundation limit, are used to validate the tsunami sediment transport model. We adopt a reference tsunami source model that can well reproduce the observed tsunami run-up heights ranging from 16 to 35 m along the steep valley during the 2011 tsunami. The simulated sand thickness distribution along the transect is consistent with the observed thickness ranging from 3 to 23 cm. The computed net erosion and deposition suggest that most of the sand deposit was originated from the near shore. The shapes of the simulated grain size distributions represented by their sorting, skewness, kurtosis, and mean at most of the sample sites are similar to the observations. The differences between the observed and simulated peak of grain size distributions are less than 1ϕ. To evaluate the sensitivity to the tsunami source model, we test five tsunami scenarios which are modified from the reference source model. While the tsunami scenario with 120% of the reference amplitude can also reproduce the thickness and grain size distribution, the scenarios with amplitudes smaller than 80% of the reference or with wave periods shorter than 50% of the reference source model underestimate the thickness and cannot reproduce the grain size distributions. Our simulation results suggest that it is possible to estimate tsunami wave amplitude and wave period from sediment deposit thickness and grain size distribution data.