Tsunami forces on onshore structures can be evaluated using drag formulas [1] based on 2D tsunami simulations or by directly computing the forces through 3D simulations [2]. In typical 2D simulations, buildings are not explicitly modeled and the vertical flow structures cannot be represented. Therefore, drag formulas based on their results may not represent actual tsunami forces. Qin et al.[3] showed that Tsunami forces estimated from 2D simulations with buildings tended to underestimate experimental forces in densely built-up areas. In this study, 2D and 3D tsunami simulations are conducted to examine the applicability of the drag formula. Tsunami forces estimated from 2D simulations and 3D simulations without buildings are compared with those directly computed from 3D simulations with buildings to clarify the effects of analysis methods and building. Furthermore, the influence of surrounding buildings on tsunami forces was evaluated for the four cases shown in Fig. 1.The simulations are conducted for Onagawa Town, Japan. The 2D model covers a wide area including the tsunami source region, while the 3D model is applied to a limited subdomain with inflow conditions prescribed from the 2D results. The 3D domain (Fig. 2, 1500 m × 1200 m) consists of approximately 4.8 million cells and required about 1.5 months of computation using 192 CPUs. The surrounding-building analysis shown in Fig. 1 was conducted using a smaller computational domain (Fig. 1, 600 m × 300 m) . The results (Fig. 3) show that the drag formula [1] based on the 3D simulations without buildings accurately reproduces the forces directly computed by the 3D simulations with buildings, whereas the 2D simulations underestimate them due to underprediction of flow velocity. Furthermore, the surrounding-building analysis indicates that upstream buildings reduce the tsunami forces acting on the target building, whereas no reduction due to downstream buildings is observed.