Various models for tsunami-induced drift objects have been developed; however, in these models, drift objects of various shapes, such as vessels and vehicles, are expressed as mass points, rectangular solids, and combinations of particles. Shibuya et al. (2012) suggested that the ability to express object shape is important for the reproduction of trapping phenomenon caused by structures for capturing drift wood in rivers. Drift objects caused by tsunamis have various shapes including vessels, containers, vehicles, and house rubble. Therefore, the ability to express shape is essential for predicting complicated phenomena in which these objects interact.

Therefore, in this study, we tried to develop a drift model which can express drift objects of arbitrary shape. In calculating drift motions, the following models are required: a fluid model which calculates water levels and flow velocity and a drift model which calculates drift motions.

Regarding the fluid model, we used the Python programming language to develop a numerical model based on Navier-Stokes equations, considering the ease of the connection with the following drift model.

Regarding the drift model, we used the Pybullet, which was the Python version of the physical engine library "bullet" and the open-source library using a zlib license, for translation and rotation of objects as well as collision, for each time step. Pybullet can express arbitrarily shaped objects by combining basic shapes including spherical, box and columnar shapes or by feeding .obj files, which is a data format for expressing 3D objects. Also, many physical engines containing Pybullet can perform high-speed calculation of 3D physical motions, and so they are used for physical calculation in games. In this drift model, we used fluid force calculated using water levels and flow velocity obtained from the fluid model and calculate the motion of drift objects at each time.

In the future, we will validate this drift model by comparing the existing experimental results. Regarding the physical engine used in the drift model, real-time performance was shown to be more important than calculation accuracy. Therefore, a quantitative assessment of the reproducibility of actual phenomena is required. In addition, we will study the following: improving the fluid force evaluation method for the drift model; improving the interaction method of the fluid model and the drift model; comparing these models with existing models; and evaluating debris dam and accumulation in inundation areas and on collision of debris with land structures.