In southern Thailand, Phuket was one of six provinces heavily affected by the 2004 Indian Ocean Tsunami. Patong is one of the most populated beaches in Phuket province. There are not only outdoor activities on the beach, but people can also enjoy several indoor activities, such as shopping, eating, massaging, etc., across the area. Therefore, population distribution can be implied by these activities related to the building occupancy classes, such as hotels, restaurants, shops, etc. Generally, tsunami evacuation simulation begins with a model of the departure locations, also known as population distribution. For example, individuals initially depart from each building uniformly distributed or based on the building occupancy classes (BDOC), such as residential, hotel, commercial, etc. The number of occupants for each building class can be assumed based on expert judgment or the field survey. In this study, we focus on the effect of population distribution on tsunami evacuation using agent-based modeling and simulation. The population in Patong, Phuket province is estimated based on the unit area of each building's occupancy class and the building’s floor area (m²). The results show that the population in commercial buildings based on the Uniform model is significantly less than in the BDOC model. The BDOC model causes faster evacuation than the Uniform model, which may be caused by the high population being closer to safe places, such as tall buildings. Population distribution significantly affected the tsunami evacuation and should be taken into consideration.