Estimating the number of evacuees during the alert period using People-Counting Data

～Case Study of the 2020 Typhoon No.10～

Abstract

When a large-scale disaster is predicted, wide-area evacuation is required in advance. Government agencies are required to secure transportation and evacuation facilities to support evacuation by residents. By grasping the number of residents in dangerous areas in real time, evacuation support activities can be planned appropriately. In 2020, a relatively large-scale pre-evacuation was carried out when Typhoon No. 10 struck. Recently data on the flow of people based on mobile phone management information has been provided nationwide in near real time. Using this data, we estimated the number of people in Kumamoto City and verified the usefulness of the data.

In analysis on the people flow data, we adopted a model that can take into account the effects of irregular holidays and hourly rainfall, as well as daily and weekly periodicity. As a result of data analysis, we were able to grasp the increase and decrease of the resident population in the flooded area. Before the typhoon hit, more residents stayed at home than in normal times. And we were able to see an increase in visitors in the area where the hotel is located. However, it was not possible to grasp that the number of residents increased in the area where the evacuation site was located. We conclude that analyzing the data separately for residents and visitors is useful for the purpose of detecting vitality in dangerous areas. In addition, we also need to use OD data in order to grasp the inflow and outflow status in more detail