Evacuation Scenario Simulator with Location-Based Social Network Data Support

Abstract

Disaster management systems apply different evacuation models to provide transportation responses to disaster situations. The main goal of the evacuation models is to reduce evacuation times to save lives.The significant increase in the use of location-based social networks (LBSN) as a major media factor can provide data for inferring the Origin-Destination matrix that represents the demands for trips during routine times. To achieve the goal of reducing overall evacuation travel times, an innovative methodology was developed that supports LBSN data for both demand prediction and decentralized personal destination recommendation.A modular multi-dimensional tool for evacuation scenario simulation (MMDT-ESS) was developed for supporting this methodology. This tool can handle multiple scenarios and compute performance evaluation values of VhT (Vehicle-Hours Traveled). MMDT-ESS can be applied using both conventional evacuation models (i.e., evacuation to predefined shelters) or decentralized personalized evacuation models and can provide different recommendations: either self-evacuation to nearest family or best friend that is situated outside the affected area or to a predefined shelter area.A test case using data from a U.S. metropolitan agency showed significant reduction in VhT when choosing evacuation scenarios with LBSN data for decentralized evacuation. Sensitivity analysis on main model parameters illustrate the robustness of the results.