Evacuation Planning on Time-Expanded Networks with Integrated Wildfire Information

TL;DR

This paper presents a scalable, modular evacuation planning method for wildfires that uses real-time hazard data integrated into time-expanded network flow models, enabling dynamic updates during evacuations.

Contribution

It introduces a novel approach combining maximum flow models on time-expanded networks with real-time wildfire hazard data for adaptable evacuation planning.

Findings

Validated on three locations with historic fires

Demonstrated viable running times and data quality

Showcased scalability and modularity of the approach

Abstract

We study the problem of evacuation planning for natural disasters, focusing on wildfire evacuations. By creating pre-planned evacuation routes that can be updated based on real-time data, we provide an easily adjustable approach to evacuation planning and implementation. Our method uses publicly available data and can be tailored for a particular region or circumstance. We formulate large-scale evacuations as maximum flow problems on time-expanded networks, in which we integrate hazard information given in the form of a shapefile. An initial flow and evacuation plan is found based on a predicted fire, and is then updated based on revised fire information received during the evacuation. We provide a proof of concept on three locations with historic deadly fires using data available through OpenStreetMaps, a basemap for a geographic information system (GIS), on a NetworkX Python script. The results validate viable running times and quality of information for application in practice. Particular strengths are the scalability and modularity of our approach and accompanying software package.