Development of a Large-Scale Traffic Simulation Model for Hurricane Evacuation—Methodology and Lessons Learned

Abstract

This paper focuses on traffic modeling of large-scale hurricane evacuation networks. Several research efforts to date have addressed the evacuation problem from a local perspective. The largest network found in the open literature on microscopic simulation of evacuation operations was of an area 45 miles long and 15 miles wide. The current study models 10 cities with a total length of approximately 2,000 miles of roadway including freeways, arterials, and local streets in two regions in Virginia--the Hampton Roads region and the city of Richmond. This research accomplished the following objectives: (1) estimate the traffic performance of evacuation routes and other major arterial streets; (2) locate the major bottlenecks, congestion, or other operational difficulties in the network; (3) estimate the total network evacuation time; and 4) recommend amendments to the traffic control plan to improve the traffic performance. The simulation results led to the following conclusions: (1) for a category 4 storm with high hotel occupancy, almost all vehicles (99%) were able to exit the network by the end of 27 h; (2) traffic demand on I-64 exceeded capacity for the section between I-64/Fort Eustis Blvd. and I-64/I-295 interchanges, significantly reducing the speed of evacuating vehicles; (3) the reversed lanes can carry more vehicles than their current assigned demand; and (4) for a category 3 storm, the throughput values for different evacuation routes are nearly the same with or without lane reversal. The modeling lessons learned from this research are also documented in the paper. For the network model development, geographic information system maps and a text processing scripting language (e.g., PERL) are shown to be useful tools. Parallel processing is recommended, especially for conducting statistical analysis of output data, given the enormously high run times (25-30 h in the current research) for the detailed simulations. The limitations of microscopic simulation in handling large-scale networks are identified and remedies are suggested.