The State of Florida uses a large scale mesoscopic dynamic traffic assignment to model hurricane evacuations throughout the state. The regional character of these evacuation studies are such that multi-county regional models covering up to half of the state are required in some cases. This can result in the need to model hundreds of thousands of evacuees travelling along hundreds of miles of roads. The computational burden on these runs can be intense, taking days to complete. Given the need to develop reliable evacuation clearance times as quickly as possible, the number of runs required to generate stable and reasonable results need to be kept to a minimum.
The State of Florida uses Cube Avenue for its evacuation models. By default, Cube Avenue uses stochastic departure times when simulating the start of each trip during a time segment. Under highly congested scenarios, the stochastic nature of the departure times can have significant impacts to model clearance times. Normally, multiple runs would be performed and average clearance times would be calculated to establish a reliable clearance time arising from the stochastic departures in the model. This variation is highest in scenarios where congestion is greatest. Testing shows that dozens of runs might be needed to establish reliable averages in the worst case scenarios. This is not practical within the planning time-frames with which many emergency management offices have to contend.
Cube Avenue allows the user to fix a random seed to ensure consistent results from run to run; however, the departure timings, while fixed from run to run, are still based on a random distribution for that seed. When comparing less congested scenarios to more heavily congested scenarios, the results can be erratic and difficult to interpret. Citilabs recently introduced an option to impose uniform departure times to its simulation algorithms, thus eliminating all stochastic elements in the run. This paper compares the differences between stochastic departure times and uniform departure times using a series of evacuation scenario runs from Florida’s evacuation model to highlight the stability and reliability of results as they pertain to emergency evacuation planning.