Roxborough Park Community Wildfire Evacuation Drill: Data Collection and Model Benchmarking

Area of Science:

• Environmental Science
• Disaster Management
• Computational Social Science

Background:

• Wildfires are escalating in severity and frequency, necessitating improved community evacuation strategies, particularly in wildland-urban interface (WUI) areas.
• Effective evacuation planning requires accurate data on community response dynamics, including pre-evacuation delays and route choices.

Purpose of the Study:

• To present a novel dataset from a community wildfire evacuation drill in Roxborough Park, Colorado.
• To benchmark two distinct evacuation models (WUI-NITY and Evacuation Management System) using this dataset.
• To analyze the sensitivity of these models to variations in input data, particularly pre-evacuation times and route usage.

Main Methods:

• A community evacuation drill was conducted in a WUI setting, collecting data via observation and surveys on population location, pre-evacuation times, route use, and arrival times.
• Two evacuation models, WUI-NITY and Evacuation Management System, were applied using the collected data across various scenarios with modified assumptions for pre-evacuation delays and routes.
• Model performance was evaluated based on sensitivity to different data inputs (observations vs. self-reporting) and the evacuation phases modeled.

Main Results:

• Evacuation model outcomes were primarily influenced by assumptions regarding pre-evacuation time inputs, especially in low-congestion environments.
• Model performance demonstrated sensitivity to the type of data used (observed vs. self-reported) and the specific evacuation phases incorporated into the models.
• The study underscored that the impact of data on model outcomes depends significantly on the underlying modeling methodologies.

Conclusions:

• The accuracy of wildfire evacuation models is highly dependent on the quality and source of pre-evacuation time data.
• Different modeling approaches exhibit varying sensitivities to input datasets, emphasizing the need to understand how data is processed.
• The released open-access dataset provides a valuable resource for future research in wildfire evacuation model calibration and validation.