Emergency evacuations are essential for protecting people from hazardous events such as wildfires, tsunamis, hurricanes, industrial accidents, and terrorist actions. One way to assess the effectiveness of an evacuation plan is through mathematical models that represent the road topology, population characteristics, and hazard conditions of a specific geographic area. Current emergency evacuation models are often based upon assumptions about spatial behavior that may poorly represent human actions during an actual evacuation. Our research challenges these assumptions with empirical data from a survey of people in Santa Barbara affected by the Jesusita Fire. The stay-or-go decision of people living in a mandatory evacuation area is of critical importance to emergency managers, thus we use survival analysis to measure the impact of risk perception, distance from the fire, and environmental conditions on this decision. Another critical decision made by evacuees is which route to take when evacuating, thus we analyze empirical evacuation route data and determine which wayfinding strategies were employed by Jesusita Fire evacuees. We develop a Rayleigh probability model and a network flow model based on the results of these analyses, which are integrated with a traffic microsimulation of our study area. This simulation allows us to generate evacuation time estimates that explicitly account for behavioral uncertainty, and to assess the effectiveness of different traffic management strategies. The objectives of this research are to develop a better understanding of spatial behavior during an emergency evacuation, and develop new models that may help reduce evacuation vulnerability.