Tsunami evacuation planning has relied on static maps generated by Least Cost Distance models, which calculate optimal routes based on street network lengths. However, by focusing on average evacuation times, these methods present limitations by ignoring the complexity of human behavior and failing to identify which variables are crucial to understand evacuation times. This suggests that current evacuation maps require improvement through the integration of simulation analysis to address underestimate risks in uncertain scenarios. This research performs a massive simulation analysis by integrating the Social Force Model with high-performance multi-threaded computing to generate comprehensive scenarios with different evacuation times.

The analysis is applied to the vulnerable coastal zone of Las Brisas de Villa in Chorrillos, Lima. Stochastic scenarios were executed by varying the initial population placement. The study focuses on analyzing the resulting distribution of high evacuation times to reveal scenarios representing low probability but high-impact events. By isolating the simulation seeds of these critical outliers, the specific spatial arrangements were reconstructed and examined.

Preliminary findings indicate that risks are driven by specific urban spatial configurations. These results demonstrate that evacuation times are dependent on the urban spatial configuration. Consequently, the identification of statistical outliers in evacuation time distributions is crucial for mitigating hidden spatial configurations that average-based models fail to detect.