- 1Department of Hydraulic and Water Resource Engineering, Universidade Federal de Minas Gerais, Belo Horizonte, Brazil
- 2Graduate program in Sanitation, Environment and Water Resources (SMARH), Universidade Federal de Minas Gerais, Belo Horizonte, Brazil
Extreme hydrological events, particularly dam-breach flooding, pose a growing challenge to risk governance worldwide. These events are characterized by short warning times, rapid flood-wave propagation, and potentially catastrophic downstream impacts. Their likelihood is rising due to interacting drivers, such as intensifying rainfall under climate change and altered runoff from land-use transitions. Especially in the Global South, these pressures converge with increased social vulnerability, making the human dimension an essential component of risk assessment. This study leverages valuable data from Brazil (2024-2025), where evacuation drills are mandated by national legislation. These exercises constitute one of the few systematic, large-scale efforts to observe human behavior during simulated dam-failure scenarios. As such, they provide rare empirical insights into how different groups interpret warnings, mobilize, and evacuate under realistic training conditions. We analyze behavioral responses from drill participants settled in the Self-Rescue Area (SRA) downstream of the Ibirité water reservoir (MG-Brazil), focusing on their mobilization performance after receiving an alert. Using ordinal logistic regression, we examine how alert responsiveness is influenced by demographic factors (e.g., gender and age), socio-cognitive variables (e.g., risk perception, emergency preparation), and experiential background (e.g., prior exposure to flood events). This approach allowed the identification of those characteristics that most strongly predict rapid or delayed evacuation initialization. The evacuation drills are characterized by low participation rates (2.4 ± 0.3%), which is a typical pattern in the Brazilian context. In consequence, statistical tests were realized using single year data from 2024 (n = 80) and 2025 (n = 65), and a combined dataset for 2024-2025 (n = 145). Demographic factors had no significant influence on mobilization. In contrast, socio-cognitive variables and experimental background shaped significantly protective actions: persons with prior drill experience took consistently longer to begin evacuating (2024: OR = 3.18 (p = 0.062) / 2025: OR = 3.01 (p = 0.057) / 2024-2025: OR = 2.61 (p = 0.015)); participation at drill-preparatory seminars were associated with shorter mobilization times (2025: OR = 0.27 (p = 0.032) / 2024-2025: OR = 0.35 (p = 0.015)); and experiential background influenced evacuation initiation positively (2025: OR = 4.11 (p = 0.039)). These outcomes suggest that evacuation drills alone may lead to a false sense of security and slower alarm responses. Educational measures and experience with real risk cues, on the other hand, can reduce reaction time during warnings. Interpreted through a human-water feedback framework, the results illustrate how behavioral responses can alter the effective consequences of extreme hydrological events. Rapid mobilization reduces the number of flood-harmed individuals, while delayed responses can exacerbate vulnerability even when warning systems operate as designed. This study demonstrates the critical value of evacuation drills as an important empirical resource for understanding human behavior during extreme hydrological events. The Brazilian context offers an important contribution from the Global South, where empirical data on human–flood interactions remain underrepresented in hydrological risk research. It is recommended to continue data collection and combine datasets of different local evacuation drills to improve the model’s performance and stability over time.
How to cite: Krause Camilo, B., Felipe Rocha Silva, A., Cardoso Eleutério, J., G. Gabrich Fonseca, M. T., and Ferreira Rodrigues, A.: Quantifying behavioral responses to dam-breach flooding evacuation drills as a function of demographic factors, socio-cognitive variables, and experiential background., EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-883, https://doi.org/10.5194/egusphere-egu26-883, 2026.