Agent-Based Models Reveal Hidden Dynamics in Flood Risk Analysis
Structural choices in ABMs markedly impact flood risk assessments, often more than parametric uncertainties, influencing population dynamics and equity considerations in flood-prone areas.
Photo by Serge Lavoie via Pexels
Floods are a growing problem, especially in cities near coasts. We use agent-based models (ABMs) to study how people interface with flood risks. These models simulate how households might decide where to reside when faced with flood hazards. The big challenge is that the results of these models can change a lot based on the assumptions we make about how people behave. Our key finding is that the way we model people's aversion to floods (like avoiding flood zones or paying for protection) greatly affects the results. This means that understanding flood risk requires careful consideration of these modeling choices.
Floods are becoming more frequent and damaging, and understanding how people perceive flood risks is crucial. Our research uses a new computer model to study how different ways of thinking about flood risks affect people's decisions about where to live. This is important because it helps us understand how people's choices can change urban development and flood risk over time. This research helps other scientists improve their models and could impact fields like urban planning, climate change, and flood risk analysis.
In our research, we explore the impact of structural versus parametric choices in agent-based models (ABMs) for flood risk assessment using the CHANCE-C model. This model simulates urban development in a flood-prone environment, focusing on how households interact with flood hazards. We introduce three structural variants of household flood aversion: disamenity, avoidance, and protection. Each variant represents a different way households may treat flood risk—either by perceiving it as a disamenity that reduces their utility for a property, avoiding flood-prone areas entirely, or adjusting housing budgets to account for flood protection costs. Our findings reveal that these structural choices significantly influence flood risk outcomes, often overshadowing parametric uncertainties. The disamenity variant tends to result in higher flood risk exposure due to population growth in flood-prone areas, while the avoidance variant can lead to population decline in these zones. The protection variant shows muted differences in outcomes, with higher-income households more likely to reside in flood-prone areas due to their ability to afford protective measures. These insights underscore the importance of carefully considering structural assumptions in ABMs, as they can fundamentally alter conclusions about flood risk and equity, highlighting the need for transparency and critical evaluation in model design.