The purpose of this session is to:
- Showcase the state of the art in agent-based and other bottom-up modelling approaches of how natural hazards affect households and firms, from direct impacts to long-run recovery.
- Advance the use of such models to assess impact on multiple dimensions, distributional consequences, and policy effectiveness
- Stimulate exchange of models, data, calibration strategies, and identify priorities for making these models more empirically grounded, credible and useful for decision-making.

Risk assessment for natural hazards has traditionally focused on direct impacts to physical assets. But disaster losses and recovery trajectories vary widely across households and firms and depend on their characteristics such as household income and savings or firm size. Moreover, at least part of the loss is also indirect and unfolds over time, through lost income, forgone care, business interruptions, and disrupted supply chains. These consequences emerge from the decisions of many interacting, adaptive agents (e.g., households, firms) and from the feedbacks between them. Agent-based models (ABMs) can represent this heterogeneity explicitly, together with adaptive behaviour and feedback loops, making them well suited to study the distributional, direct and indirect consequences of hazards. The same applies to the policies trying to mitigate losses: they do not act on the average household or a representative firm alone, and a measure that is effective in aggregate can be regressive for some or be improved upon by targeting specific subpopulations. However, how much such models can inform real decision-making depends on how well they are grounded in data, account for uncertainty and are validated.

We invite contributions that develop, calibrate, validate or apply agent-based and other bottom-up models of disaster impacts on households and/or firms. Topics include:
- Representations of household and firm decision-making, adaptation and post-disaster recovery, including behavioural and economic mechanisms.
- Empirical grounding using household survey, census, administrative, firm-level, remote-sensing or post-disaster damage data.
- Coupling with hazard and climate models, including multi-hazard, compound and repeated events.
- Methodological advances: such as sensitivity analyses, comparison with other approaches and validation.
- Conceptual contributions on what these models can and cannot do for risk assessment and policy design.