From fragmented models to transferable architectures: A two-tier architecture for agent-based models of climate adaptation

Agent-based models (ABMs) are increasingly used to study climate adaptation, yet their architectures remain case-specific, limiting comparability, cumulative learning, and the consolidation of modeling principles across applications. This paper proposes a two-tier architecture for climate-adaptation ABMs that links system-level dynamics with a staged decision architecture. Tier 1 specifies a functional system architecture in which social, economic–financial, governance, and biophysical subsystems supply social signals, market conditions, policies, and physical cues to agents, whose adaptation actions in turn update subsystem states. To bridge the abstract architecture and implementable models, we introduce a decision typology that classifies adaptation decisions by decision scope (private, collective, public) and decision mode (planned protective adaptation, emergency response, autonomous adjustment) and characterize, for each decision type, the decision factors, external conditions and decision logic that link Tier 1 to Tier 2. We then formalize one decision type, household planned protective adaptation, by extending the Protective Action Decision Model to repeated decisions over long horizons, integrating constructs from protection motivation theory and related frameworks, and embedding economic–financial drivers, social cues, and environmental signals into a staged decision pathway. The resulting formalization describes how households process cues, appraise risk, screen feasible actions, and implement measures. Its stages can be implemented as reusable building blocks and its reference formalization is documented in ODD+D. The architecture is designed to reduce fragmentation at the system and decision levels and to underpin transparency and comparability for climate-adaptation ABMs. Its transferability across cases remains to be tested.