Cascading Failures Across Interdependent Infrastructure Systems Cannot Be Predicted

Modern infrastructure systems — power grids, water networks, transportation, telecommunications, gas pipelines — are deeply interdependent, but no unified model exists to predict how failures cascade across them. When a hurricane knocks out power, water treatment plants stop, hospitals lose function, emergency services are overwhelmed, and transportation clogs — but engineers cannot simulate these cascading chains realistically. Existing models treat each infrastructure network in isolation or use grossly simplified coupling assumptions. The socioeconomic dimension is almost entirely absent: the magnitude of failure impact depends heavily on the characteristics of affected communities, but these factors are routinely omitted from resilience assessments.

Between 2020 and 2024, the U.S. experienced a billion-dollar disaster event every 16 days on average, compared to every 82 days in the 1980s. Billion-dollar severe storm events alone caused an average of $37.9 billion in annual damage. Infrastructure failures disproportionately impact low-income communities with less redundancy and fewer resources for recovery. Without realistic cascading failure models, infrastructure investments are misallocated — we harden individual components rather than addressing systemic vulnerabilities across interconnected networks.

Single-network cascade models (extensively studied for power grids) cannot capture cross-network interdependencies where different infrastructure types have fundamentally different coupling mechanisms. Physical interdependency models focus on hardening individual components rather than engineering recovery processes. Most models assume single-hazard scenarios, but real disasters involve compound, concurrent, and cascading hazards (earthquake triggers tsunami triggers fire triggers infrastructure collapse). The treatment of uncertainty in interdependent infrastructure models is surprisingly undeveloped despite post-disaster data showing wide variability of outcomes. Building inventory data needed for regional-scale simulation has remained largely unavailable. There is no unified theoretical framework describing energy-transfer pathways, interaction modes, and expansion mechanisms across multiple hazard types.

A computational framework that models infrastructure interdependencies with realistic coupling mechanisms rather than just topological connections; integration of socioeconomic vulnerability data into infrastructure resilience models; methods to model compound and cascading hazard events rather than single hazards; large-scale building and infrastructure inventory datasets that are openly available and standardized; and a shift from failure prevention to recovery process optimization that leverages positive interdependencies between systems.

A student team could build a simplified agent-based model of two interdependent infrastructure networks (e.g., power and water) for a specific small city, calibrating coupling parameters from published post-disaster reports (which are publicly available from FEMA and state emergency agencies). This is a feasible computational modeling project. Alternatively, a team could analyze publicly available post-disaster data to empirically characterize cascading failure sequences and their correlation with census-tract-level socioeconomic data, producing an empirical interdependency map.