Metro systems are exposed to various risks arising from natural and man-made disasters. The effective protection of these complex transportation systems requires a detailed understanding of how they interact with other critical infrastructure networks and how these interdependent systems-of-systems behave in extreme situations. This paper develops a bi-level modelling and simulation approach for interdependent infrastructure networks. On the first level, a Dynamic Bayesian Network is used to model damage and repair processes affecting the operability of interconnected components of different infrastructure networks. On the second level, a network flow analysis is conducted to capture network effects and assess the level of service achieved under various disruptive scenarios. This bi-level modelling approach is applied in a case study on the exposure of London’s metro system to flood risk. The results show that the modelling techniques developed in this paper can be used to analyse various phenomena of risk and resilience in complex infrastructure networks.