After natural disasters and other major disruptions, interdependencies between transport networks and other critical infrastructure systems can result in additional vulnerabilities, such as cascading failure, delayed recovery and network congestion. Empirical studies have documented such interdependency effects for various catastrophic events, including the 2001 terror attack in New York City (Lee II, Mitchell, and Wallace 2007), the 2003 electrical blackout in Italy (Buldyrev et al. 2010), various hurricanes in Florida (McDaniels et al. 2007), the 2008 winter storms in China (Rong, Han, and Liu 2010), and the 2010 earthquake in Chile (Dueñas-Osorio and Kwasinski 2012). For the development of strategies to enhance the resilience of urban transport systems, quantitative models are needed that take into account interdependencies and can evaluate the performance of different system configurations in a what-if analysis for various hazard scenarios. Previous studies on this problem have extended classical network flow theory to include interdependency effects (e.g. Lee II, Mitchell, and Wallace 2007; Holden et al. 2013; Fotouhi, Moryadee, and Miller-Hooks 2017). This paper aims to contribute to the further development of this methodology with a dynamic network flow model that captures various types of dependency relations and different degrees of anticipation. The proposed method comprises a stochastic model that generates scenarios for the operability of infrastructure network assets, and a minimum cost flow assignment model for the simulation of dynamic flows in coupled networks. Interdependency effects are taken into account both at the asset operability level and at the network flow level. The flow assignment model can simulate different degrees of anticipation, allowing the model to capture important aspects related to operational decision-making, for example, information sharing between network operators and the use of predictive models to estimate recovery times. A case study on London’s metro system demonstrates how the proposed method can be used to quantify the effects of interdependency and anticipation in a flooding scenario.