Many coastal cities around the world are becoming increasingly vulnerable to natural disasters, particularly flooding driven by tropical storm and hurricane storm surge – typically the most destructive feature of these storms, generating significant economic damage and loss of life. This increase in vulnerability is driven by the interactions between a wide number of complex social and climatic factors, including population growth, irresponsible urban development, a decrease in essential service provision, sea level rise, and changing storm regimes. These issues are exacerbated by the short-term strategic planning that dominates political action and economic decision-making, resulting in many vulnerable coastal communities being particularly unprepared for large, infrequent storm surge events. This lack of preparedness manifests in several ways, but one of the most visible is the lack of comprehensive evacuation and rescue operation plans for use after major storm surge flooding occurs. Typical evacuation or rescue plans are built using a model of a region’s intact road network. While useful for pre-disaster purposes, the immediate aftermath of large floods sees enormous swaths of a given region’s road system flooded, rendering most of these plans largely useless. Post-storm evacuation and rescue requires large amounts of atypical travel through a region (i.e., across non-road surfaces). Traditional road network models (such as those that are used to generate evacuation routes) are unable to conceptualize this type of transportation, and so are of limited utility during post-disaster scenarios. To solve these problems, this dissertation introduces an alternative network conceptualization that preserves important on-network information but also accounts for the possibility of off-network travel during a disaster. Providing this in situ context is necessary to adequately model transportation through a post-storm landscape, one in which evacuees and rescuers are regularly departing from roads and one in which many roads are completely interdicted by flooding. This modeling approach is used to automatically generate routes through a flooded coastal urban area, as well as to identify potentially critical road segments in advance of an actual storm. These tools may help both emergency managers better prepare for large storms, and urban planners in their efforts to mitigate flood damage.