Rainstorms and flooding are among the most common natural disasters, which have a number of impacts on the transport system. This reality highlights the importance of understanding resilience—the ability of a system to resist disruptions and quickly recover to operational status after damage. However, current resilience assessments often overlook transport network functions and lack dynamic spatiotemporal analysis, posing challenges for comprehensive disaster impact evaluations. This study proposes an SR-PR-FR comprehensive resilience evaluation model from three dimensions: structure resilience (SR), performance resilience (PR), and function resilience (FR). Moreover, a simulation model based on Geographic Information System (GIS) and Simulation of Urban MObility (SUMO) is developed to analyze the dynamic spatial–temporal effects of a rainstorm on traffic during Xi’an’s evening rush hour. The results reveal that the southwest part of Xi’an is most prone to being congested and slower to recover, while downtown flooding is the deepest, severely affecting emergency services’ efficiency. In addition, the road network resilience returns to 70% of the normal values only before the morning rush the next day. These research results are presented across both temporal and spatial dimensions, which can help managers propose more targeted recommendations for strengthening urban risk management.

Extreme rainfall and flood events increasingly threaten urban transportation networks, yet conventional resilience evaluations fail to capture the dynamic interplay between flooding, traffic congestion, and emergency service accessibility. To address the above gap, this study provides an integrated framework for transportation resilience evaluation with the proposed spatiotemporal accessibility analysis. Our approach combines real-time traffic simulation, flood modeling, an enhanced two-step floating catchment area (E2SFCA) method with a novel congestion-sensitive impedance function, and grid-based percolation theory to quantify network fragmentation via accessibility thresholds and largest connected component (LCC) metrics. A case study in Shanghai’s Huangpu District demonstrates that peak-hour floods induce greater accessibility losses and faster network fragmentation than off-peak events. Sensitivity analysis highlights accessibility’s higher responsiveness to congestion dynamics over facility siting, highlighting the importance of adaptive traffic management in disaster planning. The workflow provides a robust tool for planners to anticipate vulnerabilities, prioritize interventions, and enhance time-critical disaster preparedness.

Existing flood resilience assessments face a fundamental tradeoff: correlational analyses of real-world data suffer from confounding bias, while simulations of hypothetical scenarios lack empirical grounding. To address this identification gap, we propose a physically grounded causal inference framework that derives unbiased causal parameters directly from physical observation. First, we generate high-fidelity flood maps from synthetic aperture radar (SAR) imagery using a fine-tuned U-Net model adapted for dense urban environments. Second, we integrate these empirical maps into a microscopic traffic simulation model to quantify dynamic network disruptions. Third, we apply double machine learning (DML) to isolate the causal effect of flooding on transport resilience while rigorously controlling for high-dimensional urban form confounders. Applied to a flood event in a dense coastal megacity, our analysis reveals two key discoveries: a nonlinear tipping point around 25% flood intensity, distinct from recent studies assuming linear degradation, beyond which system performance collapses chaotically; and a “dual nature of vulnerability” where low-density suburbs suffer a fragility of isolation ( $2.9\times $ higher road closure rate), while dense urban cores exhibit a fragility of congestion ( $1.47\times $ greater waiting time increase). Validated across 130 sensitivity tests, this framework provides a robust, data-driven blueprint for spatially targeted resilience investments, demonstrating that causal identification in disaster contexts requires physically grounded observation, not synthetic assumptions.

Autonomous vehicles (AVs) will operate alongside human-driven vehicles for an extended transition period, during which aggressive human driving may become a major source of risk. This study proposes an integrated safety-control framework that combines real-world-data-driven behavior modeling with deep reinforcement learning to design longitudinal AV control strategies for mixed traffic. Aggressive, general, and defensive driving patterns are calibrated from the CitySim dataset, and dynamic aggressiveness is incorporated into an improved car-following model. A proximal policy optimization algorithm with a Kullback–Leibler penalty is then used to learn multi-objective strategies balancing safety, efficiency, comfort, and fuel economy in freeway and signalized-intersection scenarios. The results show that the behavior-aware strategies exhibit different strengths across traffic environments. On the freeway, the defensive-threshold strategy maintains a larger time headway, reduces positive acceleration, and lowers system-level fuel consumption, whereas the default, aggressive, and general strategies preserve higher traffic efficiency. At the intersection, signal control narrows the differences among strategies and limits the influence of longitudinal threshold settings on most evaluated indicators. These findings provide a quantitative basis for selecting behavior-aware control thresholds and designing robust AV strategies for mixed-autonomy traffic containing aggressive human drivers.