Pavements are highly susceptible to water infiltration during floods, resulting in reduced serviceability, shortened life span, and poor durability. To enhance pavement resilience and mitigate the rising risks associated with flooding events, this study aims to propose a method to predict the short-term temporal evolution of saturation levels within pavement structures resulting from inundation events. The methodology involves conducting a series of finite-element-based hydraulic simulations considering various influential factors such as pavement structure, subgrade type, groundwater table level, and flooding scenario. Time-descriptive indicators, including peak saturation time and restoration time, are calculated from the simulated volumetric water content data. To capture the complex relationships between the input parameters and output indicators, machine learning (ML) methods are employed to construct a predictive model for saturation profiles during flooding events. The results demonstrate that the 2nd-degree polynomial provides the best fit for the saturation changes within the vadose zone, and the random forest algorithm outperforms other ML methods, achieving the highest accuracy in projecting saturation changes during flooding events. The predictive model offers valuable insights for decision-making processes, including determining the optimal timing to reopen submerged roadways to traffic, and evaluating the moisture damage caused by the inundation. Overall, this research contributes to enhancing pavement resilience and enables the design and management of resilient pavements under changing climate conditions and extreme weather events.
