One important adverse impact of increased intensive rainfalls is traffic disruption due to roadway pluvial flash flooding (PFF, caused by inadequate stormwater drainage systems) and lane submersion, which can lead to injury and even loss of life. In a highly urbanized area, flood formation is a highly complex and uncertain process and its impacts on roadway mobility depend on numerous temporal and spatial variables. Despite the availability of many physics-based hydrodynamic, hydrologic, and empirical models to simulate PFF, their application on the highly localized scales at which traffic disruption occurs remains a challenge due to high computational time and lack of available data. This study detects the risk of roadway PFF based on traffic data obtained from Waze, a crowd-sourced navigation app, and a physics-based model. A hybrid model is developed that combines a Graph-based Rapid Flood Spreading Model (GB-RFSM) with machine learning (ML) to predict the risk of roadway PFF flooding at the intersection scale. The model is applied to a case study of flood-prone intersections in Dallas-, Texas. The performance of multiple ML classifiers [Random Forest Classifier (RFC), Extreme Gradient Boosting Decision Tree (XGBoost), and Support Vector Classifier (SVC)] are compared. The results showed that RFC was more precise in predicting flooded areas identified by Waze users, with a 73% recall score representing reported flood events that are identified using the hybrid model. More data, particularly on the location and configuration of the urban drainage system, are needed to improve predictions.
