Predicting Public Transport Resilience to Climate Extremes: A Hybrid Dynamical Systems Thinking Approach

Abstract

This study investigates climate change's impact on travel behavior and provides actionable insights for enhancing public transportation resilience. We employ a novel Hybrid Dynamical Systems Thinking Approach (HDSTA), integrating knowledge graphs with Machine Learning (ML) models, to predict bus ridership, traffic volume and speed. Using data from a major Israeli interstate highway, including weather reports, passenger counts, and traffic sensor inputs, our ML models demonstrated superior predictive accuracy for climate event transportation sensitivity compared to traditional methods like Structural Equation Models (SEM). Empirical analysis revealed rainfall is a more significant factor in reducing bus ridership than heatwaves. Based on these patterns, we propose a Weather Resilience Index (WRI) to quantify weather's impact on ridership at bus stops, highlighting the need for targeted adaptation strategies. These tools empower transportation stakeholders and decision-makers to analyze climate effects on transportation and implement data-driven actions.