Abstract

The growing adoption of cycling as part of the transition toward sustainable urban mobility, driven by climate change concerns and increasing congestion, has heightened the need to ensure cyclist safety in metropolitan areas. This study proposes an integrated spatio-temporal analytical framework to examine the evolution of reported bicycle–vehicle injury accidents in the Lisbon Metropolitan Area (LMA). The framework combines Geographic Information Systems (GIS)-based spatial statistics with Emerging Hotspot Analysis (EHA) to identify and track changes in accident clustering over time, across pre-, during-, and post-COVID-19 containment periods. This study contributes by applying Emerging Hotspot Analysis to bicycle accident data, an approach still largely unexplored, and by proposing a sequential and integrated framework that links traditional spatial analysis methods with dynamic hotspot detection and machine learning techniques, enabling a shift from static pattern identification to enhanced interpretation of evolving accident occurrence patterns and hotspot dynamics. Results reveal evidence of spatial consolidation and changing hotspot distributions over time, with emerging hotspots increasingly located in suburban transition zones and at the edges of existing cycling infrastructure. These patterns may reflect changes in mobility demand and infrastructure provision, although the absence of exposure data prevents a direct assessment of this relationship. Complementary analysis using forest-based machine learning models identifies key factors associated with hotspot formation and accident severity, including crash type, temporal patterns (e.g., day of the week), and environmental conditions such as slope and lighting. These findings highlight the value of combining spatio-temporal analysis with predictive modelling to support data-driven urban planning and targeted safety interventions. Lisbon provides a relevant case study for cities undergoing similar transitions toward sustainable transport systems.

IPC Classification

Keywords