Quantifying the accuracy of OpenStreetMap for longitudinal cycling-infrastructure change: A reproducible validation framework using Google Street View

Cities are rapidly expanding their cycling networks to promote sustainable mobility, yet robust longitudinal data needed to evaluate change are scarce. OpenStreetMap (OSM) provides a global, time-stamped record of urban infrastructure, but its reliability for detecting change remains uncertain. This study quantified the temporal accuracy of OSM for identifying cycling-infrastructure additions and removals, using Barcelona (2015–2023) as a case study within a reproducible open-source workflow with manual validation using street-level imagery. We reconstructed baseline and follow-up cycling-infrastructure networks from dated OSM snapshots and applied geometric differencing to detect change. A stratified sampling design based on population density and centrality was used to validate detected changes against historical Google Street View (GSV) imagery, allowing estimation of precision, recall, and F1 scores. OSM-derived estimates indicate an 80.3% increase in cycling-infrastructure length, from 130.7 km to 235.7 km. For additions, validation shows high recall (0.96) and strong precision (0.82), indicating that OSM captures most network expansions. For removals, recall is high (1.00) but precision is low (0.20; n = 5), with wide uncertainty given the small sample. Apparent losses may reflect mapping edits rather than real infrastructure change. By quantifying detection accuracy, this study provides a framework for calibrating OSM-derived measures of cycling infrastructure change. OSM snapshot differencing can serve as a scalable screening approach for identifying infrastructure additions, while apparent removals require cautious interpretation and independent verification to avoid mapping artefacts being mistaken for physical infrastructure loss. This framework supports more robust longitudinal use of OSM histories in planning and research.