A Global Dataset of Urban Road Centrelines for 2,588 Coastal Cities Using Voronoi Tessellation

This dataset provides unified road centreline networks for 2,588 coastal cities across 110 countries, derived from Overture Maps data (version 2026-02-18) using a novel Voronoi tessellation algorithm with adaptive spatial partitioning. The dataset addresses a fundamental limitation of traffic-oriented road network representations, where multi-lane roads are stored as multiple parallel segments, introducing distortions in spatial analyses such as space syntax, centrality calculations, and graph neural network applications.The dataset consolidates 68.2 million raw road segments totalling 13.35 million kilometres into 42.1 million unified centreline segments comprising 9.64 million kilometres. All cities achieved excellent geometric quality grades upon validation, with an average length preservation rate of 73.53%.Files are provided in GeoParquet format with WGS84 coordinates (EPSG:4326), compatible with Python GeoPandas, R sf, and QGIS (version 3.28 and later). The dataset is organised by city, following the naming convention {Country}_{CityName}_centerline.parquet. Summary statistics and quality validation results are provided in accompanying CSV and JSON files.This dataset supports research in urban morphology, climate resilience planning, space syntax analysis, and geospatial artificial intelligence.