Michelin Roads 1910-1930

Road Network France (1910–1930) Core Metadata Field Value Dataset Name Historical Road Network France Dataset Object Road network (vectorised segments) Owner Litvine, Alexis; Thévenin, Thomas; Starzec, Arthur; Mille, Patrick; Séguy, Isabelle; Proffit, Guillaume; Lekoumou, Cléonna Version 1.1 Documentation Fields Field Value Data Repository https://doi.org/10.5281/zenodo.19681266 Data Paper Link https://doi.org/10.17863/CAM.116780 BibTeX Citation @dataset{litvine2026roads, author = {Litvine, A. and Thévenin, T. and Starzec, A. and Mille, P. and Séguy, I. and Proffit, G.}, title = {Historical Road Networks for France, 1910–1930}, year = {2026}, doi = 10.17863/CAM.116780, publisher = {Cambridge Working Paper Series in Economic and Social History} } License CCBY 4.0 Geographic Coverage Metropolitan France Temporal Coverage c.1910 and c.1930 Key Variables road_id, geometry, road_type, surface_type, width_class, quality_class, slope, speed_coefficient File Format(s) GPKG Access Notes On our MAHDI Zenodo, not to be circulated Description This dataset provides a fully digitised and harmonised road network for France circa 1930 and raw data for France circa 1910 and Great Britain circa 1930. The network was constructed from Michelin cartographic sources (1:100,000 scale) and manually vectorised to ensure high topological accuracy. The resulting dataset includes all roads visible on the original maps, classified according to their quality, surfacing, and width. Files included The Zenodo repository contains three GeoPackage files: MICHELIN_FR_1910.gpkg MICHELIN_FR_1930.gpkg MICHELIN_UK_1930.gpkg Each file contains a single layer representing the road network at the corresponding date, with the full set of digitised road segments. MICHELIN_FR_1930.gpkg: This is the fully corrected and harmonised dataset, including derived speed information for each segment, based on the modelling framework developed by Litvine and Thévenin. MICHELIN_FR_1910.gpkg and MICHELIN_UK_1930.gpkg: These are raw digitised networks. While geometrically complete, they have not yet undergone full correction or harmonisation and should be used with caution. Coordinate Reference Systems French datasets (MICHELIN_FR_1910.gpkg, MICHELIN_FR_1930.gpkg) use Lambert 93 (EPSG:2154) UK dataset (MICHELIN_UK_1930.gpkg) uses British National Grid – OSGB 1936 (EPSG:27700) Attributes (MICHELIN_FR_1930.gpkg) Field Name Description OBJECTID Unique identifier for each road segment Shape* Geometry type (Polyline) ID Internal identifier ROAD_TYPE Road quality classification (e.g., GOOD_ROAD, AVERAGE_ROAD, POOR_ROAD) ROAD_WIDTH Width category of the road (e.g., NARROW, AVERAGE, WIDE) ROAD_SURFA Road surface type (when available) TYPE Original classification code from source data layer Source layer identifier LENGTH Length of the segment (in meters) JOIN_KEY Combined key (ROAD_TYPE + ROAD_WIDTH), used for classification TRAVEL_MIN Estimated travel time (minutes, model output) TIME1910 Estimated travel time (minutes) using 1910 parameters TIME1930 Estimated travel time (minutes) using 1930 parameters Key features A topologically correct graph structure suitable for network analysis A standardised road typology, harmonised across time to enable comparison Integration with Digital Elevation Models (DEM) to compute slope effects Derived speed coefficients (for the corrected dataset) reflecting historical vehicle performance and road characteristics Compatibility with historical transport modelling frameworks The dataset underpins the construction of historical travel-time models, enabling the estimation of accessibility, journey times, and market potential in early twentieth-century France. It is part of the broader French Historical GIS (ANR-Communes project) infrastructure. Limitations Heterogeneous Data Quality: Only the 1930 France dataset is fully corrected and enriched with speed data. The other files remain raw digitisations. Cartographic Interpretation: Road classification relies on historical map symbology, which can be ambiguous (e.g., colour variation, administrative distinctions). Manual Digitisation Bias: Although carefully validated, the network was digitised manually, introducing potential human error in geometry and classification. Temporal Approximation: The dataset represents approximate snapshots (c.1910 and c.1930), not exact yearly states. Topological Tolerance: A tolerance threshold (~100m) was used to correct dangling nodes, which may slightly alter local geometry. Road Coverage: Only roads represented on Michelin maps are included; minor or informal paths may be missing. Speed Modelling Assumptions: Derived speed variables depend on modelling choices (vehicle characteristics, slope effects, road coefficients) and should be interpreted as estimates rather than direct observations.