Maps of historical forests in France and their temporal continuity since the first half of the 19th century

Jean‐Luc Dupouey, Laurent Bergès, Nathalie Leroy, Romain Lafite, Frédéric Archaux, Vincent Augé, Raphaël Bec, Maxime Bellifa, Julien Bourguignon, Jérôme Buridant, Béatrice Burlin, Simon Caubet, Albane Chaleat, Sandrine Chauchard, Thomas Cordonnier, Guillaume Decocq, Matthieu Delcamp, J. Fleury, Sylvain Gaudin, Alain Gervaise, Grégoire Gautier, Julien Guilloux, Arnoul Hamel, Wilfried Heintz, Philippe Janssen, Sophie Labonne, P. Lair, Thierry Lallemant, Guy Landmann, Laurent Larrieu, Hilaire Martin, Claude Michel, Sylvain Mollier, Christophe Panaïotis, Benoît Renaux, Xavier Rochel, Aline Salvaudon, Marie Thomas, Thierry Touzet, Daniel Vallauri

Integrating environmental history is essential for understanding present-day ecosystem dynamics and guiding modern conservation strategies, particularly under the EU Biodiversity Strategy for 2030. This data paper presents the nationwide digitisation and vectorisation of the French General Ordnance survey map (1818–1866), capturing the country's forest cover at its historical minimum (a pivotal moment known as the "forest transition"). The original manuscript sheets surveyed at a 1:40,000 scale (976 sheets) were scanned, georeferenced and vectorised. They offer higher thematic accuracy than the older Cassini map and are far more feasible for nationwide vectorisation than the highly detailed Napoleonic cadastre. The area-weighted mean survey date is 1843, and the dataset covers 99.6% of modern mainland France. To correct for paper deformation, historical surveying errors and coordinate system transformations, a rigorous workflow was established. This shifted from a global 6-parameter affine transformation (root mean square error of 60 m) to a local elastic transformation based on thousands of control points, for 21% of the territory, which reduced the positioning error to 34 m. We assessed data quality by comparing the General Ordnance Survey maps with the Napoleonic cadastre–the standard reference for 19th-century land-use data–across more than 600 municipalities. The correlation between the two sources regarding forest cover percentages was exceptionally high (r>0.9). Because forests formed large, compact blocks of significant strategic interest to military engineers, they were mapped with high precision. Localised inaccuracies were primarily found in remote areas, most notably in the mountains. The resulting historical vector layer was intersected with contemporary forest data (BD Forêt® v2, 2005–2019). Historical polygons smaller than 0.5 ha were filtered out to comply with modern FAO forest definitions. This spatial overlay generated a new dataset detailing four distinct land-use trajectories: . ancient forests (44.8% of present day forest): land classified as forest in both the 19th century and the present day (indicating maximum temporal continuity). . recent forests (55.2% of present day forest): land that was non-forested in the 19th century but has since undergone reforestation. . deforested areas (18.9% of 19th–century forest): land recorded as forest in the 19th century but subsequently converted to other land uses. . stable non-forest areas: land that has remained unforested across both periods. Our results suggest that the forest area of mainland France at its historical minimum should be revised upward to 9.9 million ha. A preliminary analysis further indicates that current spatial variation in forest cover is explained more by land-use dynamics occurring since the forest transition than by the initial extent of forest cover. These two open-access national datasets (the 19th-century forest layer and the land-use transition map) open the door to a better understanding of present-day forests : e.g. their biodiversity, soil quality, tree growth and belowground water quality. Furthermore, they provide decision-support tools for conservation planning.