Species- and region-specific growth curves, forest age (2020), and disturbance trends for French forests at 30 m resolution

Abstract. Spatially explicit biomass growth curves data that jointly resolve tree species, stand age and disturbance regime are scarce, particularly at fine spatial resolutions. We present five consistent data products for metropolitan French forests around 2020, derived from a 30 m dominant-species map, the 30 m FORMS-B satellite above-ground biomass (AGB) map and the Landsat-based European Forest Disturbance Atlas: (1) 290 AGB growth curves for 28 species categories across 11 large ecological regions; (2) a 30 m forest age map for 2020 covering 17.01 Mha of forests, obtained by inverting these curves against the AGB map; (3) 3,752 harvest and natural disturbance rates and patch sizes, with their 1999–2024 trends; (4) 3,752 event-based disturbance severities (AGB loss per disturbed hectare); and (5) 1,874 rotation periods from National Forest Inventory (NFI) plot samples remeasurements. Products 3–5 cover 28 species categories across 86 small ecological regions. These cover the species×region combinations present in French forests, not the full combinatorial grid. Not all 28 species categories occur in every ecological region; the totals reflect strata with forest area only. All products include uncertainties and gap-filling flags. Growth curves match National Forest Inventory-derived curves for early growth (+12±15 %, 95 % CI) but give a higher asymptotic biomass (+22±18 %) and a later half-saturation year (+33±23 %), partly reflecting differences in the representation of age extremes, as NFI measurements exclude trees <7.5 cm in diameter and we aggregate stands >60 years into a single age class. The age map matches the NFI nationally aggregated age distribution (median offset −2 [−20 to +42] yr). The 2020 harvested area (44±6 kha) and the resulting AGB loss (2.5±0.3 MtC.yr−1) are of the same order as independent clear-cut estimates (62 kha from the NFI; ~3 MtC.yr−1 from satellite-mapped clear-cuts over 2018–2024). The Landsat disturbance product captures only stand-replacing and high severity events and underestimate selective logging and diffuse mortality. These products can be combined directly to model forest carbon dynamics at species and regional resolution. All data are openly available at https://doi.org/10.5281/zenodo.23499335 (Pellissier-Tanon et al., 2026).