Schwartz et al. (2026) FORMSpoT: Revealing fine-scale forest disturbances from nation-wide 1.5 m forest canopy height time series

Identification

• Journal: Remote Sensing of Environment
• Year: 2026
• Date: 2026-08-26
• Authors: Martin Schwartz, Fajwel Fogel, Nikola Bešič, Damien Robert, Louis Geist, Jean-Pierre Renaud, Jean‐Matthieu Monnet, Clemens Mosig, Cédric Vega, Alexandre d’Aspremont, Loïc Landrieu, Philippe Ciais
• DOI: 10.1016/j.rse.2026.115631

Research Groups

• Laboratoire des Sciences du Climat et de l’Environnement (LSCE/IPSL), CEA-CNRS-UVSQ, Université Paris-Saclay
• CNRS & Département d’Informatique, École Normale Supérieure – PSL
• Université Gustave Eiffel, Géodata Paris, IGN, LIF, Nancy
• EcoVision Lab, DM3L, University of Zurich
• LIGM, École des Ponts, CNRS, IPP, UGE, Champs-sur-Marne
• Office National des Forêts, Pôle Recherche et Développement Innovation
• Université Grenoble Alpes, INRAE, LESSEM
• Institute for Earth System Science and Remote Sensing

Short Summary

This study introduces FORMSpoT, a decade-long (2014-2024) country-scale mapping of forest canopy height at 1.5 m resolution over France, derived from SPOT-6/7 imagery using a hierarchical transformer model trained on LiDAR HD data. The resulting dataset, FORMSpoT-Δ, provides annual fine-scale forest disturbance polygons validated against successive airborne LiDAR acquisitions and French National Forest Inventory plot revisits.

Objective

• Investigate the effectiveness of high-resolution satellite time series in detecting forest disturbances at the scale of individual trees.
• Evaluate the contribution of small canopy gaps to the overall area of disturbed forests in Europe.
• Develop a method for extracting disturbance polygons from a height time series using a 5 m height-loss threshold and morphological filtering.

Study Configuration

• Spatial Scale: National-scale mapping of forest canopy height at 1.5 m resolution over France (551,695 km2).
• Temporal Scale: Decade-long (2014-2024) annual time series of forest canopy height maps.

Methodology and Data

• Models used: PVTv2 pyramid transformer model pretrained on the Open-Canopy dataset.
• Data sources: SPOT-6/7 imagery, LiDAR HD data, French National Forest Inventory plot measurements.

Main Results

• FORMSpoT achieves an r2 of 0.74 and a MAE of 3.85 m against NFI reference heights.
• FORMSpoT remains consistent over time with limited variability in yearly MAE (SD = 0.08 m) and r2 (SD = 0.02).
• FORMSpoT-Δ polygons closely match ALS-derived reference disturbances at low-mountain mixed forests but miss small disturbances at high-mountain sites.

Contributions

• This study provides a new approach for detecting forest disturbances at the scale of individual trees using high-resolution satellite time series.
• The resulting dataset, FORMSpoT-Δ, offers a valuable resource for researchers and managers interested in understanding forest dynamics and disturbance regimes.

Funding

• This research was funded by the French National Research Agency (ANR) through the DINAMIS initiative.

Citation

@article{Schwartz2026FORMSpoT,
  author = {Schwartz, Martin and Fogel, Fajwel and Bešič, Nikola and Robert, Damien and Geist, Louis and Renaud, Jean-Pierre and Monnet, Jean‐Matthieu and Mosig, Clemens and Vega, Cédric and d’Aspremont, Alexandre and Landrieu, Loïc and Ciais, Philippe},
  title = {FORMSpoT: Revealing fine-scale forest disturbances from nation-wide 1.5 m forest canopy height time series},
  journal = {Remote Sensing of Environment},
  year = {2026},
  doi = {10.1016/j.rse.2026.115631},
  url = {https://doi.org/10.1016/j.rse.2026.115631}
}