Leveraging GEDI and NFI to inform on forest productivity at the level of the management units
Abstract
Key message: The integration of national forest inventory data with spaceborne remote sensing data from the Global Ecosystem Dynamics Investigation (GEDI) improves the precision of forest productivity estimates in small domains via area-level Fay-Herriot models. Despite sparse spatial coverage, GEDI data enhance the reliability of localized estimates at finer spatial and temporal resolutions, thus supporting improved forest management and decision-making. Context: Forest productivity is highly sensitive to environmental changes, necessitating reliable, localized estimation for effective forest management. National forest inventories typically use design-based estimators that are unbiased but imprecise at such fine scales due to inadequate sample size. In such circumstances, model-based small area estimation provides approaches for improving estimate precision by leveraging auxiliary data. The recent availability of GEDI LiDAR data offers a promising auxiliary source to support such estimation. Aims: This study evaluated the suitability of GEDI data as auxiliary information in a Fay-Herriot model and compared estimates of forest productivity to traditional design-based direct estimators in localized domains. We also evaluated changes in basal area production and volume production (increment + ingrowth) over a 4-year period. Methods The study, conducted in the Région Bourgogne-Franche-Comté, Eastern France (45,000 km 2 ), used public forest data from the French National Forest Inventory and GEDI L2A metrics from 2019 to 2022. With area-level Fay-Herriot models, we estimated five forest attributes (quadratic mean diameter, basal area, volume, basal area production, and volume production) at two hierarchical levels of forest management. Results GEDI-augmented Fay-Herriot models achieved higher relative efficiency for forest attribute estimates than direct estimates, particularly in domains with limited field data. Precision gains exceeded 50% across all domains, with basal area production showing a slight temporal decrease that mostly remained statistically consistent with the temporally aggregated baseline, demonstrating FH’s capability to capture minor interannual dynamics within a generally stable multi-year trend. Conclusion: Our results highlight GEDI’s utility in addressing the limitations of national forest inventories to produce reliable estimates in heterogeneous landscapes where field observations are scarce. Future work should explore fusing GEDI with complementary data sources to expand spatial coverage and enable more continuous temporal tracking of forest dynamics.
Related
- Forest belowground productivity and carbon allocation predominantly driven by soil properties rather than climate
- Leaf economic strategy and stand structural complexity are associated with net woody productivity in a subtropical mixed forest
- Drought and Productivity Trends in Freshwater Flooded Forest Swamps in the North American Coastal Plain
- From bench to farm: leveraging Drosophila as a model insect for agricultural pest management.
- Mapping spatial colleague connectivity patterns from individual-level registry data to inform regional pandemic interventions
BibTeX
@article{onwunji2026leveraging,
title = {{Leveraging GEDI and NFI to inform on forest productivity at the level of the management units}},
author = {Irene Chiagozielam Onwunji and Cedric Vega and Sylvie Durrieu and Lauri Korhonen and Timo Tokola and Alexander Massey and Thomas Cordonnier and Olivier Bouriaud and Nikola Besic and Selim Behloul and Jean-Pierre Renaud},
journal = {Annals of Forest Science},
year = {2026},
month = sep,
doi = {10.1186/s13595-026-01358-2},
url = {https://hal.science/hal-05769618v1},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Leveraging GEDI and NFI to inform on forest productivity at the level of the management units
AU - Irene Chiagozielam Onwunji
AU - Cedric Vega
AU - Sylvie Durrieu
AU - Lauri Korhonen
AU - Timo Tokola
AU - Alexander Massey
AU - Thomas Cordonnier
AU - Olivier Bouriaud
AU - Nikola Besic
AU - Selim Behloul
AU - Jean-Pierre Renaud
PY - 2026
DA - 2026/09/29
JO - Annals of Forest Science
DO - 10.1186/s13595-026-01358-2
UR - https://hal.science/hal-05769618v1
AB - Key message: The integration of national forest inventory data with spaceborne remote sensing data from the Global Ecosystem Dynamics Investigation (GEDI) improves the precision of forest productivity estimates in small domains via area-level Fay-Herriot models. Despite sparse spatial coverage, GEDI data enhance the reliability of localized estimates at finer spatial and temporal resolutions, thus supporting improved forest management and decision-making. Context: Forest productivity is highly sensitive to environmental changes, necessitating reliable, localized estimation for effective forest management. National forest inventories typically use design-based estimators that are unbiased but imprecise at such fine scales due to inadequate sample size. In such circumstances, model-based small area estimation provides approaches for improving estimate precision by leveraging auxiliary data. The recent availability of GEDI LiDAR data offers a promising auxiliary source to support such estimation. Aims: This study evaluated the suitability of GEDI data as auxiliary information in a Fay-Herriot model and compared estimates of forest productivity to traditional design-based direct estimators in localized domains. We also evaluated changes in basal area production and volume production (increment + ingrowth) over a 4-year period. Methods The study, conducted in the Région Bourgogne-Franche-Comté, Eastern France (45,000 km 2 ), used public forest data from the French National Forest Inventory and GEDI L2A metrics from 2019 to 2022. With area-level Fay-Herriot models, we estimated five forest attributes (quadratic mean diameter, basal area, volume, basal area production, and volume production) at two hierarchical levels of forest management. Results GEDI-augmented Fay-Herriot models achieved higher relative efficiency for forest attribute estimates than direct estimates, particularly in domains with limited field data. Precision gains exceeded 50% across all domains, with basal area production showing a slight temporal decrease that mostly remained statistically consistent with the temporally aggregated baseline, demonstrating FH’s capability to capture minor interannual dynamics within a generally stable multi-year trend. Conclusion: Our results highlight GEDI’s utility in addressing the limitations of national forest inventories to produce reliable estimates in heterogeneous landscapes where field observations are scarce. Future work should explore fusing GEDI with complementary data sources to expand spatial coverage and enable more continuous temporal tracking of forest dynamics.
SN - 1286-4560
SN - 1297-966X
ER -