[Paper Review] Estimation of forest height and biomass from open-access multi-sensor satellite imagery and GEDI Lidar data: high-resolution maps of metropolitan France

David Morin, Milena Planells+2 more|arXiv (Cornell University)|Oct 23, 2023

Remote Sensing and LiDAR Applications4 citations

TL;DR

This study develops a machine learning framework to estimate forest height and biomass across metropolitan France using open-access satellite data (Sentinel-1, Sentinel-2, ALOS-2) and GEDI Lidar as reference. It produces high-resolution maps with a mean absolute error of 4.3 m for height and 30 m³/ha for aboveground biomass when aggregated by forest type, enabling national-scale forest monitoring and carbon accounting.

ABSTRACT

Mapping forest resources and carbon is important for improving forest management and meeting the objectives of storing carbon and preserving the environment. Spaceborne remote sensing approaches have considerable potential to support forest height monitoring by providing repeated observations at high spatial resolution over large areas. This study uses a machine learning approach that was previously developed to produce local maps of forest parameters (basal area, height, diameter, etc.). The aim of this paper is to present the extension of the approach to much larger scales such as the French national coverage. We used the GEDI Lidar mission as reference height data, and the satellite images from Sentinel-1, Sentinel-2 and ALOS-2 PALSA-2 to estimate forest height and produce a map of France for the year 2020. The height map is then derived into volume and aboveground biomass (AGB) using allometric equations. The validation of the height map with local maps from ALS data shows an accuracy close to the state of the art, with a mean absolute error (MAE) of 4.3 m. Validation on inventory plots representative of French forests shows an MAE of 3.7 m for the height. Estimates are slightly better for coniferous than for broadleaved forests. Volume and AGB maps derived from height shows MAEs of 75 tons/ha and 93 m${}^3$/ha respectively. The results aggregated by sylvo-ecoregion and forest types (owner and species) are further improved, with MAEs of 23 tons/ha and 30 m${}^3$/ha. The precision of these maps allows to monitor forests locally, as well as helping to analyze forest resources and carbon on a territorial scale or on specific types of forests by combining the maps with geolocated information (administrative area, species, type of owner, protected areas, environmental conditions, etc.). Height, volume and AGB maps produced in this study are made freely available.

Motivation & Objective

• To extend a local forest parameter estimation method to national-scale mapping across metropolitan France.
• To leverage open-access satellite imagery and GEDI Lidar data for high-resolution forest monitoring.
• To produce accurate, spatially explicit maps of forest height, volume, and aboveground biomass (AGB) for carbon and forest management applications.
• To validate the accuracy of the maps using both airborne laser scanning (ALS) data and field inventory plots.
• To enable territorial-scale analysis by integrating forest maps with geospatial data on ownership, species, and protected areas.

Proposed method

• Trained a machine learning model using GEDI Lidar data as ground truth for forest height across France.
• Integrated multi-sensor satellite data from Sentinel-1 (C-band radar), Sentinel-2 (optical), and ALOS-2 (PALSAR-2 radar) to capture forest structural and spectral characteristics.
• Applied a supervised learning approach to predict forest height from satellite reflectance and backscatter data at 10 m spatial resolution.
• Converted estimated forest height into volume and aboveground biomass (AGB) using species-specific allometric equations.
• Validated predictions against independent ALS-derived height maps and national forest inventory plots.
• Aggregated results by sylvo-ecoregion and forest type to improve accuracy and support targeted analysis.

Experimental results

Research questions

• RQ1 Can a machine learning model trained on GEDI Lidar data accurately estimate forest height across large, heterogeneous regions like metropolitan France?
• RQ2 How does the accuracy of forest height and biomass estimation vary between coniferous and broadleaved forests using multi-sensor satellite data?
• RQ3 To what extent do aggregated maps by forest type and sylvo-ecoregion improve the accuracy of biomass and volume estimates?
• RQ4 Can open-access satellite data combined with GEDI Lidar produce high-resolution, nationally consistent forest maps suitable for carbon accounting and forest management?
• RQ5 How do the model's predictions compare to independent ALS and field inventory data in terms of mean absolute error (MAE) for height and biomass?

Key findings

• The forest height map achieved a mean absolute error (MAE) of 4.3 m when validated against ALS-derived maps, approaching state-of-the-art accuracy.
• Validation on national forest inventory plots yielded an MAE of 3.7 m for forest height, with slightly better performance for coniferous forests.
• Volume and aboveground biomass (AGB) maps showed MAEs of 75 tons/ha and 93 m³/ha, respectively, when evaluated at the national scale.
• When aggregated by forest type and sylvo-ecoregion, the MAE for AGB was reduced to 30 m³/ha, indicating improved accuracy through stratification.
• The model demonstrated robust performance across diverse forest types and regions, supporting reliable national-scale monitoring.
• The resulting high-resolution maps of forest height, volume, and AGB are freely available for research, policy, and land management applications.