# Europe-wide maps of biomass density based on satellite remote sensing data for 2017, 2020, 2021 and 2023

## TL;DR

This paper presents detailed maps of forest biomass in Europe using satellite data from 2017 to 2023, providing insights into forest structure and changes over time.

## Contribution

The study provides the first Europe-wide annual maps of forest biomass variables using Sentinel-1 and ALOS-2 SAR data with a consistent processing chain.

## Key findings

- Maps of Growing Stock Volume, Aboveground Biomass, and Belowground Biomass were generated at 20 m resolution for multiple years.
- Validation showed poor pixel-level accuracy but improved results at larger spatial scales like administrative units.
- Inter-annual consistency was affected by the number of satellite observations per year.

## Abstract

Spatially explicit information on forest structure and biomass is needed to meet the monitoring and reporting requirements of several European policies. Satellite images enable mapping and monitoring of the Europe’s forest resources through operational observations from the Sentinel-1 Synthetic Aperture Radar (SAR) and the Advanced Land Observing Satellite 2 (ALOS-2) Phased Array l-band SAR 2 (PALSAR-2) instruments. Data acquired in 2017, 2020, 2021 and 2023 were used to generate annual maps of forest biomass variables, namely Growing Stock Volume (GSV), Aboveground Biomass (AGB) and Belowground Biomass (BGB), with a pixel size of 20 m × 20 m. All products are in the geometry of the Sentinel-2 tiling system. A spatially averaged map with a pixel size of 100 m × 100 m (1 hectare) in geographic projection is also supplied, for users who do not require the highest spatial resolution. The maps were generated with a fully documented processing chain that includes (i) pre-processing of the SAR data to create stacks of co-registered terrain geocoded images of the backscattered intensity and (ii) inversion of a physically-based model to estimate GSV. AGB and BGB were subsequently estimated using allometric relationships. Per-pixel standard deviations were computed for each biomass variable by propagating uncertainties from both the SAR observations and the model parameters. The maps clearly reproduce the expected spatial patterns of forest biomass across Europe and provide sufficient spatial detail to identify biomass dynamics related to, e.g., logging and regrowth. Validation against measurements collected by National Forest Inventories (NFIs) indicates poor agreement with map values at the pixel scale, with errors larger than 50% of the reference biomass. The correspondence substantially improved for spatial aggregates, such as administrative units, for which the bias was mostly negligible and the mean square error was below 30% of the reference value. The number of ALOS-2 PALSAR-2 images affected the inter-annual consistency of the maps, which was lower in regions with only one or two observations per year.