Stock di Carbonio Attuale - Biomassa Aerea (AGB)

Land Usev0

Responsible:Niccolò Calandri

Current Carbon Stock — Above-Ground Biomass (AGB)

Overview

The Current Carbon Stock — Above-Ground Biomass (AGB) layer quantifies the carbon stored in living vegetation above the soil surface, including tree trunks, branches, leaves, and bark, expressed in tonnes of carbon per hectare (tC/ha). Above-ground biomass is the most visible and directly measurable carbon pool in terrestrial ecosystems, and it typically represents the largest share of living biomass carbon in forested landscapes.

AGB is a critical metric for carbon credit verification, REDD+ monitoring, national greenhouse gas inventories, and ESG reporting. Changes in AGB directly reflect deforestation, reforestation, forest degradation, and natural regrowth processes. The dataset is derived from the global carbon density baseline of Walker et al. (2022), which maps carbon pools at ~300 m resolution for the year 2016. To achieve finer spatial detail, the baseline AGB layer is downscaled using a Random Forest machine learning pipeline that integrates current satellite-derived features.

The AGB pool specifically captures carbon in:

Tree stems and trunks — the dominant carbon reservoir in forests
Branches and twigs — both live and attached dead wood
Foliage — leaves and needles
Bark — outer protective tissue
Standing dead trees — snags that have not yet fallen

Calculation Methodology

Random Forest downscaling pipeline for AGB extraction:

Baseline extraction — The Walker et al. 2022 AGB carbon density raster is loaded from S3 (Cloud-Optimized GeoTIFF) and clipped to the area of interest. The AGB pool is isolated from the total carbon density map using the pool-specific decomposition provided by the dataset.
Feature engineering — For each pixel, three predictor variables are computed:
- NDVI (median composite) from Sentinel-2 — strongly correlated with vegetation biomass
- Land cover class from Dynamic World — distinguishes forest, cropland, grassland, etc.
- Elevation from SRTM DEM — captures altitudinal biomass gradients
Training — A Random Forest regressor learns the relationship between baseline AGB values and the auxiliary features at the coarse resolution (~300 m)
Prediction — The trained model predicts AGB density at the higher resolution of the input features, yielding maps at approximately 200 m resolution
Validation — Cross-validation metrics (R-squared, RMSE) assess the downscaling quality

Output unit: tC/ha (tonnes of carbon per hectare, AGB pool only)

Data Sources

Code	Name	Provider	Type	Coverage	Resolution	Reference Year
Walker 2022	Global Carbon Density (AGB)	Walker et al. / S3	Raster (COG)	Global	~300 m	2016
`COPERNICUS/S2_SR_HARMONIZED`	Sentinel-2 L2A	ESA/Copernicus	Multispectral	Global	10 m	2017 — present
`GOOGLE/DYNAMICWORLD/V1`	Dynamic World	Google/WRI	Land cover	Global	10 m	2015 — present
`USGS/SRTMGL1_003`	SRTM DEM	NASA/USGS	Elevation	60N-56S	30 m	2000

Quality Range and Levels

Indicator	Unit	Range	Inverted
`carbon_stock_current_agb`	tC/ha	[0, 5, 20, 60, 100, 200]	No

Inverted = No: a higher value indicates greater above-ground carbon storage.

Level	Thresholds	Interpretation
E (Very Low)	< 5 tC/ha	Bare soil, urban areas, deserts, recently cleared land
D (Low)	5-20 tC/ha	Sparse shrublands, degraded grasslands, young plantations
C (Moderate)	20-60 tC/ha	Croplands with trees, open woodlands, Mediterranean maquis
B (High)	60-100 tC/ha	Mature temperate forests, dense shrublands, managed forests
A (Very High)	> 100 tC/ha	Tropical forests, old-growth temperate forests, mangroves

Technical Notes

AGB represents the most remote-sensing-responsive carbon pool due to the strong correlation between vegetation structure and spectral indices like NDVI
In very dense tropical forests, NDVI saturation may limit discrimination among high AGB values
The Walker 2022 baseline represents circa 2016 conditions; subsequent land use changes may cause discrepancies with current AGB
AGB does not include dead organic matter on the forest floor (litter, coarse woody debris), which is accounted for in the total carbon stock pool

References

Walker, W.S. et al. (2022). "The global potential for increased storage of carbon on land". Proceedings of the National Academy of Sciences, 119(23), e2111312119.
Spawn, S.A. et al. (2020). "Harmonized global maps of above and belowground biomass carbon density in the year 2010". Scientific Data, 7, Article 112.
Breiman, L. (2001). "Random Forests". Machine Learning, 45(1), pp. 5-32.

Data Sources & References

Sources

WRD_CARBN_22

Methodology

Il layer Carbon Stock Attuale - AGB viene calcolato estraendo la componente di biomassa aerea (Above-Ground Biomass) dalla mappa globale di densita di carbonio di Walker et al. (2022), anno di riferimento 2016. La biomassa aerea comprende il carbonio presente in tronchi, rami, foglie, corteccia e alberi morti in piedi. Il raster baseline a ~300 m viene raffinato attraverso un algoritmo di downscaling Random Forest che utilizza come variabili predittive l'NDVI mediano da Sentinel-2, la classificazione di copertura del suolo da Dynamic World e l'elevazione da SRTM. Il modello viene addestrato sulla relazione tra valori AGB e variabili ausiliarie, quindi predice la densita AGB alla risoluzione di ~200 m. L'output e espresso in tonnellate di carbonio per ettaro (tC/ha). Metriche di validazione (R-quadro, RMSE) vengono calcolate tramite cross-validazione.