| Description | This dataset quantifies the aboveground live carbon density (Mg C/ha), canopy cover (%), and canopy height (m) of plants at 30 m spatial resolution. Data has global coverage and annual temporal resolution, available from 2013 onwards. Data includes pixel-level variable uncertainties (90% prediction intervals). |
|---|---|
| Usage notes | New time points are delivered 5-6 months after year end. Spatial coverage is limited to land surfaces between 76°N and 56°S. In urban areas, double-bounce radar backscatter among buildings may result in over-predictions. Circular spatial artefacts can lead to temporal differences between adjacent pixels. |
| AOI restrictions | Maximum size 500,000 ha. Must have < 1,500 vertices. |
| Licence | Commercial |
| CRS | EPSG:4326 |
| Version | 1.1 |
| Release date | April 2024 |
aboveground_live_carbon_density| Description | The density of carbon stored in live woody vegetation on a per hectare basis. Data is generated using LightGBM regression models trained to predict GEDI L4A aboveground biomass carbon from modelled canopy height, modelled canopy cover, elevation, and location. Dataset includes 90% prediction interval variables (aboveground_live_carbon_density_lower_bound, aboveground_live_carbon_density_upper_bound) |
|---|---|
| Usage notes | Calculations are restricted to live woody aboveground biomass. A relative carbon content of 0.476 was used to convert GEDI L4A biomass data from total biomass density to biomass carbon density. Pixels with canopy_cover of 0% are set to 0 Mg C/ha. |
| Type | int16 |
| Units | Mg C/ha |
canopy_cover| Description | The relative cover of trees. Data is generated using U-Net image regression models trained to predict > 9 million km² of airborne LiDAR data from a combination of spaceborne remote sensing datasets. Dataset includes 90% prediction interval variables (canopy_cover_lower_bound, canopy_cover_upper_bound). |
|---|---|
| Usage notes | Trees are defined as vegetation with canopy_height > 5 m. Buildings, bridges, and other non-vegetation classes are excluded. Values < 10% are set to 0%. Pixels with canopy cover of 0% at the start and end of the full temporal coverage are set to 0% throughout. |
| Type | uint8 |
| Units | % |
canopy_height| Description | The average height (m) of trees. Data is generated using U-Net image regression models trained to predict > 9 million km² of airborne LiDAR data from a combination of spaceborne remote sensing datasets. Dataset includes 90% prediction interval variables (canopy_height_lower_bound, canopy_height_upper_bound). |
|---|---|
| Usage notes | Trees are defined as vegetation with canopy_height > 5 m. Pixel values are averages of higher resolution point clouds, so modelled height values are shorter than the tallest individual tree in a pixel. Buildings, bridges, and other non-vegetation classes are excluded. Square spatial artefacts can result in canopy height appearing up to 2 m higher in a pixel relative to surrounding pixels, but remain within prediction intervals. Pixels with a canopy height difference of < 3 m between the start and end of the full temporal coverage are set to the median value. Canopy height for pixel values with canopy_cover of 0% is set to 0 m. |
| Type | uint8 |
| Units | m |
day_of_year