| Description | This dataset quantifies tree cover (%) in the year 2000, forest gain from 2000 to 2012, and, if present, the year of forest loss from the year 2000 onwards. Data has global coverage at 30 m spatial resolution. |
|---|---|
| Usage notes | New time points are delivered approximately annually. Ongoing improvements in Landsat sensors, land data availability, and algorithms may create inconsistencies in forest change metrics. Consider computing three-year moving averages, where appropriate. This dataset is not proven for rigorous area-based deforestation analysis (e.g. for EUDR). |
| Licence | Open |
| CRS | EPSG:4326 |
| Version | 1.12 |
| Release date | December 2024 |
tree_cover| Description | The canopy cover of trees for the year 2000. Data is generated using a decision tree trained to predict high resolution imagery and existing tree cover layers from Landsat 7 (2000+), Landsat 8 (2013+), and Landsat 9 (2022+) data. |
|---|---|
| Usage notes | Tree is defined as vegetation > 5 m height. |
| Type | uint8 |
| Units | % |
forest_gain| Description | Indicates whether a pixel shifted from a non-forest to forest over the period 2000-2012. Data is generated using a decision tree trained to predict high resolution imagery and existing tree cover layers from Landsat 7 (2000+), Landsat 8 (2013+), and Landsat 9 (2022+) data. |
|---|---|
| Usage notes | Values are shift to forest (1) or not (0). Forest is defined as a pixel with > 50% tree cover. Reported global and per-biome accuracy is > 99%. Variable has not been updated since v1.0. |
| Type | uint8 |
loss_year| Description | The year a forest pixel became non-forest, Data is generated using a decision tree trained to predict high resolution imagery and existing tree cover layers from Landsat 7 (2000+), Landsat 8 (2013+), and Landsat 9 (2022+) data. |
|---|---|
| Usage notes | Values are as either 0 (no loss) or a value of 1+ (loss in year 2001+). Forest loss is defined as stand-replacement disturbance or shift from forest to non-forest state. Reported global and per-biome accuracy is > 99%. Forest loss does not consider forest degradation (e.g. no change to non-forest state). This dataset version includes reprocessing of loss data from 2011 onwards, possibly leading to inconsistencies before and after this year. |
| Type | uint8 |
data_mask| Description | Indicates whether a pixel is land or a permanent water body for the period 2000-2012. |
|---|---|
| Usage notes | Values are no data (0), land surface (1), or permanent water body (2). |
| Type | uint8 |