This dataset is part of a Publisher Catalog, and not managed by Google Earth Engine. Contact Land & Carbon Lab for bugs or view more datasets from the Land & Carbon Lab Global Pasture Watch Catalog. Learn more about Publisher datasets.
- Catalog Owner
- Land & Carbon Lab Global Pasture Watch
- Dataset Availability
- 2000-01-01T00:00:00Z–2023-01-01T00:00:00Z
- Dataset Provider
- Land and Carbon Lab Global Pasture Watch
- Contact
- Land & Carbon Lab
- Earth Engine Snippet
-
ee.ImageCollection("projects/global-pasture-watch/assets/ggc-30m/v1/nat-semi-grassland_p")
- Cadence
- 1 Year
- Tags
Description
This dataset provides global annual probability maps of natural/semi-natural grassland from 2000 to 2022 at 30-m spatial resolution. Produced by Land & Carbon Lab Global Pasture Watch initiative, the mapped grassland extent includes any land cover type, which contains at least 30% of dry or wet low vegetation, dominated by grasses and forbs (less than 3 meters) and a:
- maximum of 50% tree canopy cover (greater than 5 meters),
- maximum of 70% of other woody vegetation (scrubs and open shrubland), and
- maximum of 50% active cropland cover in mosaic landscapes of cropland & other vegetation.
The grassland extent is classified into two classes: - Cultivated grassland: Areas where grasses and other forage plants have been intentionally planted and managed, as well as areas of native grassland-type vegetation where they clearly exhibit active and heavy management for specific human-directed uses, such as directed grazing of livestock. - Natural/Semi-natural grassland: Relatively undisturbed native grasslands/short-height vegetation, such as steppes and tundra, as well as areas that have experienced varying degrees of human activity in the past, which may contain a mix of native and introduced species due to historical land use and natural processes. In general, they exhibit natural-looking patterns of varied vegetation and clearly ordered hydrological relationships throughout the landscape.
The implemented methodology considered GLAD Landsat ARD-2 images (processed into cloud-free bi-monthly aggregates, see Consoli et al, 2024 ), accompanied by climatic, landform and proximity covariates, spatiotemporal machine learning (per-class Random Forest) and over 2.3 million reference samples (visually interpreted in Very High Resolution imagery). Custom probability thresholds (based on five-fold spatial cross-validation and balanced precision and recall values) were used to derive dominant class maps, 0.32 and 0.42 for cultivated and natural/semi-natural grassland probability thresholds, respectively.
Limitations: Grassland extent is partly under-predicted in southeastern Africa (Zimbabwe and Mozambique) and in eastern Australia (shrublands and woodlands of the Mulga ecoregion). Cropland is misclassified as grassland in parts of northern Africa, the Arabian Peninsula, Western Australia, New Zealand, the center of Bolivia, and Mato Grosso state (Brazil). Due to the Landsat 7 SLC failure, regular stripes of grassland probabilities are visible at parcel-level, particularly in the year 2012. The usage of coarser resolution layers (accessibility maps and MODIS products) introduced curvilinear macroscopic errors (due to the downscaling strategy based on cubicspline) in Uruguay, Southwest Argentina, South of Angola and in the Sahel region in Africa. Users need to be aware of the limitations and known issues; whilst considering them carefully to ensure appropriate use of maps at this initial prediction stage. GPW is working actively to collect systematic feedback via the Geo-Wiki platform, validate the current version and improve future versions of the dataset.
For more information see Parente et. al, 2024, Zenodo and https://github.com/wri/global-pasture-watch
Bands
Bands
| Name | Min | Max | Pixel Size | Description |
|---|---|---|---|---|
probability |
0 | 100 | 30 meters | Natural/Semi-natural grassland probability value derived through Random Forest. |
Image Properties
Image Properties
| Name | Type | Description |
|---|---|---|
| version | INT | Product version |
Terms of Use
Terms of Use
Citations
Parente, L., Sloat, L., Mesquita, V., et al. (2024) Global Pasture Watch - Annual grassland class and extent maps at 30-m spatial resolution (2000—2022) (Version v1) [Data set]. Zenodo doi:https://doi.org/10.5281/zenodo.13890401
Parente, L., Sloat, L., Mesquita, V., et al. (2024). Annual 30-m maps of global grassland class and extent (2000–2022) based on spatiotemporal Machine Learning, Scientific Data. doi: http://doi.org/10.1038/s41597-024-04139-6
DOIs
Explore with Earth Engine
Code Editor (JavaScript)
Map.setCenter(63.72366666, 38.30182604, 4); var nat_semi_grassland = ee.ImageCollection( "projects/global-pasture-watch/assets/ggc-30m/v1/nat-semi-grassland_p" ) var min_prob = 42 // Probability threshold var visParams = {min: 15, max: 85, palette: 'f7f1e5,af8260,803d3b,322c2b'} var nat_semi_grassland_2022 = nat_semi_grassland.filterDate('2022-01-01', '2023-01-01').first(); Map.addLayer( nat_semi_grassland_2022.mask(nat_semi_grassland_2022.gte(min_prob)), visParams, 'Natural/Semi-natural grassland prob. (2022)' ); var nat_semi_grassland_2000 = nat_semi_grassland.filterDate('2000-01-01', '2001-01-01').first(); Map.addLayer( nat_semi_grassland_2000.mask(nat_semi_grassland_2000.gte(min_prob)), visParams, 'Natural/Semi-natural grassland prob. (2000)' );