Advancing Ecosystem Monitoring: Global Annual Maps of Biophysical Vegetation Properties (LAIe, FAPAR, FCOVER) for 2019-2025

Disclaimer This dataset is currently under revision. The data is available at three resolutions: 1000 m: This deposition (global mosaics as Cloud-Optimized GeoTIFFs) 100 m: See related datasets below 20 m and full 100 m products: Available on Google Earth Engine (GEE): LAIe: ee.ImageCollection('projects/ee-speckerfelix/assets/open-earth/laie_predictions-mlp_20m_v03') FAPAR: ee.ImageCollection('projects/ee-speckerfelix/assets/open-earth/fapar_predictions-mlp_20m_v03') FCOVER: ee.ImageCollection('projects/ee-speckerfelix/assets/open-earth/fcover_predictions-mlp_20m_v03') Earth-Engine App: https://ee-speckerfelix.projects.earthengine.app/view/global-trait-maps Earth-Engine Visualisation Script: Open in GEE Code Editor Earth-Engine Analysis Script: Open in GEE Code Editor Abstract We present S2BIOPHYS, a global dataset of annual vegetation biophysical properties derived from Sentinel-2 imagery at 20 m and 100 m resolution for 2019–2025. The dataset includes effective leaf area index (LAIe), fraction of absorbed photosynthetically active radiation (FAPAR), and fractional vegetation cover (FCOVER), estimated using a radiative transfer model inversion approach calibrated with in-situ data. Annual composites are generated from peak growing-season observations and include per-pixel mean values and uncertainty estimates. Global mosaics (EPSG:4326) are provided at 100 m and 1000 m resolution as Cloud-Optimized GeoTIFFs, with full-resolution products and extended uncertainty information available via Google Earth Engine. The dataset supports applications in ecosystem monitoring, biodiversity assessment, and restoration analysis. Description This dataset provides global annual composites of vegetation biophysical properties derived from Sentinel-2 data. The Zenodo archive contains global mosaics in geographic coordinates (EPSG:4326) at 1000 m resolution (this deposition) and 100 m resolution (child datasets). Annual Effective Leaf Area Index (LAIe) (2019–2025): Ensemble mean and total uncertainty (standard deviation) Annual Fractional Vegetation Cover (FCOVER) (2019–2025): Ensemble mean and total uncertainty (standard deviation) Annual Fraction of Absorbed Photosynthetically Active Radiation (FAPAR) (2019–2025): Ensemble mean and total uncertainty (standard deviation) Each product represents an annual composite based on multiple Sentinel-2 observations during the peak growing season. Resource Description This Zenodo record provides global annual composites of LAIe, FAPAR and FCOVER for 2019–2025 as Cloud-Optimized GeoTIFFs (COGs). The data are distributed as global mosaics in geographic coordinates (EPSG:4326) at 1000 m spatial resolution (this deposition), with additional 100 m products available as separate Zenodo records. Each product contains the ensemble mean and total uncertainty derived from multiple Sentinel-2 observations during the peak growing season. These datasets are designed to be analysis-ready and easily accessible for large-scale applications. The full dataset is available on Google Earth Engine (GEE) at 20 m and 100 m resolution in Sentinel-2's native UTM projection: ee.ImageCollection("projects/ee-speckerfelix/assets/open-earth/{laie, fapar, fcover}_predictions-mlp_{100,20}m_v03") The GEE products additionally provide observation counts ([laie, fapar, fcover]_count) and a decomposition of uncertainty into within-acquisition ([laie, fapar, fcover]_stdDev_within) and across-acquisition ([laie, fapar, fcover]_stdDev_across) components, enabling more detailed uncertainty analysis and quality filtering. In addition to these annual products, we provide the gee-biophys Python package, which enables users to generate custom spatiotemporal composites. It implements both S2BIOPHYS and SL2P and allows users to define time intervals, spatial extents, and resolution using a simple configuration file, supporting reproducible and scalable generation of biophysical maps. This framework also enables access to the full set of uncertainty components at native resolution where required. Data Format and Scaling Data Types: Mean and standard deviation maps are stored as int16. NoData Values: -9999 Scaling Factors: FAPAR and FCOVER: 0.0001 LAIe: 0.001 Note: Zenodo-hosted datasets (1000 m and 100 m) include only the ensemble mean and total uncertainty bands. Additional uncertainty components and observation counts are available exclusively via Google Earth Engine. Related datasets laie Mean 2019 2020 2021 2022 2023 2024 2025 Std 2019 2020 2021 2022 2023 2024 2025 fapar Mean 2019 2020 2021 2022 2023 2024 2025 Std 2019 2020 2021 2022 2023 2024 2025 fcover Mean 2019 2020 2021 2022 2023 2024 2025 Std 2019 2020 2021 2022 2023 2024 2025 Code The code used to generate the data is available at Zenodo, and at the GitHub repository. Naming Convention We follow the Open-Earth-Monitor file naming convention: Variable: [laie, fapar, fcover] Method: rtm.mlp.v02 Statistic: [mean, std] Resolution: [20m, 100m, 1000m] Depth: s (surface) Time start: YYYY0101 Time end: YYYY1231 Extent: go (global land, excluding Antarctica) CRS: epsg.4326 Version: v03 Contact Felix Specker: [e-mail] / [e-mail] Johan van den Hoogen: [e-mail]