Dataset Description
For 2005-2012 and 2015-2024 epochs, CCI Biomass v7 provides spatially explicit estimates of AGB, expressed in tonnes (t) or Megagrams (Mg) per hectare (ha; equivalent to 100 m) and defined as the mass, expressed as the oven-dry weight of the components (stem, bark, branches, twigs; but excluding stumps and roots) of all living woody plants (primarily trees). Maps at coarser spatial resolutions of 1 km, 10 km, 25 km and 50 km are also available.
Each annual map is accompanied by a corresponding standard deviation (SD) layer that provides a per-pixel estimate of uncertainty. AGB change products are provided on an annual and decadal basis, with these expressed as the difference between AGB maps for either two consecutive years (e.g. 2020-2019) or over a decade (2020-2010). The SD of the estimated AGB change and a quality flag map is provided, which suggests the reliability of the AGB change estimate.
Usage
When working with ESA CCI’s BIOMASS products, users are advised to note the following:
- The AGB and AGB change maps are not masked for forest. Users wishing to derive forest land statistics are advised to apply their own forest/non-forest dataset to the CCI BIOMASS data products.
- A pixel with a value equal to 0 means that the AGB is equal to 0 Mg/ha.
- The AGB and AGB change values are representative of the pixel area (i.e., they do not account for partial or total forest cover within the area of the pixel). The biomass stock in a pixel is obtained from the AGB value multiplied by the area of the pixel (NOTE: the area of the pixel changes with latitude).
- AGB refers to the aboveground LIVE dry biomass. However, AGB is derived from spaceborne Synthetic Aperture Radar (SAR) backscatter observations that are sensitive to wood density and the moisture content of vegetation and hence values larger than 0 Mg ha-1 occur and may correspond to dead standing wood.
- Shrublands/grasslands are not accounted for. All AGB estimates were obtained with a retrieval model relating the radar backscatter to the AGB of woody vegetation.
- The biomass of understory and liana vegetation is not accounted for as it is assumed that the remote sensing data used to estimate biomass do not contain sufficient signal scattered from layers below the top canopy.
- Use of the AGB estimates of individual full resolution pixels should be avoided as they have very high uncertainty.
- The maps are global and may, be biased at regional level. Users intending to adopt the CCI BIOMASS data product for national or sub-national studies are advised to check the quality of the CCI product by comparing with local data (e.g., from own field survey, National Forest Inventory, airborne datasets, etc.).
- The CCI BIOMASS datasets have not been cross-checked with other CCI datasets.
- The AGB in each annual map corresponds to an average value for the corresponding year. Consequently, the AGB value in areas that have been affected by changes within a given year may not be representative of the AGB either before or after the change.
- The annual maps for 2015-2024 are equally reliable and supersede any previous release.
The annual maps for 2015 and 2016 have different qualities with respect to later years because of the smaller number of satellite observations.
- The annual map for 2022 has a different quality with respect to previous years because of the smaller number of satellite observations. Across the boreal zone, the maps for 2023 and 2024 may not be fully consistent with previous years due to a larger number of L-band observations used.
- The 2010 dataset is an improved version of older releases and of the GlobBiomass AGB dataset (http://globbiomass.org), which is superseded.
- The datasets for 2005, 2006, 2011 and 2012 are based on C-band only and might not be consistent with those of 2007-2010. In regions where the AGB estimates are based on L-band data only (e.g., tropical rainforest), the AGB estimates were extrapolated and may therefore not represent the correct trend.
- The 2005-2012 dataset may indicate AGB > 0 corresponding to non-forest areas instead of AGB = 0 as in more recent years. This is attributed to the lower quality of the EO data used to estimate AGB. Users are advised to mask out non-forest areas to avoid incorrect interpretation of temporal trends.
- Fluctuations of the AGB values from one year to the next may occur in mixed landscapes, cropland and arid land due to the inter-annual variability of the number of radar observations. Trends in the time series of AGB maps have not been analysed and may therefore need to be critically investigated, particularly when comparing data from 2005-2012 with more recent years because of the different set of EO observations.
AGB change maps
- The CCI Biomass CRDP provides the standard deviation of the change value and a quality flag to interpret the reliability of the change value.
- AGB change maps should be interpreted carefully. It is strongly advised to use the quality flag layer to understand the reliability of the AGB change values. AGB change values corresponding to pixels labelled as “improbable change” may be unrealistic. AGB change values corresponding to pixels labelled as “gain” or “loss” are realistic, although the standard deviation AGB change map should also be considered.
- Be extremely cautious when using the CCI BIOMASS maps to assess AGB changes. We strongly advise to validate changes before further analysis of the data.
Aggregated data products
-
AGB and AGB standard error values are for the pixel area at the scale of aggregation (i.e., the scale of the averaging).
-
AGB change values at coarser resolution are derived from the aggregated AGB maps by differencing. The standard error of the AGB change map is obtained by summing the square of the AGB standard error maps and taking the square root.
Methodology
A description of these data layers is available in the Algorithm Theoretical Basis Document (ATBD) of the CCI BIOMASS project. Note that the model training phase does not require in situ observations, such as AGB plot data. AGB estimates were derived from spaceborne observations comprising C-band SAR data from Envisat ASAR (2005–2012) and Sentinel-1 (from 2015 onwards), together with L-band SAR data from ALOS-1 PALSAR-1 (2007–2010) and ALOS-2 PALSAR-2 (from 2015 onwards). Initially, a single algorithm, referred to as BIOMASAR, is applied to obtain separate global maps of AGB from the C-band and L-band datasets. BIOMASAR inverts a semi-empirical model that relates forest backscatter to canopy density and canopy height. The model uses two functions relating canopy density to height (based on spaceborne LiDAR measurements) and canopy height to AGB (based on spaceborne LiDAR height metrics and global statistics of AGB produced by National Forest Inventories). The retrieval model contains three unknown parameters corresponding to ground and canopy backscatter and forest attenuation. These parameters are estimated using auxiliary datasets, including information on canopy density, microwave transmissivity and maximum biomass. A detailed description of these datasets and the retrieval framework are given in the ATBD. Note that the model training process does not require in situ AGB plot measurements
The AGB maps produced independently from the C-band (BIOMASAR-C) and L-band (BIOMASAR-L) observations are subsequently merged using a set of weighting rules designed to minimise systematic errors. Greater weight is generally assigned to L-band estimates in high-biomass forests because of the reduced sensitivity of C-band backscatter in denser and more mature forest canopies. In lower-biomass and regenerating forests, the final AGB estimate typically reflects a combination of both retrievals. Per-pixel uncertainty estimates are generated by propagating uncertainties associated with the SAR measurements, the retrieval model and the merging procedure.
Validation results at 0.1° in different AGB bins. The size of bins is based on the number of plots inside the bin. The error bars indicate the 25th and 75th quantiles of the map value.
Validation results at 0.1° in AGB bins of 25 Mg ha-1. The number of plots inside each bin is indicated by the size of the dot marking. The error bars indicate the 25th and 75th quantiles of the map value.
Uncertainty and Accuracy
Validation of the ESA CCI 2020 map for 2020 used a dedicated tool (plot2map) that allowed comparison with an extensive dataset of ground-based tree size measurements collected from small (averaging 0.15 ha; n > 70,549) and medium (0.9 - 3 ha; n > 462) to large-sized (> 6 ha; n > 20) research plots (Tiers 1-3 respectively), with adjustments made for temporal discrepancies and partial forest fractions. Comparison against Tier 1 and Tier 2 plots indicated that globally, the CCI Biomass maps at their original 1 ha resolution tended to slightly over-predict up to 300 Mg ha-1 and under-predict beyond that. These differences were attributed in part to within-pixel sampling error that occurred because the AGB of single small plots may significantly differ from the population mean in the pixel. Spatial aggregation of plot and map data to 0.1° cells (a level of aggregation suitable for most climate modellers) considerably improved the agreement. In general, between 50 Mg ha-1 and 400 Mg ha-1, mean differences between the mapped and reference AGB were well within 20% at the 0.1° cell level. Validation at 0.1° of the Biomass CCI 2020 map and comparison with the validation from previous CCI maps are shown below. Validation of the 2010, 2017 and 2018 products can be found in the project PVIR.
Dataset Sustainment
Annual products will continue to be released (e.g., for 2025) plus an investigation into the feasibility of products for past years (2005/7 and 2015/16). Change maps have been produced for 2010-18 and 2017-18, and will continue to be developed and improved.
Technical Characteristics
Spatial resolution: ~100 m at the Equator
Geographical coverage: Global
Temporal coverage: 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2015, 2016, 2017, 2018, 2019, 2020, 2021, 2022, 2023 and 2024
Update frequency: Annual
Format: NetCDF (one global file) and GeoTiff (10° x 10° tiles)
Data Policy: Creative Commons Attribution 4.0 International (CC-BY-4.0)
Associated Guidance or User Manual
Project website: Link
Product User Guide (v7.0): Link
Validation report available from: Link
CCI Biomass Version 7.0 data access:
Link
Points of contact for queries
Maurizio Santoro
GAMMA RS
Switzerland
Email: santoro@gamma-rs.ch
