Land cover and land cover change

Loss of biodiversity and pressures on ecosystem services are among the most pressing global environmental challenges. Changes in land cover and land use are the leading contributors to terrestrial biodiversity loss.

Loss of natural and semi-natural vegetated land is presented as a proxy for pressures on biodiversity and ecosystems. The indicator is defined as the percentage of tree cover, grassland, wetland, shrubland and sparse vegetation converted to any other land cover type. Gains of natural and semi-natural vegetated land are conversions in the opposite direction. The denominator used is the ‘stock' of natural and semi-natural land at the start of the period.

Changes to and from 9 individual land cover types are also presented. These include conversions from natural and semi-natural vegetated land to cropland , and conversions from cropland to artificial surfaces, among others.

Land cover 'snapshots' for a given year provide context against which the conversions detailed above can be evaluated. This multi-class dataset allows for analysis of changes in land cover consistently at the global scale. It builds on decades of Earth observation missions by different national and supranational space organisations.

Important limitations for this dataset include the temporal heterogeneity of sensor inputs (different satellites at different times) which frustrates comparison of rates of change from one period to the next, (comparison of rates of change between different periods is not recommended) and the relatively more coarse resolution of change detection (approx. 1km) which means that smaller changes are not recorded. Furthermore, summary results for smaller countries and regions are more sensitive to errors caused by mis-classification and should be interpreted carefully. Note that quality flags describing the confidence in classification and accuracy of change detection are available for the underlying data (see link under data sources below), however these are not used in producing these statistics - presented here are simple zonal statistics (essentially pixel counts) of the classified maps and changes between them for different administrative units (with some convenient aggregations of land cover classes).

For users interested specifically in urbanisation or surface water, there are indicators based on higher-resolution datasets of built-up area and surface water published alongside that may be more suitable. Users interested in a specific country or region will likely find that nationally or regionally produced datasets (e.g. CORINE for EEA countries) tailored to local contexts are more suitable.

Data sources

• Land cover: Copernicus/European Space Agency and Université catholique de Louvain Geomatics Climate Change Initiative - Land Cover (CCI-LC) Annual Maps: https://cds.climate.copernicus.eu/cdsapp#!/dataset/satellite-land-cover
• For political and administrative boundaries, the latest OECD Territorial grid geographies are used where available. For the remaining countries, the FAO Global Administrative Unit Layers (GAUL 2014) are used (see the Boundary definition section below for details). The latest OECD-EU Functional Urban Area (FUA core + commuting zone) definition is used for cities.

Background, context and method

Hašcic, I. and A. Mackie (2018), "Land Cover Change and Conversions: Methodology and Results for OECD and G20 Countries", OECD Green Growth Papers, No. 2018/04, OECD Publishing, Paris, https://doi.org/10.1787/72a9e331-en.

Specific known issues:

• Global: conversions between the wetland definition (Shrub or herbaceous cover, flooded, fresh-saline or brackish water) and the flooded forest classes (Tree cover, flooded, fresh or brackish water, Tree cover, flooded, saline water) are typically spurious
• New Zealand: High-elevation areas misclassified as cropland in the early 1990's
• Northern Europe and Scandinavia: spurious change detection from trees to cropland in mosaic agricultural areas with mixed cropland/tree cover
• Norway: Urban areas misclassified around Bergen
• Very northern urban areas (e.g. Iceland, Greenland, N. Canada) are missing from the 2018 dataset however this has been resolved for the 2019 dataset.

Note on GIS-derived indicators

These indicators are calculated by intersecting political, administrative or functional urban area boundaries with raster datasets using GIS software. They provide accessible tabular statistics of the underlying datasets for a variety of geographic output areas that can be used immediately without performing the spatial analysis that would otherwise be required.

Prospective users are encouraged to examine the underlying data for their area of interest and familiarise themselves with the methodology used in their production so as to better understand what the data show and what kinds of conclusions they can be used to support. Irrespective of the underlying data, all earth-observation derived statistics come with caveats such as scale dependence, limitations associated with classification of continuous phenomena into discrete classes, and uneven geographical and temporal accuracy.

Boundary definition

• Large regions correspond to OECD territorial level 2 (TL2) or GAUL level 1 (GAUL 1).
• Small regions correspond to OECD territorial level 3 (TL3) or GAUL level 2 (GAUL 2).

The boundary unit codes indicate the source and year of the definition used. OECDXXXX2021 means that boundary comes from the OECD's reference areas as of 2021, and the unique ID in the source dataset is XXXX. Similarly GAUL1XXXX2015 means the boundaries come from GAUL 1 (2015) and the source GAUL id is XXXX. FUAs all come from the same dataset and simply use the unique ID from the source.

The OECD aggregate includes all 37 OECD members as of March 2021 plus Costa Rica (Costa Rica had been invited but had not yet finalised its membership at the time of the most recent update). Aggregate group memberships are not adjusted to reflect changing memberships over time.