Land Cover Background
From the GCOS “Implementation Plan for the Global Observing System for Climate in Support of the UNFCCC”, 2004:
A further area where historical data could be useful is in the creation of improved records of changes in natural land cover and land use. Existing historical land-cover data sets have a limited usefulness because of the coarseness of the classification schemes employed. Inventory data could be used to refine these classifications, providing there is sufficient documentation and the data are made available in spatially disaggregated form. GTOS, the Global Land Cover Network (GLCN) and GCOS (working with the appropriate national agencies) will request that national agencies retrieve archival data and make them available through appropriate International Data Centres.
ECV – Land Cover
Land cover and the change of land cover affect the services provided to human society (e.g., provision of food, fibre, recreational opportunities, etc.), force climate by modifying water and energy exchanges with the atmosphere, and change greenhouse gas and aerosol sources and sinks. Landcover distribution is partly determined by regional climate, so changes in cover may indicate climate change.
Although land-cover change can be observed using data from Earth observing satellites, currently available data sets vary in nature, are of uncertain accuracy and use different land-cover type characterization systems. Data of similar nature are provided from different sources and at different spatial resolutions. There are currently no viable plans to establish operational satellite systems capable of providing the long-term records needed for observations of changing land cover. It is necessary, and feasible with present-day technology, to provide satellite-based optical systems at 10-30m resolution with temporal, spectral and data-acquisition characteristics that are consistent with previous systems. Commitment to fly such systems though has not yet been secured from Space Agencies.
Data sets characterizing global land cover are currently produced at resolutions of between 250m and 1km by several Space Agencies in close cooperation with the research community (especially those research groups participating in the GTOS technical panel Global Observation of Forest and Land Cover Dynamics (GOFC-GOLD)). The lack of compatibility between these products makes it difficult to measure and monitor climate-induced or anthropogenic changes in land cover. A range of approaches is adopted, e.g., centralized processing using a single method of image classification (e.g., MODLAND), or a distributed approach using a network of experts applying regionally specific methods (e.g., GLC2000). Using a single source of satellite imagery and a uniform classification algorithm has benefits in terms of consistency, but may not yield optimum results for all regions and all land-cover types. Automated land-cover characterization and land-cover change monitoring thus remains a research priority.
It is necessary that land-cover classification systems and therefore map legends adhere to internationally-agreed standards. In the long-term, such standards should be agreed upon by the appropriate intergovernmental organization (see Action T1). In the near term, however, full benefit should be taken of existing initiatives, e.g., the FAO’s Land Cover Classification System for legend harmonization and translation and the legends published by the IGBP and
the GTOS GOFC-GOLD. As a minimum, new land-cover maps should be produced annually, documenting the spatial distribution of land-cover characteristics with attributes suitable for climate, carbon and ecosystem models, and using a common language for class definitions (e.g., include wetland information describing forest peat lands (boreal), mangroves, sedge grasslands, rush grasslands and seasonally flooded forests, and area of land under irrigation), at 250m-1km resolution.
In addition to their use in Earth system models, these global products will help identify areas of rapid change, although the development of automated detection of changes in land-cover characteristics remains high on the research agenda. The production of such land-cover data sets will involve Space Agencies for processing the satellite data used in the database production, the FAO/science panels to ensure legend relevance and standards and the research community for optimizing image classification approaches. Mechanisms to fund such partnerships are emerging (e.g., the EU global monitoring for environment and security initiative) but not yet guaranteed on a sustained basis.
Global land-cover databases must also be accompanied by a description of by-class thematic/spatial accuracy. The CEOS/WGCV, working with GOFC-GOLD and GLCN is establishing agreed validation protocols, which should be used. The current protocols base accuracy assessment on a sample of high-resolution (1-30m) satellite imagery, itself validated by in situ observations wherever possible. To better quantify changes in land-cover characteristics, these high-resolution data should also be used for wall-to-wall global mapping at resolutions of 10-30m. Maps at this resolution are needed every 5 years over long time periods (several decades). Global data sets of satellite imagery at 30m resolution have been assembled for selected years (e.g., 1990 and 2000) and some regional landcover maps have been generated from these. The technologies have been developed and tested (e.g., Landsat and the Satellite Probatoire d’Observation de la Terre High Resolution (SPOT HRV)),
suitable methods for land-cover characterization on these scales exist, but institutional arrangements to ensure operational generation of land cover at these resolutions are not yet in place. Space agencies should thus assure that suitable optical sensors with 10-30m resolution are available for operational monitoring using the same data acquisition strategies as those systems in current operation.
Samples of high-resolution satellite imagery have also been used to estimate change and are proposed for example by the FAO’s Global Land Cover Network. Initiatives such as these will provide much needed capacity-building and offer a framework for acquisition of in situ
observations to support the satellite image-based monitoring. Such in situ networks will also
provide information on how land is being used (as opposed to what is covering it). Land use cannot always be inferred from cover.
Action T22 (TF1)
Action: Establish international standards and specifications for the production of land-cover characterization maps.
Who: FAO, UNEP, GLCN and ISO (TC-211) in collaboration with GOFC-GOLD.
Time-Frame: Standards and specifications by 2005. Operation continuing.
Performance Indicator: Publication of standards.
Cost Implications: Category I.
Action T23 (TF8, TF9, TF10)
Action: Produce reliable accepted methods for land-cover map accuracy assessment.
Who: CEOS WGCV, in collaboration with GOFC-GOLD and GLCN.
Time-Frame: By 2005 then continuously.
Performance Indicator: Protocol availability.
Cost Implications: Category I.
Action T24 (TF8, TF9, TF10)
Action: Commit to continuous 10-30m resolution optical satellite systems with data acquisition strategies at least equivalent to the Landsat 7 mission for land cover. Who: Space Agencies.
Performance Indicator: Operational plans, data availability.
Cost Implications: Category IV.
Action T25 (TF8, TF9, TF10)
Action: Develop an in situ reference network and apply CEOS WGCV validation protocols for land cover.
Who: Parties’ national services, research institutes and Space Agencies, in cooperation with GOFC-GOLD, CEOS/WGCV, FAO GLCN and the GTOS web-based data system TEMS. Time-Frame: Network established by 2009.
Performance Indicator: Availability of validation statistics.
Cost Implications: Category III.
Action T26 (TF8, TF9, TF10)
Action: Generate annual products documenting global land-cover characteristics at resolutions
between 250m and 1km, according to internationally-agreed standards and accompanied by statistical descriptions of the maps’ accuracy.
Who: Parties’ national services, research institutes and Space Agencies through GLCN in
collaboration with GOFC-GOLD research partners, and the IGOS land theme (IGOL). Time-Frame: By 2005, then continuously.
Performance Indicator: Data set availability
Cost Implications: Category III.
Action T27 (TF8, TF9, TF10)
Action: Generate maps documenting global land cover at resolutions between 10m and 30m every 5 years, according to internationally-agreed standards and accompanied by statistical descriptions of the maps’ accuracy.
Who: Space Agencies, in cooperation with GCOS, GTOS, GLCN and other members of CEOS.
Time-Frame: First by 2005, then continuously.
Performance Indicator: Availability of operational plans, funding mechanisms and eventually
Cost Implications: Category III.
6.5 Terrestrial Domain – Synthesis and Findings
Specific five-year milestones include:
; Characterization of land cover (the location and extent of inland water-bodies,
vegetation type, snow and ice, barren, artificial surfaces, fractional tree cover) at 250-
In the longer term (5 to 10 years):
; Global land-cover change will be mapped at 30m resolution.
Findings – Terrestrial Domain
TF8: An international body should advise on standards for the production of land-cover maps, specifically in terms of the resolution and land-type characterization to be employed. TF9: Existing land-cover data should be analyzed and/or reprocessed, wherever possible, to ensure the compatibility of maps produced for the last decade. New land-cover maps should be produced every five years.
From GCOS “The Second Report on the Adequacy of the Global Observing Systems for Climate in Support of the UNFCCC”, 2003:
Land cover describes the distribution of vegetation types and human use of the land for living space, agriculture, and forestry. Natural vegetation distributions are in large part determined by regional climate, and their changes provide a way to monitor climate change. The spatial patter of land cover is also critical information for determining the capacity of biodiversity to adapt to climate change. Land-cover changes also occur in response to changes in weather patterns and changes in land management/land use (e.g., conversion of forest to pasture or cropland). Changes in land cover force climate by modifying water and energy exchanges with the atmosphere, and by changing greenhouse gas and aerosol sources and sinks. Many climatically-relevant variables that are difficult to measure at a global scale (e.g. surface roughness) can be inferred reasonably well from vegetation type. Thus land cover can be a surrogate for other important climate variables.
Although land-cover change can be measured using data from Earth observation satellites, the currently available global land-cover data sets vary significantly, are of uncertain accuracy and use different land-cover-type characterization systems. Data are also provided from different sources and at different spatial resolutions. Whilst the efforts of some Space Agencies and other entities to produce global land-cover products are important steps towards more adequate measurement of this important variable, the lack of compatibility between the products means that there are significant difficulties in using them to measure and monitor climate-induced or anthropogenic changes in land cover.
FINDING: An international body should advise on standards for the production of land-cover maps, specifically in terms of the resolution and land-type characterizations to be employed. FINDING: Existing land-cover data should be analyzed and/or reprocessed, wherever possible, to ensure the compatibility of maps produced for the last decade. New land-cover maps should be produced every five years.
FINDING: Historical land-use data sets could be significantly improved if more and better-documented inventory data sets were made available.
Land cover can change dramatically in a matter of days (loss of forest due to fire for example, or agricultural expansion) though change is more usually apparent on time scales of years. Vegetation has its own dynamics; the diurnal cycle of plant photosynthesis and respiration, growth, and differentiation with the seasons. Our capacity to monitor key variables
quantifying these dynamics (Fraction of Absorbed Photosynthetically Active Radiation (FAPAR), Leaf Area Index (LAI) and biomass) is improving.
SPECIFIC PROGRESS SINCE THE FIRST ADEQUACY REPORT
Finding 21. Support should be provided for the implementation of more reliable procedures of land use and land cover change. Space agencies should be encouraged to ensure the acquisition and exchange of appropriate fine and medium resolution data on a regular basis. Progress: There are now several instantaneous assessments of land cover based on data collected by Earth observation satellites (research missions). Some land-cover products are being produced on a routine basis by Space Agencies. There are two data sets describing historical changes in land cover, based on combining remotely-sensed data, inventory data and historical land-use information.
From GCOS “Status Report on the Key Climate Variables”, 2003:
Variable: Land cover (including vegetation type)
Main climate application
Land cover describes the distribution of vegetation types and human use of the land for living space, agriculture and forestry. Natural vegetation distributions are in large part due determined by regional climate, and their changes provide a way to monitor climate change. The spatial pattern of land cover is also critical information for determining the capacity of biodiversity to adapt to climate change. Land cover changes also occur in response to changes in weather patterns and changes in land management/land use. Changes in land cover force climate by modifying water and energy exchanges with the atmosphere, and by changing greenhouse gas and aerosol sources and sinks. Many climatically-relevant variables that are difficult to measure at a global scale (e.g. surface roughness) can be inferred reasonably well from vegetation type. Thus land cover can be a surrogate for other important climate variables.
Contributing GCOS/GTOS baseline observations
Other contributing observations
Although land cover can be measured using data from Earth Observation satellites, the currently available global land cover data sets vary significantly, are of uncertain accuracy and use different land cover type characterization systems. Data are also provided from different sources and at different spatial resolutions. The lack of compatibility between the products means that there are significant difficulties in using them to measure and monitor climate-induced or anthropogenic changes in land cover.
International organisations, such as FAO, contribute to the land cover by providing national tabular assessments of forests, grazing and crop areas on a regular basis (e.g., the FAO forests assessment was updated every decade; crop statistics are provided on an annual basis). The locations of these land covers are not specified, which reduces the utility of the data sets.
Significant data management issues
The current land-cover data sets are well described in the scientific literature and freely available for the Internet.
Several research groups at the University of Wisconsin and RIVM have tried to enhance the resolution of the land-cover data sets by combining tabular, satellite-based and other data sources. In this way they estimated patterns of land-cover change over the last century.
The current observation capability allows that local and regional land-cover products are developed at high resolution and with high precision for dedicated projects but that global applications are rare and only done for a specific time slice. It is therefore almost impossible to get a comprehensive quantitative overview of land-cover change.
Issues and priorities
; An international body should advise on standards for the production of land cover maps,
specifically in terms of the resolution and land-type characterization to be employed. ; There is an urgent need to develop a comprehensive time series of recent land cover
changes with a high spatial resolution and a decadal temporal resolution. Existing land
cover data should be analyzed and/or reprocessed, wherever possible, to ensure the
compatibility of maps produced for the last decade. New land cover maps should be
produced every 5 years.
; Significant improvements to historical land-use data sets could be achieved if more and
better-documented inventory data sets were made available. Improved algorithms linking
in-situ data and FAO agricultural statistics with satellite-based classifications are required. ; Improved algorithms have to be developed to compare/translate the different
characteristics of the generally discrete land-cover classifications. Improved algorithms
are needed for assigning land-surface properties to the discrete classes of the available
From GCOS “Systematic Observation Requirements for Satellite-Based Products for
Climate: Supplemental details to the satellite-based component of the „Implementation Plan
for the Global Observing System for Climate in Support of the UNFCCC‟”, 2006:
3.3.5 ECV Land Cover
Land cover influences climate by modifying water and energy exchanges with the atmosphere, and changes greenhouse gas and aerosol sources and sinks. Many climatically-relevant variables that are difficult to measure at a global scale (e.g. surface roughness) can be inferred in part from vegetation and land-surface types. Thus, land cover can be a surrogate for other important climate variables. Land-cover distributions are linked to the regional climate, so changes in cover can indicate climate change on a regional scale. Current climate models operate on resolutions of 0.5 degrees to 1 degree, but land-cover information at moderate resolution (250m – 1km) is needed to correctly describe the spatial heterogeneity of the land surface within the model cells.
Although moderate-resolution land cover can meet most science needs and detect intrinsically large-scale land-cover changes, it is not sufficient in resolution to deal with all land-cover changes, e.g., involving agriculture or forestry, deforestation in particular. High-resolution measurements of land cover are needed to quantify changes in the area covered by a specific land-cover type. For example, Brazil routinely maps the whole of legal Amazonia at high resolutions to quantify the rate of deforestation; the European Union uses data at these
resolutions to monitor agricultural production, urbanization, and other land change processes. High-resolution measurements provide a key intermediary level of information that is necessary to integrate in situ and global-scale land-cover observations.
High-resolution measurements of land cover change would provide significant support for national reporting under many chapters of the GHG inventories call for by the UNFCCC (especially those linked to agriculture, grasslands, wetlands, forestry). Including historical datasets of land cover, they would allow the establishment of neutral baselines for verification of carbon trading, e.g., pre-1990 for a/reforestation, or the degree of avoided deforestation. They would also contribute to the understanding of long-term land-cover change and its link to climate processes. High-resolution land-cover mapping could also be linked to other climate products, such as maps of lake area (T.1.1.), and glaciers (T.2.1).
Historical land-cover datasets could be generated on a decadal scale form the 1970s to 2000 using available Landsat-type data (high resolution), and by comparing recent 1 km-resolution global land cover datasets with sampled historical high-resolution imagery from the Landsat and SPOT archives to reconstruct land cover. Given limited comparability between existing land cover databases, some should be reprocessed to ensure the consistency of observations.
In close relation to land cover, land-surface temperature (LST) is one of the key subsidiary variables used to observe, identify, and understand land-surface climate processes at regional and global scales. Because of the strong heterogeneity in land-surface characteristics, such as vegetation, topography, and soil physical properties, LST changes rapidly in space as well as in time. An adequate characterization of LST distribution and its temporal evolution therefore requires measurements with detailed spatial and temporal frequencies, which can only be provided by satellite remote sensing.
Two products are required for this ECV:
; Moderate-resolution maps of land-cover type (T.5.1)
; High-resolution maps of land cover type, for the detection of land cover change (T.5.2)
From “The Committee on Earth Observing Satellites (CEOS) Response to the Global Climate Observing System (GCOS) Implementation Plan (IP)”:
Land Cover and Glaciers: ensure the continuityof terrestrial climate monitoring to
monitor land cover and glacier change and to improve the ECVs derived from them.
The current gap of Landsat-class data needs to be addressed using one of several
approaches currently being investigated by CEOS agencies. Research efforts need to
be directed toward the extraction of the ECVs from this fine-resolution record. Action T-1: CEOS agencies will determine which alternative approach best fills the current Landsat-class data gap and will explore the potential of integrating high-resolution data from multiple platforms (e.g., China-Brazil Earth Resources Satellite (CBERS), Indian Remote Sensing (IRS) satellite, Landsat, Satellite Pour l’Observation
de la Terre (SPOT), and others) based on the results of a CEOS study team led by the United States Geological Survey (USGS) that will establish, by 2007, the basis for a future Land-Surface Imaging Constellation.
Action T-2: CEOS agencies will assess the feasibility of generating global historic and continuing ECVs at fine resolutions for land cover and glacier change.
Snow, Albedo, Land Cover, fAPAR, LAI, and Fire: improve the continuity of terrestrial climate monitoring through enhancements of the moderate-resolution historical record.
AVHRR data reprocessing must be undertaken to ensure a consistent data set to fulfil
historical fAPAR, LAI, burned area, active fires, land cover, land surface temperature,
and snow are requirements. Reprocessed AVHRR data also potentially will contribute
to historical albedo.
Action T-3: CEOS (led by USGS and NOAA), in cooperation with relevant stakeholders, will explore the feasibility, by 2007, of retrieving and reprocessing the 1-km AVHRR data record from various centralized archives (NOAA and High Resolution Picture Transmission (HRPT) stations).
Action T-4: CEOS will work to enhance the quality of the FCDRs and the ECVs generated from the AVHRR record to meet threshold requirements.
Moderate-resolution maps of land cover type (T.5.1)
; There are adequate FCDRs both from the US and from Europe suitable for generation
of these products. For periods earlier than 2000, the AVHRR record is of value
especially if it is available at 1-km resolution. However, it has not yet been established
that this record is sufficient to retrospectively generate global land cover products for
the 1990s when 1-km data are available. Before 1990, only 8-km data are currently
available globally. Current products generated from moderate-resolution instruments
are adequate and do not require reprocessing. Use of AVHRR and SPOT-
VEGETATION (VGT) data for pre-2000 land cover classification has not yet been
established as being adequate. There is a need for the reprocessing of the AVHRR
data record and for the establishment of the AVHRR-ATSR/(A)ATSR-MODIS-VIIRS
; Global land cover products at 250- to 500-m resolution are currently being generated
in the US with MODIS Land (MODLAND) and in Europe with Global Land Cover
service (GLOBCOVER). There are several moderate-resolution (250-m to 1-km)
multispectral instruments with adequate FCDRs. Sensors include AVHRR, ATSR,
(A)ATSR, MODIS, VIIRS, and sensor on Sentinel-3.
Adequacy of provisions
; Significant processing of these observations will be needed to turn them into adequate
FCDRs for this purpose. There must be some revisiting of the 5-year reporting of
change, particularly in tropical areas, as within 5 years significant areas can be cleared
but will then regrow so that they may be difficult to distinguish from intact forest.
High-resolution maps of land cover type, for the detection of land cover change (T.5.2)
; For the period from 2000-03 the long-term acquisition plan of Landsat 7 ensured
adequate holdings for much of the world. Some areas with very high cloud amounts
were not monitored adequately. Since 2003, problems with the ETM+ of Landsat 7
have significantly reduced its capabilities.
; The TM and ETM+ of Landsat 5 and 7 provide adequate FCDRs. These sensors have
been very well characterized, as have those of the SPOT High Resolution Visible
(HRV). For sensors with similar spatial resolutions, much less thorough
characterization has been carried out. Near-global orthorectified Landsat coverage is
available for the mid-1970s, 1990, and 2000.
; Land cover change products have been created regionally and on a country basis, but
no comprehensive global land cover change products have been generated. Scaling up
the country and regional efforts is feasible, with the principal limitation being
resources for the ECV creation.
; The current collection of satellite imagery is not sufficient to meet the requirements
for this action. Scattered regional maps at 30-m resolution exist – e.g., Coordinated
Information on the European Environment (CORINE) for Europe, Africover for
Eastern Africa, Brazil’s Programa de Cálculo de Desflorestamento de Amazonia
(PRODES), and Canada’s Earth Observation for Sustainable Development of Forests
(EOSD) – but institutional arrangements to ensure operational generation of global
land cover maps at these resolutions are not yet in place.
Adequacy of provisions
; There are currently U.S. plans to create a global, mid-decadal data set based largely on
Landsat 5 data supplemented with composited Landsat 7 data. Securing this data set is
a very high priority for the carbon community. There is likely to be a continuing gap
of high-resolution records until early in the next decade and international cooperation
should be initiated to use existing assets to fill this continuing major deficit. ; 30-m spatial resolution represents the threshold, but for areas with more spatial
heterogeneity 10-m data should be the target. For many parts of the world a 5-year
frequency would reduce many uncertainties. But higher-frequency imaging and
product generation is needed regionally.
; There is probably an over-reliance on optical sensing for land cover monitoring. Some
areas are persistently cloudy; these areas must be delimited and the resulting
information used to determine an acquisition strategy for radars. This is an addition to
the role of InSAR measurements in estimating biomass. A long-term commitment for
this type of sensing is needed but currently absent.
Supporting Product to T.5.1, T.5.2: Land surface temperature, in conjunction with land cover type (T5.3)
; Sensors that may be used for land surface temperature include AVHRR (global data
sets based on AVHRR Ch4 and Ch5 radiometric data have been available since 1982)
and (A)ATSR. With the launch of ATSR onboards ERS-1 in July 1991, the first
sensor operating in biangular-mode was available. ATSR measures at nadir and at
forward view, less than 55 degrees.
; ATSR-2, onboard ERS, and (A)ATSR, onboard Envisat, are currently the only
observing systems able to provide quasi-simultaneous multispectral (from visible to
thermal infrared) measurements at two view angels (approximately 0 and 53 at
; Global LST maps with an accuracy of ?1 degree C and emissivity maps with an
accuracy of ?0.005 are available from (A)ATSR, MODIS, and ASTER data for many
; ASTR is currently being reprocessed to version 2. This will be ready near the end of
2007 and will be continuous data record from July 1991 to present.
; MODIS (since 2000) uses coefficients in the split window algorithm that are given by
interpolation on a set of multi-dimensional look-up tables.
Adequacy of provisions
; The ECV threshold may be reached for LST. The key issue is whether an acquisition
strategy will be implemented by space agencies to ensure that historical AVHRR
observations are collected.
; Continued, long-term measurements from VIIRS and other F/O missions are crucial to
reduce uncertainty and to model future changes.