Keywords

1 GLASS-GLC—Global Land Surface Satellite-Global Land Cover

Product

LULC general

Dates

1982–2015

Formats

Raster

Pixel size

5 km

Thematic resolution

8 classes:

0 (a), 1 (ag), 4 (v), 0 (m), 1 (na)Footnote 1

Compatible legends

FROM-GLC

Extent

Global

Updating

Not planned

Change detection

Possible, although sources of uncertainty may arise

Overall accuracy

Expected to be >82%

Website of reference

Website Language English

http://data.ess.tsinghua.edu.cn/

Download site

https://doi.pangaea.de/10.1594/PANGAEA.913496

Availability

Format(s)

Open Access

.tiff

Technical documentation

Liu et al. (2020)

Other references of interest

–

Project

GLASS-GLC is the result of the research activity on LUC mapping carried out by a group of Chinese researchers. It is part of the efforts led by the Tsinghua University to map global LUC information, which also includes the FROM-GLC project, reviewed later in this chapter.

The project has delivered a series of global LUC maps at coarse resolution (5 km). This spatial resolution may limit the applicability of the dataset as, for example, it does not include information on impervious areas.

Production method

GLASS-GLC was obtained after making a supervised classification of AVHRR satellite imagery with the Google Earth Engine cloud platform. Random forest was the selected classifier. Auxiliary data, such as a Vegetation Continuous Field layer or a Digital Elevation Model, were also used in the classification.

To ensure the consistency of the maps over time, the authors applied the “LandTrendr” method and a linear regression-based algorithm. These helped to detect the LUC changes in the imagery archive used to obtain the LUC maps.

Product description

GLASS-GLC can be downloaded as a single compressed file. This file includes all the LUC maps for each year in the map series (1982–2015), as well as auxiliary data to help users understand the product.

Downloads

GLASS-GLC

– A raster file with the LUC information for each available year (.tiff)

– Word document with a technical description of the product

Legend and codification

Code

Label

Code

Label

10

Cropland

70

Tundra

20

Forest

90

Barren land

30

Grass

100

Snow/ice

40

Shrubland

0

No data

2 LC-CCI—Land Cover-Climate Change Initiative

Product

LULC general

Dates

1992 – 2018

Formats

Raster

Pixel size

300 m (150 m for water bodies and 500 m for snow condition)

MMU Changes: 1 km

Thematic resolution

37 classes:

1 (a), 2 (ag), 26 (v), 4 (m), 1 (na)

Compatible legends

PFT, FAO LCCS

Extent

Global

Updating

Updated planned (no date)

Change detection

Yes

Overall accuracy

Expected to be >70%

Website of reference

Website Language English

https://www.esa-landcover-cci.org/

Download site

http://maps.elie.ucl.ac.be/CCI/viewer/download.php

Availability

Format(s)

Open Access after provision of name, institution and email

.tiff, .nc (NetCDF4)

Technical documentation

ESA (2017)

Other references of interest

Bontemps et al. (2012), Hollmann et al. (2013), Hua et al. (2018), Mousivand and Arsanjani (2019), Vilar et al. (2019)

Project

The Land Cover-Climate Change Initiative is a project run by the European Space Agency (ESA) that seeks to create LUC products that meet the requirements of the Global Climate Observing System (GCOS) for Essential Climate Variables (ECV) and the Climate Modelling Community (CMC). It builds on the lessons learned during the GlobCover project. It also takes into account the opinion and the needs of users working in the climate and global land cover research communities, who were consulted and engaged with during the project.

The purpose of the project is to deliver a time series of land cover data that is stable, dynamic, transparent and flexible. This means: first, obtaining a historical series of land cover maps that show the changes over time, with no technical errors or instability: second, the production of a LUC dataset with a wide range of applications; and third, the provision of all relevant information regarding the quality of the dataset.

The project was launched in 2009 and has been developed in different phases. The initial idea was to create a LUC product covering three time periods (1998–2002, 2003–2007 and 2008–2012). Later, an improved yearly LUC product for the period 1992–2015 was launched, which replaced the previous one. Recently, this latter product has been updated and now includes new LUC maps for the period 2016–2018 which are consistent with the previous series.

Apart from LUC maps, other interesting products have also been created as part of the Climate Change Initiative: weekly image composites of the AVHRR (1992–1999, 1 km), MERIS (2003–2012, 300 m and 1 km) and PROBA-V (2014–2015, 1 km) sensors; a static map of open water bodies; and three global land surface seasonality products characterizing the dynamics of vegetation greenness, snow and burnt areas.

Production method

The LC-CCI LUC map series is based on a single base LUC map that is progressively updated and backdated. The base LUC layer was created by classifying a series of composite MERIS imagery for the period 2003–2012. A different classification was carried out for each year of this period, and the map finally obtained was a combination of all these classifications. This allowed them to differentiate between land cover states (i.e. those land features that remain stable over time) and land cover seasonality (i.e. natural, seasonal variability of land cover features that do not imply a change in the cover itself).

The classification method combined the GlobCover unsupervised classification chain with a machine learning algorithm. During the classification process, a series of spectrotemporal classes were identified. These were later labelled to LUC classes with the help of experts. The classification was regionalized to account for regional diversity and local heterogeneity of land cover characteristics.

Change detection for updating and backdating the base map was carried out with imagery from different sensors (AVHRR, SPOT, MERIS and PROVA), according to image availability. Changes were detected at a spatial resolution of 1 km, and since 2013 have been delineated at 300 m. Previously, delineation of changes at finer spatial resolutions had been impossible due to the lack of available images.

As a general rule, the only changes studied were those between six wide categories, which are not semantically close to each other: agriculture, forest, grassland, wetland, settlement and others. These changes had to persist for at least two years to be considered. The purpose of these rules was to try to ensure the stability over time of the LUC map series, avoiding technical changes and noise.

Product description

The LC-CCI dataset is distributed in different ways. This gives users the flexibility to download the product that best suits their needs. A single LUC map in either GeoTIFF or NetCDF4 may be downloaded for each year of the period 1992–2015. For the most recent years (2016–2018), these are only available in NetCDF4 format. Additionally, the whole time series of maps for the period 1992–2015 can be downloaded as a single raster with multiple bands, in either of the two formats available.

When downloading the LUC maps, users only gain access to the rasters with LUC information. However, other supplementary information is available on the project’s website. This includes a CSV file with the legend description; layer style files for displaying the rasters in common GIS software (ArcGIS, ENVI and QGIS); GeoTIFF files with information about the quality and uncertainty of the LUC maps time series (Quality flags); and a data package for users working with the Sen2Cor classification software.

Downloads

LC Map 2015

– Raster file with LUC map

LC maps full 1992–2015 series

– Raster file with LUC maps series

Legend and codification

Code

Label

0

No data

10

Cropland, rainfed

11

Herbaceous cover

12

Tree or shrub cover

20

Cropland, irrigated or post-flooding

30

Mosaic cropland (>50%)/natural vegetation (tree, shrub, herbaceous cover) (<50%)

40

Mosaic natural vegetation (tree, shrub, herbaceous cover) (>50%)/cropland (<50%)

50

Tree cover, broadleaved, evergreen, closed to open (>15%)

60

Tree cover, broadleaved, deciduous, closed to open (>15%)

61

Tree cover, broadleaved, deciduous, closed (>40%)

62

Tree cover, broadleaved, deciduous, open (15–40%)

70

Tree cover, needleleaved, evergreen, closed to open (>15%)

71

Tree cover, needleleaved, evergreen, closed (>40%)

72

Tree cover, needleleaved, evergreen, open (15–40%)

80

Tree cover, needleleaved, deciduous, closed to open (>15%)

81

Tree cover, needleleaved, deciduous, closed (>40%

82

Tree cover, needleleaved, deciduous, open (15–40%)

90

Tree cover, mixed leaf type (broadleaved and needleleaved)

100

Mosaic tree and shrub (>50%)/herbaceous cover (<50%)

110

Mosaic herbaceous cover (>50%)/tree and shrub (<50%)

120

Shrubland

121

Evergreen shrubland

122

Deciduous shrubland

130

Grassland

140

Lichens and mosses

150

Sparse vegetation (tree, shrub, herbaceous cover) (<15%)

152

Sparse shrub (<15%)

153

Sparse herbaceous cover (<15%)

160

Tree cover, flooded, fresh or brackish water

170

Tree cover, flooded, saline water

180

Shrub or herbaceous cover, flooded, fresh/saline/brackish water

190

Urban areas

200

Bare areas

201

Consolidated bare areas

202

Unconsolidated bare areas

210

Water bodies

220

Permanent snow and ice

Practical considerations

The project is aimed at the climate change research community and therefore provides the LUC data in the NetCDF4 raster file format commonly used by this community. However, .nc files are much heavier than .tiff files.

LUC maps for single years are easily displayed in QGIS. However, raster files storing the whole series of LUC maps for the period 1992–2015 are very heavy and are difficult to display in QGIS without a computer with good processing power.

3 GLC30—GlobeLand30

Product

LULC general

Dates

2000, 2010, 2020

Formats

Raster

Pixel size

30 m

Variable UMC depending on the category (3 × 3 to 10 × 10 pixels)

Thematic resolution

10 classes:

1 (a), 1 (ag), 4 (v), 0 (m), 0 (na)

Compatible legends

GLC30

Extent

Global

Updating

Not planned

Change detection

Yes

Overall accuracy

Expected to be >78%

Website of reference

Website Language English

http://www.globallandcover.com/home_en.html

Download site

http://www.globallandcover.com/defaults_en.html?src=/Scripts/map/defaults/En/download_en.html&head=download&type=data

Availability

Format(s)

Open access under registration

.tiff

Technical documentation

Chen et al. (2010, 2011a, b, 2012, 2014, 2016), Tang et al. (2014), Xie et al. (2015), Zhu et al. (2010)

Other references of interest

Cao et al. (2014), Chen et al. (2013, 2017), Han et al. (2015), Jun et al. (2014), Manakos et al. (2018), Shi et al. (2016a, b), Wu et al. (2016), Yang et al. (2017)

Project

GlobeLand30 (GLC30) is a project funded and promoted by the Chinese government and the National Science Foundation of China. It aims to coherently map the land uses and covers on the world’s surface at a detailed scale, using images from the Landsat satellite imagery archive.

The project initially focused on analysing the best methods and procedures to carry out such an ambitious task. It then produced a global LUC map at 30 m for the reference years 2000 and 2010. An update of the dataset for the year 2020 was recently released, in which Antarctica was mapped for the first time.

Production method

GLC30 was obtained after classifying Landsat imagery using a pixel-object-knowledge-based (POK-based) classification approach. Other sources of complementary imagery were also used for the reference years 2010 (HJ-1—China Environment and Disaster Reduction Satellite) and 2020 (GF-1—China High Resolution Satellite).

The classification was carried out independently for each of the mapped categories. Water bodies were mapped first, followed by wetlands, snow and ice, artificial surfaces, cultivated land, forest, scrubland, grassland, barren land and finally tundra. Once a LUC category had been classified, the pixels assigned to that category were masked for the following classifications.

Each category was classified according to a specific approach, adapted to the characteristics of the features being mapped. For most of the categories, the classification approach consisted of three main steps: a pixel-based classifier, image segmentation and knowledge-based verification. For this last step, different sources of auxiliary information were used via their integration in a web-based data platform.

Product description

GLC30 is distributed in tiles. Users can separately download a LUC map for each tile and year of reference. The download includes the LUC map in raster format, a metadata file and a vector file with information about the satellite imagery used to obtain the map.

Downloads

GLC30 2020

– Raster file with LUC map (.tiff)

– Shapefile file with information about the imagery source used in the LUC classification (.shp)

– Metadata file (.xls)

Legend and codification

Code

Label

Code

Label

10

Cultivated land

60

Water bodies

20

Forest

70

Tundra

30

Grassland

80

Artificial surfaces

40

Shrubland

90

Bareland

50

Wetland

100

Permanent snow and ice

Practical considerations

The GLC30 LUC maps for 2000, 2010 and 2020 can also be accessed online through the project website,Footnote 2 which also includes plenty of information about the project and various other datasets. These include the 2020 imagery used to map the latest update of the dataset and different sources of reference data used as auxiliary information in the mapping process.

There are no technical documents describing the latest update of the map for the year 2020. Methodological changes in the production of the map could have been implemented which could lead to errors when comparing with previous editions.

The project website is not always maintained. It has been unattended for many months over recent years. If the website is not maintained, it is possible that the dataset may be not accessible in the future.

4 GLC250—Global Land Cover 250 m

Product

LULC general

Dates

2001, 2010

Formats

Raster

Pixel size

250 m

Thematic resolution

25 classes:

0 (a), 6 (ag), 7 (v), 1 (m), 0 (na)

Compatible legends

FAO-LCCS, IGBP

Extent

Global

Updating

Not expected

Change detection

Yes

Overall accuracy

Expected to be >75%

Website of reference

Website Language English

http://data.ess.tsinghua.edu.cn/

Download site

http://data.ess.tsinghua.edu.cn/

Availability

Format(s)

Open Access

.tiff

Technical documentation

Wang et al. (2015)

Other references of interest

–

Project

This product forms part of the project led by Tsinghua University to effectively map land uses and covers across the world, which mainly focused on FROM-GLC and the production of thematic LUC databases. Several of these datasets were used in the production of GLC250. The classification legend for GLC250 was also taken from FROM-GLC.

Production method

GLC250 was obtained after the classification of MODIS imagery (MOD13Q1) with a random forest classifier fed with auxiliary data: slope, latitude, MODIS vegetation indexes. For each year of reference (2001, 2010), a classification was carried out for three different dates: the year of reference, the year before and the year after. For the year 2001, for example, images from 2000, 2001 and 2002 were classified.

The three probability maps obtained after the classification carried out for each year of reference were processed through a spatial–temporal consistency model (MAP-MRF) to improve the LUC classification. The final LUC map was improved in a post-classification phase through a rule-based label adjustment method using auxiliary data from MODIS Vegetation Continuous Fields (MOD44B), slope and Enhanced Vegetation Index series.

Product description

A map for each year of reference can be downloaded in a single compressed file. Each file contains all the raster files that make up the LUC map for each year of reference. To this end, the global map is split into multiple tiles following the MODIS tile grid.Footnote 3

Downloads

GLC250—2010

– Raster files with a LUC map for each tile making up the MODIS tile grid (296 files)

Legend and codification

The GLC250 classification scheme is the same as that developed for FROM-GLC. It is a two-level classification scheme, which allows the LUC map to be displayed at two different levels of detail. Only the most detailed scheme (Level 2) is displayed here. Interested users can consult the correspondence between Level 2 and Level 1 classes on the project’s website.Footnote 4

Code

Label

Code

Label

11

Rice fields

42

Other shrublands

12

Greenhouse farming

61

Lake

13

Other croplands

62

Reservoir/pond

14

Seasonal croplands

63

River

15

Pastures

64

Ocean

21

Broadleaf forests

91

Dry salt flats

22

Needleleaf forests

92

Sandy areas

23

Mixed forests

93

Exposed bare rock

24

Orchards

94

Dry lake/river bottoms

31

Marshland

95

Tidal area

32

Herbaceous tundra

101

Snow

33

Other grasslands

102

Ice

41

Shrub and brush tundra

  

5 MCD12Q1—MODIS/Terra + Aqua Land Cover Type

Product

LULC general

Dates

2001 – 2020

Formats

Raster

Pixel size

500 m, 1 km, 0.05º

Thematic resolution

18 classes (IGBP legend):

1 (a), 1 (ag), 10 (v), 2 (m), 1 (na)

Compatible legends

IGBP, UMD, LAI, BGC, PFT, FAO-LCCS

Extent

Global

Updating

Expected

Change detection

Not recommended

Overall accuracy

Expected to be >71%

Website of reference

Website Language English

https://lpdaac.usgs.gov/products/mcd12q1v006/

Download site

https://lpdaac.usgs.gov/products/mcd12q1v006/

Availability

Format(s)

Open access under registration

.hdf

Technical documentation

Friedl et al. (2002, 2010), Friedl and Sulla-Menashe (2019), Sulla-Menashe et al (2019)

Other references of interest

Fritz and See (2005), Giri et al. (2005), Hao and Gen-Suo (2014), Tchuenté et al. (2011)

Project

MCD12Q1, also known as MODIS Land Cover type, dates back to 2002, after the launch into space of the TERRA satellite carrying the MODIS sensor. The MODIS sensor provided a new source of imagery for global LUC mapping. This led to the appearance of the MODIS Land Cover project, which aimed to produce a yearly series of LUC maps that could satisfy the demands of different communities interested in climate and environmental monitoring at global or very coarse scales. At the time the dataset was launched, only a few global LUC datasets were available, usually at coarser resolutions.

MODIS Land Cover was created by a team led by the University of Boston. Since 2002, six versions of the product have been developed. The latest is MODIS Land Cover Collection 6, which has included the most important changes in the production method of the dataset since its early developments.

A complementary product at coarser resolution has been developed as part of the same project: MCD12C1 (0.05 Deg).

Production method

MCD12Q1 was obtained by means of supervised classification (Random Forests) of MODIS imagery for the period 2001–2020. Once the classification had been obtained for each year, it was adjusted with the aid of auxiliary data: C5 MCD12Q1, C6 MODIS Land Water mask, C5 MODIS Vegetation Continuous Fields (VCF), WorldClim dataset, a global urban layer and global crop type information compiled from census data.

As a result of the classification, class probability rasters were obtained for each LUC category. These inform about the probability of each pixel belonging to a specific LUC category. These probability layers provided a base on which to map LUC covers according to six different classification schemes: IGBP, UMD, LAI, BGC, PFT and FAO-LCCS. In order to ensure the consistency of the classification over time, a hidden Markov model (HMM) was applied to the adjusted classification to reduce spurious changes over time.

Product description

MCD12Q1 may be downloaded through different servers or tools: AppEEARS, Data Pool, NASA Earthdata Search, USGS EarthExplorer, OPeNDAP, DAAC2Disk Utility and LDOPE. Depending on the server or tool chosen, users can download the product as a single file for each year of reference or in tiles for specific areas of interest.

The download includes the raster file with LUC data in six different classification schemes and PDF documents with the technical specifications for the product.

Downloads

MCD12Q1 (500 m)

– Raster file with multiple bands, including LUC information in five different classification schemes and data quality (.hdf)

– PDFs with technical information about the product

Legend and codification

MCD12Q1 is distributed for six different, widely used classification schemes. The only one displayed here is the IGBP scheme, which is one of the most commonly used. However, more information about the codes and class descriptions for the other classification legends is available in the user guide for this dataset (Sulla-Menashe et al. 2019).

MCD12Q1—IGBP (International Geosphere-Biosphere Programme)

Code

Label

Code

Label

1

Evergreen needleleaf forests

10

Grasslands

2

Evergreen broadleaf forests

11

Permanent wetlands

3

Deciduous needleleaf forests

12

Croplands

4

Deciduous broadleaf forests

13

Urban and built-up lands

5

Mixed forests

14

Cropland/natural vegetation mosaics

6

Closed shrublands

15

Permanent snow and ice

7

Open shrublands

16

Barren

8

Woody savannas

17

Water bodies

9

Savannas

18

Unclassified

Practical considerations

Users can consult the dataset online through the Web Map Service (WMS) available here.Footnote 5 The dataset is also available at a spatial resolution of 0.05 : MCD12C1 (0.05 Deg).Footnote 6

This dataset is not recommended for the study of LUC change, because of the high technical variability in LUC covers from one year to the next.

6 GLCNMO—Global Land Cover by National Mapping Organization

Product

LULC general

Dates

2003, 2008, 2013

Formats

Raster

Pixel size

1 km (2003) 500 m (2008, 2013)

Thematic resolution

20 classes:

1 (a), 3 (ag), 11 (v), 2 (m), 0 (na)

Compatible legends

FAO LCCS

Extent

Global

Updating

Not planned

Change detection

No

Overall accuracy

Expected to be >75%

Website of reference

Website Language English

https://globalmaps.github.io/glcnmo.html

Download site

https://globalmaps.github.io/glcnmo.html

Availability

Format(s)

Open Access

.tiff

Technical documentation

Kobayashi et al. (2017), Tateishi et al. (2011, 2014)

Other references of interest

Hua et al. (2018)

Project

GLCNMO is a project promoted by the International Steering Committee for Global Mapping (ISCGM) in collaboration with the Geospatial Information Authority of Japan (GSI), Chiba University and national mapping organizations from different participant countries. It is part of a wider effort to create global datasets on different subjects, including land cover and land use.

The project has delivered three global LUC maps. Each one was produced at a different time and various methodological changes were introduced between the production of each map. The most evident one was the change in spatial resolution after the 2003 map. Another important difference was the number of countries taking part in each edition of the map: 40 countries took part in the production of the 2003 map, 14 in the 2008 map and 22 in the map for 2013.

The ISCGM was wound up in 2016 and its data was transferred to the Geospatial Information Section in the United Nations. We, therefore, do not expect any updates on this project.

Production method

The three LUC maps were produced at the continental level using a mixture of different methods. The maps for each continent were prepared by separate groups, with national experts providing assistance for each case.

Most of the categories (14 in 2003 and 2008 and 11 in 2013) were obtained through supervised classification of MODIS imagery. The training samples for the classifier were selected with great care using photointerpretation from sources like Google Earth and other auxiliary data. Different classifiers were used for the different maps. Whereas the map for 2003 was produced using a maximum likelihood classifier, the ones for 2008 and 2013 were based on a decision tree classifier.

The remaining categories that were not classified using the method described above were individually mapped according to different procedures adapted to the specific needs of each category. These were urban, tree open, mangrove, wetland, snow/ice and water in 2003 and 2008. In addition to those, herbaceous areas, forests and agricultural areas were also mapped in this way in 2013. The strategies used to map these categories also varied in the different editions of the map, mainly involving specific classification methods of MODIS imagery, as well as the use of additional information, such as population density datasets, thematic MODIS products and other global LUC maps.

Product description

The GLCNMO LUC map is distributed individually for each available year. The map for each year is split into four tiles, which can be downloaded in different zipped files. No other additional information is provided, except for the scientific papers presenting each map.

Downloads

GLCNMO (version 3)

– Raster LUC map covering North America, the north of South America and the west of Europe and Africa (1_1)

– Raster LUC map covering Europe, the east of Africa and Asia (1_2)

– Raster LUC map covering South America and the west of the Antarctic (2_1)

– Raster LUC map covering Africa, the south of Asia and Oceania (2_2)

Legend and codification

Code

Label

Code

Label

1

Broadleaf evergreen forest

11

Cropland

2

Broadleaf deciduous forest

12

Paddy field

3

Needleleaf evergreen forest

13

Cropland/other vegetation mosaic

4

Needleleaf deciduous forest

14

Mangrove

5

Mixed forest

15

Wetland

6

Tree open

16

Bare area, consolidated (gravel, rock)

7

Shrub

17

Bare area, unconsolidated (sand)

8

Herbaceous

18

Urban

9

Herbaceous with sparse tree/shrub

19

Snow/ice

10

Sparse vegetation

20

Water bodies

Practical considerations

As there are no auxiliary datasets or documentation, users who require more detailed information about the characteristics of the dataset should consult the scientific papers cited above (14.6 Technical Documentation).

7 GlobCover

Product

LULC general

Dates

2005, 2009

Formats

Raster

Pixel size

300 m

Thematic resolution

23 classes:

1 (a), 2 (ag), 14 (v), 4 (m), 1 (na)

Compatible legends

FAO LCCS

Extent

Global

Updating

Not planned

Change detection

Not recommended

Overall accuracy

Expected to be >78.0%

Website of reference

Website Language English

http://due.esrin.esa.int/page_globcover.php

Download site

http://due.esrin.esa.int/page_globcover.php

Availability

Format(s)

Open Access

.tiff

Technical documentation

Bicheron et al. (2008), Bontemps et al. (2011)

Other references of interest

Defourny et al. (2010)

Project

GlobCover is a project run by the European Space Agency (ESA) in collaboration with the Joint Research Centre (JRC) of the European Commission, the European Environment Agency, the FAO, the UN Environment Programme (UNEP), the Global Observations of Forest Cover Land-use Dynamics (GOFC–GOLD) programme and the International Geosphere-Biosphere Programme (IGBP). It started in 2005 and produced two global LUC maps for the reference years 2005 and 2009. The Université Catholique de Louvain (UCL) also contributed to the 2009 edition of the map.

The aim of the project was to develop global LUC maps using images from the MERIS sensor onboard the ENVISAT satellite. At the time it was launched, the 2005 GlobCover map was the first global LUC map at a spatial resolution of 300 m.

Based on the results of GlobCover, the ESA launched a new project called GlobCorine in which two new LUC maps compatible with the Corine Land Cover classification legend were created for Europe from the same imagery. The LC-CCI project from the ESA (see Sect. 2) builds on the progress made and the lessons learnt during the GlobCover project.

Production method

GlobCover maps were obtained by classifying imagery captured by the MERIS sensor. Urban and wetland areas, which are not well represented, were classified using a supervised classifier. The remaining categories were classified in a series of spectro-temporal classes through an unsupervised classifier. Once classified, the spectro-temporal classes were labelled automatically according to the information provided by the reference datasets. For the 2005 map, the reference datasets were the GLC2000 global LUC map (see Sect. 3 in Chap. “Global General Land Use Cover Datasets with a Single Date” Global General Land Use Cover Datasets with a Single Date) and other high-quality regional LUC maps. For the 2009 map, the GlobCover 2005 map was used as a reference.

The area for classification was divided into different regions, to account for the ecological and reflectance diversity of the world. Once labelled after classification, the LUC map was finally edited to account for inaccuracies in the representation of certain features.

For the 2005 version, regional maps with a more detailed legend were also produced following the same classification procedure.

Product description

A zipped file is available for each GlobCover map. It contains the raster layer with the LUC information and all the auxiliary data that users may need to correctly interpret the dataset. This includes the classification legend, technical and data quality information, and files with the layer style of the map to automatically symbolize the raster in GIS software. A complementary raster detailing the source of the LUC information for each pixel (MERIS sensor classification (value = null) or a land cover database (value = 1)) is also provided. In a separate file, users can also download a raster for a coloured version of the LUC map.

Downloads

GlobCover

– Raster file with LUC map (“GLOBCOVER_L4_200901_200912_V2.3”)

– Raster file with quality information (“GLOBCOVER_L4_200901_200912_V2.3_CLA_QL”)

– Preview image of the product

– Excel sheet with the map legend (“Globcover2009_Legend”)

– Layer style files for ArcGIS (.lyr) and ENVI (.dsr)

– PDFs with technical information about the product

GlobCover coloured

– Raster file with coloured version of LUC map

Legend and codification

Code

Label

11

Post-flooding or irrigated croplands (or aquatic)

14

Rainfed croplands

20

Mosaic cropland (50–70%)/vegetation (grassland/shrubland/forest) (20–50%)

30

Mosaic vegetation (grassland/shrubland/forest) (50–70%)/cropland (20–50%)

40

Closed to open (>15%) broadleaved evergreen or semi-deciduous forest (>5 m)

50

Closed (>40%) broadleaved deciduous forest (>5 m)

60

Open (15–40%) broadleaved deciduous forest/woodland (>5 m)

70

Closed (>40%) needleleaved evergreen forest (>5 m)

90

Open (15–40%) needleleaved deciduous or evergreen forest (>5 m)

100

Closed to open (>15%) mixed broadleaved and needleleaved forest (>5 m)

110

Mosaic forest or shrubland (50–70%)/grassland (20–50%)

120

Mosaic grassland (50–70%)/forest or shrubland (20–50%)

130

Closed to open (>15%) (broadleaved or needleleaved, evergreen or deciduous) shrubland (<5 m)

140

Closed to open (>15%) herbaceous vegetation (grassland, savannas or lichens/mosses)

150

Sparse (<15%) vegetation

160

Closed to open (>15%) broadleaved forest regularly flooded (semi-permanently or temporarily)—fresh or brackish water

170

Closed (>40%) broadleaved forest or shrubland permanently flooded—saline or brackish water

180

Closed to open (>15%) grassland or woody vegetation on regularly flooded or waterlogged soil—fresh, brackish or saline water

190

Artificial surfaces and associated areas (urban areas >50%)

200

Bare areas

210

Water bodies

220

Permanent snow and ice

230

No data (burnt areas, clouds…)

Practical considerations

Eleven regional maps with more detailed classification schemes were developed as part of the GlobCover Project for 2005. These maps were produced using the same methodology as the global GlobCover, but provided more thematic detail. Unfortunately, they are currently unavailable for download.

8 FROM-GLC—Finer Resolution Observation and Monitoring of Global Land Cover

Product

LULC general

Dates

2010, 2015, 2017

Formats

Raster

Pixel size

250 m, 500 m, 1 km, 5 km, 25 km, 50 km, 100 km (2010)

30 m (2010, 2015, 2017)

10 m (2017)

Thematic resolution

8 classes (2017):

1 (a), 1 (ag), 3 (v), 0 (m), 0 (na)

Compatible legends

–

Extent

Global

Updating

Not planned

Change detection

Not recommended

Overall accuracy

Expected to be >65%

Website of reference

Website Language English

http://data.ess.tsinghua.edu.cn/

Download site

http://data.ess.tsinghua.edu.cn/

Availability

Format(s)

Open Access

.tiff

Technical documentation

Chen et al. (2019), Gong et al. (2013), Yu et al. (2013, 2014)

Other references of interest

Ji et al. (2015), Xu et al. (2019)

Project

FROM-GLC was a project funded by Chinese research and innovation programmes that was led by Tsinghua University. It brought together researchers from Chinese and other international institutions.

The goal of this project was to produce global LUC datasets at medium to high spatial resolution. When the project started, there were no global LUC maps available at a resolution of 30 m using images from the Landsat archive. Maps at that resolution are useful for different user communities working in cross-regional and cross-national areas at that level of detail. The aim of FROM-GLC was therefore to provide new sources of data for modelling communities that required detailed global datasets. Global LUC maps at detailed scales are also useful for countries for which no other detailed LUC datasets are available.

Three global LUC maps at three different time points (2010, 2015 and 2017) were created as part of this project. Three LUC maps are available for the year 2010. The original (FROM-GLC) was successively improved by changes in the production method, producing maps known as FROM-GLC-egg and FROM-GLC-agg, the latter being the final, most updated version. It is available at the original (30 m) and 7 other spatial resolutions: 250 m, 500 m, 1 km, 5 km, 25 km, 50 km and 100 km. Unlike the maps for 2010 and 2015, the one for 2017 was produced at two spatial resolutions: 10 and 30 m.

The research team involved in the production of FROM-GLC has also taken part in related projects to produce other national, regional and thematic LUC maps, most of them at fine spatial resolutions. These maps can be accessed through the project website and include national maps of China or Chile, thematic maps about water covers and other global LUC datasets.

Production method

Each FROM-GLC map was produced using a different method. The maps for 2010, 2015 and 2017 at 30 m were produced using a supervised classification of Landsat imagery.

Four different classifiers were compared in the production of FROM-GLC for 2010. The first improved version of FROM-GLC, known as FROM-GLC-egg, included an image-segmentation method in the classification process and used two different classifiers. In addition, impervious surfaces were individually mapped. For its part, FROM-GLC-agg was obtained by combining the previous two LUC maps (FROM-GLC, FROM-GLC-egg) using a decision tree algorithm. Impervious surfaces were remapped according to the information provided by the Nighttime Light Impervious Surface Area (NL-ISA) and the MODIS urban extent (MODIS-urban) datasets. Once the FROM-GLC-agg map had been obtained at 30 m, it was then aggregated at seven other spatial resolutions through majority aggregation and proportion aggregation approaches.

The map for 2017 at 10 m was obtained through a supervised classification of Sentinel-2 imagery with a random forest classifier in the Google Earth Engine.

Product description

The FROM-GLC LUC maps are not provided as a single global file. To facilitate downloading of the product, the world is split into different tiles. Users can download the tile corresponding to their area of interest according to its latitude and longitude values.

FROM-GLC products for the year 2010 can also be downloaded through an assisted kmz layer. When uploading it in Google Earth, users can visualize their area of interest and automatically download the map corresponding to that area.

Downloads

FROM-GLC-agg (2010)

– Raster file with LUC map

FROM-GLC-agg hierarchy (2010)

– Raster file with LUC map at 30 m

– Raster file with LUC map at 250 m obtained by majority aggregation

– Raster file with LUC map at 500 m obtained by majority aggregation

– Raster file with LUC map at 1 km obtained by majority aggregation

– Raster file with LUC map at 5 km obtained by majority aggregation

– Raster file with LUC map at 5 km obtained by proportion aggregation

– Raster file with LUC map at 10 km obtained by majority aggregation

– Raster file with LUC map at 10 km obtained by proportion aggregation

– Raster file with LUC map at 25 km obtained by majority aggregation

– Raster file with LUC map at 25 km obtained by proportion aggregation

– Raster file with LUC map at 50 km obtained by majority aggregation

– Raster file with LUC map at 50 km obtained by proportion aggregation

– Raster file with LUC map at 100 km obtained by majority aggregation

– Raster file with LUC map at 100 km obtained by proportion aggregation

FROM-GLC (2015)

– Raster file with LUC map

FROM-GLC 30 m (2017)

– Raster file with LUC map

FROM-GLC 10 m (2017)

– Raster file with LUC map

Legend and codification

A specific two-level classification scheme legend was initially developed for the FROM-GLC project in 2010. This was updated with various changes for the FROM-GLC map for 2015. The map for 2017 has the simplest, least detailed classification legend (Level 1). In each case, we include the most detailed classification scheme available for each year. Users can consult the correspondence between level 2 and level 1 of the classification scheme for the years 2010 and 2015 at the project website.Footnote 7

FROM-GLC (2010)

Code

Label

Code

Label

11

Rice

62

Pond

12

Greenhouse

63

River

13

Other

64

Sea

39

Crop in urban

69

Water in urban

21

Broadleaf

71

Shrub

22

Needleleaf

72

Grass

23

Mixed

81

High albedo

24

Orchard

82

Low albedo

29

Forest in urban

91

Saline-Alkali

31

Managed

92

Sand

32

Nature

93

Gravel

39

Grass in urban

94

Bare Cropland

40

Shrubland

95

Dry river/lake bed

49

Shrub in urban

96

Other

51

Grass

99

Bareland in urban

52

Silt

101

Snow

59

Wetland in urban

102

Ice

61

Lake

120

Cloud

  1. * Categories only available in the FROM-GLC-Hierarchy product are shown in italics

FROM-GLC (2015)

Code

Label

Code

Label

11

Rice paddy

41

Shrubland, leaf-on

12

Greenhouse

42

Shrubland, leaf-off

13

Other

51

Marshland

14

Orchard

52

Mudflat

15

Bare farmland

53

Marshland, leaf-off

21

Broadleaf, leaf-on

60

Water

22

Broadleaf, leaf-off

71

Shrub and brush tundra

23

Needleleaf, leaf-on

72

Herbaceous tundra

24

Needleleaf, leaf-off

80

Impervious surface

25

Mixed leaf, leaf-on

90

Bareland

26

Mixed leaf, leaf-off

92

Bareland

31

Pasture

101

Snow

32

Natural grassland

102

Ice

33

Grassland, leaf-off

120

Cloud

FROM-GLC (2017)

Code

Label

Code

Label

1

Cropland

6

Water

2

Forest

8

Impervious

3

Grass

9

Bareland

4

Shrubland

10

Snow/ice

Practical considerations

The project website, where all the information is stored and available for download, is not user-friendly. It is not easy to find the information the user is looking for. Users may also struggle to download datasets for their area of interest according to latitude and longitude information. When available, we recommend using the kmz file with Google Earth for this purpose.

There is little additional information. For a complete description of the characteristics of the different maps, we recommend users to read the scientific papers cited in the introduction to this dataset above (14.8. Technical Documentation).

9 CGLS-LC100—Copernicus Global Land Service Dynamic Land Cover Map

Product

LULC general

Dates

2015 – 2019

Formats

Raster

Pixel size

100 m

Thematic resolution

24 classes:

1 (a), 1 (ag), 18 (v), 2 (m), 1 (na)

Compatible legends

FAO LCCS

Extent

Global

Updating

Yes, every year

Change detection

Possible, although sources of uncertainty may arise

Overall accuracy

Expected to be >80%

Website of reference

Website Language English

https://land.copernicus.eu/global/products/lc

Download site

https://lcviewer.vito.be/download

Availability

Format(s)

Open Access

.tiff

Technical documentation

Buchhorn et al. (2020a, b, c), Tsendbazar et al. (2019, 2020)

Other references of interest

–

Project

CGLS-LC100 is one of the deliverables produced as part of the Copernicus Global Land Service (CGLS), which aims to provide a series of bio-geophysical products to monitor land surface at a global scale. In addition to this LUC package, the programme produces other relevant variables, such as the Leaf Area Index (LAI), the Fraction of Absorbed Photosynthetically Active Radiation (FAPAR), the Land Surface Temperature, soil moisture and other vegetation indices.

The first version of CGLS-LC100 was released in 2017, mapping LUC for Africa. Since then, several updates of the product have improved the production methodology and extended its temporal and geographical coverage. The last version of the product (Collection 3), released in 2021, covers the whole world for the period 2015–2019. It includes a method for detecting land cover change that addresses the main sources of technical uncertainty when studying change in a time series of LUC maps.

In addition to the LUC map described here, the product also includes a series of continuous field layers or “fraction maps” for the basic LUC classes mapped. Future updates of the product are expected on an annual basis, using the imagery provided by the Sentinel satellite missions.

Production method

The Copernicus Global Land Service Dynamic Land Cover map is produced through a multistep processing framework. First, PROBA-V satellite images are pre-processed and merged following a Sentinel-2 tiling grid to create a 3-year epoch mosaic for each reference year. Second, a series of metrics (spectral and textural metrics, descriptive statistics) are extracted from each epoch mosaic. Third, imagery for all the epochs is classified using a regression algorithm, which delivers a cover fraction layer for each basic LUC class and reference year, and a supervised classification algorithm, which delivers a LUC map for each reference year.

Various auxiliary data sources are used in the classification phase, i.e. seven different data masks and three extra datasets: biome clusters, water cover fractions and built-up cover fractions.

In order to ensure the temporal consistency of the LUC map series, it was decided to include a temporal post-processing phase in the production of the dataset. This consists of a BFAST break detection algorithm and a Hidden Markov Model. The former is used to detect changes in an independent time series of MODIS NIRv imagery, while the latter is used to rule out technical changes in the classified epoch images.

Product description

CGLS-LC100 is distributed in tiles, following the Sentinel-2 tiling grid (110 × 110 km). For each tile, users can download many different layers: the discrete classification containing the LUC map for the selected area; a layer with the classification probability; layers of cover fractions for each of the basic LUC classes mapped; a layer showing the level of confidence for the change measured between the different years in each pixel; and two extra layers: forest types and input data density.

The download of the LUC map only includes the raster file with the LUC information. Each reference year must be downloaded separately.

Downloads

Land Cover classification—discrete classification

– Raster file with LUC map

Cover fractions—bare and sparse vegetation

– Raster file with the cover fraction for the land cover under consideration

Land Cover changes—change confidence

– Raster file indicating the reliability of the change in the discrete class

Others—forest types

– Raster file indicating for all pixels with a cover fraction >1% the type of forest represented in the pixel

Legend and codification

Land Cover classification–discrete classification

Code

Label

Code

Label

0

No input data

113

Closed forest, deciduous needle leaf

20

Shrubs

114

Closed forest, deciduous broad leaf

30

Herbaceous vegetation

115

Closed forest, mixed

40

Cultivated and managed vegetation/agriculture (cropland)

116

Closed forest, unknown

50

Urban/built up

121

Open forest, evergreen needle leaf

60

Bare/sparse vegetation

122

Open forest, evergreen broad leaf

70

Snow and ice

123

Open forest, deciduous needle leaf

80

Permanent water bodies

124

Open forest, deciduous broad leaf

90

Herbaceous wetland

125

Open forest, mixed

100

Moss and lichen

126

Open forest, unknown

111

Closed forest, evergreen needle leaf

200

Open sea

112

Closed forest, evergreen, broad leaf

113

Closed forest, deciduous needle leaf

Cover fractions—bare and sparse vegetation

Code

Meaning

0–100

Percentage of the pixel (0–100%) covered by the land cover under consideration

200

Masked sea

Land Cover changes—change confidence

Code

Change confidence

Code

Change confidence

0

No change

2

Medium confidence

1

Potential change

3

High confidence

Others—forest types

Code

Forest type

Code

Forest type

0

Unknown

3

Deciduous, needle leaf forest (DNF)

1

Evergreen, needle leaf forest (ENF)

4

Deciduous, broad leaf forest (DBF)

2

Evergreen, broad leaf forest (EBF)

5

Mixed

Practical considerations

Because of the large number of datasets available through this project, users are encouraged to make use of the different layers of LUC information available. This will give them a better understanding of the uncertainties and limitations of the product.

Users can download the product covering the whole globe, which is distributed through the files in the Zenodo repository.Footnote 8