Land-use and land-cover change (LULCC) impacts local energy and
water balance and contributes on global scale to a net carbon
emission to the atmosphere. The newly released annual ESA CCI (climate change initiative) land
cover maps provide continuous land cover changes at 300
LULCC (Land-use and land-cover change) is the essential human
perturbation on natural ecosystems (Klein Goldewijk et al., 2016) and
one of the main drivers of climate change (Alkama and Cescatti, 2016;
Bonan, 2008) through biophysical (e.g. albedo and transpiration
change; Peng et al., 2014; Zhao and Jackson, 2014) and biogeochemical
effects (e.g. carbon emissions from gross deforestation and carbon
sinks in secondary forest regrowth; Houghton and Nassikas,
2017). Forest loss from 2003 to 2012 was found to have caused a local
increase in air temperature of about 1
Accurate, well defined and spatially explicit gridded LULCC data are
a prerequisite for calculating
The newly released annual ESA CCI land cover maps from 1992 to 2015
partly overcome these challenges with 300
The annual ESA CCI LC maps cover a period of 24 years from 1992 to
2015 at a spatial resolution of 300
This unique long-term land cover time series was achieved by combining
the global daily surface reflectance of five different observation
systems while aiming to maintain a good consistency over time. This
was identified as a key requirement from the modeling community
(Bontemps et al., 2012). Each of these global daily measurements of
multispectral radiance recorded from 1992 to 2015 have been
pre-processed to complete radiometric calibration, geometric and
atmospheric correction, and clouds and cloud shadow
screening. The full archive of MERIS (2003–2012) providing
15 spectral bands at 300
The accuracy of ESA CCI LC products was evaluated on a global scale according to international standards, using an independent validation dataset to produce a confusion matrix and derive overall an accuracy figure. An object-based validation database of 2600 primary sampling units was built by a panel of international experts to specifically assess the accuracy of both the LC classes and changes (ESA, 2017). Research is currently ongoing to find how to address the new challenges underlying this database, i.e. following a per-object approach and interpreting not just a unique land cover class but also a distribution of land cover classes within a primary sampling unit. The uniqueness of these two concepts in the framework of global land cover validation requires more time to derive reliable figures about LC classes and LC change accuracy. It will also prevent any comparison with previous validation figures.
In this respect, for the sake of comparison, the accuracy of the ESA CCI LC product from 2010 was assessed using the GlobCover 2009 validation database (Bontemps et al., 2011). Using all the points interpreted as “certain” by the experts, whether “homogeneous” (i.e. made of a single LC class) or “heterogeneous” (i.e. made of several or mosaic LC classes), the overall accuracy was found to be 71.5 %. Accounting for only the “homogeneous” and “certain” points, the overall accuracy raised to 75.4 % (ESA, 2017). The highest user accuracy values were found for the classes of rainfed cropland, irrigated cropland, broadleaved evergreen forest, urban areas, bare areas, water bodies and permanent snow and ice. Conversely, mosaic classes of natural vegetation were associated with the lowest user accuracy values, as well as the three classes of lichens and mosses, sparse vegetation and flooded forest with fresh water.
The overall accuracy of the ESA CCI LC products was also assessed by independent studies over specific regions (e.g. Tsendbazar et al., 2015, over Africa and Yang et al., 2017, over China), which can give valuable insights for specific applications.
The original 37 ESA CCI LC classes were first aggregated into
Description and comparison of different land-use/land-cover datasets used in this study. The term of “dataset” in this study can also involve some model output (e.g. forest area from Land Use Harmonization dataset, LUH2v2h).
Gross changes need to be considered differently because it is only
possible to derive the net change by comparing the annual
300
In order to derive the gross transitions, all possible transitions
(506 in total) between the 23 original LC classes with gross changes
were calculated at 300
Three land-use and land-cover datasets (Table 1) were used for comparison: (i) forest, grassland and cropland area from the Land Use Harmonization (LUH2v2h) dataset (Hurtt et al., 2011); (ii) forest cover data from Hansen et al. (2013) and (iii) national forest area data from Houghton and Nassikas (2017). The cropland and pasture areas in the LUH2v2h dataset are from HYDE 3.2 (Klein Goldewijk et al., 2016), in which the ESA CCI epoch LC map in 2010 (representing 2008–2012) was used as a spatial reference map for the area allocation and the national cropland and grazing land were adjusted to match the FAO STAT data (FAOSTAT, 2015) as close as possible. The national forest areas from Houghton and Nassikas (2017) are based on FAO Forest Resources Assessment (FRA) data (FAO, 2015; also see Keenan et al. ,2015, for the main findings of FAO FRA 2015). Thus, these two additional sources of data, HYDE 3.2 (Klein Goldewijk et al., 2016) and FAO FRA (FAO, 2015) were not shown in the figures.
It should be noted that land use data are not necessarily the same as land cover data, and the exact definitions and categorization of forest (cropland and grassland) are different for each dataset (see details in Discussion). Nevertheless, these represent the best datasets available for the use in LSMs for comparison, and we have tried to harmonize the definitions where possible (see below), but to some degree this is an ongoing discussion between the modeling and data communities. Furthermore, all the LSMs have to use these datasets for deriving PFT changes back through time, so it is a very worthwhile exercise to determine if the broad groupings differ, and to what extent.
Absolute areas, net changes and gross transitions from 1992 to 2015 in the LUH2v2h dataset (Hurtt et al., 2011) were used for comparison. Forest used in this study from LUH2v2h (Hurtt et al., 2011) refers to the total of primary and secondary forest, cropland refers to all crop types and grassland refers to the total of pasture and rangeland. Because LUH2v2h data use cropland and grazing land areas from HYDE 3.2 as an input (Hurtt et al., 2011), the spatial distributions are mainly determined by HYDE 3.2. The gross transitions in LUH2v2h data are calculated from the Global Land Use Model (Hurtt et al., 2006) that tracks sub-grid cell loss and gain in land use categories. They first determined the urban area in each grid cell proportionally from cropland, pasture and secondary lands, and if these areas cannot fulfill the urban increase, primary lands were cleared. The minimum transition rates between cropland, pasture and other (sum of primary and secondary) lands were then calculated to identify the gross transitions between these land use categories (Hurtt et al., 2011). Transitions related to shifting cultivation and wood harvest were determined last (Hurtt et al., 2011).
Only annual gross forest loss each year during 2000–2014 and total gross forest gain during 2000–2012 are available in the dataset of Hansen et al. (2013). Thus, the net forest area change from this dataset only refers to the period of 2000–2012. The national forest area data from 1992 to 2015 in the dataset of Houghton and Nassikas (2017) were used to calculate the forest area changes.
A land mask with nine regions (Fig. 1) defined by Houghton (1999) was used to derive the regional values.
Global and regional areas of forest, cropland and grassland PFTs in the year 2000 in comparison with data from LUH2v2h (Hurtt et al., 2011), Hansen et al. (2013) and Houghton and Nassikas (2017). Different colours indicate different PFTs.
After translating the original ESA CCI LC classes into PFTs using the
cross-walking table (Table S1), the global and regional areas of forest,
cropland and grassland PFTs in the year 2000 are shown in Fig. 1. Global
areas of forest (excluding shrub), cropland and grassland PFTs are 30.4, 19.2
and 35.7 million
Forest area from ESA CCI is lower than that from Hansen et al. (2013) in all
regions except North Africa and Middle East and the Pacific developed region.
Here, the regions refer to the defined regions in Fig. 1.
Forest area from LUH2v2h (Hurtt et al., 2011) is larger than
that from ESA CCI in most regions except in South and Central America,
tropical African and the Pacific developed region. Forest area from Houghton
and Nassikas (2017), however, is systematically higher than that from ESA CCI
in all regions. Cropland area from ESA CCI matches that from LUH2v2h (Hurtt
et al., 2011) in North America but is higher in all the other regions.
Although the global grassland area is similar between ESA CCI and LUH2v2h
(Hurtt et al., 2011), larger differences are seen on a regional scale.
Grassland area from ESA CCI was found to be much higher than that from
LUH2v2h (Hurtt et al., 2011) in North America and the former Soviet Union
(4.0 and 3.5 million
After translating all of the 422 gross transitions detected between the
original ESA LC classes into PFTs, the time series of gross changes in PFTs
are shown in Fig. 2, and the mean annual change rates are shown in Table S2.
Generally, the gross changes are related to the net, i.e. where there are
more gross changes, more net changes can be found. Major gross changes occur
in forest, cropland and grassland PFTs, with a global gross gain of 0.91, 1.2
and 1.1 million
Gross changes in PFTs from 1992 to 2015 after translating gross transitions between original ESA land cover classes. Gross changes from LUH2v2h (Hurtt et al., 2011) and Hansen et al. (2013) are also shown for comparison. The red line indicates the zero line.
The temporal correlations of gross and net changes between ESA CCI
PFTs, Hansen et al. (2013) and LUH2v2h (Hurtt et al., 2011) are not
significant (
Spatial distributions of net and cumulative gross changes in forest, cropland and grassland PFTs between 1992 and 2015 derived from the ESA CCI data. The color scales indicate the changed fraction in each half-degree grid cell.
Gross changes in shrub and bare soil are also detected over the whole period, and the net change in these PFTs is generally a loss in area. The magnitudes of gross water body area changes are small compared to other PFTs. There is a relatively large net increase during 1995–2000 and a moderate net decrease during 2000–2010. Urban areas keep expanding over the whole period, and the increasing rates are high during 2001–2004 and 2012–2014.
The spatial distributions of net and cumulative gross changes in forest, cropland and grassland PFTs between 1992 and 2015 are shown in Fig. 3, and the distributions of the other PFTs are shown in Fig. S1. Intensive gross forest loss and sparse gross forest gain in South America result in a strong net decrease in forest area (Fig. 3). There are also considerable gross and net forest losses in South and Southeast Asia and in some regions of tropical Africa. Gross forest gain occurs pervasively in boreal regions. Some regions of intensive gross forest gain were found in south Asia, tropical Africa and South America, but with a small extent. Gross cropland gain occurs all over the world, and especially in South America, tropical Africa (particularly in the Sahel), South and Southeast Asia, and central Asia. By contrast, gross cropland loss is only observed in Europe and across the North China Plain. The cropland loss in these two regions is mainly caused by urbanization and thus an increase in urban area was found (Fig. S1). Overall, the net cropland change is an increase in most regions except Europe and the North China Plain. Grassland in temperate and tropical regions experienced extensive gross gain and gross loss, but the gross gain and loss are not fully coincident, leading to a pattern of coexisting net gain and loss everywhere (Fig. 3). The changes in grassland are relatively small in boreal regions.
The changes to shrubs are largely distributed in tropical regions, with a net
gain in South America and net loss in tropical Africa and south Asia
(Fig. S1). Intensive gross changes in bare soil were found in north China,
central Asia, Australia and the south edge of the Sahara, mainly caused by the
gross transitions between original ESA LC classes “200” (bare areas) and
“150” (sparse vegetation; tree, shrub, herbaceous cover
The global and regional net area changes in forest, cropland and grassland
PFTs from ESA CCI LC maps since 1992 are shown in Fig. 4 (solid lines).
Global net forest loss and net cropland gain between 1992 and 2015 are 0.60
and 0.67 million
Global and regional net area changes in forest, cropland and grassland PFTs derived from ESA CCI land cover maps since 1992. Data from LUH2v2h (Hurtt et al., 2011), Hansen et al. (2013) and Houghton and Nassikas (2017) are also shown for comparison. Note that net forest area change from Hansen et al. (2013) is corresponding to the period of 2000–2012, and thus the forest area change between 1992 and 2000 from ESA CCI was added as Hansen et al. (2013) data in the plot.
The magnitudes of net forest area change from LUH2v2h (Hurtt et al., 2011) are much smaller than those from ESA CCI, mainly because the forest area decrease between 1992 and 2009 (Fig. 4) is not reflected in the LUH2v2h dataset (Hurtt et al., 2011). Although the net increased cropland areas from 1992 to 2015 are similar between ESA CCI and LUH2v2h (Hurtt et al., 2011), the temporal trajectories are rather different. The increase in cropland in ESA CCI data happened between 1992 and 2004, while cropland area in LUH2v2h (Hurtt et al., 2011) mainly increased since 2007 (Fig. 4). Grassland area changes in LUH2v2h (Hurtt et al., 2011) display more variations than those from ESA CCI. There was an increase in grassland in LUH2v2h (Hurtt et al., 2011) in the earlier period (1992–2004) where ESA CCI had the increase in cropland. Globally, net forest area loss between 1992 and 2015 from both Hansen et al. (2013) and Houghton and Nassikas (2017) is much larger than that from ESA CCI and LUH2v2h data (Hurtt et al., 2011).
Consistent with the spatial distributions of net forest change in Fig. 3, net forest loss in South and Central America dominates the global net forest loss (Fig. 4), accounting for 75 % of the global total. The magnitude of net forest loss is close to that observed by Hansen et al. (2013) in this region. However, the magnitudes of net forest loss from ESA CCI PFTs in other regions are generally smaller than those from Hansen et al. (2013). Net forest area change from Houghton and Nassikas (2017) also shows a stronger loss in all three tropical regions than that in other datasets, especially in South and Central America and tropical Africa. It should be noted that the net forest loss in South and Southeast Asia is consistent between LUH2v2h (Hurtt et al., 2011), Hansen et al. (2013) and Houghton and Nassikas (2017), and all these datasets have much larger net forest area loss than ESA CCI data. All datasets demonstrate net forest gain in North America, except Hansen et al. (2013) which has a strong forest loss. The forest area in LUH2v2h data (Hurtt et al., 2011) and inventory-based data from Houghton and Nassikas (2017) shows a net increase in the China region and western Europe. In contrast, forest area in the satellite-based datasets of ESA CCI PFTs and Hansen et al. (2013) is stable or slightly decreasing.
Net area changes in forest, cropland and grassland PFTs derived from ESA CCI land cover maps since 1992 in countries with largest net forest area loss between 1992 and 2015. Data from LUH2v2h (Hurtt et al., 2011), Hansen et al. (2013) and Houghton and Nassikas (2017) are also shown for comparison.
South and Central America, tropical Africa and the former Soviet Union are the regions with the largest contributions to the global total net cropland increase, representing 37, 33 and 11 % of the global total. The regional patterns of temporal net cropland area change are rather different between ESA CCI PFTs and LUH2v2h (Hurtt et al., 2011) although the global net changes from 1992 to 2015 are similar. Cropland from LUH2v2h (Hurtt et al., 2011) expands more in tropical regions but decreases more in other regions than in ESA CCI PFTs (Fig. 4).
Grassland area from ESA CCI PFTs slightly increases in South and Central America and South and Southeast Asia, and slightly decreases in North America, the former Soviet Union, North Africa and Middle East. Differences in grassland change are large between ESA CCI PFTs and LUH2v2h (Hurtt et al., 2011) in all regions other than tropical regions.
The countries with the largest net forest PFT area loss between 1992 and 2015
from ESA CCI maps are shown in Fig. 5, and countries with the largest net
forest PFT gain in Fig. 6. Brazil, Bolivia and Indonesia are the three
countries with largest net forest losses during 1992–2015 with a net loss of
0.28, 0.044 and 0.042 million
Net area changes in forest, cropland and grassland PFTs derived from ESA CCI land cover maps since 1992 in countries with largest net forest area gain between 1992 and 2015. Data from LUH2v2h (Hurtt et al., 2011), Hansen et al. (2013) and Houghton and Nassikas (2017) are also shown for comparison.
The overall net cropland gain from 1992 to 2015 between ESA CCI and LUH2v2h (Hurtt et al., 2011) is similar in Bolivia but is rather different in all other countries in Fig. 5. Larger cropland gain from LUH2v2h (Hurtt et al., 2011) compared to ESA CCI was found in Brazil, Indonesia, Argentina and Paraguay, while lower cropland gain was found in Cambodia and the DRC. The cropland area change in China and Russia from LUH2v2h (Hurtt et al., 2011) even shows a net loss rather than gain. Grassland area increased in Argentina, Paraguay, Russia, Cambodia and the DRC in LUH2v2h (Hurtt et al., 2011), which was not captured by ESA CCI maps.
The magnitudes of forest change in the countries with the largest forest gain
in Fig. 6 are much smaller than those with largest forest loss (Fig. 5). For
example, the net forest gain from 1992 to 2015 is
0.019 million
The forest, cropland and grassland areas from different datasets do not match
on global or regional scales (Fig. 1), mainly caused by the differences in
land cover definitions and data sources (Table 1), as well as the
uncertainties in the cross-walking table used for translating original ESA
CCI LC classes into PFTs. The canopy cover of forest varies in different ESA
CCI LC classes with defined ranges such as
Forest area estimates in LUH2v2h (Hurtt et al., 2011) are based on aboveground biomass density from the Miami-LU ecosystem model (Hurtt et al., 2006), and cropland and pasture areas are based on HYDE 3.2 (Klein Goldewijk et al., 2016). HYDE 3.2 uses the cropland and pasture areas from FAO STAT (FAOSTAT, 2015) as the main land-use input data and the ESA CCI epoch LC map of 2010 as a spatial reference map (Klein Goldewijk et al., 2016). Thus, the grasslands in LUH2v2h refer to the sum of intensively managed pastures and less intensively used rangelands (Klein Goldewijk et al., 2016), while the grassland PFT from ESA CCI maps also includes natural grassland, which may be the reasons for less grassland in LUH2v2h (Hurtt et al., 2011) than ESA CCI, especially in the former Soviet Union, western Europe and North America (Fig. 1).
Similarly, the underestimate in cropland area in ESA CCI maps (Fig. 1) is likely due to differences in definitions of what constitutes a cropland based on remote sensing datasets used to derive the ESA maps vs. land use statistics and country-dependent reporting used to derive FAO statistics that are used to define croplands in HYDE 3.2 (Klein Goldewijk et al., 2016), in addition to differences in spatial resolution. For example, the attribution of oil palm plantations is an important factor for the differences in area changes between different datasets, especially in Indonesia. Oil palm is taken as cropland rather than forest in the FAO definitions (FAOSTAT, 2015) but detected as tree covers from the remote sensing (Carlson et al., 2012, 2013; Hansen et al., 2013; Koh et al., 2011; Tropek et al., 2014), including in the CCI LC products. This partly explains that the larger cropland increase in LUH2v2h (Hurtt et al., 2011) and larger forest decrease in Houghton and Nassikas (2017) than those in ESA CCI PFTs and Hansen et al. (2013) in Indonesia (Fig. 4). Furthermore, the classification of cropland in ESA CCI is also based on remote sensing temporal analysis. In the ESA CCI algorithm, for example, spectral features at key moments during the year were used to optimize the discriminations between all major crop classes: differentiating between cropland and natural vegetation (typically harvesting dates). Cropland in LUH2v2h that is essentially from FAO statistics (Klein Goldewijk et al., 2016), on the other hand, depends on country reporting and therefore comprises different definitions and data sources.
Pérez-Hoyos et al. (2017) provide an extensive comparison of multiple
cropland datasets, including ESA CCI epoch and annual maps, for the purposes
of cropland monitoring, and they found that the ESA CCI 2015 annual map is
more suitable for cropland monitoring than the epoch map because of the
reduction in cropland area over the Congo basin. They also showed that spatial
resolution is a key driver of product suitability for agriculture monitoring
(Pérez-Hoyos et al., 2017). However, the specific framework of their
study is the suitability for agriculture monitoring for early warning and
they focus on a limited number of countries selected to be “with high risk of
food insecurity” (Pérez-Hoyos et al., 2017). The issues of cropland area
from the ESA CCI LC maps discussed in their study is fully justified for
a study addressing the challenge of cropland monitoring, but it does not
allow generalizing the conclusion to all domains (e.g. to derive PFT maps for
use in global land surface modeling in this study). As Pérez-Hoyos
et al. (2017) and our study show, the agreement (or lack thereof) is
country-dependent, further implying that more consistent definitions of LC
classes are required and/or regional LC satellite mapping algorithms (or
cross-walking table, see below) are needed. Cropland mapping issues,
including those discussed in Pérez-Hoyos et al. (2017) are being
addressed in upcoming versions of the ESA CCI maps. Additionally, Waldner
et al. (2016) have produced a product that aims to combine the “fittest” LC
maps on the country level into a unified 250
The final spatial area of each PFT in this study is derived from
a combination of the ESA LC map and the cross-walking table (Table S1) used for
translating original ESA LC classes into PFTs. The range in tree cover canopy
openness (as discussed above) and percent of each type of vegetation for
mosaic LC classes in the LC description contributes to uncertainty in the
conversion fractions used to translate the LC classes into PFTs in the
cross-walking table. Thus, uncertainties in the cross-walking table
contribute to the differences in forest, cropland and grassland PFT areas
when comparing with other datasets. Only one value is used to prescribe the
fraction of each PFT for a given class, e.g. class “50” corresponds to
90 % of broadleaf evergreen trees in Table S1. This hinders an explicit
representation of spatially heterogeneous tree cover fractions. In the
absence of other information, the approximate mid-point of the range in the
LC class description is used when calculating the fraction of forest PFT from
a given LC class. For example, class “61” represents a closed canopy (
Likewise, an explicit regional classification is required for cropland. For example, class “10” (cropland, rainfed) is separated well in North America, i.e. mainly partitioning into class “11” (herbaceous cover), and thus the cropland area in this region is highly consistent with LUH2v2h data (Hurtt et al., 2011; Fig. 1). In tropical Africa where class “10” is not separated into a more detailed classification, the difference in cropland areas between these two datasets are large (Fig. 1). This is because if most of the cropland in this region belongs to class “12”, using the corresponding value for class “10” in the cross-walking table (90 % for class “10” vs. 30 % for class “12”, Table S1) overestimates cropland areas.
Hartley et al. (2017) also investigated the uncertainty in simulations of carbon, water and energy fluxes from three LSMs as a result of cross-walking table uncertainty. This study found that the spread in model outputs due to cross-walking uncertainty was higher than uncertainty due to the underlying LC maps (mapping algorithm; Hartley et al., 2017). Despite these uncertainties, satellites provide the only plausible way to derive the global maps of vegetation distribution needed to drive LSMs and validate dynamic global vegetation models. Future efforts by the ESA CCI LC project and collaborators will focus on reducing the uncertainty introduced when translating from LC to PFT, including using optimized and regionally based cross-walking tables.
The ESA CCI LC magnitudes of gross changes for all PFTs are lower than those
of all three products considered. This is explained by the effect of spatial
resolution combined with a change consolidation approach. Using Earth
observation time series
of 1
The large magnitude of gross changes in forest and cropland in LUH2v2h (Hurtt
et al., 2011; Fig. 2) mostly distributes in the tropical regions (Figs. S3
and S4) where gross changes reflect shifting agriculture (Heinimann et al.,
2017). The gross gain and loss of forest (or cropland) in the tropics from
LUH2v2h maintains a similar constant rate with other small variations
(Figs. S3 and S4). This is because the gross changes in LUH2v2h are
mainly generated from the shifting cultivation in the tropics by assuming
a turnover rate of 6.7
The discrepancies in temporal PFT net area changes between ESA CCI maps and FAO data (cropland and pasture area changes in LUH2v2h, Hurtt et al., 2011, and forest area changes in Houghton and Nassikas, 2017; Figs. 4–6) are mainly caused by the different approaches for estimating LC change used by different countries in FAO reports (FAO, 2015; FAOSTAT, 2015). Some countries like Canada distinguish land use and land cover when compiling forest statistics. For example, a forest cleared for wood harvest is not taken as a forest loss because new secondary forest will be planted on this land, thus no change in land use (Keenan et al., 2015). However, remote sensing can easily detect such land cover change and treat it as forest loss. Cropland and pasture in HYDE 3.2 (Klein Goldewijk et al., 2016) adopted the FAO categories for “Arable land and permanent crops” and “Permanent meadows and pastures”, respectively, as the main data source. In the ESA CCI LC maps, pastures are mapped as grassland and translated into 100 % “Natural Grass” PFT (Table S1). The different trajectories of temporal cropland changes between ESA CCI and LUH2v2h (the former shows increasing from 1992 to 2004 while the latter increases after 2007; Fig. 4) are probably caused by the time lag between the real changes and country reporting to FAO. Finally, the trends of cropland area change from FAO STAT data may contradict those from national statistics (Li et al., 2016), e.g. comparing FAO STAT data (FAOSTAT, 2015) with USDA estimates (Nickerson et al., 2011) for the United States or with NBSC estimates (NBSC, 2015) for China.
There are also many other land cover and land use datasets that can be used
for comparisons to assess the accuracy of land cover or land cover change in
ESA CCI LC products. However, they are either regional maps (e.g. the maps
for Europe from Fuchs et al., 2015) or global epoch maps (e.g. the
GlobeLand30 maps for 2000 and 2010; Chen
et al., 2014) and not suitable for the application in LSMs. Thus, we did not
include them in this study. In fact, there have already been studies on the
detailed comparisons of different datasets in a region (e.g. Fuchs
et al., 2015, for Europe; Yang et al., 2017, for China; and Achard
et al., 2014, for the tropics). In addition to the accuracy assessments
conducted in the ESA CCI project (ESA, 2017), a systematic comparison with
all other land cover datasets in future will help to validate the land cover
classification and land cover change detection in the ESA CCI LC products.
Instead of using a single dataset, combining a sample of several datasets is
reported to be considerably more efficient and accurate to estimate land
cover area and change (Olofsson et al., 2014; Sannier et al., 2016) and has
been adopted as technical guidelines (GOFC-GOLD, 2016; GFOI, 2016) in the
remote sensing community, especially for forest monitoring to reduce
emissions from deforestation and forest degradation in developing countries
(the REDD
The ESA CCI LC maps can be viewed online using
In this study, we compare the absolute areas and areal changes between PFTs from annual ESA CCI LC products and other datasets. In the intensive LULCC regions like South and Central America, both forest area and net forest change are consistent with those from other datasets. The detection of LC changes has significantly improved from the last version of 5-year epoch ESA CCI maps (Li et al., 2016). The detailed annual cropland changes from 1992 to 2000 fill the gaps of HYDE 3.2 data for this period, in which only decadal changes are available (Klein Goldewijk et al., 2016).
Considering the discrepancies, advantages and disadvantages among different datasets (Table 1), we propose different
choices of these datasets for application in LSMs depending on research
purposes. For example, if we would like all LSMs to share the same historical
and future maps in a model intercomparison project (e.g. using LUH2v2h data
in CMIP6), annual ESA CCI data products should be cautiously harmonized
considering the large differences between ESA CCI and LUH2v2h (Hurtt et al.,
2011). On the other hand, if we want to analyse recent carbon and water
budgets with LSMs, ESA CCI maps are definitely an appropriate choice. The
detailed LC classes in ESA CCI products provide a valuable reference map for
modellers to partition land covers into PFTs, e.g. separating the generic
forest in LUH2v2h dataset (Hurtt et al., 2011) into different forest PFTs
(Table S1). LSMs can also benefit from the 300
The supplement related to this article is available online at:
The authors declare that they have no conflict of interest.
Wei Li was supported by the European-Commission-funded project LUC4C (grant no. 603542). Wei Li and Philippe Ciais were supported by the European Research Council through Synergy grant ERC-2013-SyG-610028 “IMBALANCE-P”. Natasha MacBean, Sophie Bontemps, Céline Lamarche and Pierre Defourny were supported by the Climate Change Initiative program supported by the European Space Agency. Edited by: David Carlson Reviewed by: two anonymous referees