this update we are taking a similar approach that is informed by the same basic principles.
We are obtaining best estimates of the same 16 basic land cover types: Open Ocean,
Estuaries, Seagrass/Algae Beds, Coral Reefs, Shelf, Tropical Forest, Temperate/Boreal
Forest, Grass/Rangelands, Tidal Marsh/Mangroves, Swamps/Floodplains, Lakes/Rivers,
Desert, Tundra, Ice/Rock, Cropland, and Urban.
The state of the art of global mapping of land cover has progressed significantly since
the mid 1990’s which has spawned a diverse assortment of land cover data products in a
diverse assortment of classification schemes and spatial resolutions. A brief perusal of the
USGS land cover institute’s web site provides a sense of these developments over the last 15
years (http://landcover.usgs.gov/landcoverdata.php). Our determination of the best global
representation of land cover relevant to this inquiry was the GlobCover data set produced by
the European Space Agency in partnership with the United Nations Food and Agriculture
Organization (http://www.esa.int/esaCP/SEMZ16L26DF_index_0.html). The GlobCover
classification scheme is a good match to the 1997 classification scheme and it includes
improved measures of the spatial extent of wetlands, water bodies, and urban areas. One
issue for utilizing the GlobCover measures of land cover was the lack of a category for
‘Tundra’. The GlobCover categories of ‘lakes/rivers’, ‘grassland/rangeland’ and ‘ice and
rock’ were all significantly higher than the areal extents used in the 1997 paper. We chose to
hold the areal extent of ‘ice/rock’ and ‘lakes/rivers’ constant and attribute the difference to
the ‘tundra’ category. This still represents a ‘loss’ of ‘tundra’ that is probably a classification
issue captured in the ‘grass/rangelands’ category.
This update of the areal extent of land cover would ideally only represent true
changes to the actual land surfaces of the earth that have taken place over the time span.
Nonetheless, some of the differences can undoubtedly be attributed simply to improvements
in our ability to map, classify, and measure the surface of the earth. The ‘urban’ category
provides a case in point. The land cover classification ‘urban’ can be discussed and argued
about at great length. In any case, the 1997 number of 337 million hectares represents an
estimate of terrestrial urbanization at roughly 2.2%. We know urban extent has increased
over the past 15 years despite the great discrepancies in measurements of urban extent (5).
Conservative MODIS based measures of urban extent are on the order of 65 million ha
(~0.5% of the land), the GlobCover dataset is 31 million ha (~0.24% of the land), and the
Global Rural-Urban Mapping Project (6)( estimates urban extent at 352 million hectares
(~2.74% of the land). We used the GRUMP number to show a modest (~4%) increase in
urban extent over the time period in question. The importance of the ‘urban’ category relative
to measures of the economic value of ‘natural’ capital manifests primarily in the idea that
human well-being is increased via the interaction of social, natural, built, and human capital
and ‘urban’ is the spatial location of a significant fraction of built, human, and social capital.
Conservative estimates of ‘urban’ consequently can dramatically minimize the nature of the
spatial interactions that occur between natural, human, social, and built capital.
Figure S1 shows global land cover converted to ecosystem service value using the 2011 unit
values shown in Table S1.