United States Environmental Protection Agency	Office of Research and Development

National Exposure Research Laboratory
Research Abstract

Government Performance Results Act (GPRA) Goal 8
Annual Performance Measure 74

Significant Research Findings:

Produce Land-cover/Land-use Digital Database for Contiguous U.S.

Scientific	Many of today's environmental problems, such as nutrient pollution, habitat loss,

Problem and	climate change, and urbanization, are regional to national in scale and require

Policy Issues	land-cover data to study their causes and consequences. Such environmental

problems transcend administrative boundaries; national-scale production of the
MultiResolution Land Characteristics (MRLC)-National Land Cover Data (NLCD)
overcomes a fundamental obstacle that has plagued all previous land-cover
mapping efforts — lack of data beyond a given jurisdictional boundary. The
MRLC Consortium, of which The U.S. Environmental Protection Agency (EPA)
is a member, sponsored the production of the first national-scale, consistently
classified NLCD to provide information (land cover) to address these and many
other issues. The NLCD provides, free-of-charge through Internet access
(http://www.epa.gov/mrlc), generalized land-cover data for the conterminous U.S.
(e.g., deciduous forest, evergreen forest; cropland, hay/pasture; high- and low-
density residential). The data are widely used within EPA to address issues
related to many Government Performance Results Act (GPRA) goals (e.g., 2.2.3 -
clean water), and the data are used by many outside EPA.

MRLC-NLCD land-cover mapping had two primary objectives: 1) produce land-
cover data for the contiguous U.S., and 2) evaluate the accuracy of that data. The
first objective, land-cover mapping for the 1990s, was accomplished using
Landsat Thematic Mapper (TM) data collected using a target year of 1992.
Conversion of the Landsat TM spectral data to land cover was accomplished
using minimum-distance-to-mean Euclidean clustering and post-clustering logical
modeling. The post-clustering logical modeling used a suite of ancillary data,
including Digital Elevation Models (DEM) and TIGER road and census data. The
second objective, evaluation of land-cover thematic accuracy, was accomplished
using two-stage cluster sampling. The two-stage sampling design was
implemented to meet four criteria of a statistically rigorous sample: 1) satisfy
protocols defining a probability sample, 2) provide sufficient sample size for each
mapped land-cover class to estimate accuracy with reasonable precision, 3)
maintain reasonable cost, and 4) achieve a spatially well-distributed sample.

Results and	Land-cover mapping was completed during the 2001 calendar-year and was made

Impact	available the through the Internet. The land-cover data can be obtained free-of-

Research
Approach

National Exposure Research Laboratory — October 2003


-------
charge at http://www.epa.gov/mrlc. The data are made available by state. In
addition, numerous landscape indicators were developed from these data and are
Internet-available (http://www.srs.fs.fed.us/4803/landscapes). Thematic accuracy
evaluations have been completed for eight of the 10 EPA federal regions, and are
in progress for the other two. The accuracy evaluations completed to date
indicate that the thematic accuracy of the MRLC-NLCD land cover meet the long-
standing nominal standard of 85% agreement (Anderson et al. 1976) at the
Anderson Level I thematic detail (e.g., forest, agriculture, urban, water, barren,
grass/shrub). At the Anderson Level II of thematic detail, overall accuracies are
about 60%. The Anderson Level II land-cover classes are: water, perennial snow
and ice, low-intensity urban, high-intensity urban,

commercial/industrial/transportation, bare rock/sand/clay, mining, transitional,
deciduous forest, evergreen forest, mixed forest, shrubland, orchards/vineyards,
"semi-natural" grassland, hay/pasture, row crops, small grains, fallow, open
urban areas, woody wetland, emergent wetland.

EPA's research collaborations can be grouped into three areas: land-cover
mapping, accuracy evaluation, and landscape indicator development. U.S.
Geological Survey (USGS) EROS Data Center (EDC) is the principal collaborator
for land-cover mapping. USGS EDC and Steve Stehman (SUNY-ESF) are the
principal collaborators for accuracy assessment, and the U.S. Forest Service is
the principal collaborator for landscape indicator development. Relevant
research publications are listed below in four categories: program overview and
land-cover mapping, accuracy evaluation, landscape indicator development, and
applications.

Program overview and land-cover mapping

Vogelmann, J. and Wickham, J. Implementation Strategy for Production of National Land-cover
Data (NLCD) from the Landsat 7 Thematic Mapper Satellite. US EPA/600/R-00/051, Office of
Research and Development, U.S. Environmental Protection Agency, Washington, DC, 2000.

Vogelmann, J., Howard, S., Yang, L., Larson, C., Wylie, B., and Van Driel, N. Completion of the
1990s National Land Cover Data Set for the Conterminous United States from Landsat
Thematic Mapper data and ancillary data sources. Photogrammetric Engineering and Remote
Sensing, 67:650-662, 2001.

Wickham, J., Hegge, K., Homer, C., and Morissette, J. National land-cover database 2000 update.
The Earth Observer, 14(2): 13-15,2002.

Accuracy evaluation

Smith, J., Wickham, J., Stehman, S., and Yang, L. Impacts of Patch Size and Land Cover

Heterogeneity on Thematic Image Classification Accuracy. Photogrammetric Engineering and
Remote Sensing, 68:65-70, 2002.

Wickham, J., Stehman, S., Smith, J., and Yang., L. Submitted. Thematic Accuracy of MRLC-
NLCD land cover for the Great Lakes, Midwest, and Pacific Northwest. Remote Sensing of
Environment.

Yang, L., Stehman, S., Smith, J., and Wickham, J. Short Communication: Thematic accuracy of

Research
Collaboration and
Research
Products

National Exposure Research Laboratory — October 2003


-------
MRLC land-cover for the eastern United States. Remote Sensing of Environment, 76:418-422,

2001.

Landscape indicator development

Riitters, K.H., Wickham, J.D., Vogelmann, J.E., and Jones, K.B. National land-cover pattern data.
Ecology', 81(2): 604 and Ecological Archives [www.esapubs.org/archive/ecol/E081/004].i 2000.

Applications

Jones, K., Neale, A., Wade, T., Wickham, J. , Cross, C.., Edmonds, C., Loveland, T., Nash, M.,
Riitters, K., and Smith, E. The consequences of landscape change on ecological resources: an
assessment of the US mid-Atlantic region, 1973-1993. Ecosystem Health, 7:229-242, 2001.

Riitters, K., Wickham, J., O'Neill, R., Jones, K., Smith, E., Coulston, J., Wade, T., and Smith, J. in
press. Fragmentation of continental United States forests. Ecosystems.

Smith, J., Wickham, J., Norton, D., Wade, T., and Jones, K. Utilization of landscape indicators to
model pathogen impaired waters. Journal of the American Water Resources Association, 37:1-
10, 2001.

Wickham, J., O'Neill, R., Riitters, K., Smith, E., Wade, T., and Jones, K. Geographic targeting of
increases in nutrient export due to future urbanization. Ecological Applications, 12:93-106.,

2002.

Future Research Completion of the first NLCD product was based on Landsat TM data collected
on a target year of 1992. Over the next three to five years, the principal emphasis
of MRLC-NLCD will be to produce a second land-cover database from Landsat
ETM+ collected on a target year of 2000. There are two major methodological
changes in the production of the MRLC-NLCD 2000 land-cover data: 1) mapping
by ecoregion-like mapping zone and 2) use of decision-tree software for
modeling of Landsat spectral data into land-cover classes. In addition there are
three research areas supporting MRLC-NLCD 2000 land-cover production: 1)
accuracy evaluation of the MRLC-NLCD 2000 land-cover product, 2) sampling
design and accuracy evaluation of a level I land-cover change product derived
from comparison of the 1992 and 2000 data, and, 3) development and analysis of
landscape indicators from the MRLC-NLCD 2000 land-cover data. Collaborators
for accuracy evaluation of the 2000 products include the U.S. Department of
Agriculture (USDA)-SCS, USGS EROS Data Center, and SUNY-ESF.
Collaborators for sampling design and accuracy evaluation of change include
SUNY-ESF and USGS EROS Data Center. The USFS is the principal
collaborator for development and analysis of landscape indicators. Completion
of MRLC-NLCD 2000 is anticipated to be FY2008. The supporting research will
be ongoing as data become available for selected regions.

National Exposure Research Laboratory — October 2003


-------
Contacts for	Questions and inquiries can be directed to:

Additional	James D. Wickham, MRLC Director

Information	U.S. EPA, Office of Research and Development

National Exposure Research Laboratory
Environmental Sciences Division
Landscape Ecology Branch
Research Triangle Park, NC 27711
Phone: 919/541-3077
Fax: 919/541-1438
E-mail: wickham.iames@epa.gov

National Exposure Research Laboratory — October 2003


-------