banner literature about projects tools meetings search links BBS GAP home GAP home
home

 

<-Previous | First | Next ->

Land Cover Characterization in Gap Analysis:
Past, Present, and Future

COLLIN HOMER,  EROS Data Center, Sioux Falls, South Dakota
and PATRICK CRIST, National GAP Office, Moscow, Idaho

Introduction

The evolution of land cover characterization in GAP over the last
13 years has been truly remarkable. Although relatively coarse by
today’s standards, early GAP land cover mapping was innovative
and for its time represented the first detailed, statewide mapping
effort. It was driven by biologists’ requirements to map vegetation
diversity and animal habitat covering large areas and to use exist-
ing or emerging technology to characterize the landscape. This
mind-set has evolved into an ongoing theme in GAP—a constant
push to characterize the landscape at improved spatial and thematic
scales, resulting in rapid development of methodologies and proto-
cols.

The first land cover mapping effort conducted for GAP was a pilot
project in Idaho, which began in 1987. This effort created a synthe-
sized vegetation map based on information from existing local, re-
gional, and state vegetation maps. The map was compared to and
refined based on Landsat Multispectral Scanner (MSS) satellite
image prints (Scott et al. 1993, Scott and Jennings 1997). In 1989,
mapping for the Oregon GAP project utilized visual photo-inter-
pretation of satellite image prints to locate boundaries of vegeta-
tion classes (Scott et al. 1993). Subsequent third-generation projects
such as California, Nevada, and Utah mapped vegetation using some
combination of digital image classification, photo-interpretation of
satellite imagery, and reference to existing maps and ancillary data
(Davis et al. 1995, Homer et al. 1997, Scott and Jennings 1997).
By this time GAP was on its way to becoming a national program,
and the state-based business model was instituted for political and
practical reasons. Consciously, however, it was believed that em-
ploying the best people in each state to develop this large-area map-
ping process would result in significant innovation and achieve-
ment. This indeed has appeared to be the case (Eve and Merchant
1998).

Regional and national products were always envisioned and, in 1997,
the 11 western states’ land cover products were joined (Wright et

al., in preparation). The effort was limited to cross-walking the
classification schemes and joining the maps into a common data
structure and projection. Despite the high variability in mapping
methods and spatial and thematic detail (Eve and Merchant 1998),
there were few serious boundary problems between the states. Still
the variation, particularly in thematic level of detail, required much
aggregation of types across the region and suggested the need for
improved methods and regional cooperative approaches in future
mapping.

Present

Continuing evolution of technique and technology has led to cur-
rent GAP land cover characterization efforts, which are: digitally
based on multi-date Landsat TM 30 m data, involve image stratifi-
cation, special clustering, some form of pre- or post-classification
modeling, and accuracy assessment. Each project still retains a
unique mix of technical expertise, capabilities, cooperators, land
cover, and goals (Eve and Merchant 1998).

Current tool and data set availability reflect a rapid evolution. Early
projects constantly suffered hardware and software limitations. For
example, the first digital mosaic of Utah had to be created, stored,
and classified in pieces because of inadequate disk storage space
and software. Agreements such as the Multi-Resolution Land Char-
acteristics Consortium (MRLC) have helped greatly in developing
base TM imagery and ancillary data sets outside of the project. Pre-
MRLC projects had to spend large portions of their budgets and
time in acquiring and rectifying their own imagery and building
ancillary data layers from scratch before mapping could even com-
mence.

Most importantly, expertise in land characterization, remote sens-
ing (RS), and GIS through GAP has developed into a remarkable
talent pool. GAP land cover mapping efforts have created an im-
pressive training ground for hundreds of people (Eve and Merchant
1998). The program has served as the catalyst to introduce a wide
variety of private and government collaborators to RS/GIS tech-
niques, data sets, and capabilities. This legacy of GAP will con-
tinue to have profound positive effects in pushing the science of
land cover characterization forward. Not all innovations have been
in remote sensing, however. For some time, GAP researchers have
 

recognized the difficulty in translating between ecologically-based
classification systems like UNESCO, followed by the National Veg-
etation Classification System (NVCS) (FGDC 1997), and the reali-
ties of remote-sensing limitations. Partnerships between GAP and
The Nature Conservancy plant ecologists are increasingly address-
ing this issue through the addition of new map objects that are eco-
logically consistent with the NVCS, such as Ecological Complexes
and Compositional Groups (Pearlstine et al. 1998).

The newest innovation in GAP addresses consistency by conduct-
ing mapping updates by multistate regions. The first such project
is Southwest Regional GAP (SW-ReGAP) that includes Arizona,
Colorado, Nevada, New Mexico, and Utah and covers an area 93%
of the size of Alaska. This project will take advantage of methods
developed by other projects such as the use of multi-date TM imag-
ery that helps distinguish similar vegetation by phenological differ-
ences. It may also explore the use of airborne imagery such as
videography, digital imaging, and low-elevation, multispectral cam-
eras. The most dramatic change, however, is that mapping will be
a coordinated effort where preprocessing and clustering of imagery
is done regionally by EROS Data Center and Utah State University.
Then state labs will label and model the land cover types based on
“mapping zones” defined primarily by Bailey’s section-level
ecoregions rather than state boundaries. 

Figure 1

Figure 2

The objective is a seamless, thematically consistent land cover
map of the entire region.

 

 

 

Figure 1. A diagram illustrating the land cover mapping process for
the Southwest Regional Gap Analysis Project. An R following a task
indicates Regional Lab (RS/GIS or EROS) responsibility. An S
indicates State Lab responsibility.

 

Future

The increasing momentum of new data sources, refined tools and
processes, and developing expertise should converge for a bright
future in land cover characterization. Future efforts will be driven
by improvements in several key areas:

Data - It is an exciting time as we watch the continued expansion
of air- and space-borne remote sensing platforms. Several success-
ful recent satellite launches (e.g., Landsat 7, Terra, Ikonos) will
greatly expand remote sensing data availability. They provide ad-
ditional spectral and spatial information at lower cost, creating new
possibilities in cover-type characterization. More importantly, this
increase in data will improve the ability to choose both the timing
and type of data acquisition, thus enabling a deeper understanding
of climate patterns, vegetation seasonality, and closer real-time com-
parison of imagery and field data.

However, spectral and spatial improvements will not automatically
create more meaningful cover-type classes. There will always re-
main ambiguities between spectrally meaningful categories and
human-defined meaningful categories. This means that ancillary
data will continue to play a large role in successful characteriza-
tion. The increasing availability of nationwide, regionally consis-
tent information on wetlands, topography, soils, cultural features,
and other data will still prove critical to successful efforts in model-
ing vegetation community distributions and their attributes.

Tools - The continued rapid development of software and hardware
tools will allow expanded capabilities hardly imagined now. Soft-
ware developments will increase functionality in visualization and
manipulation of data sets, automated interpretation processes, data
integration, standardized file formats, and real-time application.
Hardware developments will allow increasing miniaturization, port-
ability, capability, and affordability. Future development of other
tools in spatial navigation, digital imaging, and data capture prom-
ise to revolutionize ground data analysis with remote sensing.

Process - Bringing data and tools together to create more mean-
ingful information is probably the most critical future challenge.
GAP land cover efforts, to date, have had only the resources to map
simple, discrete categories of land cover. These serve as only coarse
surrogates for representing wildlife habitat and vegetative diver-
sity. Mapping efforts typically only focus on creation of the final
labeled land cover layer, with intermediate, ancillary image and
clustered data layers often viewed as “throw-away” steps. A key
shift of the future will bring more focus on building a “data sand-
wich” with layers of information seen as components of a database
rather than simple intermediate steps (Estes et al. 1999). These
components will be applied in different combinations to supply the
needed land cover derivatives of the flexible database. Derivatives
can be discrete (a labeled land cover category) or continuous (a
biophysical measurement). Combinations of data set derivatives
could ultimately be tailored to represent each animal and plant spe-
cies’ unique habitat needs. For example, a generalist mammal spe-
cies (i.e., coyote) might be adequately represented with broad, dis-
crete cover-type categories, while a neotropical bird (i.e., Kirtland’s 

warbler) with specialized seasonal requirements might need high-
resolution habitat definitions, based on habitat structure and com-
position continuums. Likewise, individual plant species distribu-
tions and status may be modeled (see Fertig et al. 1998) based on
NVC community distribution combined with other ancillary data
of land characteristics. The component-built flexible database will
offer greater utility, expanded flexibility, ease of update (replace
the component, not the entire set), data layer independence (com-
ponents stand alone), and manageability.

Regionalization of completed, state-based land cover maps will
continue, and updates in regionally coordinated projects will be-
come the norm. Mapping zones may become institutionalized, and
updates for them may become distinct from state or multistate GAP
projects. Partnerships will also increase. SW-ReGAP is already
planned to have an equal or greater share of the cost paid by other
partners. Future mapping efforts will be planned as highly robust
projects incorporating a wider variety of data sets, tools, and objec-
tives that will involve multiple partners.

 

Conclusion

The convergence of better data, tools, and processes will result in
capabilities that will continue to spur innovation in GAP land cover
characterization. This ability to increasingly combine the tools and
data in processes that characterize the landscape in more diverse
and meaningful ways will provide further insight into the interplay
of biodiversity and landscape ecology. Future characterization ef-
forts will focus more on creation of a database “sandwich” capable
of flexible derivatives to meet the wide array of wildlife habitat
characterization needs. Components in this database can poten-
tially be assembled and used independently. More and more of the
time-consuming components of current types of GAP land charac-
terization can be “out-sourced,” leaving GAP investigators more
time and resources to focus on biology-related issues and research.
Because general land characterizations will be more commonly
available, the real challenge of the future will not be the limitations
of the tools but development of science and methods to more accu-
rately characterize the distribution and condition of those features
that represent biodiversity on the ground.

 

Literature Cited

Davis, F.W., P.A Stine, D.M. Stoms, M.I. Borchert, and A.D. Hol-
lander. 1995. Gap analysis of the actual vegetation of Califor-
nia 1. The southwestern region.
Madroño 42:40-78.

Estes, J.E., A.S. Belward, J.O. Justice, T.R. Loveland, J. Scepan, J.

Townshend, and A. Strahler. 1999. The way forward. Photo-
grammetric Engineering and Remote Sensing
65:10891093.

Eve, M.D., and J.W. Merchant. 1998. A national survey of land
cover mapping protocols used in the Gap Analysis Program.
Published on the Center for Advanced Land Management In-
formation Technologies (CALMIT) Internet page at http://
www.calmit.unl.edu/gapmap.

Fertig, W., W.A. Reiners, and R.L. Hartman. 1998. Gap analysis
for plant species.
Gap Analysis Bulletin 7:24-25.

FGDC. 1997. National Vegetation Classification Standard. Fed-
eral Geographic Data Committee, Vegetation Subcommittee,
U.S. Geological Survey, Reston, Virginia.

Homer, C.G., R.D. Ramsey, T.C. Edwards, Jr., and A. Falconer,

1997. Landscape cover-type modeling using a multi-scene The-
matic Mapper mosaic.
Photogrammetric Engineering and Re-
mote Sensing
63:59-67.

Pearlstine, L., A. McKerrow, M. Pyne, S. Williams, and S. McNulty.

1998. Compositional groups and ecological complexes: A
method for alliance-based vegetation mapping.
Gap Analysis
Bulletin
7:16-17.

Scott, J.M., F. Davis, B. Csuti, R. Noss, B. Butterfield, C. Groves,

H. Anderson, S. Caicco, F. D’Erchia, T. Edwards, J. Ulliman,
and G. Wright. 1993. Gap Analysis: A geographic approach to
protection of biological diversity.
Journal of Wildlife Manage-
ment
57(1) supplement, Wildlife Monographs No. 123.

Scott, J.M., and M.D. Jennings. 1997. A description of the Na-
tional Gap Analysis Program. Published on the National Gap
Analysis Program Internet home page at http://
www.gap.uidaho.edu/About/Overview/GapDescription. Bio-
logical Resources Division, U.S. Geological Survey, March
1997.

Wright, G., J.M. Scott, S. Mann, and M. Murray. In preparation.

Identifying unprotected and at-risk plant communities in the
West. To be submitted to
Biological Conservation .

<-Previous | First | Next ->