of the Intermountain Semi-Desert Ecoregion
Vegetation Mapping Approach
Management Status Mapping
The Biological Resources Division of USGS coordinates the Gap
Analysis Program (GAP), a proactive conservation assessment
to identify "gaps" in the protection of native biodiversity
(Scott et al. 1993). Maps of plant cover types are compiled over
political units (i.e., individual states), but analysis must also
be conducted over ecologically meaningful regions. Although GAP
has standards for mapping land cover, the individual states often
generate products with different levels of spatial and taxonomic
detail, due in part to the evolution of the standards over the duration
of the program. The objective of a regional gap analysis is to evaluate
the conservation status of cover types (following the Driscoll classification
schema derived from that of UNESCO and now proposed as the federal
Vegetation Classification System standards). Initial examination
suggested that simply mosaicking the state maps together would not
provide the consistent product needed. Remapping the region from
scratch was unacceptable given the investment in producing the state
maps. The task then was to compile a regional map from the existing
GAP state maps, while improving the spatial and taxonomic consistency.
This project was funded by the IBM Environmental Research
Program and the national
Gap Analysis Program (GAP).
The land cover maps for Bailey et al.'s (1994) Intermountain Semi-Desert
Province (ISD) (see map or shaded
relief) (Washington, Oregon, Idaho, Nevada, California, Utah,
Colorado, and Wyoming) were cross-walked into consistent cover types
according to the GAP/UNESCO system at their finest level of the
hierarchy (see the state maps at the alliance
level-revised September 26, 1996). Four main obstacles were
1) not all states distinguished forest from woodland which is at
the highest division of the UNESCO classification.
2) Some cover classes were labeled as mosaics of more than one type,
often from two different UNESCO classes or subclasses. Furthermore,
four states (CA, WA, WY, and CO) labeled map units as landscape
mosaics comprised of two or three cover types.
3) Taxonomic (floristic) detail varied between states and between
4) The spatial grain tended to be much finer in states which used
supervised classification of digital imagery from those that used
a photointerpretation approach.
The potential consequences of using the map data as given was that
the areal extent and locations of cover types could be inaccurate
which could affect their apparent management status. Even the esthetic
appearance of this intial regional map could create scepticism among
policy makers and the public.
Vegetation Mapping Approach
A workshop was held at UCSB in June, 1996, to plan the mapping
strategy for the ecoregion. The workshop (members listed below)
reviewed the cross-walk to alliance types, the map-guided classification
approach, and the potential publication alternatives. The following
persons attended the workshop:
Frank W. Davis, California GAP
David M. Stoms, California GAP
Michael J. Bueno, California GAP
Ken Driese, Wyoming GAP
Blair Csuti, Oregon GAP
Jimmy Kagan, Oregon GAP
Michael Murray, Idaho GAP
Chris Grue, Washington GAP
Kelly Cassidy, Washington GAP
The first phase of the regional mapping process was to cross-walk
the state GAP land cover types into a consistent set of alliances.
This has been done (see cross-walk table)
iteratively, with review by the project collaborators from the state
A map-guided classification technique was developed to classify
multi-temporal AVHRR data so that plant phenology and surface temperature
variation can be incorporated in the classification. This utilizes
an independent image dataset while making maximum use of the existing
GAP land cover map information. First, daily AVHRR images for the
1990 growing season were processed into ten-day images to remove
cloud cover (Stoms et al, in review). AVHRR datasets were generated
for four periods throughout the 1990 growing season (April, June,
July, and September) to capture the main intra-annual variation
of the phenology of the semiarid vegetation of this region. For
each of the four periods, the derived datasets included Normalized
Difference Vegetation Index (NDVI), brightness temperature of band
4, and near infra-red reflectance of band 2. Principal
components were then derived from the four sets of AVHRR composites
to further reduce noise. Image classification is being done using
the state GAP land cover maps, cross-walked to alliances, as training
data. The process is iterative, in that strong associations between
spectral classes and cover types are assigned first, and further
iterations only work on the remaining unclassified pixels. Some
polygons and classes such as water bodies and riparian forests that
were better mapped from TM data by the state GAP projects will be
retained in the new map. The map was then
reviewed by the project collaborators and modified as suggested.
The state land stewardship/management maps were obtained and assigned
to the four GAP management status classes according to the program
standards. Many of the states already had made these assignments
(Washington, California, and Wyoming), while others had ownership
but not specific status assignments (Montana, Colorado, Utah and
Nevada). Idaho had a coverage for status 1 and 2 but not for 3 and
4. Oregon had a 1-10 rating and rules for cross-walking to the 4
class GAP system. UCSB reviewed the state maps, made a comparative
table of assignments, and compiled a regional
map for the ISD Ecoregion (revised September 12, 1996). Some
state assignments were modified to increase regional consistency,
such as all water bodies being assigned to status 0. A few questions
remain to be resolved such as Department of Energy sites. The Idaho
National Engineering Lab was assigned to status 2, while the Hanford
Lab in Washington was assigned partly to 2 but mostly to 4.
An iterative, "map-guided"
classification approach was developed to compile a spatially and thematically
consistent, seamless land-cover map of the entire Intermountain Semi-Desert
ecoregion from a set of semi-independent subregional maps derived
by various methods. A multi-temporal dataset derived from AVHRR data
was classified using the subregional maps as training data. The resulting
regional map attempted to meet the guidelines of the proposed National
Vegetation Classification Standards for classification at the alliance
level. The approach generally improved the spatial properties of the
regional mapping, while maintaining the thematic detail of the source
maps. The methods described may be useful in many situations where
mapped information exists but is incomplete, compiled by different
methods, or is based on inconsistent classification systems.
A conservation gap analysis was conducted for the Intermountain
Semi-Desert ecoregion to assess the representation of land-cover
types within areas managed primarily for biodiversity objectives.
Mapped distributions of plant communities were summarized by land
management status categories. The total amount of land permanently
protected in the ecoregion is less than 4% and most types that are
characteristic of the region have less than 10%. Of 48 land-cover
types, twenty were found to be particularly vulnerable to potential
loss or degradation, because of low level of representation in biodiversity
management areas and the impact of expected land use activities.
The gap analysis data and findings will be useful in providing a
regional perspective in project impact assessment and future conservation
planning within this ecoregion.
Two manuscripts have been recently published in scientific journals,
one on the gap analysis of the ecoregion, the other on the mapping
technique. The manuscript citations are:
Stoms, D. M., F. W. Davis, K. L. Driese, K. M. Cassidy, and M.
P. Murray. 1998.
Gap analysis of the vegetation of the Intermountain Semi-Desert
Ecoregion. Great Basin Naturalist. 58: 199-216.
Stoms, D. M., M. B. Bueno, F. W. Davis, K. M. Cassidy, K. L. Driese,
and J. S. Kagan. 1998.
Map-guided classification of regional land-cover with multi-temporal
AVHRR data. Photogrammetric Engineering and Remote Sensing
The Nature Conservany is using these GAP data from the Columbia Plateau portion of this
ecoregion as the coarse-filter component in developing a regional
portfolio of conservation sites.
comments to David Stoms: firstname.lastname@example.org
to UCSB IBM ERP home page