USGS Gap Analysis Program, Moscow, Idaho
GAP has long recognized that post-modeling regionalization of state-based predicted animal
distributions would be problematic (Csuti and Crist 1996; Crist and Jennings 1997). In a gap
analysis of terrestrial vertebrates in the western U.S., Crist and Karl (in preparation) joined
existing state GAP maps and graphically assessed them for consistency among states. The
results for herpetofauna and some mammals were fairly good (Figure 1), but many widely
distributed mammals and most birds showed only a fair to poor match among states (Figure 2).
This suggests that for regional analysis, many species will have to be re-modeled using
regionally consistent base GIS coverages (e.g., land cover, soils, hydrography, topography) and
regionalized habitat association models that can account for true regional differences in
associations. Such a task will be time-consuming and expensive.
In response, four multistate regions within GAP have initiated coordinated approaches to
regional animal distribution modeling. These range from fairly loose cooperation to obtain
consistency in range limits to full coordination of all modeling tasks. In only one case
(Southwest ReGAP) was regional modeling part of the original project objective. In the other
regions (the Great Plains, Southeast, and Upper Midwest) state GAP projects are voluntarily
coordinating their efforts. Following are reports on the methods, status, and outlook for each effort.
Fish & Wildlife Information Exchange, Virginia Tech, Blacksburg
The Southeast Regional Working Group has discussed coordination of animal modeling at
several times in recent years and developed Web-based tools to aid coordination on range extent
mapping, model sharing, and discussion.
Coordination of the animal range maps was initiated by the Virginia GAP Project. An ArcView
Internet Map Server application was developed that can read lists of EMAP hexagons and
display them interactively over the Web. This approach would allow animal modelers to look
for disjunct ranges and other anomalies prior to applying the habitat relationship data to the
model. A similar effort was successfully completed for the mid-Atlantic GAP states a few years
ago. This system is still available, but to date few states have established databases of EMAP
hexagons that can be accessed by the regional mapping system. The NC-GAP Project has
received hex species range data from TN, KY, LA, NC, and VA and is planning on combining
and comparing these ranges.
Steve Williams of NC-GAP developed a matrix approach (Microsoft Access) for coordinating
animal modeling and proposed it to the group approximately three years ago. To date this
approach has been adopted by individual GAP projects, but it has not been used to compile and
compare modeling data, a useful step towards edge-matching potential distribution maps.
Although not intended as a coordination mechanism, several GAP projects got their start in
animal modeling by acquiring habitat relationship data from other GAP projects or surrounding
state fish and wildlife agencies and/or natural heritage programs. These sources were typically
modified with local information, new references, and local expert opinion. By using the same
base data, states in close proximity to others probably have very similar animal models.
Finally, the Southeast group has been using a listserver set up specifically for the group for some
time. This mechanism has made it possible for the group to easily communicate regarding
coordination issues. To further communication, they hold regular coordination meetings where
animal modeling issues are addressed.
We feel that coordination of the animal modeling information is crucial for edge-matching the
species potential distribution maps regionally and to allow for comparison of gap
recommendations among states. As with most other data coordination efforts, this will require
time and funding committed to the process. One of the major problems in coordination is the
widely varying start dates for various states in the region. The status of state projects in the
Southeast region varies from AL starting in 2000 to TN, VA, FL, and others being completed
this year. This has led to differential degrees of participation among the states and prohibited
true regional modeling of animal distributions.
USGS Upper Midwest Environmental Sciences Center, LaCrosse, Wisconsin
The Upper Midwest Gap Analysis Program (UM-GAP) was organized to avoid duplicating
efforts while meeting the diverse information needs of the participating state and federal
cooperators. Coordination efforts have gone well with land cover classification efforts that are
now nearing completion. This regional land cover mapping approach has always been considered a necessary prerequisite to a uniform and consistent predicted vertebrate distribution mapping effort.
Regional coordination of predicted vertebrate distribution modeling began in fall 2000 with a
meeting of partners at the Upper Midwest Environmental Sciences Center in La Crosse,
Wisconsin. In attendance were representatives from the Illinois Natural History Survey, the
Michigan Department of Natural Resources, Michigan State University, and the Minnesota
Department of Natural Resources.
Proposed state species lists were compared, and the criteria for inclusion into a uniform regional
list were discussed. Initial discussions also explored ideas such as stratifying species models by
Baileys Ecoregion Provinces. Along those lines, Minnesota is planning to collaborate with the
Great Plains regional animal modeling coordination effort for the western or Great Plains portion
of the state. Although this was only the initial meeting, a general consensus was reached that the
regional partners will collaborate to produce regionwide species range maps based on EPA
hexagons. Individual species distribution modeling responsibilities will probably then be shared
among states that share the range for that species.
UM-GAP has always been organized with the belief that cooperation in the development of
uniform and consistent landscape-scale GIS layers will, in the long run, prove to be more cost-
effective if applied at the planning and development stage of the data layers rather than as an
after-the-fact effort to edge-match incompatible GIS layers.
USGS Cooperative Fish and Wildlife Research Unit, Manhattan, Kansas
At the 1998 Great Plains Regional Gap Analysis Program meeting, the Kansas Gap Analysis
Project (KS-GAP) illustrated the database expert system that we were developing for use in
vertebrate modeling in Kansas. The expert system was an Access database program that tracked
the literature and other sources of information used to attribute land cover classes and range
distributions to each vertebrate species within the state. Following the presentation and
workshop that employed the expert system, Nebraska, South Dakota, North Dakota, and Iowa
decided to adopt the modeling. KS-GAP also had developed a partial list of vertebrates for the
Great Plains, which suggested that regional modeling could considerably reduce the number of
species that had to be modeled by each state. In addition, this regional effort would reduce
future problems of edge-matching predicted vertebrate species distributions across state
Initially, regionalization required extra work, because species lists had to be made for all
vertebrates across the five state region, and lists of vegetation association descriptions had to be
found and compiled across the five states. Following these steps, the database expert system had
to be expanded to include both vertebrate species that occurred in the Great Plains and vegetation
alliances found in the five-state region. Additionally, this information needed to be searchable
and linkable to habitat descriptions in the literature for all developed species models. Because
the vegetation association descriptions were not standardized across the five-state region, we had to develop criteria by which the vegetation association could be cross-walked to each states land cover categories. This was further complicated because the five states were using different mapping methods for their land cover layers. Compiling lists across the region and cross-
walking the land cover categories across the states were the most complicated aspects of our
The development of the regional expert system required approximately one year of time, but we
believed the effort was justified by reducing our share of species modeling by 40%. The original
concept that distributing the species modeling responsibilities would reduce everyone's costs
could not be realized, however, because the five participating states were at different levels of
completion, and because KS-GAP was scheduled to finish before the other states. South Dakota,
North Dakota, and Iowa were in similar stages relative to vertebrate modeling when the Great
Plains Vertebrate Database was completed, so these states are using the database to develop
interstate models for vertebrate species they share. In the meantime, Nebraska decided to try a
different statistical approach to model its vertebrates where adequate specimen records are
available. However, Nebraska will be able to use the database and completed models from
Kansas, Iowa, and South Dakota to help model those species that could not be modeled by using
We believe that the initial costs in time spent both to develop the database program and the
interstate cooperation will be repaid easily during the current phase of gap analysis on the Great
Plains. We also are in an excellent position to proceed with GAP II as the five states plus
Minnesota finish their initial GAP projects. Because we have a common methodology and a
common language built into the database, we can proceed to develop truly regionalized
vertebrate distribution and habitat models. With encouragement from the National Gap Analysis
Program, we hope to see a regional land cover map and stewardship layer developed with the
same level of interstate cooperation as has occurred in modeling vertebrates. When these
coverages are available, we will have a truly integrated regional gap analysis of the northern
USGS Cooperative Fish and Wildlife Research Unit, Las Cruces, New Mexico
Animal modeling in SW-ReGAP is in a very early stage because initial effort is devoted
primarily to land cover mapping (through 2001). Nonetheless, several early coordination steps
are under way in the regional animal modeling lab at the New Mexico Research Unit. Early
development emanates from a January 2000 Animal Modeling Workshop. At that workshop the
group decided that two key concepts would guide the project. First, habitat association and
model development for each species would be distributed among the states according to
expertise or largest proportion of species range and second, GIS modeling tasks would be
conducted at a central lab at the New Mexico Research Unit. The first two steps of the work
have been initiated.
For the first step, we are preparing a taxonomic decision rule base to guide all projects in jointly
selecting taxa that are to be modeled across the region. The draft rule was distributed to all members of the regional project in December 2000 for review, revision, and ultimate acceptance across the region.
The second step is to assign individual species modeling responsibilities to specific projects
within SW-ReGAP. This approach ensures that species models evolve from available
information and enhances the likelihood that species model information are derived from local
expert knowledge. Further, this approach will realistically distribute effort to provide critical
help to the regional lab.
A preliminary assignment of at least 865 species among projects will be coordinated with all
projects after the taxonomic rule base is completed. The assignment is based on establishing
model development responsibilities with the state that encompasses the primary range or
distribution of key habitat for each species. Those species with wide distribution or habitat
associations will be the responsibility of the regional lab throughout. Other projects will
contribute their model development to the regional lab for review and revision across the region
prior to final model and map preparation by the regional lab. We believe this approach will
facilitate numerous other facets of the modeling process associated with literature review, habitat
association databasing, and production of seamless predictions.
The requirements for regional modeling are straightforward: (1) establish a common list of
species to be modeled, (2) establish a consistent view of range extent for each species, (3)
develop seamless regional modeling base layers (e.g., land cover, topography, hydrography), and
(4) develop a regional habitat association database that accounts for subregional changes in
associations. As we have seen, other regional efforts (except the Southwest) have had limited
success due to differential start dates and inconsistent methodologies. They have, however,
achieved much in the way of information sharing, communication, and development of useful
tools and can serve as a guide for future updates in GAP. Success then is a function of the
degree that the modeling integration is systematic rather than opportunistic. The Southwest
ReGAP approach holds great promise in maintaining the tie to local knowledge of species
location and habitat associations while increasing efficiency and assuring seamless products by
centralized integration and processing. A national analog may be a national database of habitat
associations by ecoregions to retain local specificity, as well as occurrence locations and range
limits based on EMAP hexagons or other fixed units. A well-maintained system, accessible for
input by local researchers and naturalists, would allow dynamic modeling using the most current
data. Standardization of GAP models, habitat association databases, and GIS layers may be the
first step to achieving national animal modeling capability.
Crist, P.J., and M. Jennings. 1997. Regionalizing state-level data. Pages 15-16 in E.S.
Brackney and M.D. Jennings, editors. Gap Analysis Bulletin No. 6. USGS Gap Analysis
Program, Moscow, Idaho.
Csuti, B., and P.J. Crist. 1996. Methods for developing terrestrial vertebrate distribution maps
for Gap Analysis. In A Handbook of Gap Analysis. USGS Gap Analysis Program, Moscow,