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

Previous | First | Next ->

Experience and Trends in Animal Distribution Modeling

1 PATRICK CRIST
and 2 J. MICHAEL SCOTT

1National Gap Analysis Program, Moscow, Idaho
2Idaho Cooperative Fish and Wildlife Research Unit, Moscow, Idaho

Introduction

Modeling the predicted distribution of thousands of animal species
over large regions, but at detailed spatial resolution, represents as
significant a challenge as land cover mapping, if not more so. I
recall during my first few months with New Mexico GAP, that an
academic scientist challenged our ability to accomplish such a task.
His argument was that “you’d never be able to get the GIS data [of
land characteristics] needed for the habitat associations.” Our re-
sponse was that we already had all the layers; what was lacking,
and continues to be, is the information on species associations with
those land characteristics. As I will discuss, technological progress
will continue to greatly improve our ability to model animal pres-
ence/absence and even allow us to begin dealing with concepts such
as abundance, viability, movement, and metapopulations. How-
ever, technology is not a replacement for knowledge, and this re-
mains the greatest challenge for all modelers.

Past

As described in the land cover article above, Idaho was the first
pilot project for animal distribution mapping in 1987. At that time,
it was recognized that the two biggest challenges were gathering
information on each species’ habitat associations (Karl et al., in
press [a]) that were mappable, and testing the accuracy of the re-
sulting distributions. A decision was made early in the program’s
history to begin mapping terrestrial vertebrates and possibly butter-
flies (though little was done for the latter in most states). There
were several reasons for this decision: they are the better studied
taxa, making the modeling more tractable; they are popular targets
for conservation action; and there was some suggestion that they, in
combination with vegetation, would be a suitable surrogate for
biodiversity in general. The latter assumption is still a largely un-
tested hypothesis. The early state projects tended to restrict model-
ing to a species association with land cover, but by the early ’90s
projects were beginning to add other variables such as topography,
hydrology, soils, climate, etc. (see, for example, Thompson et al.
1996, Edwards et al. 1995). In 1996, I had the pleasure of working
with one of GAP’s originators, Blair Csuti, to compile GAP’s knowl-
edge and experience to date in the form of the GAP Handbook chap-
ter on animal distribution modeling (Csuti and Crist 1996). The
qualitative process we presented is summed up by Figure 1.

Figure 1. The animal modeling process diagram from the GAP
Handbook (Csuti and Crist 1996).

Present

As evidenced from Figure 1, GAP has traditionally used a qualita-
tive rather than quantitative process to model animal distributions.
The fundamental difference between these approaches is that GAP
uses known species habitat associations from the literature, exist-
ing databases, and expert knowledge to place species where the
conditions are mapped to exist, whereas the quantitative process
uses mapped land characteristics at known occurrence points to place
species wherever such conditions exist. GAP researchers have gen-
erally believed that sampling for species occurrence has been bi-
ased and grossly incomplete; therefore, the points used in the quan-
titative approach may give incomplete or biased distributions. Some
limited study has suggested this to be true. For relatively rare spe-

cies with good sampling, the quantitative process provides a more
precise result than the GAP process, but for more widely distrib-
uted species (with which GAP is more concerned), the qualitative
method provides a better model (K. Smith, pers. comm.). It is safe
to say that tests of the various modeling methodologies are insuffi-
cient to draw any operational rules, but it appears likely that em-
ploying a suite of methods that can be tailored to each taxonomic
group or guild will generate the most robust results.

The process described by Figure 1 has remained largely unchanged,
which in some ways is disappointing. Virtually all completed as-
sessments of GAP projects’ distributions show an 80% or better
accuracy when compared to checklists for managed areas (Scott et
al. 1993, Edwards et al. 1996); this may suggest the process is do-
ing an acceptable job. Unfortunately, so few independent field-
based data sets exist that assessments of accuracy at finer levels of
spatial resolution have been difficult; however, see Thompson et al.
(1996), Llewellyn and Peterson (in preparation), and Beard et al.
(1999). Work done in Idaho used field observations at 1,628 sites
testing the independent influence of rarity on accuracy of model
predictions. A second effort tested the change in model accuracy
with increasing model complexity (Karl et al., in press [b]). Simi-
lar work is being conducted by GAP researchers in Maine (Boone
and Krohn 1999).

Anecdotal evidence suggests that GAP is indeed doing well in this
revolutionary first effort to map so many species over vast areas
and at a cost of $100-200/species/state. Those using the traditional
GAP qualitative approach are probably doing as well as quantita-
tive modelers with species associated with well-defined environ-
mental constraints such as minimum or maximum elevation. All
modelers have problems finding realistic limits of distributions for
generalists and usually do poorly with species associated with mi-
crohabitat features we cannot map, such as rock outcrops or forest
snags. Species with particular movement and patch-size require-
ments or high sensitivity to habitat quality are also increasing our
errors of commission; however, GAP has a conscious desire to err
on the side of commission rather than omission.

Some GAP researchers are beginning to tackle these problems with
new quantitative methods and data sets (Edwards et al. 1996) or
with the use of landscape ecological principles as part of the model
(Drobney 1999). It is also important to address the diversity of the
taxa presently being mapped by GAP. The original decision to limit
taxa to terrestrial vertebrates has generally been continued by the
states, but only because of financial limitations and not lack of in-
terest. GAP intends to be a complete biodiversity program and
ultimately include all taxa. The inclusion of aquatic taxa has been
a long-standing dream with some false starts and, to date, an inabil-
ity to garner base funding. We have gradually been overcoming
this, first through the New York watershed pilot project (Bain and
Meixler 1998, Meixler and Bain 1999) followed by the Missouri
statewide aquatic program (Sowa 1998, 1999). (Also see updates
on these projects in this issue.) These projects have been building
the methods and protocols to guide future efforts, and many GAP

projects are currently including some limited aquatic aspect and
garnering partner funding to do so. This bodes well for the pro-
gram, but only if base funding can be provided to ensure national
consistency in the methods, systems, and products.

Some GAP projects have begun incorporating other terrestrial taxa
such as ants. In Florida, Craig Allen has mapped geographic distri-
bution of ant species at the county level as determined primarily
from published sources (Allen et al. 1998). In South Carolina, a
sample-based approach to mapping ant diversity was taken. From
the results of these sampling efforts both the county-level distribu-
tion and habitat affinity of each ant species will be determined (Allen
et al. 1998).

Many of these issues were addressed at a recent symposium in Snow-
bird, Utah, sponsored by the USGS Idaho Cooperative Fish and
Wildlife Research Unit in partnership with the U.S. Forest Service,
Bureau of Land Management, U.S. Fish and Wildlife Service, and
Boise Cascade and Potlatch corporations. The symposium, titled
“Predicting Species Occurrence: Issues of Accuracy and Scale” at-
tracted 325 participants from 14 countries. While most papers fo-
cused on vertebrates, other taxa such as plants, fungi, and inverte-
brates were well represented. Areas identified as needing more
work were predicting abundance and viability. The majority of the
presentations dealt with species presence/absence despite aggres-
sive solicitation of papers on predicting abundance and viability.
Tests of accuracy of model predictions using independent field data
were uncommon. Finally, there is a bias worldwide towards pre-
dicting bird and large mammal occurrences. However, there is some
pioneering work on predicting occurrences of plants going on in
Australia. Papers from this symposium will be published by Island
Press in the spring of 2001 (J.M. Scott and P.J. Heglund, eds.).

Future

I’ll now use the experience from the past and present to organize
some thoughts on where GAP is, or should be, heading in the next
ten years to develop better distributions for more taxa.

Taxa: Clearly, an effective biodiversity program must address a
broader representation of taxa (Ricketts et al. 1999) including habi-
tat representation (aquatic) and size/life history representation (in-
vertebrates). The research of Ricketts et al. (1999) suggests, how-
ever, that the mix of taxa used by GAP may likely be a suitable
surrogate of species richness for a much broader group.

The modeling work to date has set the stage for new techniques and
energized both our researchers and data users to desire such im-
provements. Funding is the greatest constraint, and while current
interest is high for increasing GAP’s budget, any foreseeable in-
creases will still fall far short of including the breadth of taxa that
should be addressed. This suggests that GAP must become even
more aggressive in partnering and leveraging funds to achieve greater
representation of taxa in our assessments. Alternatively, at the 1999
national GAP meeting, Malcolm Hunter noted that many countries
are choosing to rely on the nonbiotic surrogate of “enduring fea-
tures,” believing that mapping individual taxa is impractical, too

slow, and/or a waste of resources (Hunter 1999). When asked how
one determines what size or configuration to make a reserve based
on these surrogates, he responded, “for that you need the animals.”
While it may be important for gap analysis to include enduring
features, we need to expand our biotic surrogates and continue ad-
dressing the needs of individual biotic elements to ensure their sur-
vival and thus save their ecosystems.

Methods: Above I introduced the qualitative versus quantitative di-
chotomy, but it is certainly a false one. GAP has never had a re-
quired method for animal distribution modeling; rather we strongly
advocate the use of the best available methods for producing usable
products, and we certainly encourage a research dimension in ev-
ery project. In the near future we will fund comparative studies,
hopefully engaging other researchers and our critics to identify
which models work, for which taxa, and under what conditions.
We also need to pursue more robust models to incorporate viability,
abundance, and metapopulation dynamics (see Vilella and Minnis,
this issue). In this we will again find ourselves limited by the lack
of actual knowledge of species’ life histories, but just as we once
believed GIS data was our limiting factor, it is important to have
the tools ready when the information becomes available.

Biotic knowledge: Despite inroads in quantitative methods, quali-
tative information on species’ life histories is critical to effective
modeling, and we may find this even more so when we learn more
about incorporating concepts like viability and metapopulation dy-
namics. GAP projects consistently find a dearth of published knowl-
edge on species’ basic habitat associations, let alone life histories
required for more robust modeling. Work in Idaho indicated that
the habitat association information in the literature is biased to-
wards large vertebrate species, especially game species (Karl et al.
1999). Some projects have begun to do basic field study to estab-
lish range extents within their states (see Wall et al., this issue), but
the U.S. desperately needs a systematic program for not only cap-
turing information from ongoing field studies but new programs to
study species in their habitats, particularly the poorly studied taxa.
GAP is already playing an important role by compiling nearly the
sum of knowledge of many species into modeling databases and
the maps themselves.

Conclusion

GAP researchers and cooperators have much to be proud of. We
have undertaken the largest, broadest effort in U.S. history to com-
pile our sum knowledge of thousands of species and convert this
knowledge into spatial renditions of their distributions. Assess-
ments, though limited, are demonstrating a consistently high level
of confidence for most species. In this respect the initial GAP ob-
jectives are being fulfilled, and we have the enviable position of
focusing now on expanding the breadth and depth of our endeavor.
Substantial success will not come as easily, however. Significant
funding and partnership increases are required as well as a whole-
sale rethinking of the way data are collected, archived, and distrib-
uted by the entities charged with doing so. Any GAP researcher
can recount visiting agency offices in search of data and being

pointed in the direction of banks of filing cabinets. This must change.

Literature Cited

Allen, C.R., L. Pearlstine, and D.P. Wojcik. 1998. Gap analysis
for ant species.
Gap Analysis Bulletin 7:10-14.

Bain, M.B., and M.S. Meixler. 1998. Making Gap Analysis work
for New York waters: A state perspective on aquatic GAP.
Gap
Analysis Bulletin
7:22-23.

Beard, K.H., N. Hengartner, and D.K. Skelly. 1999. Effectiveness
of predicting breeding bird distributions using probabilistic mod-
els.
Conservation Biology 13:1108-1116.

Boone, R.B., and W.B. Krohn. 1999. Modeling the occurrence of
bird species: Are the errors predictable?
Ecological Applica-
tions
9:835-848.

Csuti, B., and P.J. Crist. 1996. Methods for developing terrestrial
vertebrate distribution maps for Gap Analysis. A handbook for
conducting Gap Analysis. USGS Gap Analysis Program, Mos-
cow, Idaho. http://www.gap.uidaho.edu/handbook/
VertebrateDistributionModeling. Version 2, March 12, 1998.

Drobney, R.D., T. Haithcoat, and D. Diamond. 1999. Missouri

Gap Analysis - Final Report. University of Missouri-Colum-
bia. 198 pp.

Edwards, T.C., Jr., C.H. Homer, S.D. Bassett, A. Falconer, R.D.

Ramsey, and D.W. Wight. 1995. Utah Gap Analysis: An envi-
ronmental information system. Final Project Report 95-1, Utah
Cooperative Fish and Wildlife Research Unit, Utah State Uni-
versity, Logan, Utah.

Edwards, T.C., Jr., E.T. Deshler, D. Foster, and G.G. Moisen. 1996.

Adequacy of wildlife habitat relation models for estimating spa-
tial distributions of terrestrial vertebrates.
Conservation Biol-
ogy
10:263-270.

Hunter, M.L. 1999. Banquet presentation at the 9th Annual Na-
tional Gap Analysis Program Meeting. August 3, 1999. Duluth,
Minnesota.

Karl, J.W., N.M. Wright, P.J. Heglund, and J.M. Scott. 1999.

Obtaining environmental measures to facilitate vertebrate habi-
tat modeling.
Wildlife Society Bulletin 27:357-365.

Karl, J.W., L.K. Bomar, P.J. Heglund, and J.M. Scott. In press (a).

Species commonness and the accuracy of habitat relationship
models. In J.M. Scott, P.J. Heglund, M. Morrison et al. Pre-
dicting species occurrences: Issues of accuracy and scale. Is-
land Press, Washington, D.C.

Karl, J.W., P.J. Heglund, E.O. Garton, J.M. Scott, N.M. Wright,
and R.L. Hutto. In press (b). Sensitivity of species habitat rela-
tionship model performance to model complexity, data resolu-
tion, and scales of application.
Ecological Applications .

Llewellyn, R.L., and C.R. Peterson. In preparation. Testing gap
analysis models at multiple spatial scales: The distribution of
amphibians and reptiles on Craig Mountain, Idaho.

Meixler, M.S., and M.B. Bain. 1999. Application of Gap Analy-
sis to New York State waters. Final project report by the New
York Cooperative Fish and Wildlife Research Unit, Cornell
University, Ithaca, New York, to the U.S. Geological Survey,
Gap Analysis Program.

Ricketts, T.H., E. Dinerstein, D.M. Olson, and C. Louks. 1999.
Who’s where in North America.
BioScience (49)5:369-381.

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

H. Anderson, S. Caicco, F. D’Erchia, T.C. Edwards, Jr., J.
Ulliman, and G. Wright. 1993. Gap analysis: A geographic
approach to protection of biological diversity.
Wildlife Mono-
graphs
123.

Sowa, S.P. 1998. Gap Analysis in riverine environments. Gap

Analysis Bulletin 7:18-20.

Sowa, S.P. 1999. Implementing the aquatic component of Gap

Analysis in riverine environments. Training workbook. Mis-
souri Resource Assessment Partnership, Columbia, Missouri.

Thompson, B.C., P.J. Crist, J.S. Prior-Magee, R.A. Deitner, D.L.

Garber, and M.A. Hughes. 1996. Gap analysis of biological
diversity conservation in New Mexico using geographic infor-
mation systems. New Mexico Cooperative Fish and Wildlife
Research Unit, Las Cruces, New Mexico.

 

Previous | First | Next ->