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

Previous | First | Next ->

Landscape Diversity as the Basis for a Reserve
Design Initiative in Vermont

Spatial Analysis Laboratory, School of Natural Resources, University
of Vermont, Burlington

The Vermont Biodiversity Project has been conducted hand-in-hand
with the Vermont/New Hampshire Gap Analysis Project but at finer
resolution and with an objective of using measures of landscape
diversity as indicators of community and species diversity. There
is ample rationale in the published literature for this approach. One
line of support comes from evidence that biological diversity is pre-
dicted well by physical diversity, and that this relationship holds at
various scales (e.g., Lapin and Barnes 1995, Burnett et al. 1998,
Nichols et al. 1998). Another argument for physical diversity is
that land conservation efforts should consider temporal scales that
acknowledge shifts in ranges of species and communities in response
to ecosystem processes and changes in climate (Hunter et al. 1988,
Hunter 1991).

In the Vermont Biodiversity Project, our goal was representation.
We sought to delineate a system of potential reserves that included
all elements of landscape diversity. We began with an analysis of
biological (tree species) and physical data (climate, geology) that
led to the delineation of seven biophysical regions in the state. Within
each region we then employed a methodology that assured repre-
sentation by most or all elements of physical diversity.

We were constrained by the need to use data that were available for
the entire state. Soils or ecological land types would have been
ideal, but neither has been mapped statewide. The only consistent
sources of data were digital elevation models (DEMs), bedrock
geology, surficial geology, hydrography, and wetlands. From these
spatial data, we derived Landscape Diversity Units (LDUs) and
sampled for richness and representation.

Landscape Diversity Units
The most significant component of landscape diversity was the deri-
vation of landforms (Fels and Matson 1997). Using 30-meter DEM
data, we first derived five slope classes. We then calculated a land-
scape position index (LPI) for each DEM point using a focal func-
tion. This function assigns to each cell the mean of the distance-
weighted differences in elevation between that point and all other
elevation points within a specified radius. The LPI was “calibrated”
for each region and divided into discrete classes that described rec-
ognized landforms. A matrix of slope classes by LPI classes pro-
duced 14 landform descriptions that could be mapped. Then we
added four other categories from existing data, one for wetlands
and three for classes of surface waters, for a total of 18 landform

Also from DEM data, we defined five elevation zones with recog-
nized links to the distribution of forest types in Vermont. Addi-
tional data came from digital maps of bedrock geology (1:250,000)
and surficial geology (1:62,500). In both cases, we developed an
ecologically-based crosswalk that reduced the number of classes to
nine bedrock categories and eight surficial units. The final step
was to overlay these four GIS coverages and to label each pixel of
a 30-meter by 30-meter grid with codes for landform, elevation zone,
bedrock type, and surficial class. This code—a product of four
descriptors of the landscape—is referred to as a Landscape Diver-
sity Unit (LDU). Although there are several thousand potential
LDU labels, there were only several hundred actual labels in each
biophysical region. The mean size for LDUs was 3.9 ha; the range
was from 0.09 ha to 1,957 ha. The smallest size represented single
cells that were anomalies of GIS overlays and were filtered from
the database.

Representative Landscapes
We used grids of hexagonal cells to sample LDUs in each biophysi-
cal region. Each sample cell represented 5% of the area of the
region, a decision based on a suggested minimum size for a func-
tional reserve. Grid cells were custom-built for each biophysical
region and were shaped to best approximate configuration of the
region; they ranged from 26,000 to 60,000 ha among the regions.
We used an algorithm based on richness and complementarity to
select a set of hexcells that most efficiently represented landscape
diversity. The first cell selected was the richest, followed by five
more cells that represented the best complements, such that the six
cells selected gave the most efficient representation of LDUs (Fig-
ure 1). The decision to stop at six sample cells was again related to
practicalities of reserve design; here the rule was based on the no-
tion that it would not be realistic to identify more than 30% (six
cells) of the landscape as potential reserve.

GapBulletin853-00.jpg 274x356
Figure 1. Six grid cells in Northeastern Highlands, Vermont, sampled to
maximize richness of landscape diversity units. In this example, 477 of
586 LDUs are represented.

GapBulletin853-01.jpg 271x354
Figure 2. Representative landscapes of Vermont.

 Hex grids overlapped boundaries of biophysical regions, so some
cells were effectively smaller than others. We compensated for this
by shifting the sample grid twice; thus we derived three solutions
of complementary sets of cells for each region. We resolved these
different solutions–and imposed a reality check on the results–by
producing transparent overlays of cells selected in different runs of
the algorithm and aligning these with paper maps of each of the
layers comprising LDUs. That is, we sequentially overlaid differ-
ent efficient hexcell solutions with maps of landforms, bedrock,
surficial materials, and elevation zones. At each stage, we manu-
ally outlined polygons of significant features on these map layers;
these were later digitized on screen and smoothed along bound-
aries of landscape features. We refer to these polygons as Repre-
sentative Landscapes (RLs; Figure 2).

RLs were quite efficient in accounting for the diversity of physical
features in each of the seven biophysical regions. By drawing land-
scape-based polygons rather than maintaining sample hexcells, we
selected from 17 to 26% of each region rather than the 30% repre-

sented by six sample cells. When results from the seven regions
were combined, 22% of the area of the state was featured as RLs.
This area accounted for 89% of LDU richness and 95% of propor-
tional representativeness, or similarity. By comparison, a random
sample of 22% of the state, using hexcells sized as the average of
RLs, accounted for only 77% of LDU richness. Fifty percent of the
state would have to be sampled randomly to equal the LDU rich-
ness captured in our RLs.

Biological Representation and Applications
We have not yet been able to compare results of our landscape di-
versity analysis to the distributions of vertebrate species predicted
by Gap Analysis and assess the degree to which physical diversity
predicts diversity of vertebrate habitat. However, we have done a
biological assessment using an extensive collection of atlas data on
a number of plant and animal taxa. Although many of these atlas
surveys were incomplete or showed sampling biases, we were able
to compare statewide records for 1,617 species referenced to the
251 towns in Vermont. Eighty-three percent of the species were
represented by records that corresponded with representative land-
scape polygons; we captured only 76% of the species records by
random samples of an equivalent area.

We believe that there are a number of ways that analyses of physi-
cal diversity such as we have described might be useful in Gap
Analysis. In the eastern United States, where land cover is often
shaped more by historical land use than by natural processes, land-
forms—as we have derived them—might serve as better predictors
of habitat diversity than maps of land cover derived from Landsat
imagery. Furthermore, landforms can be used as a means of en-
hancing interpretation of such imagery or used in combination with
land cover maps to predict the location and extent of natural com-
munities. Some preliminary efforts at such predictions of natural
communities in New Hampshire have been promising (M. Ander-
son, TNC, Boston, personal communication).

We hope this note will encourage others to further investigate the
use of physical diversity as a means of representing biological di-
versity. Given the dearth of information on the distribution of spe-
cies and communities, the uncertainty of making predictions of these
distributions, and the importance of considering long-term processes
in the establishment of nature reserves, landscape diversity may be
a cost-effective means for planning conservation of biological di-

Literature Cited
Burnett, M.R., P.V. August, J.H. Brown, Jr., and J.T. Killingbeck.

1998. The influence of geomorphological heterogeneity on
biodiversity: I. A patch-scale perspective.
Conservation Biol-

Fels, J.E., and K.C. Matson. 1997. A cognitively-based approach
for hydrogeomorphic land classification using digital terrain
models. Proceedings, Third International Conference on Inte-
grating GIS and Environmental Modeling. National Center for
Geographic Information and Analysis, Santa Barbara, Califor-

Hunter, M.L., G.L. Jacobson, and T. Webb III. 1988. Paleoecol-
ogy and the coarse-filter approach to maintaining biological di-
Conservation Biology 2:375-385.

Hunter, M.L. 1991. Coping with ignorance: The coarse-filter strat-
egy for maintaining biodiversity. Pages 266-281 in K. Kohm,
editor. Balancing on the brink of extinction. Island Press, Wash-
ington, D.C.

Lapin, M., and B.V. Barnes. 1995. Using the landscape ecosys-
tem approach to assess species and ecosystem diversity.
servation Biology

Nicholls, W.F., K.T. Killingbeck, and P.V. August. 1998. The in-
fluence of geomorphological heterogeneity on biodiversity. II.
A landscape perspective.
Conservation Biology 12:371-379.

<-Previous | First | Next ->