Physical Diversity

Painstakingly compiled species atlas databases for vascular plants, reptiles and amphibians, breeding birds, and other taxa, were available for analysis, but how useful would they be in a statewide assessment of biodiversity? Not very useful, as it turns out. There are two major problems. First, data are usually spottily distributed across an area as large as a state, and questions of sampling bias can be difficult to dismiss. Second, although data for a given taxon may be of high quality, biodiversity studies have shown that diversity hotspots for different taxa do not often overlap. Only rarely does providing conservation protection for one taxon offer similar protection to a wide range of others.

In the absence of adequate biological information, the VBP made two important assumptions: 1) that the protection of natural communities serves as a "coarse filter" conservation measure for species and genetic resources; 2) that a diversity of natural landscape features best insures a diversity of natural communities. These two assumptions were based on the strong link between the physical environment - bedrock, soils, climate, and topography - and ecosystem pattern and process. Together they provided a theoretical basis for the use of physical diversity as a surrogate for biological diversity.

The biophysical regions of Vermont offered a spatial framework for the analysis of physical diversity. These regions were defined on the basis of broad-scale differences in elevation, climate, topography, soils, bedrock, and vegetation. Disturbance regimes, hydrological processes, soil development, nutrient cycles, and other ecological processes in which species and assemblages of species have evolved vary across these regions. Our objective was to represent as fully as possible the genetic and ecological variation of species, communities, and ecosystems throughout the state. Only by conducting the diversity analysis in each of the biophysical regions could we hope to reach this objective.

Quantifying Landscape Diversity

What is landscape diversity, and how can it be quantified? We used a geographic information system to assemble units of landscape diversity based on elevation, bedrock and surficial geology, and landform. The process is illustrated here for the Northeastern Highlands biophysical region shown at right.


Climate has a determining effect on vegetation distributions at all scales: continental, regional, and local. Climatic sub-regions are defined on the basis of variation in precipitation and temperature, which is itself largely the result of change in elevation and latitude. Furthermore, elevation has been shown to be a powerful predictor of forest composition in Vermont (Siccama, 1971). We used USGS 7.5 minute digital elevation models, grids of elevation values at 30 meter intervals (over 27 million such values define the surface of Vermont), to define five discreet elevation zones with relevance to the distribution of Vermont forest types. Three of those five zones occur in the Northeastern Highlands.

Bedrock Geology

Local bedrock type is a principal determinant of soil chemistry, texture, and nutrient availability. Differences in the erodability of local bedrock types also contribute strongly to the formation of a region's characteristic landforms. We grouped lithological units appearing on the geologic map of Vermont (Doll et al., 1961) into nine classes, each of which may be expected to have particular biological and ecological implications. Most of the Northeastern Highlands is underlain by calcium-bearing phyllites and schists of the Gile Mountain and Waits River Formations (light and dark blue map colors), mostly non-calcareous rocks of the Albee Formation (grey), and plutonic granitic rocks (green).

Surficial Geology

In the absence of statewide soils data, we used information on surficial deposits (Doll et al., 1969) to model differences in moisture retention and nutrient availability. We grouped surficial units into nine classes that reflect the geomorphic origin and texture of the substrate. Notable in the Northeastern Highlands are the wetlands and low kamic rises of the Nulhegan Basin; sand and alluvium deposits along the Connecticut River; broad wetland complexes along several of the region's streams; and extensive areas of outwash and ice-contact features that characterize several other of the regions' stream valleys. All but one of Vermont's nine surficial classes occur in the region.


We also used the USGS elevation data to model topographic units such as ridges, sideslopes, and coves on the landscape. As landform varies, so does ecological process; one author called landform "the anchor and control of terrestrial ecosystems." Ancillary hydrographic and wetland data were incorporated into the landform layer to model water bodies and wetlands. You can examine landforms in an area of varied topography in the central part of the Northeastern Highlands by clicking on the small image at right [or something like that].

These four layers were constructed as raster data sets; just as with the digital elevation models, information is attached to each cell in a 30-meter grid that covers the state. We combined the layers to make a grid of "landscape diversity units" (LDUs). For each 30-meter cell in the LDU grid there is a five-digit code that represents some combination of elevation zone, bedrock type, surficial class, and landform. Contiguous cells with the same five-digit label make up a discrete LDU.

Analyzing Landscape Diversity

In order to identify areas of high physical diversity, and to uncover patterns of regional diversity, we used the LDUs in the same way that we might have used data on species distributions. We laid a grid of hexagonal sampling cells over each biophysical region. Examples here are the Northeastern Highlands and the Valley of Vermont. Most of Vermont's biophysical regions are elongate in the north-south direction, and grids were custom-built to best fit region shapes. Each cell represented 5% of the area of the region it overlaid. We programmed the geographic information system to select the six cells that together accounted for the highest number of LDU labels that occur in the region. The objective was to capture as much physical diversity as possible within the least area. A closer look at the way the selection algorithm works is available by clicking on the small image at right [link to algorithm figure].

We ran the selection algorithm three times in each region, using three offset sampling grids. This ensured complete sampling coverage near region borders, and performed a check on the stability of possible cell selection solutions.

Converting hexcell selections into a set of polygons that efficiently captured physical diversity gradients was largely an analog process. A sheet of drafting film was laid over each of the four LDU component grids in turn, and lines in four different colors were drawn around natural features that appeared to have driven the selection of a given cell. All the different colored polygons were then synthesized into a set of polygons that optimally represented physical diversity in the region. We called these polygons "representative landscapes" (RLs).

Recognizing the need for thorough documentation, we recorded primary information components for each representative landscape polygon, as well as details on the bedrock, surficial, elevation, and landform information each incorporated. A sample of polygon documentation is available by clicking in the shaded area of the map at right.