PENNSYLVANIA GAP ANALYSIS:
A Progression

Gap analysis is a preliminary process of prospecting for habitats which accommodate numerous species but have not been accorded commensurate concern for conservation. These constitute "gaps" in commitment to conservation where sustainability is suspect. GAP is also an acronym for the national Gap Analysis Program orchestrated by the U.S.G.S. Biological Resources Division, which is a state-by-state series of gap analysis projects having parallel protocols.

The national venture originated with the U.S. Fish & Wildlife Service from initial work by J. M. Scott in Hawaii followed by a prototype effort in Idaho. It was then shifted to the transient National Biological Survey/Service, and thence to the U.S. Geological Survey. Gap analysis protocols have developed in an evolutionary fashion by accommodating adaptation on the part of projects. Although not among the first wave of state projects, Pennsylvania's gap analysis initiative was started early enough to have substantial latitude for experimentation. Innovation has been a hallmark of the Pennsylvania project, whereby it has taken a trajectory that entails elements not typically encountered elsewhere. It is thus appropriate to outline the nominal components of a gap project as a backdrop for sketching specifics regarding Pennsylvania, and why Pennsylvania's approach is seen more as a progression than a particular package of products. The national GAP homepage http://www.gap.uidaho.edu/gap/ offers further insights regarding the scope of GAP across the country, and also has links to homepages for other state gap analysis projects.

A poster entitled; Systematic Analysis of Habitat Availability for Conservation Planning in Pennsylvania is being presented at the National GAP Conference.

After a brief outline of the GAP process, we then give an overview of present and prospective Pennsylvania products including:

• information on ranges of species;
• maps of conservation lands;
• satellite PHASEviews of landscapes;
• land cover/vegetation;
• habitat profiles;
• habitat mappings;
• habitat diversity;
• a framework for landscape ecological mapping.

The PASDA web-based repository for Pennsylvania spatial data is likewise introduced.

The core corps of investigators for the Pennsylania gap analysis project has been:

Dr. Wayne L. Myers, Principal Investigator
Dr. Robert Brooks, Co-Principal Investigator
Dr. Gerald Storm, Co-Principal Investigator
Joe Bishop, Graduate Research Assistant

Several others have been involved in the project for more limited periods of time in special roles. W. Myers serves as a contact point, and will relay referrals as appropriate. Contact details are:

Nature of the GAP Approach to Nature

GAP generalizes geographically via GIS generated maps and mappable layers of information. GIS stands for Geographic Information Systems which are computerized capabilities for managing and manipulating many mappable datasets that might be logical layers in a stack of map transparencies or overlays. Layers are also called coverages or themes. While the mapping mode lends spatial specificity, GAP tends to generalize by finding features that are particularly prominent. It has thereby been seen as a "coarse filter" strategy seeking to spin a thin conservation safety net that could capture outstanding opportunies to stabilize shaky sustainability for a spectrum of species over sizeable sectors of countryside. It complements conventional conservation of critically constrained components of habitat for species already at risk. GAP thus works at synoptic spatial scales in hopes of husbanding habitat to foster functional depth of diversity in every ecosystem instead of persistently prioritizing panic preservation of biota on the brink. GAP is rendered relatively robust to incidental irregularities in the data by virtue of its broad scope.

Certain coverages are crucial to gap analysis. Since circumstances of current conservation are central to gauging gaps, a so-called STEWARDSHIP LAYER must be compiled. The stewardship layer links location with land tenure, legal status of provisions for protection, and mode of management relative to habitat.

Land cover is the single most important consideration regarding habitat for a host of species, so a LAND COVER LAYER plays a major role in a gap analysis process. General cover types such as forest, herbaceous, and water lend overall character to the habitat. Species composition and stand structure may or may not be major considerations, depending on the species in question. Other physiographic features can likewise have greater or lesser importance for particular species.

Knowledge of WILDLIFE HABITAT RELATIONS makes land cover and related environmental features interpretable in terms of habitat for species under study. WHR models specify minimal environmental circumstances for supporting each species, and thus enable mapping of areas that are seemingly suitable for the species. Habitat determinations via WHR models are typically tentative because they often assume the presence of some factors for which detailed data is difficult to obtain.

RANGE MAPS are required to constrain WHR models at the known limits of distribution for species. This is because WHR models typically examine vicinities, whereas determinants of range may be regional or historical. It has become a GAP convention to cast ranges in terms of 635 sq. km. hexagons used by the U.S. Environmental Protection Agency (EPA) in its EMAP work.

GAP treats the range-constrained, WHR-modeled habitats as surrogates for breeding occurrence of species. These surrogates are overlaid in GIS for analysis of diversity as determined by (apparent) habitat. Major habitat complexes are considered both in terms of total diversity and complementarity. Important areas of uncertain stewardship are designated as gaps in the regional conservation framework. GAP has had an initial focus on vertebrates, with a presumption that major habitats for these taxa would also support a spectrum of other taxa.

Pennsylvania Progression in Scale and Detail

Scale and detail have been prominent among the topics of concern for gap analysis, and they are not readily separable. The scale of mapping originally promoted by national GAP was 100 hectares. This means that 100 hectares (ha) or 1 sq. km. is the minimum mapping unit (mmu), so that areas smaller than this are merged with adjacent ones rather than being separately delineated on a map. This is slightly less than 250 acres, and does indeed constitute a coarse look (or coarse filter), especially for much of the eastern portion of the United States where land management typically deals in smaller parcels. While it is clear that more spatial detail (larger scale/finer grain) is needed for Pennsylvania, it is also of interest to know what sorts of patterns emerge at such a coarse scale. Toward this end, Pennsylvania has chosen to conduct investigations at more than one scale. These scales range from EPA hexagon units encompassing 635 sq. km. to the 0.09 ha size of resolution elements in a Landsat satellite image. Notably, interest on the part of national GAP is also no longer restricted to the 100 ha level.

Pennsylvania Range Mapping

Pennsylvania initially started compiling vertebrate distribution information by digitizing published range information. That work had only begun when national GAP formed a partnership with EPA and The Nature Conservancy (TNC) to compile range (distribution) information on the basis of EMAP hexagons for vertebrates and selected other taxa. Pennsylvania and Oregon were the first two states to be done under this partnership. There being no point in duplication of effort, the TNC work on EPA hexagons has become the (range) distribution database for Pennsylvania gap analysis.

TNC reviewed available evidence and made a determination regarding apparent presence or absence for each organism in each hexagon according to weight of evidence for breeding occurrence. Three weights of evidence in support of occurrence were designated as confident, probable, or possible. Confident corresponds to a subjective certainty of 95% or greater, probable to a range from 80% to 95%, and possible to a range from 10% to 80%. In Pennsylvania, we have dubbed this as "hexpectation" of occurrence. TNC provided this information to Pennsylvania in a dBase compatible format with each record representing occurrence of a given species in a particular hexagon. Several tabulations by species and level of evidence are contained in a Pennsylvania Gap Analysis 1995 interim progress report. Hexagon data for Pennsylvania mammals was analyzed by Kyle Joly in a master of science (ecology) thesis titled "Mammalian Biodiversity in Pennsylvania at the USEPA 635 Square Kilometer Hexagonal Scale." The occurrence data for Pennsylvania can be made available by special request.

Satellite-Based PHASEviews of Pennsylvania Landscapes

The national GAP guidelines indicate that remotely sensed image data from a thematic mapper (TM) sensor on the NASA/EOSAT Landsat satellite should serve as a basis for mapping land cover. National GAP joined with several other agency interests in an MRLC consortium to cofund purchase of these data for use by state projects. There are 10 scenes or frames needed to cover Pennsylvania. Each scene is a grid of cells (or pixels) that are 30 meters on a side, with nearly 50 million cells per scene. Due to the 30 meter resolution element, one cannot discern individual trees or even most buildings in such imagery. However, landscape level things such as forests and fields along with major roads and rivers are evident. The satellite data as originally obtained is proprietary, and therefore cannot be made available directly to the public. The original data also requires very high capacity storage media.

The Khoros Group at Los Alamos National Laboratory pioneered a way of condensing multiband satellite image data as "hyperclusters" for use with special Spectrum computer software. The EROS Data Center at Sioux Falls, SD distributes the Spectrum hypercluster version for a nominal charge, but it takes substantial savy of computer systems to install and operate the Spectrum software or convert those hypercluster data for use with other software systems. A Pennsylvania innovation is a more comprehensive concept of hypercluster compression called PHASES. PHASE compression produces a condensed image version that is directly usable with common GIS facilities and for which freeware viewers are available to use with Pentium-based PC computers. The full suite of PHASES software facilities is also made available without charge via the World Wide Web for those who may have access to original multiband image data that they might like to condense for redistribution.

PHASE-condensed renditions of ten Landsat (TM) based scenes covering Pennsylvania have been produced for distribution in three formats. These were labelled as "TerraByte" products before it was discovered that TerraByte has also been used as a label elsewhere. A TerraByte "professional" CDROM containing the scenes in "grid" format for ESRI's commercial ArcInfo and ArcView software was first produced with promotial sponsorship by Digital Equipment Corporation (now merged with Compaq). A second "PaExplorer" edition was produced with similar sponsorship having the scenes in BSQ image format for use with ESRI's ArcExplorer freeware viewer (included on the CDROM) as well as for ArcView. Either of the CDROMs is available for $5.00 cost of shipping and handling. Still a third version is available via the WorldWide Web by county in GeoTIFF format from PASDA. Both CDROMs (as well as PASDA) also have supplementary information such as roads for viewing the images in landscape context.

Having an image backdrop against which to map other features is important to conservation considerations because it provides a context of landscape that is otherwise largely lacking. The central ideas of the emerging science of landscape ecology can then be brought into the picture informally without undertaking a protracted computational campaign involving multiple metrics of landscape pattern. The human eye is adept at picking up patterns in images, and spatial patterns involving connectivity and corridors are fundamental to landscape ecology.

Pennsylvania Land Cover Comparatives

Pennsylvania gap analysis maps land cover in different ways and differing levels of spatial and categorical detail even from the same data source. This Pennsylvania "comparadigm" addresses ecological hierarchy theory that expects environmental processes operating over different distances to find expression in landscape patch patterns at particular scales. Comparing different dates and sources of data entails both spatial and temporal components of change. Even the perception of what constitutes land cover changes with scale, and thus differing mobility of species demands commensurate scaling. Degree to which pattern differs between sources, scales, and modes of mapping is perhaps more enlightening than supposedly absolute measures of map accuracy which are really relative to issues like minimum mapping unit and category definitions.

The Pennsylvania "broad brush" or "first cut" at land cover is a split into "naturalistic" versus "humanistic" types with 100 hectares (one square kilometer) as a minimum mapping unit. The underlying assumption is that Pennsylvania soil and climatic regimes lead naturally to forests and waters, with herbaceous and otherwise nonvegetated areas being transient disturbances that are currently caused and/or perpetuated largely by humans. Complete consistency in this respect would have treated reservoirs as being humanistic as well, but this would have entailed investigating original streambed locations within impoundments. Major waterbodies of Pennsylvania are well mapped, and can be segregated by computer overlay if desired. This coverage was created by displaying the satellite imagery on a computer screen and using a mouse to delineate large humanistic holes in a naturalist landscape matrix or vice versa. GIS software was then invoked to "dissolve" or obliterate any islands/slivers less than 100 hectares imbedded in a contrasting matrix. This coverage is included on the "Terrabyte Professional" CDROM (see section on Satellite-Based PHASEviews).

Ecologically, this coarse coverage divides Pennsylvania into regions having a forested landscape matrix versus regions having a nonforest landscape matrix. Thus naturalistic regions have nonforest patches imbedded in forest, whereas humanistic regions have forest patches imbedded in nonforest. This coverage is therefore considered to be LANDSCAPE MATRIX theme. A salient finding from mapping in this very generalized mode is that a single naturalistic polygon encompasses more that 65% of Pennsylvania. It contains much of northern Pennsylvania, extending to the southern border through the central mountains and projecting into the southeast along the fingerlike ribs of the "ridge and valley" region. There are thus major orders of patch patterning in Pennsylvania. At this level, "fragmentation" of Pennsylvania is quite concentrated. There is predominant connectivity in all but the southeast, major urban areas, and large agricultural valleys. Even some of the major urban (Philadelphia and Pittsburg) and urbanizing (Poconos) areas have considerable tree cover in a "bird's-eye" view.

Given availability of waterbody/watercourse and floodplain coverages either on CDROM or via PASDA, an obvious finer scale question concerns urbanization. Mapping of high intensity (commercial/industrial) and low intensity (strip and suburban/residential) urbanization would allow assessment of urban influence on forested landscapes, and would further allow landscapes with nonforest matrix to be broken into rural versus urbanized subregions. Accordingly, a second mapping was conducted for this purpose. Delineations were again done with a mouse on a computer display of satellite imagery, but a roads coverage was superimposed to facilitate interpretion. This URBANIZATION map theme is contained on the PaExplorer CDROM in shapefile format.

When the LANDSCAPE MATRIX and URBANIZATION themes are coupled analytically in various ways and superimposed on a PHASEview backdrop of satellite imagery, they provide a Pennsylvania landscape framework within the context of which finer scale information can be more meaningfully considered. The next concern is for finer scale land cover data other than urbanization. In Pennsylvania this has been addressed primarily in terms of vegetation physiognomy. The PHASE approach to multiband satellite image data condenses each scene into 255 types of cells with respect to strength of reflected energy in the several spectral "bands" as seen by the satellite-borne sensor. The "magic number" of 255 corresponds to the number of distinctions than can be made by using one byte (8 bits) of computer media per cell, which is the way such media is conventionally parceled. Each of the 255 types (called clusters) is distributed in a patchy pattern across the scene. Since sectors of the terrain having similar cover will also have similar spectral "appearance," each of these anonymous clusters tends to represent a more or less specific kind of cover. Classes of clusters can therefore be recognized that correspond to "meaningful" categories of land cover.

Our Pennsylvania progression uses eight categories of land cover corresponding to presence or absence and physiognomy of vegetation. These are:

Water surface

Conifer forest

Mixed conifer & broadleaf forest

Broadleaf forest

VegPlex/transitional (admixture of several types on microscale)

Perennial herbaceous

Annual herbaceous (nonforage crops)

Terrestrial unvegetated (rock rubble, pavement, structures, etc.)

corresponding to a spectral "ordination" of increasing aggregate reflectance. The spectral ordering causes more propensity to confuse categories that are listed successively than otherwise. For instance, conifer forest on the sunny side of a ridge may have similarities to mixtures of conifer and broadleaf on the shady side.

Each of the 255 clusters in each of the ten scenes has been tentatively assigned to one of the eight classes, although some scenes have clouds that are distinctive enough to be segregated (cloud 9; cloud shadow or image anomaly 0). There are thus 2550 clusters in all. On the "Terrabyte Professional" CDROM, each of the 10 scenes has a companion table that assigns each cluster (row) to a cover class. Those having relational GIS capability for grids can use the table to render cluster maps in terms of cover class. When the cluster assignments have been checked for scene-to-scene consistency along with spot checks in the field, then a separate (30 meter) cover grid will be generated as a set of tiles. In the interim, one would be obliged to use the tables and do tiling or seaming as appropriate for their area of interest. Just by examining fidelity of cover pattern to landscape pattern of the image, we feel that these assignments are superior to ones made by MRLC on essentially the same images. This contention does, however, remain to be documented. Aerial videography has been collected for purposes of checking accuracy of the land cover grid.

It is desirable to be able to work with land cover themes in either grid or polygon format. This is especially true for conducting habitat modeling and analyzing habitat richness and complementarity. Although our computer systems and software are reasonably sophisticated by most standards, the 30 meter resolution has proven to cause information overload when it comes to converting the grid version to polygon form. Feasibility studies have been conducted relative to generalizing the grid data spatially. It is feasible to do polygon conversion after degrading to a 2-hectare level of resolution, but the polygons for a single scene still number in the millions. If some of the categories are collapsed, then polygons become much less numerous. For example, combining the perennial herbaceous and annual herbaceous categories gives patches consisting of several farms as opposed to segregating by groups of fields within farm. For purposes of our habitat modeling, this degree of spatial grouping is of little consequence. We may likewise combine conifer and mixed forest categories, since there tends to be gradation and confusion between these categories anyhow. It has been determined that there will be a 2-hectare polygon (vector) product, but a final decision is yet to be made with regard to category combinations.

A further level of cluster classification is pending that would attempt to break down forest cover into compositional groups with respect to species. Existing state forest maps and videography would be used to guide this effort. Such a breakdown is, however, viewed as being premature relative to the Pennsylvania progression. Our habitat models are configured for the more general physiognomic level, and thus do not require the compositional breakdown. Further, a new generation of state forest type mapping from higher resolution airphotos is now underway along with a companion ecological zonation (ECOMA) for Pennsylvania. These will provide for much more incisive compositional mapping in a matter of a year or two, besides which the most recent of our satellite data is already five years old. Consequently, the prospective compositional mapping in conjunction with first generation gap will be provisional at best and soon obsolete anyhow.

Pennsylvania Land Stewardship Status

The Pennsylvania land stewardship status layer showing current conservation areas is in the final stages of compilation and metadata creation. This is a fairly conventional compilation that was started earlier by others as a prelude to site selection for a low level nuclear waste disposal facility, and has been carried forward for gap purposes. Pennsylvania cadastral information is generally of a rather low accuracy order by virtue of its "metes and bounds" legacy and the preliminary nature of many GIS ventures in the Commonwealth. Boundary maps for adjacent ownerships often contain inconsistencies that are not evident until someone does a digital comparison in the emerging GIS environment. A compiler such as the gap group can notify the concerned parties of inconsistencies detected and can do cosmetic harmonization for mapping purposes, but has no legal authority to even undertake resolution of conflicts.

Pennsylvania has vast areas of public forest holdings by various agencies, with abutting jurisdictions being quite common. Boundary inconsistencies involving public agencies are typically not high on the priority list for agency action. It is usually only when an agency undertakes to upgrade its GIS using global positioning systems (GPS) that minor but noticeable inconsistencies get resolved. Several Pennsylvania agencies are now engaged in such upgrades, but the timetables for this work tend to be indefinite. Several municipalities in Pennsylvania have protected forest areas in which the watershed role is paramount. Boundary information for many of these consists of paper maps that lack georeferencing entirely. Using GPS to establish georeferencing for these as a preliminary to digitizing would be a major undertaking in itself. There are thus numerous reasons why the current stewardship layer should be considered provisional, and it is very important to find a custodian for this layer that is willing to do updates as the Commonwealth develops better cadastral infrastructure.

Models of Wildlife Habitat Relations (WHR)

Pennsylvania gap analysis addresses species of vertebrates that breed in the state. Our basic intent is to refine the TNC distribution database spatially in a manner that removes portions of "occupied" hexagons that are unsuitable in terms of one or more major habitat requirements. Character of land cover is the main criterion for most species, with special features such as streams, lakes, or elevation settings being taken into account as appropriate. Generalization of land cover will be at the 2-hectare level for most species, since smaller segments become more likely to constitute debilitating sinks of wasted (mostly unsuccessful) reproductive potential. Patches of habitat at this level can be considered relative to landscape matrix situation at the 100-hectare level for species that are purported to be particularly area sensitive. Habitat can then be mapped directly, or one can consider units such as hexagons or counties in terms of habitat proportion. This permits at least regional rankings for individual species when deliberating on conservation priorities, and provides a framework for more intensive synoptic studies such as are currently being conducted on bobcat and woodcock.

Avian habitat models are essentially complete and ready for review. Birds account for roughly half of Pennsylvania vertebrates. Mammals are quite well known with respect to habitat requirements. Herps are somewhat less well documented, and thus merit somewhat more attention than mammals. Being exclusively aquatic, Pennsylvania fishes are the most problematic. National GAP is also engaged in several special studies to investigate promising protocols for dealing with aquatics. A major database on occurrence of fishes has been assembled by scientists at Penn State University. A geographic database on small watersheds (numbering in the thousands statewide) has also been recently completed for Pennsylvania. Gap analysis of fishes for Pennsylvania will be exploratory, and involves overlaying small watersheds on the fish occurrence database to determine major areas of occurrence in terms of small watershed units. Adjoining watersheds in the same major drainages will then be tagged more speculatively based on stream and/or waterbody characteristics. Land cover proportions for the small watersheds will also be considered as a way of helping to inform the speculation. Later studies of particular drainages can then build upon this base. Pennsylvania is particularly interesting with regard to aquatics by virtue of being split on a continental basis by Atlantic and Mississippi drainages, as well as having both Great Lakes and Chesapeake influences.

PASDA: Spatial Data Access

Pennsylvania gap analysis has a ready repository for its data products. PASDA is a World Wide Web based system that supports search, display, and retrieval of FGDC (Federal Geographic Data Committee) standard systems (GIS) data, and imagery related to Pennsylvania's Environment. PASDA is a collaborative effort of the Pennsylvania State University and the Pennsylvania Department of Environmental Protection (DEP). The Pennsylvania State University is developing PASDA with major funding from the Pennsylvania Department of Environmental Protection. PASDA is being developed in accordance with guidelines established by the Federal Geographic Data Committee for nodes of the National Spatial Data Infrastructure (NSDI). Access to additional Pennsylvania datasets is available through the Maps Room of Pattee Library, Penn State, University Park, and from DEP. Three Penn State units are involved in the development of the PASDA system. Penn State's Environmental Resources Research Institute (ERRI), has the lead responsibility for project management; the Maps Room of Pattee Library, Penn State, University Park performs metadata preparation; and the Deasy GeoGraphics Laboratory, a unit of Penn State's Department of Geography, is contributing web interface design, and development of data search and delivery capabilities. Please visit PASDA at http://www.pasda.psu.edu and do a little browsing.

Echelon Analysis of Diversity

Compiling different habitat maps in an overlay mode on a cellular grid gives a "topography" of diversity that constitutes a virtual surface corresponding to a digital elevation model. The usual way of depicting such a surface is by contours, and analysis is most often done in terms of absolute highs and lows. Geostatistics could be used, but entails some major assumptions about regionalization in the surface. One can also just ignore the surface nature and look for combinations of cells that capture the most diversity in various senses of "most." The Pennsylvania approach to this involves a novel concept that we call "echelons" which focuses on hillforms of the virtual surface in terms of saddles, fallzones, and peaks.

Echelons objectively divide the "surface" into a genealogy of hills giving rise to other hills that capture "mountain ranges" of diversity. This approach reflects the Pennsylvania project's subtitle, which is "biotopa" for BIOtic TOpologies of PA. Kyle Joly has taken the echelon approach to distribution data for mammals of Pennsylvania in a master's project. The approach continues to evolve both in terms of software and presentation, and much of Pennsylvania's final analysis will unfold in this manner. We have an ECHELON WEB-BASE that gives details of the concept and allows software to be downloaded without charge.

Biogeography and Pennsylvania ECOMAP

Where physiography strongly influences both land cover and land use as it does in Pennsylvania, the ultimate gap analysis seems to lie in a mapping of terrain units that have natural biogeographic integrity to the degree that entire landscape patterns are characteristic of an area which enables prediction of biotic propensities for many taxa of both plants and animals. Persisting with progression, Pennsylvania gappers and partners are providing leadership in an ecomapping endeavor for Pennsylvania (ECOMAPPA). Wayne Myers has taken a fall 1998 sabbatical to work with the Pennsylvania Department of Conservation and Natural Resources (DCNR), Bureau of Forestry in extending the national ECOMAP effort promoted by the US Forest Service to landscape levels and training district foresters to map the two million acres of state forest land in this manner. The Allegheny Nation Forest (ANF) has pledged cooperation with this endeavor, which encompasses hundreds of thousands more acres. With progress of this order, we should be able to enlist the help of other agencies and interests in extending the landscape ECOMAP thrust across the remainder of Pennsylvania.

Creating a Cooperative Framework for Conservation

Gap analysis is intended to further the conservation agenda among its clientele. Accordingly, Pennsylvania gap analysis is one of the convenors for the 1998 Pennsylvania Biological Survey (PABS) Conference entitled "Conserving Pennsylvania's Natural Diversity: Creating a Cooperative Framework for Action." Among other things, this conference sets the stage for reception of the Pennsylvania gap analysis report and utilization of gap analysis products with a view to their upkeep. Preconference workshops provided input from agencies regarding their conservation purviews as backdrop for an array of workshops. This sort of conference is about focusing, energizing, and synergizing conservation efforts in the Commonwealth. Very importantly, conservation is considered in the context of economics and development rather than as an alternative. Pennsylvania's landscapes have been extensively humanized over the course of centuries, even before colonization. Natural diversity of the present attests to its resilience. Concern for the new millenium is to further harmonize humanization with natural processes at scales that will sustain functional diversity and ecosystem health so that temporary setbacks to one component can continue to be compensated by another.

Pertinent Papers

Joly, K. 1996. Mammalian biodiversity in Pennsylvania at the USEPA
635 square kilometer hexagonal scale. Master of Science thesis,
The Pennsylvania State University, The Graduate School, Inter-
college Graduate Degree Program in Ecology, University Park, PA.

Myers, W., R. Brooks, G. Storm and J. Bishop. 1995. Pennsylvania
gap analysis report - 1995. Technical Rept. ER9602, Environmental
Resources Research Institute, The Pennsylvania State University,
Univ. Park, PA.

Myers, W., G. P. Patil and K. Joly. 1997. Echelon approach to areas
of concern in synoptic regional monitoring. Environmental and
Ecological Statistics 4:131-152.

Myers, W., G. P. Patil and C. Taillie. 1995. Comparative paradigms
for biodiversity assessment. pp. 67-85 in: T. J. B. Boyle and
B. Boontawee (eds.). Measuring and Monitoring Biodiversity in
Tropical and Temperate Forests. Proceedings of IUFRO Symposium
held at Chiang Mai, Thailand, Aug. 27-Sept. 2, 1994. CIFOR Center
for International Forestry Research, Bogor, Indonesia.

Scott, J. M., F. Davis, B. Csuti, R. Noss, B. Butterfield, C. Groves,
H. Anderson, S. Caicco, F. D'Erchia, T. C. Edwards, J. Ulliman,
and R. G. Wright. 1993. Gap Analysis: a geographic approach to
protection of biological diversity. Wildlife Monographs, No. 123.

Thorne, S., K. C. Kim, K. C. Steiner and B. J. McGuinness. 1995.
A heritage for the 21st century: conserving Pennsylvania's native
biodiversity. A report by the Pennsylvania Biodiversity Technical
Committee. Published by the Pennsylvania Fish & Boat Commission,
Bureau of Education & Information, P.O. Box 67000, Harrisburg, PA
17106-7000.

White, D., A. Kimerling and W. S. Overton. 1992. Cartographic and
geometric components of a global sampling design for environmental
monitoring. Cartography and Geographic Information Systems 19(1):
5-22.