Mapping Wildlife Habitat Protection Priorities in Delaware
John Mackenzie, Kitt Heckscher and Lynn Broaddus


Background:

Delaware's piedmont and coastal plain ecosystems have endured a long history of disruption by human activities. Delaware has experienced more extensive landscape alteration than any other region in the U.S (Broaddus, 1996), and historic rates of species loss are likely the highest in the continental U.S. There is continued strong development pressure on the land base, with rapid transitions of forest and farmland acreage to residential, commercial, industrial and infrastructure development. Between 1974 and 1984, Delaware experienced a 15.4 percent net loss of its coniferous forest and a 7.1 percent net loss of deciduous forest to both farming and development; a 1.2 percent net loss of farmland (with large gross losses of farmland to development mostly offset by gross gains of farmland from forest); and a 4.5 percent net loss of coastal wetlands (Mackenzie, 1989). The proportion of forested wetlands lost during this period is uncertain. Since 1984 net losses of both forest and farmland to development appear to have accelerated, while regulatory controls on wetland conversions appear to have slowed rates of wetland loss.

The few published studies of land-use/land cover (LULC) transitions in Delaware have merely quantified aggregate net acreage transitions between different LULC categories. There has been little attempt to quantify historic declines in contiguities or spatial densities of habitat areas within the framework of a geographic information system (GIS), and the effects of declining contiguities and densities on ecosystem health are not well understood. Published synoptic LULC mappings of Delaware suffer from inconsistencies: they are based on various classification systems, and various data sources with differing spatial resolutions. A recent LULC transition analysis by Carter (1997) comparing 1984 LULC data based on 10+ acre minimum mapping units versus 1992 LULC data derived from digital orthophotography and generalized to an equivalent spatial resolution indicates continuation of the same trends Mackenzie identified in the 1974-1984 period: Delaware is still experiencing rapid forest loss and moderate cropland loss to development. Unfortunately, classification inconsistencies between the 1984 and 1992 mappings preclude accurate quantification of these transitions.

Several habitat inventory efforts are now underway in Delaware. The U.S. Fish and Wildlife Service's gap analysis project is the most ambitious habitat mapping project, but will require one to two years to complete. Delaware's Natural Heritage Program, one of various cooperators in the gap analysis project, has accumulated a significant body of wildlife inventory data over the last 10 years, much of which has been digitized and incorporated into the USF&WS gap analysis project. The Delaware Coastal Management Program (DNREC) is developing various mappings in support of planning efforts in Kent and New Castle Counties. The Center for Remote Sensing at the University of Delaware's College of Marine Studies has been involved in related coastal zone analyses.

By the end of 1997 the State of Delaware expects to have formulated a comprehensive statewide land use plan based on land use plans submitted by the counties. These county planning efforts are relying on geographic information system (GIS) technologies to integrate "layers" of geographic information representing growth needs and constraints, e.g., population and housing densities, transportation and other infrastructure, land-use/land cover, zoning, land ownership, topography, drainage, soils and geology. We are concerned that Delaware's wildlife habitat protection needs are unlikely to receive adequate consideration in the planning process unless they are articulated as GIS layers and readily integrated into the GIS analyses which are driving the comprehensive planning process.

The Spatial Analysis Lab is supporting a collaborative multidisciplinary effort, involving the principal investigators of the various habitat inventory projects underway in Delaware, to distill these inventory data into an aggregate mapping of habitat protection priorities which can be directly incorporated into county and state-level land-use planning processes.

The maps will be based on a habitat quality scoring system to be developed by a technical committee of wildlife biologists, ecologists and GIS experts. We note that scientific data on wildlife population distributions in Delaware are incomplete. Furthermore, any habitat scoring system will necessarily involve gross simplifications of complex ecosystem interdependencies. We emphasize, however, that the target audience for this project will be county and state planners rather than expert biologists. The maps must identify critical habitat areas in clear terms to non-experts, and they must clearly convey expert opinion regarding habitat protection priorities so that these priorities are readily included in the overall planning process. By publicizing current habitat inventory efforts and demonstrating their relevance to Delaware's comprehensive planning effort, this project can help motivate more reliable funding for continued habitat analyses.

Technical Committee. We are working with a Technical Committee of GIS practitioners and biologists from various public and private wildlife-oriented agencies including the U.S. Fish and Wildlife Service, the Delaware Natural Heritage Program, the Delaware Department of Agriculture (Forestry Section), the Nature Conservancy's Delaware Field Office, the Delaware Nature Society and any other interested parties. This committee's charge is to

Forest Fragmentation Analysis.

Digital land-use/land cover mappings of Delaware are available or readily derivable for the years 1984 and 1992. In the first part of this project we characterize aggregate land-use/land cover changes by county during that period, attempt to quantify aggregate forest acreage losses, and discuss changes in statistical distributions of residential, agricultural and forest polygon densities (polygon area/perimeter ratios).

1984 LULC mappings of each county were developed by Earth Satellite Corp., Chevy Chase, MD, for the Delaware Dept. of Agriculture, using air photos and/or Landsat MSS imagery. These mappings are based on a 15-acre minimum mapping unit, and use a modified 2-digit Anderson land-use/land cover classification system. Forested wetlands are assigned to one of three forest categories (deciduous, coniferous or mixed).

1992 LULC mappings of each county were developed by Photo Science Inc., Reston, VA, for the Delaware Dept. of Natural Resources and Environmental Control, using a 4-acre minimum mapping unit, and use a different modified 3-digit Anderson land-use/land cover classification system. Forested wetlands are assigned to a generic wetland category.

The different spatial resolutions of these mappings makes accurate quantification of LULC changes quite complicated. First, it is difficult to determine whether the small polygons (< 15 acres) in the 1992 mapping were present or absent in the 1984 mapping. Second, since the 1984 mappings were digitized at 1:100,000 scale, they exhibit significantly more spatial error than the 1992 mappings, which were digitized on 1-meter resolution digital orthophotography. Equivalent LULC polygons in the two mappings thus have poor edge consistency.

We re-projected the 1984 LULC data (originally Delaware State Plane NAD 1927, in feet) and 1992 LULC data (originally Delaware State Plane NAD 1983, in meters) to a consistent projection (UTM Zone 18, NAD 1983), and then rasterized the maps for each county at 30x30 meter resolution.

To correct a major inconsistency in forest/wetland classification, we identified wetland pixels in the 1992 data which were assigned to one of the three forest categories in the 1984 data, and reclassified these to their 1984 forest categories. Residential and commercial, industrial and other urban categories were aggregated into two generic categories (residential and other urban). Multiple agriculture and water categories were also aggregated into generic agriculture and water categories.

New Castle County

1984-92 LULC Transitions

Change SummaryBetween 1984 and 1992 New Castle County experienced 33% growth in residential acreage and about 60% growth in acreage in commercial, industrial and infrastructure categories.  Some of this occurred as infill between already-developed areas in the northern part of the county.  About 46% of the acreage converted to residential use during this period is located south of the Chesapeake and Delaware Canal.  The county lost approximately 16% of its agricultural land and 8% of its forest land to development.  Due to inconsistencies in forest and wetland categorizations, losses by forest type could not be computed.

The side-by-side comparison of LULC mappings gives a general idea of where the major transitions occurred. The accompanying histograms indicate relative acreages in each LULC category in the two years, and acreage changes occurring over the 8-year interval.

Fragmentation:  The spread of residential and commercial development and the associated fragmentation of forest and agricultural land is more easily visualized on separate mappings of residential, forest and agricultural polygons (below). The color scheme of each map pair represents the natural logarithm of polygon size. Each map is accompanied by a histogram showing total acreages in each polygon size category. Note that the vertical axes of these histograms use different scales.

The histograms of residential polygon areas show an increasing bi-modal pattern consistent with progressive residential infill in the north and extensive suburban development around the Chesapeake & Delaware Canal.

The histograms of forest polygon areas show a leftward shift reflecting the fragmentation of major forest polygons in the south of the county and a proliferation of very small forest polygons.

The histograms of agricultural polygon areas show a similar leftward shift reflecting equivalent fragmentation of farmland, particularly in the Middletown-Odessa-Townsend area just below the Chesapeake & Delaware Canal.

Residential densitiesForest densitiesAgricultural densities




Kent County

1984-92 LULC Transitions

Change Summary: Between 1984 and 1992 Kent County experienced a 50% net increase in acreage in residential use, a 29% net increase in acreage in commercial, industrial or infrastructure use, a 1% net decline in agricultural acreage and a 12% net decline in forest acreage.  As noted above, inconsistent forest categorizations precluded a more detailed breakdown of forest losses by forest type.  Gross losses of agricultural land to development were substantially higher than the net loss due to significant conversions of forest acreage to agriculture.

Fragmentation:  Compared to New Castle Co., new development in Kent County is dispersed rather than infill, as indicated by large percent increases in frequencies of small residential polygon.  The forest fragmentation analysis shows breakup of some major forest polygons west and south of Dover associated with scattered development in those areas.  Farmland breakup does not appear to be significant.

Residential densitiesForest densitiesAgricultural densities


 

Sussex County

1984-92 LULC Transitions

Change Summary: Between 1984 and 1992 Sussex County experienced an 83% growth in acreage in residential use, a 103% net increase in acreage in commercial, industrial or infrastructure use, a 7% net decline in agricultural acreage and an 8% net decline in forest acreage.  The major development occurred along the Atlantic Coast, reflecting strong demand for both primary and vacation residences; As with Kent Co., gross losses of agricultural land to development were substantially higher than the net loss due to some conversions of forest acreage to agriculture.

Fragmentation:  Much of the residential development in Sussex Co. represents infill along the Atlantic coast, reflected in an increased frequency of large residential polygons.  Forest fragmentation is evidenced by a decline in large forest polygons and a notable increase in frequency of small forest polygons.  Agricultural fragmentation appears to be relatively minor.

Residential densitiesForest DensitiesAgriculture densities

Delaware Natural Heritage Inventory

Natural Heritage Inventory Data Analysis:

The Delaware Natural Heritage Program provided approximately 10 years worth of rare species inventory data georeferenced by lat-long. The database consists of over 2,700 individual records representing field discoveries of animals, plants and invertebrates which are rare in Delaware or globally rare. The records include state and global rarity indices.

Review and Implementation Strategy. Once the mapping effort is completed and approved by the Technical Committee, the Spatial Analysis Lab will solicit external review of the maps by county and state planning officials, and obtain input on how these habitat protection priorities can be most effectively implemented as policy. What practical development constraints or mitigation requirements should be imposed in areas we define as moderate- or high-priority habitat?  We will also be asking the College of Agricultural Sciences to sponsor larger public meetings in early 1998 to obtain public review of the priority maps and gauge political support for them.
 

References:

Anselin, Luc. 1995. SpaceStat version 1.80 (software). Regional Research Institute, West Virginia University, Morgantown, WV.

Bissell, S.J., K. Demarest and D.L. Schrupp. 1986. The use of zoning ordinances in the protection and development of wildlife habitat. Integrating Mand and Nature in the Metropolitan Environment. Proceedings of the National Symposium on Urban Wildlife, Chevy Chase, MD, 4-7 November 1986. National Institute for Urban Wildlife, Columbia, MD.

Carter, D. 1997. Land-Use/Land Cover Transitions Analysis (unpublished?). Delaware Coastal Management Program, DNREC, Dover, DE.

Mackenzie, J. 1989. Land-Use Transitions in Delaware. Agricultural Experiment Station Bulletin #483, College of Agricultural Sciences, University of Delaware, Newark, DE.