Introduction

This study summarizes land-use/land cover (LULC) changes occurring in Delaware between 1984 and 1992.  It is, in part, a follow-up to AES Bulletin 483, Land Use Transitions in Delaware, 1974-1984, published September, 1989, by Mackenzie.  Complete results from this prior study are available in hardcopy.

The present analysis was undertaken by the Spatial Analysis Lab without external funding, and without any contractual commitments to statistical accuracy.  Data inconsistences between the 1984 and 1992 LULC layers, and corrections of some of these inconsistencies, are discussed below.

An analysis of LULC changes between 1992 and 1997 has been completed by the Office of State Planning.  The 1992 and 1997 data were developed from equivalent imagery sources using a consistent classification scheme, and have only minor registration errors.

1984 and 1992 Data

We are distributing the data used in this analysis in both a UTM-18N (NAD83) projection (based on GRS80 spheroid), and unprojected lat-lon decimal degrees.  All data are Arc Export (.E00) format, WinZipped.  The change analysis was performed on the UTM data.  The lat-lon files were processed by Jian Chen, UD Research Data Management Service.

The 1984 LULC series were developed by Earthsat Corp. under contract with the Delaware Dept. of Agriculture at resolution consistent with Landsat MSS satellite imagery (79-meter pixel size), using a modified 2-digit Anderson LULC classification scheme, with a 15-acre minimum mapping unit.  The original data were in Delaware State Plane Projection, NAD 1927, in feet, based on the Clarke 1866 spheroid.

The classification codes for the 1984 LULC series are:

The 1992 and 1997 LULC series was developed by Earth Data (formerly PhotoScience) under contract with the State of Delaware. LULC polygons were digitized on 1-meter resolution digital orthophotos using a different modification of the Anderson LULC classification scheme, with a 4-acre minimum mapping unit. The original data were in Delaware State Plane Projection, NAD 1983, in meters, based on the GRS80 spheroid.

The classification codes for the 1992 and 1997 LULC series are:

There are various inconsistencies between these data sources.

1. spatial error: The 1992 data, digitized on 1-meter resolution digital orthophotos, have minimal spatial error. The 1984 series, digitized from county maps at 1:100,000 scale, exhibit significant positional error relative to the 1992 data. This analysis made no attempt to correct spatial error in the 1984 data. The spatial error precluded any attempt to develop LULC transition matrices quantifying acres changing between individual LULC categories.
2. inconsistent resolution: The 1992 series are based on a 4-acre minimum mapping unit, while the 1984 series are based on a 15-acre minimum mapping unit. Therefore prior LULC categories for 1992 polygons smaller than 15 acres are indeterminate. This study treats such small polygons appearing in the 1992 mapping as valid LULC change areas. Although this imparts some upward bias to estimated increases in developed acreage categories, the aggregate acreage in small (4 to 15 acres) residential polygons is relatively minor.
3. inconsistent classification: The major classification inconsistency involves reassignment of forested wetlands from forest categories in the 1984 mappings to a generic wetland category in the 1992 mappings. There are also discrepancies in the classifications of brushland, individual forest categories, wetland, water, beach and barren land and some individual urban land categories. Some water areas in the 1992 mapping (e.g., the Chesapeake & Delaware Canal) were not included in the 1984 mapping, thus generating some inconsistencies in county acreage totals.

Data Processing

Both series were re-projected to UTM Zone 18 (meters) in ArcInfo, and then imported to GRASS for analysis. All maps were rasterized at 30-meter resolution. Wetland cells in the 1992 maps which were assigned to any forest category in the 1984 maps were re-assigned to that same forest category for 1992. The map series for both years were then reclassified to the following simplified scheme:

Analysis

A graphical frequency analysis of LULC categories in the two series was performed for each county, as well as for the northern and southern halves of New Castle County.  The thumbnail images below access full-screen side-by-side LULC mappings for 1984 and 1992.  Accompanying histograms are included to indicate aggregate acreages in each LULC category in the two mappings, and acreage changes occurring over the 8-year interval between them.  These images are intended to provide a general idea of where the major transitions occurred in each county.

The following tables summarize acreage estimates in each aggregated category.  Some percent change figures reflect significant classification inconsistencies.  Acreage totals do not match mainly because of inconsistent treatment of water features.

These numbers indicate continuing rapid rates of conversion of agricultural and forest lands to residential and other urban uses in all three counties.

• Sussex County had the largest gross gain and the largest county-wide percent gain in residential acreage (+17,797 acres, +80%). New Castle County below the C&D Canal experienced the most rapid localized rate of residential development (+6,300 acres, +221%).
• New Castle County had the largest gross gain in other urban uses (+12,649 acres). The portion of New Castle County above the C&D Canal had the largest overall acreage gain (+11,398, +56%) while the area below the C&D Canal had the most rapid rate of gain (+1,251 acres, +210%). Sussex County had the largest county-wide percent gain in other urban uses (+103%).
• Sussex County had the largest gross loss of agricultural land (-19,746 acres), while New Castle County had the largest percent loss (-16%), with the most dramatic rate of farmland loss occurring north of the C&D Canal (-44%).
• Sussex County had the largest gross loss of forest acreage (-17,000 acres), while Kent County had the largest county-wide percent loss (-12%). New Castle County below the C&D Canal experienced a 16% net loss of forest acreage.

Change calculations for brushland, individual forest categories, water, wetland and beach/barren categories reflect various classification inconsistencies between the two mappings.

AGGREGATE LAND-USE/LAND COVER CHANGES, 1984 AND 1992
DELAWARE, BY COUNTY