Introduction

The first major component of IL-GAP, the vegetation or land cover layer, is a necessary and intensive step towards completion of the entire Gap Analysis Project. The method used to create a land cover map is remote sensing. Remote sensing is the science of deriving information about an object or phenomenon at or near the surface of the earth through the analysis of data acquired by a camera or sensor system located in an aircraft or orbiting satellite. One of the ways in which remote sensing imagery has commonly been used is to derive information regarding surface, or land cover characteristics.

At the site level and for small regions, land cover data has typically been obtained from the interpretation of aerial photography, and more recently from digital ortho-photography. At the statewide level, land cover information is usually produced from the analysis of satellite imagery and the resulting inventory can provide accurate, regional level information for natural resource applications.

Vegetation acts as an integrator of many physical and biological properties of an area. IL-GAP seeks to map the extent and distribution of existing vegetation in Illinois in order to: 1) determine species and natural community representation within areas being managed for biodiversity conservation; 2) provide spatial data that can be used to model wildlife habitat distributions; and 3) provide a single temporal data set for monitoring trends.

Background

In late 1999, the U. S. Department of Agriculture National Agricultural Statistics Service (NASS), the Illinois Department of Agriculture (IDA), and the Illinois Department of Natural Resources (IDNR) began a cooperative, interagency initiative to produce statewide land cover information on a recurring basis. The Illinois Interagency Landscape Classification Project (IILCP) completed Cycle 1 of this initiative in the summer of 2002, resulting in the Land Cover of Illinois 1999-2000 inventory and associated database. The Illinois Gap Analysis group, located at the Illinois Natural History Survey, became a part of the interagency agreement, with the main focus of improving the wetlands and forests classifications of Illinois.

Illinois Gap 1999-2000 Classification

The IL-GAP Land Cover data layer is one of the products from the IILCP initiative and is the result of integrating the directed supervised classification of agricultural lands and the unsupervised classification and spectral class labeling of non-agricultural lands. The primary source information for the computer classification is Landsat 5 TM and Landsat 7 ETM+ satellite imagery acquired for three dates (triplicates) during the spring, summer, and fall seasons of 1999 and 2000. Ten TM Path/Row scene areas are required to cover Illinois, and same year imagery was acquired for the TM/ETM+ triplicates to ensure seasonal consistency for the computer classification. The satellite imagery for four of the TM scene areas was acquired in 1999, and the remaining six scene areas were acquired in 2000 (see 1999-2000 TM Coverage Map).

The primary goal of the Illinois Gap Analysis was to perform a detailed classification in the vegetated areas of Illinois. The Illinois Gap Analysis layer used the exact methodology of the Land Cover of Illinois 1999-2000, with a few specific changes in the wetland, forested, and urban areas. Within each of the ten TM scenes, spectral signatures for wetland, forest, and urban were extracted. After these categories were delineated, a mask was made for each. The wetland mask was used to extract wetlands, which continued as categories in the Land Cover of Illinois 1999-2000.

All of the TM/ETM+ satellite imagery were geometrically corrected and co-registered to a transverse Mercator projection with a UTM zone 16 grid and NAD83 datum. The 1999 TM/ETM+ imagery is a mixture of simple geometric correction and terrain correction, while all of the 2000 imagery was terrain corrected using USGS 3 arc-second digital elevation model (DEM) data to correct relief displacement. The TM/ETM+ multispectral imagery possesses a 30 x 30 meter (98.4 x 98.4 feet) ground spatial resolution, which means that the resulting Land Cover of Illinois 1999-2000 Classification data is suitable for GIS and mapping applications at a scale of approximately 1:100,000 (1" = 8,333') or smaller.

The computer classifications of the geometrically corrected and co-registered TM/ETM+ triplicates were conducted on a scene-by-scene basis. Spectral signatures were extracted from each TM/ETM+ triplicate data set utilizing an Isodata K-means clustering procedure (Duda and Hart 1973) and experimentation indicated that 200-250 spectral classes should be derived for each data set. The performance of several standard classifiers including minimum Euclidean distance, minimum Mahalanbois distance, and non-thresholding maximum likelihood have been evaluated and the results indicated greater improvement in classification accuracy can be achieved from the use of a maximum likelihood classifier, and this is supported by other research (Luman and Ji 1995; and Gong and Howarth 1990). The labeling and reduction of the spectral classes from the unsupervised classification procedures into the final information classes (water, wetland, etc.) was accomplished by photo-interpretation of NAPP 3 aerial photography. The supervised classification of agricultural lands, which used USDA-NASS ground reference data as training data collected the same year for 425 sites across the state, directly resulted in information classes and therefore no additional class labeling was necessary.

The final IL-GAP Land Cover data layer contains 29 land cover categories, which are as follows:

Upon completion of the integrated classifications for each of the ten TM Path/Row scene areas, a seamless mosaic was created using image processing software developed by the USDA-NASS. Careful attention was made to develop the mosaic along county boundaries to ensure that the land cover information within each county area was derived from the same Landsat TM/ETM+ triplicate. Because of the orientation and overlap of adjacent Landsat TM scene areas, a few counties are a combination of 1999 and 2000 TM/ETM+ imagery. Lastly, a formal assessment of the accuracy of the final integrated classification was conducted, and the detailed explanation of that procedure is contained in the Accuracy Assessment Section on the web site.

For more information and to download the IL-GAP Land Cover data see the Department of Agriculture website or click the following download icon.

Last updated Thursday, 04/17/2003 10:41 AM