Mapping Prairie Pothole Communities with Multitemporal Ikonos Satellite Imagery

We evaluated the ability of Ikonos imagery from August and October 2000 to classify prairie pothole community types of the Missouri Coteau of North Dakota. Classification tree analyses were conducted to create land-cover maps at three levels of detail. The analyses successfully distinguished broad cover types (potholes including emergent vegetation versus upland vegetation) at 92 percent overall accuracy. Overall accuracy dropped to 80 percent when upland vegetation was segregated into woody and grassy communities and to 71 percent when we attempted to classify at the species or near-species levels. The use of two image dates was of importance in the classifications; the failure to acquire early season imagery, therefore, might have impaired our results. Introduction The advent of readily available high spatial resolution commercial satellite imagery (Petrie, 2001) presents important new opportunities for land managers and researchers needing classifications of landscapes that are heterogeneous at fine scales. Vegetation, for example, often varies at spatial resolutions finer than are detectable using widely available moderate resolution imagery, such as Landsat-based imagery. We examined plant communities of the Missouri Coteau, the terminal moraine of the Wisconsin Glacier, which reaches from north central Montana to Iowa. The Missouri Coteau’s pothole and hilltop topography provides repeated examples of a moisture gradient occupied by vegetation reaching from aquatic through aspen, snowberry, tall grass prairie, and mixed grass prairie, to short grass prairie (Smith, 1998). The Missouri Coteau is valuable for range, agriculture, and wildlife, including migrating waterfowl (Murphy, 1993). The ability to classify accurately both grassland communities and associated wetlands, such as the prairies of the Missouri Coteau and their potholes with emergent vegetation, is of vital importance. More than one-fourth of the Earth’s land surface and over 60 percent of the United States is classified as grassland (Williams et al., 1968; Holechek et al., 1989; Laurenroth, 1979). Grasslands are critical for wildlife habitat, plant species diversity, hydrologic functions, ecosystem nutrient cycling, and grazing (Campbell and Mapping Prairie Pothole Communities with Multitemporal Ikonos Satellite Imagery Rick Lawrence, Rebecca Hurst, T. Weaver, and Richard Aspinall Lasley, 1969; Pearse, 1971). The importance of discriminating among grassland types with remote sensing has been noted as particularly important because of the vast extent of these ecosystems (Price et al., 2001). Wetlands, such as are found with the potholes of the Missouri Coteau, are similarly critical habitat for species including migratory waterfowl, and mapping such features is critical to land-use decisions (Muller et al., 1993; Semlitsch and Bodie, 1998). Satellite imagery has been used extensively to map grassland vegetation. Moderate resolution imagery, however, has been almost the exclusive tool for such mapping, thereby limiting such efforts to either broad vegetation categories or areas of homogeneous cover types at the resolution of the imagery. Landsat imagery has been used to discriminate between cooland warm-season grasses in eastern Kansas (Price et al., 2002), four grassland habitat types in North Dakota (Jensen et al., 2001), rough fescue grassland in western Canada (Thomson et al., 1985), ten plant communities in southwestern Idaho (Clark et al., 2001), and eight major grassland and shrubland groups in southwestern Idaho (Knick et al., 1997). Classification accuracies ranged from 60 percent to over 90 percent, indicating that Landsat imagery has substantial potential for mapping grasslands where the vegetation communities occur in sufficiently homogeneous areas to be detectable at 30 m resolution. Commercial high spatial resolution satellite-based sensors, including Ikonos and Quickbird, can provide classifications at resolutions of 4 m or less. Imagery from these sensors has been used for many applications, including monitoring prairie dog colonies (Sidle et al., 2002), building extraction (Lee et al., 2003), water monitoring and analysis (Huguenin et al., 2004; JiQun et al., 2004), site-specific agriculture (Metternicht, 2004; Vina et al., 2003), documenting vegetation degradation in mountainous environments (Allard, 2003), measuring tree mortality (Clark et al., 2004), estimating leaf area index (Colombo et al., 2003; Johnson et al., 2003), and assessing coral-reefs (Maeder et al., 2002; Palandro et al., 2003). Few reported studies, however, have used these sensors for classification of undeveloped land-cover (but see, e.g., Carleer and Wolff, 2004; Quinton et al., 2003; Sawaya et al., 2003), and the use of these data to examine grassland communities does not seem to be well explored. One possible reason for this lack of application might be that these sensors, having sensitivity in the visible and near-infrared portions of the spectrum (Goetz et al., 2003; Thenkabail et al., 2004), have less spectral resolution than Landsat, which also has sensitivity in the middle and thermal infrared (NASA, 2004), PHOTOGRAMMETRIC ENGINEER ING & REMOTE SENS ING Feb r ua r y 2006 169 Rick Lawrence is at the Department of Land Resources and Environmental Sciences, Montana State University, Bozeman, Montana 59717 ([email protected]). Rebecca Hurst and T. Weaver are at the Ecology Department, Montana State University, Bozeman, Montana 59717 ([email protected], [email protected]). Richard Aspinall is at the Geography Department, Arizona State University, Tempe, Arizona 85287 ([email protected]). Photogrammetric Engineering & Remote Sensing Vol. 72, No. 2, February 2006, pp. 169–174. 0099-1112/06/7202–0169/$3.00/0 © 2006 American Society for Photogrammetry and Remote Sensing 04-101.qxd 1/18/06 5:17 AM Page 169