Submitted to: Journal of Spatial Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 2/15/2010
Publication Date: 6/2/2010
Publication URL: http://handle.nal.usda.gov/10113/54044
Citation: Laliberte, A.S., Browning, D.M., Herrick, J.E., Gronemeyer, P. 2010. Hierarchical object-based classification of ultra-high-resolution digital mapping camera (DMC) imagery for rangeland mapping and assessment. Journal of Spatial Science. 55(1):101-115. Interpretive Summary: Many land management agencies are required to monitor and assess vegetation conditions across millions of acres of rangelands. Field-based assessments can be costly and inefficient, and remote sensing offers the potential to increase the number of monitoring locations, assess larger areas, and complement ground-based measurements. We investigated object-based image analysis techniques for classifying vegetation in southwestern U.S. arid rangelands with 4 cm resolution digital aerial photography. We were able to automate the object-based image analysis approach by applying algorithms from one vegetation community to another with relatively small changes. While the level of detail was less than that of ground-based measurements, high R-square values were obtained for image- and ground-based measures of percent cover for shrubs, bare ground, total vegetation, and grasses (R-square values: 0.82-0.92). The ability to classify multiple images efficiently can increase the precision of monitoring efforts by increasing sample locations; can lower costs by reducing the number of field visits; and can provide image baseline information for potential change analysis. This object-based image analysis approach with ultra high resolution digital imagery exhibits promise for nationwide application for monitoring grazing lands.
Technical Abstract: Ultra high resolution digital aerial photography has great potential to complement or replace ground measurements of vegetation cover for rangeland monitoring and assessment. We investigated object-based image analysis (OBIA) techniques for classifying vegetation in southwestern U.S. arid rangelands with 4 cm resolution digital aerial photography. We obtained high r-square values for image- and ground-based measures of percent cover (r-square values: 0.82-0.92). OBIA enabled us to automate the classification process and demonstrated potential for quantifying fine-scale land cover attributes with ultra high resolution imagery. This approach exhibits promise for nationwide application for monitoring grazing lands.