Submitted to: Journal of Photogrammetric Engineering and Remote Sensing
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 10/31/2005
Publication Date: 2/1/2007
Citation: Laliberte, A.S., Fredrickson, E.L., Rango, A. 2007. Combining decision trees with hierarchical object-oriented image analysis for mapping arid rangelands. Journal of Photogrammetric Engineering and Remote Sensing. 73:197-207. Interpretive Summary: The accuracy of arid land vegetation classifications derived from remotely sensed imagery can often be low due to high reflectance of the soil background, mixtures of green and senescent grasses, and the prevalence of shrubs in grasslands. Decision trees typically improve classification accuracies, because they allow for incorporation of ancillary variables and they are an effective data reduction tool. We developed a unique approach, using high resolution imagery, a multi-scale analysis, and object-based rather than pixel-based image information as input for a classification tree for mapping arid land vegetation. A QuickBird satellite image was segmented into spectrally homogeneous objects at four different scales. Classification trees were derived for each scale and accuracy assessments were used to determine the best input variables and optimal scale for analysis. The classification tree reduced 118 input variables to 14, and the classification accuracy was 80%. Variables chosen by the decision tree included many features not available or as easily determined with pixel based image analysis. The combination of multi resolution image segmentation and decision tree analysis facilitated the selection of input variables and helped in determining the appropriate image analysis scale. This approach proved to be an effective method for mapping arid rangeland vegetation at the pasture-level scale, but has the potential to be applied at coarser scales as well, and therefore would allow land management agencies to map vast public lands in a timely manner and with relatively high classification accuracies.
Technical Abstract: Decision tree analysis is a statistical approach for developing a rule base used for image classification developed a unique approach using object-based rather than pixel-based image information as input for a classification for mapping arid land vegetation. A QuickBird satellite image was segmented at four different scales, resulting in hierarchical netowrk of image objects representing the image information in different spatial resolutions. This allowed differentiation of individual shrubs at a fine scale and delineation of broader vegetation classes at coarser scales. Input variables included spectral, textural and contextual image information, and the variables chosen by the decision tree included many features not available or as easily determined with pixel based image analysis. Spectral information was selected near the top of the classification trees, while contextual and textural variables were more common closer to the terminal nodes of the classification tree. The combination of multi resolution image segmentation and deceision tree analysis facilitated the selection of input variables and helped in determining the appropriate image analysis scale.