Articles | Volume 14, issue 6
Hydrol. Earth Syst. Sci., 14, 847–857, 2010
https://doi.org/10.5194/hess-14-847-2010

Special issue: Quantitative analysis of DEMs for hydrology and Earth system...

Hydrol. Earth Syst. Sci., 14, 847–857, 2010
https://doi.org/10.5194/hess-14-847-2010

  01 Jun 2010

01 Jun 2010

The application of GIS based decision-tree models for generating the spatial distribution of hydromorphic organic landscapes in relation to digital terrain data

R. Bou Kheir, P. K. Bøcher, M. B. Greve, and M. H. Greve R. Bou Kheir et al.
  • Department of Agroecology and Environment, Faculty of Agricultural Sciences (DJF), Aarhus University, Blichers Allé 20, P.O. Box 50, 8830 Tjele, Denmark

Abstract. Accurate information about organic/mineral soil occurrence is a prerequisite for many land resources management applications (including climate change mitigation). This paper aims at investigating the potential of using geomorphometrical analysis and decision tree modeling to predict the geographic distribution of hydromorphic organic landscapes in unsampled area in Denmark. Nine primary (elevation, slope angle, slope aspect, plan curvature, profile curvature, tangent curvature, flow direction, flow accumulation, and specific catchment area) and one secondary (steady-state topographic wetness index) topographic parameters were generated from Digital Elevation Models (DEMs) acquired using airborne LIDAR (Light Detection and Ranging) systems. They were used along with existing digital data collected from other sources (soil type, geological substrate and landscape type) to explain organic/mineral field measurements in hydromorphic landscapes of the Danish area chosen. A large number of tree-based classification models (186) were developed using (1) all of the parameters, (2) the primary DEM-derived topographic (morphological/hydrological) parameters only, (3) selected pairs of parameters and (4) excluding each parameter one at a time from the potential pool of predictor parameters. The best classification tree model (with the lowest misclassification error and the smallest number of terminal nodes and predictor parameters) combined the steady-state topographic wetness index and soil type, and explained 68% of the variability in organic/mineral field measurements. The overall accuracy of the predictive organic/inorganic landscapes' map produced (at 1:50 000 cartographic scale) using the best tree was estimated to be ca. 75%. The proposed classification-tree model is relatively simple, quick, realistic and practical, and it can be applied to other areas, thereby providing a tool to facilitate the implementation of pedological/hydrological plans for conservation and sustainable management. It is particularly useful when information about soil properties from conventional field surveys is limited.