Relief analysis is a tool to analyse a landscape based on a Digital Elevation Model (DEM). One of the simplest parameters might be the elevation itself, or slope or the exposition of a given point in a landscape. Moore et al. (1991) state that the spatial distribution of topographic attributes can often be used as an indirect measure of the spatial variability of hydrological, geomorphologic and biological processes. The advantage compared to other information such as soil parameters or biomass production estimates is based on the relatively simple and fast techniques to model processes in large areas and the complex spatial patterns of environmental systems as seen by Moore et al. (1993b). Another relief parameter relevant for this work is landforms or relief units. Each of these contains certain characteristic physical, chemical, and biological processes and parameters (see Dehn et al., 2001). Milne (1936) was one of the first scientists, who recognised the catena principle of soil formation in a hilly terrain (Ruhe, 1960). Several authors followed on that concept (Ruhe, 1956, Conacher and Dalrymple, 1977), which eventually is the basis for all landform classification systems found today, however, in the catena concept the third dimension is missing. Hugget (1975) presented a soil landscape system (not only a landform system), which overcomes the two-dimensional character and simulated whole systems behaviour. Following on that, Willgose et al. (1991b) have developed their SIBERIA model, to understand evolution of landforms over geomorphologic time scales. Before the introduction of DEM, manual delineation of landforms was performed using field surveys or by interpretation of stereo aerial photographs. McBratney et al. (1992) showed that surveyors delineate a complex landscape according to their personal bias. On the other hand, Burrough et al. (2000) stated that a global set of rules found at a national or international level, does not provide the most sensible divisions at the local level; therefore, different types of landform classification algorithms were developed.
Discontinuous versus continuous classification
Pennock et al. (1987) used values published by Young (1972) to classify seven/nine three dimensional landform elements based on the following relief parameters: profile curvature, planform curvature, slope and catchment size. Park et al. (2001) used surface curvature and a terrain characterisation index to delineate a five landform element model. The terrain characterisation index is based on the multiplication of surface curvature and the logarithm of the upslope contributing area. Park et al. (2001) concluded, that the values of their four criteria are rather arbitrary and may need several trials to achieve a reasonable approximation of landform delineation. MacMillan and Pettapiece (1997) used the percentage of a location in relation to the minimum and maximum in a given landscape, to assign landform elements. Discontinuous landform classification as shown in the three examples above separates a landscape into definite facets with hard boundaries, whereas a continuous classification assigns the strength, or degree of membership of each cell in each of n defined classes (MacMillan et al., 2000) either based on expert judgment (Brown et al., 1998) or statistical analysis (Burrough et al., 2000, Irvin et al., 1997). An example for a continuous classification is the work by MacMillan et al. (2000), who performed a fuzzy classification, however work by Burrough et al. (2000) used expert knowledge to identify landform facets via a semantic import (SI) model (Dehn et al., 2001). The general limitation of the approaches above is that they are based on human expert knowledge. MacMillan et al. (2000) states that statistical classifications based on ordination of numerical data in n-dimensional feature space (Burrough et al., 2000) can identify the optimum number of natural classes for a given landscape. The drawback of this approach is that classes and class definitions will never be the same for any two sites and are optimised for a given landscape. Still, an advantage is that classifications are performed based on statistical evidence.
General problems on the European Scale
Researchers have been frustrated by computational issues related to the scale of investigation. MacMillan et al. (2000) described the three major problems for an automated landform classification algorithm for identifying different types of landscapes using DEMs as follow:
- Selecting and computing an appropriate suite of terrain derivatives from DEM data
- Identifying an appropriate number of meaningful different landform classes and describing their salient or defining characteristics
- Selecting and applying a classification procedure capable of using the available terrain derivatives to produce the required classes.
To give a first approximation for landform/terrain unit classification on a global scale we applied two algorithms on global DEM datasets. The first algorithm works with static thresholds based on previous knowledge of landscape classification for elevation levels and relief roughness following Meybeck et al (2001). Roughness and elevation are classified based on a DEM according to static thresholds, with a given window size (e.g. 20 in the dataset shown). This is an example of a static (in terms of thresholds) landform classification method.
The second algorithm performs a landform classification following Iwahashi and Pike (2007). It presents relief classes which are classified using an unsupervised nested-means algorithms and a three part geometric signature. Slope gradient, surface texture and local convexity are calculated based on the SRTM30 DEM, within a given window size (e.g. 10 in the dataset shown) and classified according to the inherent data set properties. This means the classification thresholds depend on the derived slope values, surface texture and convexity values. This is an example for a dynamic (in terms of thresholds) landform classification method.
- The landform classification following Meybeck et al. (2001) presents relief classes, which are calculated based on the relief roughness. Roughness and elevation are classified based on a DEM according to static thresholds, with a given window size ( e.g. 20 in the dataset shown). This is an example of a static (in terms of thresholds) landform classification method.
- The landform classification following Iwahashi and Pike (2007) present relief classes which are classified using an unsupervised nested-means algorithms and a three part geometric signature. Slope gradient, surface texture and local convexity are calculated based on the SRTM30 DEM, within a given window size (e.g. 10 in the dataset shown) and classified according to the inherent data set properties. This is an example for a dynamic (in terms of thresholds) landform classification method.
|Landform Classification - Iwahashi (Format PDF)
Landform Classification - Iwahashi (Format JPEG)
Landform Classification - Iwahashi (Format JPEG, 300 dpi High Resolution)
Landform Classification - Meybeck (Format PDF)
Landform Classification - Meybeck (Format JPEG)
Landform Classification - Meybeck (Format JPEG, 300 dpi High Resolution))
- Dehn,M., Gärtner,H., and Dikau,R., 2001. Principles of semantic modeling of landform structures. Computers and Geoscience, 27: 1005-1010
- Ruhe,R.V., 1960. Elements of the soil landscape. Madison. 165-170
- Ruhe,R.V., 1956. Geomorphic Surfaces and the nature of soils. Soil Science, 82(6): 441-455.
- Conacher,A.J. and Dalrymple,J.B., 1977. The nine-unit landsurface model: an approach to pedogeomorphic research. Geoderma, 18: 1-154.
- Hugget,R.J., 1975. Soil Landscape Systems: A model of soil genesis. Geoderma, 13: 1-22.
- Willgose,G., Bras,R.L., and Rodriguez-Iturbe,I., 1991a. A coupled Channel Network Growth and Hillslope Evolution Model 1 Theory. Water Resources Research, 27(7): 1671-1684.
- Willgose,G., Bras,R.L., and Rodriguez-Iturbe,I., 1991b. Results for a new model of river basin evolution. Earth Surface Processes & Landforms, 16: 237-254.
- McBratney,A.B., de Gruijter,J.J., and Brus,D.J., 1992. Spatial prediction and mapping of continuous soil classes. Geoderma, 54: 39-64.
- Pennock,D.J., Zebarth,B.J., and De Jong,E., 1987. Landform Classification and Soil Distribution in Hummocky Terrain, Saskatchewan, Canada. Geoderma, 40: 297-315.
- Young,A., 1972. Slopes. Edingburgh
- Irvin,B.J., Ventura,S.J., and SLater,B.K., 1997. Fuzzy and isodata classification of landform elements from digital terrain data in Pleasant Valley, Wisconsin. Geoderma, 77: 137-154.
- Milne,G., 1936. A provisional soil map of east africa. Eastern Africa Agriculture Research Station, Amanin Memoirs, 1-34.
- Burrough,P.A., van Gaans,P.F.M., and MacMillan,R.A., 2000. High -resolution landform classification using fuzzy k-means. Fuzzy Sets and Systems, 113: 37-52.
- MacMillan,R.A. and Pettapiece,W.W., 1997. Soil Landscape Models: Automated landscape characterization and generation of soil-landscape models. Research Report No. 1E.1997.
- Lethbridge, Agriculture and Agri-Food Canada, Research Branch.
- Moore,I.D., Gessler,P.E., Nielsen,G.A., and Peterson,G.A., 1993a. Soil Attribute Prediction Using Terrain Analysis. Soil Science Society of America Journal, 57: 443-452.
- Moore,I.D., Nortin,T.W., and Williams,J.E., 1993b. Modelling environmental heterogeneity in forested landscapes. Journal of Hydrology, 150: 717-747.
- Moore,I.D., Grayson,R.B., and Ladson,A.R., 1991. Digital Terrain Modelling: A Review of Hydrological,
- Meybeck, M., P. Green and C. J. Vorosmarty (2001), A New Typology for Mountains and Other Relief Classes: An Application to Global Continental Water Resources and Population Distribution, Mount. Res. Dev., 21, 34 - 45.
- Iwahashi, J. and R. J. Pike (2007). "Automated classifications of topography from DEMs by an unsupervised nested-means algorithm and a three-part geometric signature." Geomorphology 86(3-4): 409-440.