An automated method to extract fluvial terraces from digital elevation models: The Vesdre valley, a case study in eastern Belgium

Fluvial terraces are a powerful tool for unraveling the combined tectonic and climatic conditions that controlled, directly or indirectly, the Quaternary incision of rivers. Terrace long profiles are usually retrieved from sparse traces of ancient floodplains preserved in the present topography. However, when these traces classically collected from topographic maps, aerial photographs, and field analyses are too few, the inferred profiles may be questionable. Yet the now available high quality and high resolution digital elevation models (DEMs) offer an opportunity to increase greatly the quantity of information usable to reconstruct terrace profiles. Therefore, the purpose of this study was to develop a new DEM-based method of terrace recognition in order to create a larger database and better constrain the profile reconstruction. Moreover, particular procedures of image and numerical processing were defined to fully automate the analysis. Basically, our method relies on the production of bivariate scatter plots depicting the relation between slope and relative altitude (i.e., the altitude above the current alluvial plain) for all pixels of successive sections of the valley. For each scatter plot, the curve of the lowest slope values observed at every relative altitude is smoothed and its minima are assumed to locate the altitudes of the "terrace" elements preserved in the section. We successfully tested this method in the Vesdre valley, incised in the NE Ardenne massif (E Belgium), notably identifying fault deformation of the profiles. The main advantages of our approach are its objectivity, exhaustiveness, and rapidity, allowing fast and coherent analysis of many rivers over extended regions. (c) 2007 Elsevier B.V. All rights reserved.