N. Thommeret1, J.S. Bailly2, C. Puech2
1 UMR 8591- CNRS, Laboratoire de Géographie Physique,
1 place Aristide Briand, 92195 Meudon Cedex
2 UMR TETIS – AgroParisTech-Cemagref-Cirad, Maison de la Télédétection,
500 rue JF Breton, 34093 Montpellier Cedex 5
Email: bailly, firstname.lastname@example.org
To study the link between thalweg networks in badlands and hydrological functioning over eroded areas, digital indices describing the 3D network geometry and topology must be computed for different extents and resolutions. Thalwegs correspond to the line connecting the lowest points of the gully. In badlands, these networks are characterised by tree structure topologies. In case of large extents, grid DTM acquired by airborne or high resolution satellite sensors seem to be the most appropriate source of data for thalweg network extraction. To extract and characterise thalweg networks from a grid DTM, various drainage algorithms (O’Callaghan and Mark 1984, Fairfield and Leymarie 1991, Lea 1992, Tarboton 1997) offer possibilities of computing drainage networks all over the grid surface. However, the transition from one drainage flow path to a vector thalweg network is not obvious (Martz and Garbrecht 1995, Tarboton 2001, Turcotte et al. 2001). Most of the time, the process uses a unique and arbitrary drainage surface threshold beyond which thalwegs, valleys or hydrographic networks are depicted. Since this criterion has no physical significance, where to start thalwegs upstream? How to get thalweg networks fitting to the amount of information really contained in a DTM? How to limit drainage algorithm artefacts in wide valleys disrupting the resulting network topology? In other words, how to compute robust thalweg network from a grid DTM?
This paper presents a method that combines existing drainage algorithm and a morphological index, first, to map gully floor pixels and then to define a continuous and tree-structured thalweg network. To run the method on a given DTM, only the spatial and statistical distribution of altimetrical error data is needed. This method aims to extract thalweg network objectively considering the significant landforms included in a DTM. This method has been applied on two DTMs simulating virtual landscapes and on an LiDAR DTM acquired on the Draix badlands catchments (French Alps). Results are visually compared to those obtained with the usual drainage area threshold criteria. On the Draix test site, network topological index distributions are computed and compared together.