Laura Poggio1, Pierre Soille2
1 The Macaulay Institute Aberdeen AB158QH (UK)
2 Joint Research Centre, European Commission, Ispra (Italy)
Digital elevation models (DEMs) provide us with a digital representation of the continuous land surface. DEMs often contain depressions that result in areas described as having no drainage, referred to as sinks or pits. These depressions disrupt the drainage surface, which preclude routing of flow over the surface. Sinks arise when a connected component of pixels occurring at the same elevation level is surrounded by pixels of higher elevation, or when two cells flow into each other resulting in a flow loop, or the inability for flow to exit a cell and be routed through the grid. Hydrologic parameters derived from DEMs, such as flow accumulation, flow direction, upslope contributing area and river network detection require sinks to be removed (Maune, 2001). Naturally occurring sinks in elevation data with a grid cell size of 100 m2 or larger are rare in terrains modelled by fluvial erosion processes. They could occur more frequently in glaciated or karst topographies. Various algorithms have been proposed to detect and remove surface depressions, such as elevation-smoothing method (Mark 1984), depression-filling algorithms (Jenson and Domingue, 1988; Soille and Ansoult, 1990; Tarboton et al., 1991), breaching (Martz and Garbrecht, 1998) carving method (Soille et al., 2003) or hybrid method combining carving and depression filling (Soille, 2004). For a detailed review it is possible to refer to e.g. Reuter et al. (2008) or Wang and Liu (2006). Lindsay and Creed (2005) compared the performance of different algorithms on various slope classes and landforms using a very high resolution dataset. However there is few information on the influence of the various algorithms on the accuracy of the positioning of the extracted networks. The aim of this study was to assess the effects of three pit removal methods on the position of river networks extracted from the SRTM dataset.