Mean-curvature watersheds: A simple method for segmentation of a digital elevation model into terrain units

Abstract

Graphical Abstract AbstractTerrain segmentation is the process of subdividing a continuous terrain surface into discrete terrain units. If the resulting units represent meaningful geomorphic objects the approach may facilitate studies of not only landforms and land forming processes, but also the interaction among surface form, soil, vegetation, hydrology and topoclimatic regimes. Commonly used methods for terrain segmentation fail to produce terrain units with a potentially large, but cyclic variation in topographic attributes, such as uniformly curved areas bounded by topographic break-lines, although this topographic characterisation is common for a number of landforms. This paper describes a new method for terrain segmentation using meancurvature (MEC) watersheds. The method produces objects that contain a cycle of MEC values. Thus the topographic variation within each object may be large, but due to the cyclic nature of the MEC variation a geometric simplicity is ensured. In a case study we show how the resulting terrain units correspond well with a number of landforms and surface types observed in the field, and conclude that the method can be expected to be of great value for a number of applications within geomorphology and related disciplines.

Suitability of spaceborne digital elevation models of different scales in topographic analysis: an example from Kerala, India

“…Topographic information was conventionally gathered from maps of varying scales, which has restricted their applications in geomorphology and hydrology due to tiresome, biased, labor-intensive and expensive processes of data collection (Speight 1977;Dragut and Blaschke 2006;van Niekerk 2010). On the contrary, in recent years, digital elevation model (DEM) has been widely used for 3D representation of topography and derivation of various topographic attributes in terrain analysis (Pike 2000;Wilson and Gallant 2000;Oguchi et al 2003;Masoud and Koike 2011;Romstad and Etzelmuller 2012), soil and vegetation mapping (Horsch 2003;Dobos and Hengl 2009;Jelaska 2009;Cavazzi et al 2013), mapping and modelling of natural hazards (Huggel et al 2008;Gruber et al 2009;Demirkesen 2012), hydraulic, hydrologic and geomorphologic modelling (Tarboton et al 1992;Hancock et al 2006;Peckham 2009;Gichamo et al 2012;Wang et al 2012;Schwanghart et al 2013; Barnes et al 2014), watershed modelling and erosion assessment Valeriano et al 2006;Park et al 2011;Kinsey-Henderson and Wilkinson 2013), morphometric analysis (Ozdemir and Bird 2009;Guth 2011;Thomas et al 2012) and stream assessment (Vocal Ferencevic and Ashmore 2012).…”

Marine turtles are not fussy nesters: a novel test of small-scale nest site selection using structure from motion beach terrain information

“…Terrain analysis was performed using SAGA GIS 2.0 (Cimmery, 2010) to obtain profile curvature (Romstad & Etzelmüller, 2012) and terrain ruggedness (Sappington, Longshore & Thompson, 2007). In total, we measured 11 attributes for each nest and reference site; a full description of all metrics used to define nest attributes is given in Table 2.…”

Supervised classification of landforms in Arctic mountains

“…The reliance on expert interpretation subjects this method to varying degrees of accuracy . Remotely sensed methods provide an alternative and there is a significant body of work dedicated to the study of geomorphometry as reviewed by Romstad and Etzelmüller (2012) . This work is supported by improvements in the quality of remotely sensed data that have allowed more detailed analysis into remote areas and at regional scales .…”

“…13 Remotely sensed methods provide an alternative and there is a significant body of work dedicated to the study of geomorphometry as reviewed by Romstad and Etzelmüller (2012). 14 This work is supported by improvements in the quality of remotely sensed data that have allowed more detailed analysis into remote areas and at regional scales. 13,[15][16][17] These automated classification methods allow for larger areas to be mapped more quickly while reducing human error (although introducing machine error), and facilitating comparable results and model transferability.…”