R. K. Straumann
Department of Geography, University of Zurich, Winterthurerstrasse 190, CH-8057 Zurich, Switzerland
Telephone: +41 44 635 51 98
Fax: +41 44 635 68 48
The aim of this paper is to briefly review some of the work in landform-related research, to highlight the need for a sound ontological basis to such efforts and to present and discuss an approach for gathering domain knowledge and the problems encountered. There is a considerable volume of geomorphometric literature centred on topographic eminences, specifically publications regarding the delineation or extraction of hills or mountains; e.g. Fisher et al. (2004), Chaudhry and Mackaness (2007) and (less spatially) Greatbatch et al. (2007). However, many of the approaches to landform delineation/ extraction start to rapidly delineate crisply or extract fuzzily the desired objects. The authors rely to a certain degree on some common knowledge about what a mountain, a hill or a range is and, to a certain degree, assume that these concepts match what they extract. But there is fundamental work which acknowledges that these and similar landform concepts are not clear at all and that research into concepts and their formalisation is needed (e.g. Brändli 1996, Schmidt and Dikau 1999). A strain of research tries on a fundamental level to elucidate the ontology of geographic objects and with these, of landforms (e.g. Smith and Mark (2001, 2003), Smith and Varzi 2000, Mark and Smith 2004, Mark and Sinha 2006). Furthermore Mark et al. (2007) set out clearly why conceptualisations of landforms are not alike for people of different cultures or language groups.
In Geographic Information Science and geomorphometry there are a range of researchers who have developed methods to describe earth surface forms from digital elevation models (DEMs). Some of this interest has turned away from the mere description of predefined areas of land through DEM derivatives such as hypsometry, gradient or aspect to semantically richer characterisations of surface form. There is a huge breadth of publications related to the extraction of units that are homogeneous in relation to some surface properties. These units are termed (among others) “landform elements”. Besides these landform elements, there has been growing interest in recent years concerning landforms, i.e. larger regions of similar form character.
Besides, geographic information science has seen some effort to render geographic information systems more usable for lay-persons (e.g. Mennis et al. 2000). This has spawned a considerable amount of work on fuzzy spatial relations, such as something “being near” something else (e.g. Robinson 2000). Given this background, we think a similar case can be made that in the long run it would be valuable to enable GISs to make sense of landform terms such as “valley” or “mountain”.
We thus argue that – although there are already many extraction and classification algorithms especially for what are termed landform elements – there is a need to further strengthen the ontological basis such approaches implicitly rely upon. It might be insightful to review the breadth of landforms, to characterise their properties and interrelationships before pondering about ways to extract them from DEMs. We feel such an endeavour may serve well in order to strengthen efforts towards their automatic extraction, specifically, and towards landscape characterisation from DEMs in general.