O. Chaudhry, W. Mackaness

Institute of Geography, University of Edinburgh


Being part of continuous phenomenon, real world natural features do not have exact or precise spatial extents (in the way that anthropogenic features do). Nevertheless it is common to categorise landscape into different morphological structures such as mountains, hills, valleys for the purpose of analysis or visualisation.

There is a scale dependency to the precision of their boundaries. Current Geographic Information Systems lack data schema to deal with this fuzziness. Their discrete boundaries are objectively defined, loosely connected with our prototypical views on what is meant by a mountain or a mountain chain. This paper presents a quantitative technique that automatically identifies landscape features such as mountains, hills and ranges, and creates the partonomic structure that links membership between these phenomena, and stores these discrete objects.

This paper examines the relationship between scale and landscape features in the context of their representation in map form. The research is premised on the idea that large scale features are comprised and defined by the smaller features that comprise them (that mountain chains are a collection of clustered yet individually identifiable mountains). Before we can begin identifying the higher order features, we must first develop techniques for automatically discerning the smaller features. A mountain is defined by its prominence (relative height among surrounding features), extent, and a variety of morphological characteristics. The algorithm presented here uses derivatives of elevation and the density of morphological variation in order to automatically identify individual mountain and hill extents. This database transformation process is viewed as an essential prerequisite to cartographic portrayal of these features at a range of scales. The approach was applied to the identification of hills and ranges in and around Edinburgh and Fort William Scotland, UK. The results obtained were evaluated against a name database generated by cartographers. The discussion focuses on the importance of results in terms of automatic text placement and in terms of cartographic generalisation and also their use for spatial analysis. The research reflects the subjective nature in defining what is hill or mountain, but reminds us that a map seeks to capture the essence and characteristic form of the landscape something that is necessarily fuzzy and scale dependent.