The Importance of Modelling Pattern and Structure in Automated Map Generalisation

We argue that there is special value in viewing any given map as a unique collection of patterns. Symbolisation acts to give emphasis to specific qualities among those patterns. The paper builds on the idea of being able to place any given pattern within a multi dimensional ‘pattern space’ (borrowing from Edelman’s ideas on shape space). The paper also borrows from the ideas of Zacks and Tversky’s ideas on event structures, and argues that patterns can be modelled as objects within an object oriented paradigm. The authors are interested in how patterns manifest themselves and ‘morph’ as we transition scale. A focus of interest is how patterns behave as we transcend large changes in scale – leaping Muller’s conceptual cusps that arise from large changes in scale. In exploring qualities of pattern, we seek to highlight those qualities that should be invariant over large changes in scale and argue that they provide a framework for the meaningful evaluation of representations at small scales (low levels of detail).