S. Perikleous, W.A. Mackaness, O. Chaudhry

Institute of Geography, The University of Edinburgh


The extent of forested parcels of land is captured at fine detail typically from mapping at 1:1 250 or 1:10 000 scale (large scale). There are various situations in which a more synoptic view of a whole region is required, in which a more general view of the extent of forest is required. With the current emphasis on capture once use many times, research is focusing on developing map generalisation methodologies that enable the more general view to be automatically derived from the large scale view. In particular this research is concerned with the automatic derivation of forest patches for representation at 1:250 000 (or an equivalent level of detail).

The methodology of the generalization algorithm is loosely based on an idea from Muller and Wang but with significant differences.

Input came from Ordnance Surveys MasterMap(1:1250/1:10,000) Forestry layer which is a vector-based topography layer. Example output compared favourably with hand drawn maps at equivalent scale. The algorithm has removed regions of forestry inconsequence and given additional emphasis and regionality to broader forested areas. The output obtained was evaluated by comparing with OS Strategi dataset (1:250,000) and results were also evaluated by a Ordnance Survey cartographers. Though favourable results were achieved the research pointed to additional work that could usefully be undertaken.

The research is part of the long term research in pursuit of systems capable of generating multi scaled maps from a single detailed database with nominal human intercession in the map compilation process. The algorithm was implemented in Java and Eclipse. Java Topology Suite (JTS) and Jump libraries were used. The algorithm is available via the Ordnance Survey map generalisation webpage.