AUTOMATED GENERALISATION AND REPRESENTATION OF ORDNANCE SURVEY POLYGONAL LANDCOVER DATA AT 1:10 000 SCALE
Ordnance Survey Great Britain, Research Labs, Southampton, UK
Ordnance Survey has very detailed polygonal data for landcover in Great Britain. The classification includes: Coniferous Trees, Non-coniferous Trees, Scrub, Orchard, Coppice, Heath, Rough Grassland, Boulders, Rock, Scree, Marsh, Shingle, Sand and Mud. Within a polygon, combinations of these different landcover types can occur, such as: “Heath + Rough Grassland + Scrub”. Each valid combination has a feature code, and in total there are nearly 500 possible combinations.
Ordnance Survey 1:10 000 scale specifications include less than 100 of these landcover combinations, and no more than three landcover types are permitted within each combination. Combinations which do not conform to the target specification are highlighted and the cartographers reclassify them manually. A finishing process automatically applies vegetation and rock symbol “wallpaper” patterns to the polygons. The “wallpaper” does not look cartographically pleasing when the symbols are broken by the perimeters of the polygons. Legacy hand-drawn rock detail is also shown at 1:10 000 scale. The cartographers often have to reclassify the polygon data to prevent “wallpaper” rock symbols from conflicting with the hand-drawn rock.
This paper presents research into a user interface which allows landcover reclassification rules to be developed and verified. A reclassification rule comprises a condition, followed by a list of landcover types to remove (and occasionally a list of landcover types to add) if the condition is fulfilled. The interface is not just for developing rules for the current 1:10 000 scale; it also allows new specifications and rules to be designed for any target scale. The interface additionally facilitates integration of third-party landcover data with the source polygonal data, for example hand-drawn rock. The output from the interface is a reclassification table which forms an input to a model generalisation process.
Following reclassification, adjacent polygons may have the same feature code; therefore the following step is to merge such polygons. Holes below the minimum size for 1:10 000 scale are subsequently removed. Next, polygons below the minimum size are either merged with the neighbour with the largest area or the neighbour with the largest shared boundary. Lastly the polygons are simplified whilst maintaining the topological consistency of shared boundaries.
The final step places vegetation and rock symbols automatically according to the classification of each polygon. The symbol placement algorithm is designed to avoid conflicts between the symbols and the polygon boundaries. The result is of a high cartographic quality and is achieved with no manual intervention.