An integrated approach to land cover classification: an example in the Island of Jersey

A land cover map of Jersey was created using remotely sensed images recorded by satellite. This map brought together a range of disparate techniques, developed in isolation and mostly applied experimentally, integrating multisensor, multitemporal, enhanced spatial resolution data within an object-oriented integrated Geographical Information System (GIS) for an applications-driven, operational programme. It was developed under the Classification of Environment with Vector- and Raster-Mapping (CLEVER-Mapping) project: this improved approach to operational land cover mapping used information on the subdivision of the landscape into land parcels to help classify remotely sensed images on a per-parcel basis. The object-oriented approach allowed the use of remotely sensed information which relates directly to ground features, and the application of improved knowledge-based corrections using a range of external data. Unlike a conventional map, the parcel-based approach produced a GIS database containing classified land parcels which could also be used as a storage framework and analysis tool for other datasets in later analyses. The GIS recorded 21 land cover types. Validation against reference land parcel data gave a correspondence of between 85% and 95% depending on the level of class aggregation.