Urban land classification and its uncertainties using principal component and cluster analyses: A case study for the UK West Midlands

An urban land-cover classification of the 900 km2 comprising the UK West Midland metropolitan area was generated for the purpose of facilitating stratified environmental survey and sampling. The classification grouped the 900 km2 into eight urban land-cover classes. Input data to the classification algorithms were derived from spatial land-cover data obtained from the UK Centre for Ecology and Hydrology, and from the UK Ordnance Survey. These data provided a description of each km2 in terms of the contributions to the land cover of 25 attributes (e.g. open land, urban, villages, motorway, etc.). The dimensionality of the land-cover dataset was reduced using principal component analysis, and eight urban classes were derived by cluster analysis using an agglomeration technique on the extracted components. The resulting urban land-cover classes reflected groupings of 1 km2 pixels with similar urban land morphology. Uncertainties associated with this agglomerative classification were investigated in detail using fuzzy-type analyses. Our study is the first report of a quantitative investigation of uncertainty associated with a classification of this type. The resulting classification for the UK West Midland metropolitan area offers an impartial basis for a wide range of environmental and ecological surveys. The methods used can be adapted readily to other metropolitan areas where generic urban features (e.g. roads, housing density) are gridded.