Comparing the GLC2000 and GeoCover LC land cover datasets for use in economic modelling of land use

This paper compares two recent near‐global land cover (GLC) datasets, GLC 2000 and GeoCover LC, with differing resolutions (1km versus 30 m) for their potential usefulness in economic analyses of the determinants of land use. The comparisons are for two ecologically distinct parts of the world—Sumatra and the Horn of Africa. Cross‐tabs that provide different views of the datasets are developed and reported. We expected to find heterogeneity in 1km categories that were identified as mosaic and did. More surprising is the degree of heterogeneity of the categories one might expect to be homogeneous. For both regions, a 1km pixel often appears to include a wide range of economic activities that are likely to respond to economic and social forces in different ways. Hence, using the 1km data as representative of land use in models of location‐specific socioeconomic response, such as to climate change or land use policies is especially problematic, even if the classification accuracy were very high.