| Abstract: | The increasing sophistication of classification techniques used in land use and land cover analysis has not been matched by attention to the origin and effects of land cover categories. While classifications appear unproblematic and self-evident, they carry with them their own histories, meanings and effects, which remain largely unexamined. In an effort at such scrutiny, we examine the origins of land cover categories deployed in remote sensing and conclude that categories are theory-laden metaphors and occur epistemologically prior to any clustering algorithm, no matter how sophisticated. We describe the problematic effects that the imposition of classification systems in place of in situ knowledge of the landscape can have, especially in a colonial or post-colonial context. As an alternative to imposed classification, we propose and demonstrate an empirical technique based upon a growing body of work in participatory GIS. The method compares image classifications based on local and expert knowledge, using a case study from Rajasthan, India, concluding that differing metaphors of landscape lead to divergent measures of land cover. |
