Interrogating Land Cover Categories: Metaphor and Method in Remote Sensing
Robbins, Paul, Maddock, Tara, Cartography and Geographic Information Science
A khaki-clad Indian Forest Service Range officer, sitting in the shade of the guardhouse awning, flips through some photographs of local trees and grasses. He pauses when he gets to a picture of Prosopis juliflora, an exogenous plantation species introduced from Mexico around the turn of the century, locally called Angrezi (English) Babul. "When you look at a picture of the area from the air," he says, "you see 30 percent of the land in forest. Without the Angrezi Babul tree, you would find no forest at all. The tree does our work." In the following conversation, however, he expresses some frustration that such areal photos are hard to obtain and difficult to compare with one another. The common problems faced by the Indian forester are the poor inventories of these "forests" and the lack of tools for detailed mapping of cover.
The promise of remote sensing is to perfect such applications and provide the necessary tools for the forester. Coupled with classification techniques in GIS, including iterative cluster analysis, supervised classification algorithms, and texture analysis, remote sensing renders accurate pictures of the earth's coverage and enables rational planning in the face of very real problems, including deforestation in India. The challenges in this case are to obtain and analyze images of the area and to reveal the causes and areas of deforestation and reforestation. Environmental change research using GIS addresses these challenges, by using GIS to demonstrate and explain human impacts (Turner et al. 1993; Koch et al. 1995; Turner et al. 1995; Geoghegan et al. 1998). Related research in GIS and image processing has focused on the technical problem of producing classifiers that will better reveal differences between cover types on the ground (De Fries et al. 1998).
The Indian Range Officer responds enthusiastically on hearing that remotely sensed data may be available and that these techniques are increasingly possible. Yet a problem is forgotten in the process. In adopting and codifying the forester's notion of forest, other forest-like coverages, each with their own local names and categories, are either collapsed with this invasive canopy cover into a generic category, or are excluded altogether. Specifically, for the beleaguered state forester charged with turning the tide against deforestation in the region, the growing cover of juliflora is heralded as a measurable, verifiable, and quantitative environmental victory. The decline of other forms of tree cover and mixed savanna is entirely overlooked as the forester views the landscape; the coverage of the juliflora tree constitutes "forest" for him, and its spread marks management success.
By naming the monoculture of foreign juliflora, "forest," an objective measurement of land cover change confirms that a third of the land is "recovering" from degradation. In contrast, analysis ordered using local categories, which do not define juliflora coverage as forest, might well reveal a decrease in desirable forest cover. New incentives arise for the continued plantation of juliflora and for the transformation of complex landscapes into simplified, monocultural stands, since to plant juliflora is to increase "forest." The forester anticipates that remote sensing and GIS techniques can demonstrate that forests, now defined as juliflora, are blanketing the land and remote sensing can help to identify new areas for planting or enclosure. Categories on paper in this way lead to new ecologies (Robbins 1998a). This practice of classification, whereby new ecologies are formed through codifying and mapping by state authorities, seems unproblematic to the forester and marks the analytical power of categories that proliferate under what Scott (1998) calls state simplification. "Some level of abstraction is necessary for all forms of analysis," Scott explains, but historically state simplification is notable for "the narrowness of its field of vision, the degree of elaboration to which it can be subjected, and above all . …