What is land cover?

Much geographic information is an interpretation of reality and it is possible for multiple interpretations to coexist. This is unproblematic for the research community but, as the numbers of users increase through initiatives resulting in data integration on an unprecedented scale, such as E-science and GRID, issues of information meaning and conceptualisation become more important. We explore these issues through the mapping of land cover and the variety of conceptions of land-cover features that may be held by actors in the creation, distribution, and use of the information. Current metadata do not report the wider meaning of the information categories in terms of the decisions that were made and by whom in specifying class conceptualisations. DOI:10.1068/b31135 }Present address: Environment Systems, ADAS, Wolverhampton WV6 8TQ, England; e-mail: [email protected]. #Corresponding author and new address: Department of Information Science, City University, Northampton Square, London EC1V 0HB, England; e-mail: [email protected]. (2) Many users are interested only in the (digital) map. Fisher (2003) documents the shift away from extensive reports accompanying the mapped information as metadata and comments that `̀ fewer than ever [users] are even aware of the existence of the survey report'' (page 15). (3) Potential users do not have to go through lengthy processes of data selection involving dialogue with the providers, nor do they often have to go through the time and expense of capturing the data through abstraction and digitising. Rather they are able to transfer the data to their local system over the Internet or from local highcapacity storage devices. Therefore, strong financial incentives exist to use the readily available digital data in preference to any other source. If the data are shown to be completely unsuitable for a particular analysis, then the user can search for another source. (4) Current metadata standards (ISO, 2003) are adequate to guide assessment of technical constraints on data integration caused by structure (raster to vector) or scale (generalisations to lower level classes); but they convey nothing about the organisational (cultural) or epistemological context which gave rise to the data in the first place. The net result of reducing the effort required to obtain the data also reduces the incentive for users to understand those data in a wider sense. One of the consequences of this whole situation is that extensively manipulated information is treated as data by users who do not fully understand what it represents: its meaning or semantics. They assume that it fits their conceptualisations because of familiar class names and labels that apparently match their prototypical categories with those names. Unfortunately for almost all users the available information can only be a surrogate for the specific information they actually requireöa situation of which they may be unaware. The consequences of not fully understanding the conceptualisations and specifications hidden beneath familiar class labels are na|« ve and flawed analyses, a situation that many users may not be prepared to acknowledge, and one which is hard to document. Integration benefits can only be properly realised if the differences in data meaning are overcome. Even the simplest concepts can cause problems. For example, in the CORINE land-cover map of Europe a `beach' is strictly above the highest astronomical tide, not below it, whereas in Britain, according to Forest Enterprises and a `forest' might not even have any trees on it, and, in both Scandinavia and Eire, land covered in slow-growing trees might not be forest at all. Bennett (2001) discusses the enormous complexity involved in analysing conceptualisations of forest. Figure 1, constructed from the data of Lund (2004), illustrates different conceptualisations of forest from around the world, based on the physical parameters of tree height and crown canopy cover. Note that, in their definition of forest, many countries include land that could be under trees, or where there is a probable intention to replant in the foreseeable future; also that many countries include bamboo and palms in their definition of what constitutes a `tree' and hence a forest, although others do not, even where those species are common. In this paper we explore the generic problem of relativism in spatial information by using land-cover mapping from satellite imagery as an example. The distinction between geographic data and information is described and the link made between information and conceptualisation (section 2). In section 3 we review some aspects of category theory. In section 4 we describe the origins of differences in the meaning of land-cover classes and their roots in the different communities within remote sensing and in section 5 we discuss how various meanings are inferred by different communities of users. The social construction of land cover is discussed (section 6) before some concluding comments (section 7). 200 A Comber, P Fisher, R Wadsworth