A methodology for landscape characterisation based on GIS and spatially constrained multivariate analysis

Abstract

Landscape is about the relationship between people and place and in 2000 was defined by the European Landscape Commission (ELC) as "an area as perceived by people whose character is the result of natural and human actions and interactions”. In the 70s the reason for studying the landscape was because of the necessity of attributing a value to it. Nowadays the motivations behind managing, conserving and enhancing the landscape is because the landscape is the place where people belong to and, consciously or not, recognise themselves. In addition, people identify different landscapes on the basis of the particular combinations of the elements in the landscape. As a consequence a landscape can be distinguished from another on the basis of its character which, according to the Landscape Character Assessment (LCA) guidance for England and Scotland (C. Swanwick and Land Use Consultant, 2002), is defined as “a distinct, recognisable and consistent pattern of elements in the landscape that makes one landscape different from the other rather than better or worse”. This definition was the starting point of a PhD research project aimed at developing and implementing a methodology able to identify and quantify the character of the Scottish landscape through the application of GIS and statistics. The reason for doing this research was to provide the landscape architects and practitioners with a tool that could help them to define the landscape character types in a more consistent, objective, and scientifically robust way. One of the objectives of the research was to identify the spatial patterns formed by the landscape elements by taking into account the influence of the spatial location. The first law of geography, which states that "everything is related to everything else but near things are more related than distant ones" (W Tobler, 1970), was transposed in the assumption of the presence of spatial autocorrelation amongst the data which contributes to form spatial patterns within the data. Since landscape comprises of many elements, data were also multivariate, thus the analysis required a method of calculation able to deal simultaneously with multivariate and spatial autocorrelation issues. MULTISPATI-PCA, a spatially constrained Principal Component Analysis, was the statistical technique applied for the analysis of the data whose results showed that it was possible to detect the spatial structure of the data and that each spatial pattern corresponded to a distinct landscape. Despite their importance in forming the character of the landscape, aesthetic and perceptual aspects were not inlcuded in MULTISPATI-PCA analysis. It was preferred to test the technique only on data that were quantifiable in a more objective way. Perhaps taking into account the human perception of the landscape can be the starting point for future investigation.