Holistic aspects of suburban landscapes: visual image interpretation and landscape metrics

Holism deals with extremely complex structures that are difficult to analyse. The holistic approach of landscape research was stimulated by the introduction of aerial photography. Carl Troll called “aerial photo-interpretation as a high degree of landscape ecology,” which illustrates the intimate link between the bio-philosophical theory of holism and the Gestalt-theory of human perception, which is important in image interpretation. Aerial photography proved to be a valuable instrument, not only to make thematic inventories and monitor changes, but also to describe holistic aspects of complex landscapes. Contrarily, the analytical approach aims to describe transcendent characteristics of complex structures by quantitative variables. Many landscape metrics have been proposed and used, such as entropy as an indicator of fragmentation and heterogeneity. A comparison was made between a typology and classification of a highly complex suburban landscape with aid of a holistic-based approach for visual interpretation of aerial photographs and a quantitative approach using landscape metrics of the patches formed by the land use. The study area covers the suburban transition of the city of Ghent (Flanders, Belgium) to the rural countryside to the west and is characterised by a large diversity of land use and a severe fragmentation of infrastructures. The visual interpretation of aerial photographs allowed rapid and easy classification into 10 distinct and landscape types, which could be mapped. The landscape metrics used were patch area, patch shape, fractal dimension of the patch borders and two forms of information entropy. The metrics were derived from the land use map based upon a supervised classification of satellite data and improved by ancillary data with a resolution of 20 m. Overall, the fractal dimension and entropy are significantly different for the landscape types, while those for the patch area and shape are not. Landscape metrics are very sensitive to data quality, particularly in complex and very heterogeneous landscapes. Interpretation of the results is also made more difficult due to the effects caused by the algorithms of spatial analysis used to determine patches, and due to the noise added by index calculation. These effects are reflected on maps representing the distribution of the landscape metrics. The patterns revealed here depend largely upon the method of classification and visualisation. The summed entropy corresponds most closely to the landscape units defined by visual image interpretation. It can be used as a quantitative characteristic of holistically defined landscape units.