An analysis of semi-natural vegetation from remotely-sensed data
This thesis examines the potential of remote sensing to characterise heathiand vegetation. A series of ground radiometer and airborne multispectral scanner experiments were conducted to investigare relationships between remotely-sensed data and the species composition of heathland vegetation. Particular reference was made to spatial characteristics of the vegetation. The results refuted the hypothesis that a classification of remotely-sensed data accounts for the variation in heathland vegetation communities. A classification was an inaccurate and inappropriate representation of the vegetation. However, it was cautioned that any assessment of remote sensing was dependent on the accuracy and method of processing of ground data. An investigation of alternative methods of analysis found that multispectral data were related to continuous variations in the abundance of dominant species, and it was inferred that classification underestimated the potential of remote sensing. Vegetation continua were successfully identified from remotely-sensed data acquired throughout the season, and a transformation of the data generated spectral components that helped to explain the interaction between species composition and multispectral reflectance. A particular criticism of classification had been the inability to represent the spatial characteristics of vegetation due to the nominal nature of the output. Abrupt boundaries, continuous transitions and homogeneous stands were identified from the vegetation data. Alternative methods, using ordinal-level remotely-sensed data, were able to represent several spatial characteristics of the vegetation. The full potential of remote sensing to characterise heathland vegetation will only be realised once the processing of remotely-sensed data is coupled to ecological models. The methods evaluated in this thesis are a first step in that direction, but they require a better understanding of spectral and spatial properties in order to meet the routine information needs of environmental scientists at local to global scales.