ABSTRACT
According to the recommendations of the Environmental Protection Agency, changes in the ecological landscape may be characterised by three metrics (measures). These are: the measure of diversity, the measure of dispersion and fractal dimensions. The latter is especially interesting with respect to potential use to classify land cover. It can be roughly assumed that the fractal dimension describes the extent of self-similarity of image texture. Proper selection of methods of teledetection data acquisition and dimension calculation may greatly affect its value. The authors have applied what is known as the method of triangular prisms to calculate fractal dimension. They used aerial panchromatic photographs – coloured in natural colours and black and white in IR radiation. The photographs were taken during one flight on 17 September 2007. Moreover, panchromatic photographs taken in 1980 were used to analyse changes of vegetation in the littoral zone of the lake. The effect of ground sample distance on fractal dimension was analysed. The analyses were performed on photographs scanned at various levels of optical resolution. The effect of emulsion graininess and impulse noise (emulsion defects) on the dimension size was also examined. The results suggest that both factors affect the size of fractal dimension. Increasing the ground sample distance results in increasing the fractal dimension. The tendency shows that the texture of images of the natural Earth surface covers are not “pure” fractals. Due to graininess and impulse noise, textures presented in 3D space are very rough. The results of the analysis are also affected by the choice of the analysing window’s dimensions. If local sizes are calculated on small surfaces, the measure may bear a large error. In such cases, noise-reducing or smoothing filters should be used. Credible and repeatable results can be achieved by observing the adopted standard of acquisition and preparing remote sensing data. There are no such standards nowadays and are prepared ad hoc for specific tasks. According to the study, the fractal dimension does not have absolute values and it should be regarded relatively.