Automatic determination of imperviousness in urban areas from digital orthophotos

Determination of impervious areas in urban regions is the most labour-intensive part of data acquisition for rainfall-runoff modelling in urban hydrology. This paper presents an automatic determination method of imperviousness from aerial photographs. The colour, CIR (colour infrared) aerial photographs and orthophotos used have a ground resolution of 25 to 75 centimetres. A maximum likelihood classification algorithm was applied to assign each pixel to a surface category. Classification results were then then overlaid with the subcatchments to determine the imperviousness of each subcatchment. Classification and overlay were carried out with the raster-based GIS IDRISI. The method was tested on various catchment areas, and the results compared with data obtained from manually digitised surfaces. Accuracy of the estimated imperviousness for the entire catchment areas was within 10 %. The deviations for individual subcatchments were much higher. Equivalent results were obtained for colour and CIR photograplhs. A combination of both spectral ranges resulted only in a slight improvement. Consequently, this does not justify the additional costs of the second image. The developed method is an interesting alternative for use on large catchment areas where manual digitisation is very time-consuming and, thus, expensive.