SATELLITE IMAGE DATA FOR HYDROLOGICAL MODELLING OF URBAN AREAS

摘要

Rainfall-runoff modelling of urban areas requires precise information on the condition of the ground surface (pervious or impervious surfaces). Land cover classification of high resolution satellite image data could be used as a robust tool for information extraction of pervious and impervious surfaces. The main goal of this paper is to study the applicability of using the land cover classification results derived from high resolution satellite image data to define land surface condition for the rainfall-runoff modelling. Two different highly urbanized areas (residential area and industrial area) within the Greater Toronto Area (GTA) were selected for this study. The extracted surface information was utilized as an input data for the rainfall-runoff hydrological modelling for the 2, 10, and 100-year rainfall events. A pair of high resolution stereo IKONOS satellite images was geo-referenced in order to extract topographic information and generate an ortho-rectified image for the study areas. The previous and impervious surfaces were extracted manually and used as a reference. The ortho-rectified image was classified using the supervised Maximum Likelihood classification technique for both study areas. An assessment of the classification results was performed for the two urbanized areas. The results revealed that there were 7% differences in the area of vegetation calculated automatically in the industrial area, and 44% for the residential area. It was found that the difference in the model-generated runoff from the manual and automatic classification did not vary significantly for the industrial area. However, for the residential area it varied significantly for higher frequency events; e.g. 24% for the 2-year event, and 12% for the 100-year rainfall event.