Determining Effective Impervious Area for Urban Hydrologic Modeling

Abstract

This paper presents a two-step process to estimate the fraction of an urban watershed covered by a hydraulically effective impervious area. The first step applies maximum likelihood classification of fine-scale multispectral satellite imagery to derive urban land cover. The second step uses an automated macro in a geographic information system to trace the water flow path from pixels classified as impervious and subclassify them as noneffective or effective. The two steps were verified independently, with verification of the second step using idealized data. The two-step process was then tested with a small watershed study of model calibration and rooftop connectivity impact on runoff. At the watershed scale the land cover classification differences were approximately 6%, while at the pixel scale matches of 50, 60, and 83% were achieved for the rooftop, asphalt/concrete, and vegetation land covers, respectively. The effective impervious area was estimated to comprise 16% of the watershed surface, which was close to the actual value of 22%. At the pixel scale, the effective impervious area match was less accurate at 48%. Differences in both land cover and effective impervious area classification at all scales are attributed to high land surface heterogeneity, data limitations and errors, and tree canopy covering impervious surfaces. The verification tests and runoff simulations validate the method as a useful means to rapidly estimate with reasonable accuracy an essential urban hydrologic model parameter.