Analysis of the utility of remote sensing data for urban hydrologic modeling

Abstract

In this thesis analysis, a methodology is presented for evaluating uncertainty m hydrologic predictions that are based on remote sensing data for parameter estimation. The methodology is applied to the HEC-1 model for a highly developed basin in Scottsdale, Arizona to compare three remote sensing data sources; NSOO 1, Landsat, and SPOT. Hydrologic parameters are estimated using the three remote sensing data sources and the uncertainty in those estimates is determined by a procedure incorporating three sources of uncertainty; image misclassification, error in parameter assignments for a particular landuse class, and aggregation of image pixels to subbasins. The parameter uncertainty is then propagated to model output uncertainty by several different uncertainty analysis methods in order to assess the accuracy of methods more efficient than Monte Carlo Simulation. The results of the analysis were compared for (1) the remote sensing images (2) the different sources of uncertainty in each image, (3) two uncertain parameters, and ( 4) the different uncertainty analysis methods. The results showed that spatial and spectral image resolution was important in identifying model parameters and in the prediction of peak flow.