Error propagation in distributed hydrologic modeling

This research quantifies the effects of grid cell resolution and attribute errors on hydrograph simulations; utilizes the standard error method to assess the magnitude and spatial distribution of infiltration parameter uncertainty; and then tests a distributed calibration and validation method for three storm events. The overall goal of this work is to advance our fundamental and applied understanding of distributed hydrologic modeling. Grid cell resolution error and attribute error are two important aspects of calibration and validation of distributed models. The largest grid cell resolution that does not cause significant errors in simulation is the critical resolution, below which, the essential variability is captured; above which, significant error is introduced into the simulation.; Using informational entropy it is shown that hydraulic roughness can be sampled at any interval without affecting simulation results. On the other hand, infiltration parameters derived from maps of soil properties for two basins the Blue and the Little Washita exhibit a critical resolution at 900m and 1200m respectively. Further, surface runoff simulations using aggregated infiltration parameters are erratic at resolutions greater than the critical resolution.; The effects of attribute error of soil maps on runoff modeling and calibration of a distributed hydrologic model, r.water.fea are presented. Using the standard error equation, bulk density is identified as the soil property with the most uncertainty. The effect of parameter uncertainty on hydrographs is also simulated and hydraulic conductivity is identified as the infiltration parameter to which the model is most sensitive. Finally, NEXRAD rainfall estimates are used in the calibration and validation of r.water.fea. Hydraulic conductivity and hydraulic roughness are adjusted in the calibration process. The model is calibrated to within 10% of observed values for a storm occurring on May 29, 1994 and validated with similar results for two other storms. The importance of the spatial distribution of rainfall and infiltration in controlling runoff is identified through lumping studies.; The advantage of distributed approaches over lumped methods is that the relative contributions of various parameters are present. Though the spatial pattern is preserved at a particular resolution, the exact values are not known. This uncertainty propagates into the hydrograph simulation as attribute error. Further, before a distributed hydrologic model may be calibrated and validated, the proper resolution must be selected for each parameter affecting the process. This research advances the understanding of grid cell resolution and attribute errors impact on distributed hydrologic simulation.