Doctor of Philosophy (PhD)


Civil and Environmental Engineering

Document Type



This study identified important uncertainties affecting watershed-scale flow and water quality simulations and recommended modeling options for reducing the uncertainties by means of the Hydrological Simulation Program-FORTRAN (HSPF) model. Specifically, parameter-induced uncertainties in HSPF model were estimated along with uncertainties associated with DEM resolutions and sources. National Elevation Dataset (NED) DEMs with original and resampled resolutions of 3.5, 10, 30 and 100m were utilized for manual and automatic delineation of three watersheds with different gradients in HSPF model to identify the effects of DEM resolution, DEM resampling, delineation method and watershed gradient on the simulation of streamflow, nitrate (NO3), dissolved phosphorus (P) and total suspended sediment (TSS). Furthermore, the Generalized Likelihood Uncertainty Estimation (GLUE) method was applied to estimate the parameter-induced uncertainties in HSPF model. Uncertainties originated from DEM sources were compared to DEM resolution-induced uncertainties using 30m original DEMs and 100m resampled DEMs from National Elevation Dataset (NED), SRTM (Shuttle Radar Topography Mission), and ASTER (Advanced Space-borne Thermal Emission and Reflection Radiometer) sources.

Results from this study indicated that the simulation of flow and water quality parameters via HSPF model is sensitive to DEM properties (particularly resolution and source). It was found that sediment was the most sensitive and NO3 was the least sensitive parameters to the variation in DEM resolutions and sources. Greater DEM resolution-induced uncertainties were involved in watershed delineation and subsequent simulation of flow, NO3, P and sediment for the study watershed with a lower gradient. The effect of DEM resolution on flow, NO3 and P simulations was lower using resampled DEMs and was negligible when the manual delineation method was used. For watersheds with higher slopes, parameter uncertainties were shown to be substantially greater than resolution-induced uncertainties, meaning that the calibration of HSPF parameters within their feasible ranges can alleviate the resolution-associated uncertainties to a great extent. It was also found that uncertainties propagated into HSPF simulations via DEM sources are more notable than the uncertainties from DEM resolution. In the study watershed with a lower gradient, higher uncertainties were observed due to high sensitivity of extracted topographic features and subsequently simulated flow to DEM sources.

Committee Chair

Deng, Zhiqiang