Date of Award


Document Type

Campus Access Thesis


Civil and Environmental Engineering

First Advisor

Jonathan L Goodall


Watersheds are complicated systems that require sophisticated computer models to support environmental and water resource management decision making. For example, once a watershed model has been built and calibrated for a particular watershed system, it can be used to simulate different hypothetical scenarios such as the impact of increased urbanization on runoff, ground water and streamflow, as well as water quality. One challenge in creating an accurate and useful watershed model is obtaining input datasets for building and calibrating the model, and one of the most important input datasets required is estimates of precipitation volume over the watershed area. Precipitation estimates are measured at gauging stations and by radar instruments. Both approaches, gauges and radar, have benefits and weaknesses in their ability to estimate precipitation. The objective of this thesis was to improve precipitation estimates by combining data from gauge and radar precipitation estimates in an attempt to improve the accuracy of a watershed model. Using the Eno River watershed located in Orange County, NC as the study area, three different precipitation datasets were tested to estimate streamflow for the watershed over the period 2005-2010: (1) estimates based on only precipitation gauging stations, (2) estimates based only on gauged-corrected radar observations, and (3) a new dataset that is the combination of estimates from the gauge and radar data sources selected in a way to better capture observed streamflow at the watershed outlet. The hypothesis is that for different storm events, it is possible that one of the two data sources better captured that storm event, and therefore if precipitation estimates from gauged and radar based sources are combined into a single time series, that combined time series will yield more accurate streamflow estimates. The results from the work show that the combined precipitation significantly improves streamflow estimates (R2 = 0.80; E = 0.74) when compared to the gauged estimates only (R2 = 0.46; E = 0.31) and the radar based estimates only (R2 = 0.62; E = 0.48). Therefore, the results support the hypothesis that combining precipitation from multiple sources improves watershed model accuracy. The experiment was limited to one study watershed, however, and therefore it did not control for factors such as climate, ecology, and hydrogeology that will likely influence the results of the experiment.