MS, Master of Science
Thesis - Open Access
Department of Earth, Ocean and Atmospheric Sciences
Committee Chair - Dr. Henry E. Fuelberg
Committee Member - Dr. Guosheng Liu
Committee Member - Dr. Paul H. Ruscher
This study evaluates the impacts of two types of rainfall input on simulated streamflow using a specialized, fully-distributed hydrologic model—the Watershed Assessment Model (WAM). We compare gauge-only Thiessen polygon input data with the gridded 4 × 4 km Florida State University (FSU) version of the National Weather Service (NWS) Multi-sensor Precipitation Estimator (MPE) scheme. Streamflow results are compared to observed amounts over a six year period (2000-2005) at two U.S. Geological Survey (USGS) stream gauge sites in the greater Florida Suwannee River basin. One catchment has an area of 1505 km2, while the smaller catchment is 500 km2. Previous comparisons have been made between the two different precipitation data types using mean areal precipitation calculations over several Florida basins. This study of streamflow expands on those findings.
Results show significant differences in simulated streamflow when the higher-resolution FSU MPE rainfall data are input to WAM. However, the FSU MPE dataset does not always provide better results with this model configuration. The improvements in WAM simulated streamflow depend on a combination of factors, including the desired type of comparison with observed amounts (volume or correlation), rainfall pattern characteristics, and individual event scenarios.
The accumulations of FSU MPE WAM streamflow generally are found to be more accurate than those from Thiessen polygons. During drought periods, MPE-derived streamflow provided more accurate accumulations, but coefficients of determination were not always improved. During years with more average rainfall events, FSU MPE produced greater underestimates of accumulation amounts, and thus a better approximation by the Thiessen polygon input. Seasonal results emphasized the weaknesses of each data source. Rain gauges usually are not able to capture the small scale spatial variability of summer rainfall events. And, radar-derived precipitation generally is underestimated during relatively low top stratiform winter events.
When simulating streamflow with a hydrologic model using rain gauge input, it is apparent that gauge locations are very important. Generally speaking, increasing the spatial density of gauges will produce a better representation of rainfall. Our small basin was found to be prone to significant underestimates of accumulations and lower coefficients of determination regardless of the rainfall input. However, statistical differences between our larger and smaller basins are not as dramatic with the FSU MPE data.
Current results are based on the WAM model as configured for this study. Results from other models and/or other configurations may be different. Although there appear to be errors in both WAM’s ability to utilize the rainfall data properly and in the rainfall data measurements themselves, the results highlight areas where both can be improved.
Sullivan, John Lawrence Jr., "Modeling Streamflow Using Gauge-Only Versus Multi-Sensor Rainfall" (2008). Electronic Theses, Treatises and Dissertations. Paper 403.