The effects of input data degradation on hydrological model performance for a snowmelt dominated watershed
Date
2006
Authors
McKim, Scott D., author
Fassnacht, Steven R., advisor
Journal Title
Journal ISSN
Volume Title
Abstract
The quality and quantity of hydrometeorological data used as input to a hydrologic model is varied and the output compared to observed historical flows. Temperature and precipitation data were used to feed the National Weather Service River Forecast System (NWSRFS); this hydrologic model outputs streamflow and is used daily throughout the country to forecast streamflows. NWSRFS is a lumped empirical model developed in the 1970s for the NWS and is calibrated in this study to model a portion of the snowmelt dominated Yampa River watershed in northwest Colorado. An analysis scheme is followed to capture the model's dependence on representative meteorological stations located in an around the modeled basin. Many regions in the United States experience meteorological and hydrological data scarcity issues. Operationally this becomes important when the available data is insufficient enough to produce reliable model outputs. Similar to Tsintikidis et al. (2002) concluding that the installation of additional rain gauges in a modeled basin would decrease the error of precipitation measurements in the model, we sought to find if increasing data input into a model, both the quantity and quality given by site representivity, will increase the accuracy of our model runs. The study basin was chosen for its snowmelt dominance characteristic. Mean areal precipitation and temperature values for the modeled zones are developed individually in each analysis scheme by the arrangement of stations used in each sensitivity analysis. A statistical analysis of the relative difference between model runs and archived observed values is performed in an effort to illustrate the effect of different model input data arrangements on model simulations. This study aimed at testing the tenable assertion that subtracting hydrometeorological data from a model's dataset would decrease the accuracy of forecasted stream flows from that model. Stream flows and snow water equivalence are analyzed to test the model's sensitivity to the amount of data used. Since the NWSRFS uses predetermined weights to determine MAPs, the number of stations used does not significantly affect model output. The usage of predetermined weights maintains a consistent year-to-year MAP. Varying the MAT station configuration showed a more sizeable effect than the MAP scheme illustrated. Though this procedure could and should be replicated for other hydroclimates and for basins with different sizes, the specific results are not transferable to other basins. The basin modeled is very heavily snowmelt dominated; this quality, as well as it size, climate, topography, and available hydrometeorological stations all influence model results; altering any of these would change the model performance.