GSA Annual Meeting in Denver, Colorado, USA - 2016

Paper No. 295-7
Presentation Time: 3:00 PM

AN ANALYSIS OF PRECIPITATION DATA SOURCES FOR THE DEVELOPMENT OF A SWAT MODEL FOR THE CHEMUNG RIVER WATERSHED, NEW YORK AND PENNSYLVANIA, USA


KAUSHIK, Pankaj R., Department of Geology, Kurukshetra University, Kurukshetra, India and NOLL, Mark R., Department of the Earth Sciences, SUNY College at Brockport, 350 New Campus Dr, Brockport, NY 14420, pankajkaushik1993@gmail.com

The application of hydrologic models such as SWAT has become common place. This model and similar ones rely on quality sources of precipitation data. Several sources of data are now commonly available including rain gauge networks, gridded reanalysis data and radar precipitation estimates. Questions remain, however, regarding the best source of data. In this study, several sources of data were evaluated with no other changes made to the SWAT model in an effort to evaluate the best source of precipitation data for the case of the Chemung River watershed. The Chemung River watershed stretches over seven counties in western New York and three counties in Pennsylvania, and includes the cities of Corning and Elmira, NY in the southeastern portion of the basin. The Chemung is major tributary for the upper Susquehanna River. Results found that two common sources of data, rain gauge networks and gridded reanalysis data give similar results for discharge, but varying results when evaluated using the Nash Sutcliffe test. Reanalysis data appears to most closely resemble observed stream discharge data, but has a large amount of variability. This produces a Nash Sutcliffe value of -0.08, which is not acceptable. Rain gauge data produces a better model with a Nash Sutcliffe value, 0.25, but typically underestimates discharge. Neither of these produce a significant correlation with actual discharge data. Combining the two data sets produce superior results. When rain gauge precipitation values are averaged with reanalysis precipitation values from the grid points nearest the rain gauge locations, the Nash Sutcliffe value for the resultant model is increased to 0.38, and the correlation between model and actual discharge data is found to be significant. Evaluation and optimization of the precipitation input should be evaluated in areas with reasonable data resources to produce more accurate SWAT models.