STREAM FLOW MODELING IN THE NACUNDAY RIVER BASIN (PARAGUAY, SOUTH AMERICA) USING SWAT MODEL

Abstract

Paraguay is a very rich country in water resources, owing to generous rainfall regime, very favorable weather conditions and has significant water resources development potential. To promote planning and development of water resources in the country, systematic studies are required to understand stream flow behavior. The present study is therefore, undertaken to study the stream flow behavior in the Nacunday river basin using the SWAT Model. The Nacunday river basin is predominantly agricultural watershed with soybean as main crop which is grown under no tillage practices and can be considered as representative watershed in the region. Available hydrological data (January 1999 to September 2009) is split into two groups, one group (i.e. Jan 1999 to Dec 2005) is used for the model warm up and calibration of model parameters. The daily data of 1999 and 2000 is considered as model warm up period and the data from January 2001 up to the end of December 2005 is used for calibration of model parameters. The other group (i.e. Jan 2006 to Sep 2009) is used for validation of the model. During the calibration, the model performance is adjudged by comparing SWAT outputs with measured stream flow using graphical comparisons and three statistical measures viz. Percent Bias (PBIAS) on annual basis, Nash-Sutcliffe Efficiency (NSE) and Coefficient of Determination (R2) both for daily and monthly basis. During the calibration period the PBIAS varied between ±16% which could be considered as good overall annual simulation. The NSE and Rz on monthly data series were 0.7 and 0.73 respectively which indicate very good performance of the model. On daily data series the NSE and R2 were 0.61 and 0.65 respectively indicating satisfactory model performance. The performance of the SWAT model was validated using daily data from January 2006 to September 2009. During the validation period, the PBIAS remained within ±163% except for year 2006 where the annual PBIAS of 33.4% was obtained indicating under simulation of discharge from model. Therefore on the basis of PBIAS the model performance can be rated as good except for year 2006. On monthly basis the NSE and R2 were 0.61 and 0.45 respectively and for daily basis NSE and R2 0.55 and 0.58 respectively indicating satisfactory model performance. iv ABSTRACT Although the model performance was influenced by uncertainty of the available observed data, the simulated discharge from SWAT model, both for daily and monthly basis in Nacunday watershed can be rated within acceptable range of errors, so future use of the SWAT model for various scenario testing is reasonable. v