Evaluating uncertainty estimates in distributed hydrological modeling for the Wenjing River watershed in China by GLUE, SUFI-2, and ParaSol methods

• Published in: Ecological engineering Vol. 76; pp. 110 - 121
• Main Authors: Wu, Hongjing, Chen, Bing
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.03.2015
• Subjects: algorithms; China; Coefficients; Estimates; Freshwater; GLUE; Hydrological modeling; Hydrology; Mathematical models; ParaSol; rivers; Runoff; Soil and Water Assessment Tool model; SUFI-2; SWAT; Uncertainty; Uncertainty analysis; water management; watershed hydrology; Watersheds; China, People's Rep; China; SWAT; SUFI-2; Hydrological modeling; GLUE; ParaSol; Uncertainty analysis
• ISSN: 0925-8574; 1872-6992
• DOI: 10.1016/j.ecoleng.2014.05.014

Hydrological models always suffer from different sources of uncertainties. As the distributed hydrological models play a very important role in water resource management, reliable quantification of uncertainty in hydrological modeling results is quite necessary. The purpose of this study is to apply three uncertainty analysis methods to a distributed hydrological modeling system, quantify the impact of parameter uncertainties, and examine their performance and capability. Due to the important location and typical hilly features, the upper reaches of the Wenjing River watershed in Western China were selected as the study area. The soil and water assessment tool (SWAT) model was applied to simulate the surface runoff during 1998–2002 and validated by the observed data. After global sensitivity analysis and modeling calibration, the Nash–Sutcliffe coefficient (NSE) and coefficient of determination (R2) values of surface runoff for calibration are 0.75 and 0.80, and for verification periods were up to 0.74 and 0.87, respectively. Three uncertainty analysis methods were further conducted and compared within the same modeling framework: (1) the sequential uncertainty fitting algorithm (SUFI-2), (2) the generalized likelihood uncertainty estimation (GLUE) method, and (3) the parameter solution (ParaSol) method. Through the comparison of a set of proposed evaluation criteria for uncertainty analysis methods in this study, including R-factor, P-factor, the ratio of P-factor and P-factor, computation efficiency, and performance of best estimates (NSE and R2), the SUFI-2 method was able to provide more reasonable and balanced predictive results than the other two methods.