Sensitivity analysis and identification of the best evapotranspiration and runoff options for hydrological modelling in SWAT-2000

• Published in: Journal of hydrology (Amsterdam) 2007-01, Vol.332 (3), p.456-466
• Main Authors: Kannan, N., White, S.M., Worrall, F., Whelan, M.J.
• Format: Article
• Language: English
• Subjects: Colworth; Curve number (CN); Earth sciences; Earth, ocean, space; Exact sciences and technology; Freshwater; Hydrological modelling; Hydrology. Hydrogeology; Sensitivity; Stream flow; SWAT
• ISSN: 0022-1694; 1879-2707
• DOI: 10.1016/j.jhydrol.2006.08.001

Distributed models used in hydrological modelling, have many parameters. To get useful results from the model, every parameter is required to have a sensible value. Usually a calibration is undertaken to reduce the uncertainties associated with the estimation of model parameters. To ensure efficient calibration, a sensitivity analysis is conducted to identify the most sensitive parameters. This paper describes simple and efficient approaches for sensitivity analysis, calibration and identification of the best methodology within a modelling framework. For this study, the SWAT-2000 model was used on a small catchment of 141.5 ha in the Unilever Colworth estate, in Bedfordshire, England. Acceptable performance in hydrological modelling, and correct simulation of the processes driving the water balance were essential requirements for subsequent pesticide modelling. SWAT gives various options for both evapotranspiration and runoff modelling. Identification of the best modelling option for these processes is a pre-requisite to achieve these requirements. As a first step, a sensitivity analysis was conducted to identify the sensitive parameters affecting stream flow for subsequent application in stream flow calibration. Hydrological modelling has been carried out for the catchment for the period September 1999 to May 2002 inclusive using both daily and sub-daily rainfall data. The Hargreaves and Penman-Montieth methods of evapotranspiration estimation and the NRCS curve number (CN) and Green and Ampt infiltration methods for runoff estimation techniques were used, in four different combinations, to identify the combination of methodologies that best reproduced the observed data. In addition, as the initial calibration period, starting in September 1999, was substantially wetter than the following corresponding validation period, the calibration and validation periods are interchanged to test the impact of calibration using wet or dry periods.