A global sensitivity analysis tool for the parameters of multi-variable catchment models

Over-parameterisation is a well-known and often described problem in hydrological models, especially for distributed models. Therefore, methods to reduce the number of parameters via sensitivity analysis are important for the efficient use of these models. This paper describes a novel sampling strategy that is a combination of latin-hypercube and one-factor-at-a-time sampling that allows a global sensitivity analysis for a long list of parameters with only a limited number of model runs. The method is illustrated with an application of the water flow and water quality parameters of the distributed water quality program SWAT, considering flow, suspended sediment, total nitrogen, total phosphorus, nitrate and ammonia outputs at several locations in the Upper North Bosque River catchment in Texas and the Sandusky River catchment in Ohio. The application indicates that the methodology works successfully. The results also show that hydrologic parameters are dominant in controlling water quality predictions. Finally, the sensitivity results are not transferable between basins and thus the analysis needs to be conducted separately for each study catchment.