Uncertainty analysis of a spatially-distributed hydrological model with rainfall multipliers

Abstract

This paper has developed an input error model to account for input uncertainty, and applied the rainfall multiplier approaches to the calibration and uncertainty analysis of Soil and Water Assessment Tool (SWAT), a spatially-distributed hydrological model. The developed input error model has introduced the season-dependent rainfall multipliers to the Bayesian framework and reduced the dimension of the posterior probability density function. The method is applied to a watershed located in Southwestern Ontario, Canada. The results of the developed method are compared with two other methods. The SWAT model parameters and the input error model parameters are jointly inferred by a Markov chain Monte Carlo sampler. The results show the measured precipitation data overestimates the true precipitation values for the study area. The uncertainty in model prediction is underestimated for high flows and overestimated for low flows. There is no significant change in the estimation of parameter uncertainty and streamflow prediction uncertainty in the developed method from those in the other methods. The study emphasizes that the rainfall multiplier approaches are applicable to spatially-distributed hydrological modelling for accounting of input uncertainty.