Sensitivity analysis and auto-calibration of an integral dynamic model for river water quality

Abstract

ESWAT--Extended Soil and Water Assessment Tool--was developed to allow for an integral modelling of the water quantity and quality processes in river basins. ESWAT is a physically based, semi-distributed model, with a moderate-to-large number of parameters and input and output variables (depending on the desegregation scheme). An auto-calibration procedure was implemented for the optimisation of the process parameters. The procedure is based on a new approach for multi-objective calibration and incorporates the algorithms of the Shuffled Complex Evolution Method. The optimisation uses a global optimisation criterion, whereby several output variables can be taken into account simultaneously. A statistical method enables the aggregation of the objective functions for individual variables, hereby avoiding the weighting problem. To select the important parameters for the optimisation, a sensitivity analysis precedes the calibration. The latter analysis is based on the One-factor-At-a-Time (OAT) design approach. The sensitivity analysis and the calibration procedure are applied to the river Dender in Belgium. The river is characterised by high pollution loads and long residence times in summer periods.