Calibration of an in-river phosphorus model: prior evaluation of data needs and model uncertainty

A simulation model of in-stream phosphorus mobilisation and transport, developed to predict monthly total phosphorus export from the upper Hun River (northeast China), is described. To evaluate alternative programmes for collection of in-river calibration data, a set of a priori computational experiments are devised. Initially, assuming that the model and boundary condition data are error-free, daily in-river phosphorus concentration data are synthesised, representing an idealised system response. Scenarios of error in the calibration and boundary condition data, and in the system conceptualisation are then imposed, as a hypothetical representation of the conditions under which the model will actually be calibrated and evaluated. The model is calibrated under the various conditions of error, and its predictive reliability is tested using an independent idealised data set. These controlled experiments allow evaluation of sampling programme design, of other controls on model reliability, and of needs for additional a priori investigations. The results indicate that the value of the calibration data is seriously compromised by the presence of error in the pollution load data, error in the model structure, and inherent parameter equifinality. While these controls are not generally very detrimental to model reliability under calibration conditions, in some cases they cause serious misrepresentation of forecasted phosphorus export. For the case study, it is concluded that the resources for monitoring need to be increased in order to achieve the sought modelling capability. In light of the limited resources, expectations of model performance should be reviewed, and a more robust approach to model uncertainty estimation adopted.