Evaluation of uncertainty propagation into river water quality predictions to guide future monitoring campaigns

To evaluate the future state of river water in view of actual pollution loading or different management options, water quality models are a useful tool. However, the uncertainty on the model predictions is sometimes too high to draw proper conclusions. Because of the complexity of process based river water quality models, it is best to investigate this problem according to the origin of the uncertainty. If the uncertainty stems from input data or parameter uncertainty, more reliable results are obtained by performing specific measurement campaigns. The aim of the research reported in this paper is to guide these measurement campaigns based on an uncertainty analysis. The practical case study is the river Dender in Flanders, Belgium. First an overview of different techniques that give valuable information for the reduction of input and parameter uncertainty is given. A global sensitivity analysis shows the importance of the different uncertainty sources. Further an analysis of the uncertainty bands is performed to find differences in uncertainty between certain periods or locations. This shows that the link between periods with high uncertainty and specific circumstances (climatological, eco-regional, etc.) can help in gathering data for the calibration of submodels (e.g. diffuse pollution vs. point pollution).