Brief description

The Water Framework Directive of the European Union aims at a good ecological status for all natural waters by 2015. In order to accomplish this task, specific measures need to be taken to enhance water chemical and ecological quality. For the assessment of proposed measures and their cost-effectiveness, integrated river basin assessment is indispensable. Hydrological and nutrient transport models are important tools in river basin management and planning. However, model predictions need to be interpreted with due care as they are always tied up with uncertainty which comprises both input data and model uncertainty.

If the uncertainty associated with model results is not explicitly stated, decision makers tend to be overly confident in model results. Recognition of uncertainty of model results can serve as an additional criterion for selection of alternative management strategies. Therefore, an important aim of uncertainty assessment is to raise awareness of the inherent imperfection of all modelling and the specific uncertainties associated with each modelling exercise. Moreover, uncertainty analysis can serve as a guide for the efficient direction of future work in improvement of data sets and models.

Data availability and quality are often serious limitations for the use of distributed hydrological and nutrient transport models. Because data collection is a time-consuming and expensive task, frequently there is not enough information to fully characterize the variability of catchment properties and the available information is tied up with considerable uncertainty that come in the sampling and measurement process as well as in data transformation, interpolation and aggregation.

On the other hand, models are inherently imperfect as they are always simplified representations of reality and can only take a limited number of processes into account. Also, in hydrological, substance transport and water quality modelling, models are often applied at much larger scales than they were originally developed for, making use of effective parameters. Therefore, models used for prediction are also a source of considerable uncertainty.