To model the impact or immission of various sources of water and pollution in receiving surface waters, those waters have to be modelled in an integrated way together with the different physical and natural systems that are connected to the different sources. Examples of such systems are sewer systems, waste water treatment plants and hydrographic catchments. The holistic model has to be practically useful and accurate.
Most frequently applied system models cannot be used for that purpose. They are too sophisticated and the linking of many sophisticated models leads up to unacceptable high calculation times and memory needs. The uncertainty in the holistic model results is furthermore very high because of the large extent of the model. It is therefore necessary to take the uncertainties involved in the modelling into account.
In this study, a holistic modelling methodology is worked out, which uses simplified (but partly physically-based) models for the systems and sources considered: rivers, urban drainage systems, watershed runoff, agricultural and industrial pollution sources. The simplified models have both a hydrodynamic and a water quality description, and are applied in a complementary way to the more detailed standard system models. For the calibration of the simplified models, physically based and transparent methods are applied. They are based on available measurements and/or the detailed models. To quantify the different uncertainties, which are classified in input uncertainties, parameter uncertainties and model-structure uncertainties, a step-wise procedure is followed. As the rainfall input is a major source of uncertainty, its uncertainty takes a detailed description in this study. All uncertainty sources are represented by stochastic terms, which transfer the deterministic model to a probabilistic one.
The different steps of the methodology are illustrated for different case-studies. To one of those case-studies (Witte Nete brook catchment, Belgium), all steps are applied and a holistic probabilistic immission model is derived for the full hydrographic catchment. With this model, long-term simulations are performed using long-term rainfall input for the different modelled systems. To simulate accurately the spatial variability and correlation of the rainfall, a spatial rainfall generator is developed .The probabilistic simulation results are evaluated using statistical tools, including a recently developed efficient extreme value analysis technique. After analysing the contribution of different uncertainty sources to the total uncertainty in the model results, priorities are determined for model improvement and future research in the field of integrated watershed modelling. |
