The MIA-CSO project is currently led by the department “Point and non.point source pollution control of surface water” in the KWB. Its overall objective is to develop a model-based planning instrument for impact based CSO control. To that aim, an integrated monitoring in the river Spree and in the Berlin sewer system will be carried out in order to calibrate two numerical models before they are used for CSO impact assessment. A module shall be developed to allow the statistical analysis of the model results. This module will be based on Uncertainty and Sensitivity Analysis (UA and SA). The principle of such analysis is to investigate the effects of model input uncertainties on model outputs: UA establishes a mapping of model assumption into model inferences and SA is the study of how uncertainty in a model output can be apportioned to different sources of uncertainty in the model inputs. This study focuses on Sensitivity Analysis. First of all, four of the main SA techniques are described and explained mathematically. The first technique is a screening method called the Morris algorithm. This qualitative method allows one to classify the inputs in order of importance. The most important ones can then be selected for further study. Furthermore, this technique gives an idea of the linearity or nonlinearity of the effects for every input. The second technique, based on Regression Analysis, works under the assumption that the model is linear or monotonous. It gives a qualitative indication of the relative effects of each input. The third and fourth techniques, called Fourier Amplitude Sensitivity Test and the Sobol’ method, aim at calculating indicators of the relative effects of each input, called the sensitivity effects. They work without any assumption on the model and they can traduce the effects of interactions between the model inputs. Using the free software SimLab, three methods are tested in this study: Morris, FAST and Sobol’. They are applied on catchment Berlin XII in Friedrichshain, under the framework of ATV-A 128, a linear, empirical model used for designing storage tanks in combined sewer systems. After a calibration step, rules are expounded in order to define how to use these techniques and how to have reliable results. Then, the Sensitivity Analysis itself is performed for Berlin XII. Among the nine inputs of interest, the Morris screening allows to choose the four most important ones (the CODconcentration in the wastewater, the COD-concentration in the rainwater, the specific water need and the total impervious area). While the five other inputs are considered constant, FAST and Sobol’ are performed and give the exact relative effects of the four inputs. It appears that two inputs are more important than the two others. For these inputs (COD-concentration in the rainwater and specific water need), further uncertainty study should be done and the lack of data should be corrected.

Sensitivity Analysis Using SimLab: Application for the German Standard ATV-A 128