Chapter 7 : Conclusions and recommendations 7.1 Conclusions The main goal of this research was to investigate the modelling and rainfall data requirements for the design of combined sewer systems and the impact of combined sewer overflows on the receiving waters. A large variety of models are available, varying from detailed pipe models including the de Saint-Venant equations to simplified conceptual models as e.g. reservoir models. Also for the rainfall input a large variety of approaches are used, going from continuous simulations with long rainfall series to IDF-relationships with mean rainfall. In practice, all these different types of models and rainfall input are used, often not because they are the optimal choice, but for instance just because they are available, they are easy to use, they have already been used for years, etc... However, driven by the rapidly improving computer technology, also modelling techniques, tools and possibilities change equally fast. Therefore, the models and rainfall input were investigated in order to balance their degree of detail to the accuracy of the results. As the model and rainfall input requirements can differ for different applications, this research was applied to design calculations for combined sewer systems as well as emission predictions for combined sewer overflows. In a first step the rainfall characteristics were investigated using IDF-relationships, which were determined using recent and powerful analysing techniques. Special attention has been paid to the aggregation method, the independency criterion and the extreme value estimations. Partial duration series are used, which allow the determination of more accurate IDF-relationships than using yearly maxima. The new IDF-relationships that were developed, showed considerable differences with the previously (in Flanders) used design rainfall. Therefore, based on these IDF-relationships, new design storms were built, i.e. the composite storms. These composite storms have been successfully in use for sewer system calculations in Flanders since 1996. In specific cases the mean rainfall information that is incorporated in the composite storms will not be sufficient in order to obtain accurate modelling results, because of the high intrinsic variability of the rainfall. Two methodologies were developed to tackle this problem, i.e. the selection of short rainfall series and the incorporation of source control measures in modified composite storms. For storage upstream of the combined sewer system (source control) in for example rain water storage tanks, a simplified model is used to process the runoff. The effect is then incorporated as equivalent rainfall in modified composite storms. Chapter 7 : Conclusions and recommendations 7.1
The selection of short rainfall series has been investigated for the two applications, i.e. design calculations as well as impact calculations. It is important that this selection is based on the characteristics of the system which will be modelled. Only in this way the appropriate short rainfall series will be selected that will lead to the studied effects. For the application of design calculations IDF-relationships are used as the selection tool, while for the application of impact calculations a reservoir model is used. In a second step the combined sewer system models are investigated. In order to find how far the combined sewer system and the model structure can be simplified, the most extreme simplification into a reservoir model is investigated. As for design applications the hydraulic parameters have to be determined in every pipe, only the application for the prediction of combined sewer overflow emissions is investigated in this way. First, a methodology is set up to determine the main characteristics of the combined sewer system. It has been found that the sewer system behaviour can be expressed quite accurately using a limited number of parameters. The main parameters for the routing part of the model are storage, throughflow and concentration time. The storage can be split into static and dynamic storage and extra storage during the overflow event. Furthermore, not only the maximum storage and throughflow are important, but also the instantaneous relationship between these two parameters. Therefore, this relationship should not be restricted to a linear variation as is often implemented in reservoir models. Based on these findings the reservoir modelling system Remuli was built. Several combined sewer systems with various behaviour were successfully modelled with the reservoir modelling system Remuli resulting in very accurate quantitative overflow results which are very close to the results of the detailed model. For design calculations the results obtained with the new composite storms were compared with continuous simulations for several combined sewer systems with various behaviour. As in every pipe the discharge and piezometric level must be calculated, no model simplification can be applied in this case. It has been shown that the peak discharges can be predicted well with these composite storms. However, for sewer systems with a non-linear behaviour significant differences can be found. Except for highly capacitive systems, these differences are not larger than the intrinsic uncertainty on the results caused by the high uncertainty on the rainfall for the high return periods used. In case the differences become too large, the intrinsic variability of the rainfall can be incorporated using short selected rainfall series or using modified composite storms which take into account upstream storage via a reservoir model and continuous simulations. It has been found that the instantaneous relationship between storage and throughflow can give an accurate indication of the linearity of the system. This makes it possible to assess the reliability of the results obtained with single storm events, which are based on equality of the return period of the storm and the resulting effect. 7.2 The influence of rainfall and model simplification on combined sewer system design
For impact calculations, more combinations of models and rainfall input can be considered. It has been found that for linear systems the composite storms can predict in most cases the frequency distributions for the combined sewer overflow emissions relatively well. However, for non-linear systems the deviation can be large. In some cases extremely large antecedent rainfall periods have to be applied. The results of simulations with the composite storms are more deviating from the results of the continuous long term simulations when the sewer system behaviour becomes non-linear and the overflow frequency is increasing. It can therefore be concluded that it is absolutely necessary to incorporate the real variability of the rainfall for the prediction of emissions at combined sewer overflows. This can be performed using selected short rainfall series in a detailed model, but the computational times for this are still large (of the order of days). Reservoir models using long term simulations run much faster (of the order of seconds). Using a good (physically based) calibration methodology, as has been worked out in this study, good simulation results are obtained with a reservoir model. A physically based conceptual model is an ideal tool for sensitivity and scenario analyses. The main disadvantage of simplified models is the fact that the calibration stage requires some time and experience. However, this calibration stage can yield interesting information on the sewer system behaviour, which can be a great help for the designer. Much effort is spent in this work to make this calibration as user friendly and unambiguous as possible. For impact calculations the following gradation of methodologies can be proposed as a function of the accuracy of the results (highest accuracy first) : S detailed model using long term simulations S detailed model using (accurately) selected short rainfall series S simplified model using long term simulations S detailed model using composite storms Unfortunately, the ranking for the computational time consumption is reverse. The optimal methodology, taking into account the current computer technology, is the use of a simplified model and long term simulations. The use of composite storms in a detailed model cannot be considered as a good methodology, because no guarantee can be given on the accuracy of the results. Besides the routing of the flow through the combined sewer system and the sewer system parameters, also the rainfall input into the combined sewer system is very important. The runoff from the catchment into the combined sewer system is often one of the weakest parts of modelling work. This is understandable, because good data and measurements are difficult to acquire. Although the modelling results are highly dependent on the input into the sewer systems, the runoff part is often underrated. It has been shown for instance that an underestimation of the water inflow with 10 % reduces the design return period of 5 years to 3.5 years. For impact simulations the influence of the runoff model becomes even more important than for design calculations, because of the effect on the antecedent conditions. Chapter 7 : Conclusions and recommendations 7.3
Finally, the criteria were investigated which are in use in Flanders for design and impact calculations. It must be stated that the criteria do not always follow the rapidly changing technology and methodologies. Often methodologies and calculation tools changed, without reflection of the effect of these new modelling approaches on the modelling results. The way the criteria are defined can have a large influence on the modelling results and must comprise a dynamic entity with the modelling approach. In this study no water quality aspects are taken into account, although the final goal of emission calculations is to assess the pollutant concentrations and loads that are spilled into the receiving waters. It is of prior importance that the water quantities are predicted well, taking into account the variability of the rainfall and based on the effect of this uncertainty on the modelling results. If the combined sewer overflow hydrographs cannot be estimated accurately, accurate pollutant emission cannot be estimated either. It is clear from this study that long term simulations are necessary in order to obtain a good probabilistic estimation of the water quantities. Long term simulations can easily be performed using simplified models, if the sewer system parameters are identified and incorporated accurately. Furthermore, the calibration of the runoff model is extremely important. The main achievements in this work can be summarised as : The rainfall at Uccle, which can be assumed as representative for Flanders, has been investigated and characterised with new IDF-relationships (in cooperation with Patrick Willems). The application of new analysing techniques and computer technology have led to a more accurate estimation of the IDF-relationships. Based on these new IDF-relationships new (composite) design storms were built, which are more user friendly and more accurate than the previously used design rainfall in Flanders. These composite storms are commonly in use in Flanders since 1996. A non-linear reservoir modelling system was built for the emission prediction at combined sewer overflows, named Remuli. This modelling system can take into account static and dynamic storage and also extra storage during the overflow event. As the model structure is open, extra attention is paid to set up a physically based calibration methodology. The smoothing and peak shift due to the flow through the system are linked to the (variable) concentration time. Other special features as return flow and overflow basins are included. 7.4 The influence of rainfall and model simplification on combined sewer system design
A methodology has been developed for the selection of short rainfall series, which can be used if the intrinsic variability of the rainfall is needed to obtain accurate results. Therefore, the basic parameters of the system are taken into account in order to select the corresponding rainfall to the effect which has to be modelled. This has been worked out for design applications based on the new IDF-relationships. For the selection of short rainfall series for impact assessment the reservoir modelling system Remuli has been used as the selection tool. For source control applications a conceptual model was set up (using continuous long term simulations) to incorporate the effect of this source control into modified composite storms. Finally, the effect of rainfall and model simplification was investigated using these developed tools. This was performed for design and impact applications with special attention to the Flemish design practice. It has been found that the need to incorporate the intrinsic variability of the rainfall is highly correlated with the non-linearity of the sewer system behaviour. With the developed sewer system characterisation tool this degree of non-linearity can be determined. Furthermore, it has been found that detailed models are not required for impact assessment. Conceptual models can achieve very accurate results in a fraction of the computational time. Therefore, the combined sewer system has to be characterised accurately (and physically based), which requires that the model structure is not fixed a priori. Chapter 7 : Conclusions and recommendations 7.5
7.2 Recommendations It is obvious that there is still a long way to go on the path to accurate prediction of the flow (and water quality) into and out of the combined sewer systems. For design calculations more attention has to be paid to the possible non-linear behaviour of the system, especially caused by storage which is built in into the system (including source control). Furthermore, more attention must be paid to the spatial variability of the rainfall. There is still further research needed on the spatial variability and how this could be implemented in design calculations, but the first test with spatial varying storms moving over the catchment show significant effects. Finally, the Flemish design criteria should be reconsidered in order to increase the design safety and to take into account the uncertainties on the results. For the impact assessment at combined sewer overflows even more research is still needed. Considering only the water quantities is certainly not the final step. In the future, water quality models should be used. However, based on the experience with the water quantity modelling for emission predictions, the question rises whether this could be implemented in a conceptual (simplified) way too? Many signs point in that direction. Water quality models are often very complex and many aspects are still not understood. However, the mass balances have to be respected anyway. If the main parameters for the transport, dispersion, erosion and sedimentation can be identified, the water quality aspects can be easily incorporated in a simplified model. Moreover, there is a large uncertainty on the sediment and pollutant input into the combined sewer systems. The sediment and pollutant input is even more difficult to measure than the water quantities that run off to a combined sewer system. If the input cannot be predicted accurately, why should the model itself be so detailed and accurate? Further research on these water quality aspects has to point out what the optimal model detail is as a function of the application and the available data. Corresponding to the water quantity calculations, long term simulations will be necessary for an accurate assessment of the water quality, especially because the antecedent conditions become more important for water quality calculations. Besides the water quality modelling of the combined sewer system, special attention has to be paid to the modelling of ancillary structures at overflows. The sediment and pollutant transport through these structures differ considerably from those in combined sewer systems. Therefore, separate modules for ancillary structures have to be coupled with the emission models. More and more of these ancillary structures (e.g. improved high side weirs, storage sedimentation basins, etc...) will be built in order to limit the spilled pollutant loads into the receiving waters. The modelling of the behaviour of these structures on long term is the only way to estimate their efficiency and to compare alternatives in advance. 7.6 The influence of rainfall and model simplification on combined sewer system design
The final goal in the modelling of combined sewer systems is not the assessment of the overflow emissions, but the prediction of the water quality in the brooks and rivers (i.e. the immisions). The emission models play however an important role in this problem. Once it is possible to predict the emission from combined sewer overflows well, these models can be coupled with models for other subsystems, as there are the waste water treatment plant, the river, the direct runoff and other point pollution discharges into the brooks and rivers. For these subsystems also the water quantity aspects are very important and must be modelled accurately in the first place. This is especially true for the direct runoff to the brooks and rivers. Further research is necessary on these subsystems in order to determine the optimal degree of detail in accordance with the required accuracy of the results. Corresponding to the emission calculations, also for an accurate immision assessment long term simulations will be necessary. The detailed models that are currently used in most cases require too much computational time to be used in such an integrated approach. In a last stage the models for the different subsystems should be combined in one modelling system to obtain a real integrated model. This can only be implemented easily if the submodels for the different subsystems have about the same degree of detail. As there is a high uncertainty on input data and model parameters, this uncertainty should be assessed and taken into account in the models. To achieve this, probabilistic models can be built, which incorporate these uncertainties. As model parameters are often not fixed values but correspond with a distribution, Monte Carlo simulations are necessary to incorporate these stochastic sewer system characteristics into the model results. A large number of simulations is necessary for this, which certainly requires very simple models. Further research is necessary to develop these probabilistic models, because a probabilistic approach is certainly the final stage in order to predict the impact of urban storm drainage on the receiving waters. Although it is interesting to have accurate and fast modelling tools, they are useless if insufficient consideration is given to the specified criteria. Founded on new technologies and methodologies, consistent design and impact criteria should be developed. Water quality criteria are possibly not appropriate for practical application in the near future, but this does not mean that one must stick to the simple criteria of decades ago. Corresponding to new technologies and methodologies, criteria have to improve gradually. Especially for impact calculations, the currently used Flemish criteria are not adequate anymore. New criteria should be defined based on the statistical characteristics of the emissions at combined sewer overflows. In anticipation of water quality and immision standards, more elaborated water quantity criteria should be developed to evaluate the emissions at combined sewer overflows. At the same time, the different methodologies used, should be tested whether they can predict the necessary output within the required accuracy in order to evaluate these criteria. This is maybe a more important and more difficult task than the modelling work. No model, methodology or criterion will be ever defined so that it will be Chapter 7 : Conclusions and recommendations 7.7
applicable in all cases. Finally, the methodologies must be generally applied in practice. It is the design engineer who has to do the job. In this work almost all attention was paid to the modelling of combined sewer systems. However, the use of combined sewer systems themselves can be questioned. Recent years, more attention is paid to source control, because building larger combined sewers and ancillary structures at combined sewer overflows are not able to solve all problems with combined sewer systems. This is certainly a positive evolution. Source control is a very powerful tool to solve problems with too large water quantities which flow in a short time to small valleys where they cause flooding problems. However, the effect of source control on the combined sewer systems must be modelled too. The major constraint with source control is that it must be applied on a large scale to be effective. This is not the task of design engineers alone. It requires an effort of decision makers and all individual citizens. Although recently more and more stimuli are initiated, it will take decades to transform all combined sewer systems into (partially or improved) separate systems. In the mean time we will have to live with the burden and the blessings of our combined sewer systems. Modelling is an ideal tool to limit the burden. 7.8 The influence of rainfall and model simplification on combined sewer system design