AN A PRIORI EVALUATION OF MEASUREMENT CAMPAIGNS IN URBAN DRAINAGE SYSTEMS ABSTRACT

Transcription

1 AN A PRIORI EVALUATION OF MEASUREMENT CAMPAIGNS IN URBAN DRAINAGE SYSTEMS Tanja Vonach 1,2, Manfred Kleidorfer 2, Wolfgang Rauch 2, Franz Tscheikner-Gratl 3 1,2 University of Innsbruck, Technikerstr. 13, 6020 Innsbruck, Austria 3 Delft University of Technology, P.O. Box 5048, 2600 GA Delft, The Netherlands 1 tanja.vonach@uibk.ac.at ABSTRACT The calibration of hydrodynamic models of urban drainage systems got more and more important as the usage of such models as basis for planning increased considerably over the last years. Still the effects originating in the choice of data taken for model calibration are little known and advice for planning measurement campaigns for model calibration is limited, especially for small and medium-sized municipalities. The choice of measurement sites (number and location) within a sewer system to collect data for model inputs as well as outputs is affecting the calibration and finally also the assessment of the modelled system behaviour. This paper discusses the calibration of a hydrodynamic model using the representative example of a small municipality. Different calibration scenarios were estimated on a model based approach, focusing on a varying availability of in-sewer measurement data. To assess the performance of different scenarios and validate the respective models, different model outputs were compared. The calibration scenarios proved to result in high deviations in the model performances. With an increasing number of used calibration points, model performance increases significantly. Predicted CSO volumes deviate from the measured volumes in ranges between -1% and +189% for one, -42% to 27% for two and -2% to 46% for five used calibration points, depending on the used rainfall input. Consequently, the design of measurement campaigns for calibration data is a very sensitive decision in the modelling process. The model performance further influences design and decision-making processes, which are then perceptible in economic and functional aspects. Keywords: calibration, hydrodynamic models, measurement campaigns 1 Introduction The usage of hydrodynamic models, not only for flooding forecast but also as a planning tool in urban drainage, increased considerably over the last decades and with it, the importance of understanding a model s ability to reproduce the system behaviour. To ensure that the model performance is sufficient to be a reliable foundation for any planning procedure, the calibration process is a crucial and fundamental component of the model development process [1, 2]. Consequently, the process of model calibration has been the topic of many research activities and publications. For example, Di Pierro et al. (2005) [3] investigated the development of calibration algorithms, Kleidorfer et al. (2009) [4] highlighted the impact of data accuracy and Deletic et al. (2009) [5] focussed on the sources and propagation of uncertainties. However, uncalibrated or insufficiently calibrated models are still in use in engineering practice, with data availability often being the limiting factor. Calibration usually requires measurement campaigns, which in turn can increase the economic cost of the projects up to an unachievable level,

2 especially for smaller operators [6]. Calibration uncertainties relate to the data used for calibration and their selection [7] and to the calibration methods [8]. They stem of measurement errors for both, input and output data, the selection of appropriate calibration and validation datasets, the applied calibration algorithms and the objective functions used during the calibration process [9]. Another possible deficit of urban water management studies is that the case studies in scientific literature are often large, prestigious cities, which have the financial and human resources to participate in research projects. They are selected for providing a good data background, e.g. measurement data over longer periods of time, and/or the required infrastructure for further data collection and management. Such case studies are not always representative for the entire situation of the living environment in a country. At least, there is the risk that research outcomes are biased towards large municipalities [1]. Apparently, there are a manifold of factors influencing these numerous variations in data availability. In this regard, we are always limited to a certain extent by given restrictions or conditions. In this paper, the influence of data availability on calibration performance is investigated for the hydrodynamic drainage model of a small Austrian municipality. For this purpose, different scenarios of data availability for calibration are simulated. Scenarios for varying input data (different number of calibration events, different rainfall input) have already been considered in Tscheikner- Gratl et al. (2016) [1]. However, this work left open the question of the influence of using only one measurement site for calibration. Although the small size of the case study and the limitation of funds mimicked nicely an engineering approach, a wider spread measurement campaign may lead to a more differentiated outcome and consequently a better representation of the case study. This paper now shows these effects by using different in-sewer measurements as calibration input. For this, simulation results from a reference model are taken as assumed measurements, due to limited data availability. Existing studies, e.g. Kleidorfer et al. (2009) [10], investigated the influence of an increasing number of measurement stations for the calibration of conceptual sewer models. The effects of locating calibration points for hydrodynamic models were investigated in Vonach et al. (2017) [11], concluding with the intention of further research by improving the used methodology. With this work, we want to enhance the understanding of these effects by investigating the influence of a different amount and different samplings of calibration points for hydrodynamic drainage models. 2 Methods The basis for the used methodology is an existing hydrodynamic model of the case study s urban drainage network [1, 12-14]. The performance of different scenarios was assessed using the Storm Water Management Model (SWMM) software tool [15, 16], which has also been used e.g. by Yazdi (2017) [17] and Gong et al. (2017) [18]. 2.1 Case study The analysed case study Telfs is a small municipality with 15,000 inhabitants in Tyrol, Austria. Located at an altitude of about 630m above sea level in the valley of the river Inn, the investigation area is reaching from the river up to the footlets of the mountain chain Karwendel. It can be seen as the representation of a typical Tyrolean urban settlement with an average annual rainfall of about 1,000mm.

3 The urban drainage network of Telfs consists of 14km of combined sewers, 67km of wastewater sewers and 24km of stormwater sewers. These stormwater sewers have 28 outfalls (in the following figures referred to as RW for rainwater) into the receiving water bodies, while in comparison only three combined sewer overflows (CSO) exist. In total, a catchment area of approx. 95 hectare is connected to the combined sewer system. For model calibration and validation, precipitation was measured over the period of one year at three sites (rain gauges RG 1-3) within the catchment area and the water level at one site near the inflow to the wastewater treatment plant. This measurement setup also represents limited data availability, inherent to smaller operators due to limited budget. The ten rain events with the highest occurring intensities, which surpassed an event threshold of 3mm for all three rain gauges with an inter-event time of 24h, are consolidated to one continuous rain series while avoiding interconnection of the events by adding sufficient dry-weather periods of 4h [19] between the individual events (see figure 1). Due to the case study s alpine environment, a high spatial variability between the three rain gauges can be observed. The closer a measurement station is to the hillside (RG3), the more intense the rain falls and the more distinct the peak events are. This already gives a hint on the different results to expect when using the measurements of different rain gauges for calibration. Figure 1. Consolidated rain series with 10 measured rain events (RE01-RE10) A wastewater treatment plant (WWTP) is located southeast of the town. This plant additionally treats the wastewater of four nearby communities and its capacity is designed for 40,000 population equivalents. Accordingly, the case study s drainage network has to cope with conveying also the wastewater of the other association members to the WWTP. Regarding the model, it is important to mention that there are two different and independent outlets connected to the WWTP, see figure Calibration procedure For the initial model calibration, a measurement campaign was executed in Data of the flow depth level was collected in one point of the sewer system, while three rain gauges (distributed over the catchment area, see figure 3) logged precipitation data for the same period. Calibration scenarios regarding the influences of different input data were already shown in Tscheikner-Gratl et al. (2016) [1]. There, the model was calibrated with different rainfall events

4 (see figure 1) and data sampled according to empirically grounded measurement campaigns. Additional scenarios also considered uncertainties in the measurement data, by assuming systematic errors in the collection of water level monitoring data. While the influences of the usage of different model input data (i.e. rainfall recordings) for calibration were evaluated in a very detailed way, effects of the varying spatial distribution of the calibration data (in-sewer measurements) were only marginally examined. This is due to a restricted data availability. All these scenarios used one single measurement data set, which is a water level measurement at one point of the system. To supplement to these studies about varying (model) input data, this work now enhances the understanding of the possible effects due to the usage of different calibration data, obtained at different locations in the network. This paper describes three scenarios (scenario I, II and III) to show up the influences coming from the design of in-sewer measurement campaigns. As there is real data from only one measurement point in the network available (at the catchment outflow), the calibration method in this paper is purely model based. We use a reference scenario as source for assumed measurements in every point of the model. Scenarios I, II and III were established by considering weak points in the uncalibrated model (pipes, where the agreement of simulated water level courses between the uncalibrated and the reference model is low) and taking into account the operator s empirical knowledge. Basic considerations regarding this heuristic scenario development approach can be found in Vonach et al. (2017) [11]. Figure 2 exemplifies, which abstractions were made during the procedure. The basis is the real system with the real measurement. A first abstraction is made by calibrating a model to this measurement. For this, input data is varied in different calibration scenarios [1]. One of these scenario models, the model which resulted in the best overall agreement with the measurement (expressed as the Nash-Sutcliffe Efficiency NSE [20, 21]), is then used as the reference scenario. Figure 2. Necessary abstractions for the model based approach The reference model is again simulated with a consolidated rain series of the three rain events RE03, RE06 and RE09 (see figure 1). These rain events turned out to be favourable, as models calibrated with only one of those delivered a high model performance compared to the reference scenario in terms of their ability to predict the CSO and flooding volumes [1]. The used rainseries again consists of the measured rain events and artificially made intermediate dry weather periods to avoid interaction in the sewer between the events. In a further step of abstraction, the simulated model outputs are taken to serve as assumed measurement data. By this means, measurement data and system performance is available in any

5 form (e.g. water levels, CSO and flooding volumes, etc.) and in every point of the system. The models in scenario I, II and III are then calibrated to the values resulting from the reference scenario. A similar approach but for a different objective is also used and described in Kleidorfer et al. (2009) [10]. The locations of the measurement sites for the assumed measurement campaigns for the three scenarios I to III are shown in figure 3. The investigated calibration point for scenario III is the same point, where also the real data is available. Nevertheless, the artificial water levels from the reference model were used for scenario III to keep the scenarios comparable on the same level of abstraction. Figure 3. Model of the case study Telfs with assumed measurement points for scenario I, II and III In principle, only subcatchment related parameters concerning the runoff concentration and the total runoff volume are varied. The runoff model implemented in SWMM mainly uses two parameter, on which the runoff from a subcatchment is depending, the width and imperviousness. The width of a subcatchment determines the shape of the area and consequently the temporal runoff concentration while the imperviousness relates to the total runoff volume. As there are several points to calibrate the model to, calibration to each assumed measurement station is performed in a downstream order. By this means, only subcatchments lying upstream of the current and downstream of the previous calibration point are adapted. This methodology increases the number of calibration parameters to a high extend compared to a calibration to all points simultaneously. As an example, the subcatchments varied for each calibration step of scenario I are shown in figure 4.

6 Parameter adaptation is implemented and automated with R (R Development Core Team 2008). We used an optimization algorithm based on a Nelder-Mead simplex [22, 23]. The Nash-Sutcliffe efficiency NSE was chosen as the objective function to compare measured and predicted water levels. NSE is a measure to compare time series. It ranges from - (no agreement) to 1 (perfect match). For the calibrations in this paper, a threshold of NSE=0.9 is chosen. As soon as the optimization algorithm finds a parameter set, which leads to an exceedance of this threshold at the calibration point, the algorithm terminates and the model is considered as calibrated for the currently regarded calibration point. Figure 4. Systematic order of adapted subcatchments for calibration scenario I A calibration to systems outlets can lead to a loss of information about high-resolution system behaviour. To highlight the possible extent of this loss, a sensitivity analysis is performed for scenario II (a calibration to measurements at the outlets to the WWTP) as an addition to the calibration scenarios. Therefore, calibration parameters were varied 300 times randomly within defined ranges. The so created models were then simulated three times with three different rainfall inputs. Two of the rainfall inputs are design storms Euler II with a return period of 5 and 10 years, respectively, prepared according to Austrian design guidelines [24]. More information about the application of these design storm events can be found e.g. in Mikovits et al. (2017) [25] or De Toffol et al. (2006) [26]. They are characterized by a high peak (here 12.9 and 15.7mm/5min) and a short duration (here 120min). Data for these rain events are available from the Austrian rainfall

7 database (ehyd) [27]. The second simulation is done with the measured rain event RE03, where the highest intensities of the measurement series occur (9.2mm/5min). 2.3 Model validation and evaluation Influences caused by the choice of calibration points are evaluated with scenarios I, II and III. The models of the intermediate calibration steps are again simulated with the entire rain series (our own measurements), a design storm event of type Euler II with a return period of 10 years and the precipitation data sets ZAMG1 and ZAMG2. ZAMG1 and ZAMG2 are data sets of the Austrian Central Institute for Meteorology and Geodynamics (ZAMG) and are chosen from the nearest measurement sites available (ZAMG1 10km from the catchment and ZAMG2 30km). A spatial distribution of the occurring rainfall is only taken into account with the entire rain series, which has been measured with three different gauges (RG1, RG2 and RG3) within the case study s catchment. Subsequently, model outputs are compared to the reference model with regard to the occurring flooding and CSO volumes. To show the effects of generalizing the system behaviour, accompanied by information loss due to a smaller number of measurement stations, simulated CSO and flooding volumes (derived from the randomly created models of the sensitivity analyses for scenario II) were compared with the results from the reference model. Only those models were considered which resulted in a good agreement (NSE>0.8) for the water level courses in the pipes right before both outlets to the WWTP. So the occurring effect of information loss about the upstream system s behaviour can be highlighted while maintaining a good agreement at the calibration points. 3 Results and discussion The final models from the calibration scenarios were again validated with rainfall series lasting 200 days. Out of all available measured rainfall series (RG1, RG2, RG3, ZAMG1, ZAMG2), ZAMG1 caused the most flooding volume in the reference scenario. Therefore, figure 5 shows the results for models simulated with the measured rainfall of ZAMG1. It shows the deviations of predicted flooding and CSO volumes from the reference scenario not only for the scenarios evaluated in this paper but also in comparison with the calibration scenarios with different rainfall input and systematic errors in the water level measurements [1]. The other rain series (RG3 and RG2) result in very low flooding volumes (5 and 30m³) for the reference scenario and the relative deviations are therefore even higher but accordingly less significant. Rain series ZAMG2 and RG1 did not elicit any flooding in the reference scenario, therefore deviations are not possible to evaluate. The chosen reference scenario has been calibrated to the available measurement right before the most downstream CSO (in the south of the network, the same point as the calibration point of scenario III in figure 3) before the WWTP. This scenario has been calibrated with the entire rain series and the differences of all the three rain gauges were taken into account [1]. By subdividing the scenarios in terms of deviation of the flooding and CSO volume for the rain set that elicits the most flooding ZAMG1, the effect of calibration is easy to see (see figure 5). Scenarios I, II and III show all less than 25% deviation from the reference scenario, both for CSO as well as for flooding volume. As a comparison, only one of the scenarios with varying rainfall data

8 sets deviates not more than 25% for both volumes. Even scenario III with the same amount of calibration points as the other calibration scenarios (1 point, except I and II) shows a very good performance. Five of the scenarios exceed deviations of 100%. This shows that these scenarios perform even worse than an uncalibrated model or at least not better for the rainfall ZAMG1. The good agreements of scenarios I, II and III with the reference scenario could elicit from an advantageous sampling of model input data. A spatial distributed rainfall as well as different rain events were used for these calibrations. This is in contrast to the majority of the other scenarios, where mostly either only one rain gauge or only one rain event is used for calibration. Looking at the resulting models themselves, scenario II agrees best in terms of the connected impervious area and the mean imperviousness. The reference scenario has a connected impervious area of 52.4ha in total with a mean imperviousness of 65.17%. Scenario II almost hits the same value with a mean imperviousness of 65.24% and a contributing impervious area of 50.6ha, whereas scenarios I and III overestimate both the connected impervious area and the mean imperviousness with 58.7ha and 75.00% (scenario I) and 56.2ha with 71.54% (scenario III). Figure 5. Flooding volume and CSO volume deviation for measured 1 year rain series ZAMG 1

9 3.1 Uncertainties due to a spatial difference in calibration data availability Concerning the necessary amount of measurement stations to gather data for calibration, figures 6 and 7 show the change in the model behaviour with a different number of calibration points. For this, the final model and the intermediate models during calibration to subsequent calibration points are simulated again with a design storm event Euler II, with the measured rainfall series ZAMG1 as well as our own measurement data from the three rain gauges RG1-3. Again, only ZAMG1 and RG2 elicited significant flooding in the reference model. To be able to compare all simulated rainfalls, CSO volumes are chosen instead of flooding volumes to illustrate the model performances. The highest deviations for the final models of these scenarios occurr for scenario II and the rainfall series of ZAMG2 with an underestimation of -79% for the CSO volume. Scenario III simulated with the rainfall series RG3 results in the highest overestimation of 64% for the occurring CSO volumes. Using the reference model to produce fictitious measurement data also enables us to compare the resulting water level courses in all other pipes. Figures 8a to 8h exemplify how the model fitness changes elsewhere. For this, the Nash-Sutcliffe efficiency is calculated in selected pipes of the network (in about one third of the pipes contained in the model). Figure 6. CSO volume for design storm event Euler II, measured rainfall from ZAMG1 and own gauges RG1-3 for different measurement campaigns and calibration steps The fifth calibration point of scenario I (at the south western outlet to the WWTP) is the same as the second from scenario II. For both scenarios I and II, the additional consideration of this point has a high impact on the model s agreement to the reference scenario. All the scenarios led to a reduction of impervious area and therefore the total CSO volumes during their calibration procedure. Models with less calibration points overestimate the total volumes consistently, see figures 6 and 7.

10 The last calibration point of scenario I and II is located at the end of a collector sewer. Between the major connections of the catchments and this calibration point, a CSO is interposed. CSO s moderate the impact of higher amounts of water, as peaks in the flow would be intercepted by an overflow event. Consequently, the comparison of water levels right after an overflow structure is not able to characterize high flow rates within the system and a certain loss of information has to be expected. This may also be the reason for scenario II to constantly underestimate the CSO volumes. The effects of CSO structures is also emphasized by the good agreement of scenario III, where only one assumed measurement is considered for model calibration. For this scenario, water level data is assumed to be available directly before the previously mentioned CSO. Except one outlier (RG1), the agreements after a calibration to this one point show up to lie within tolerable ranges (189% for RG1, 64% for RG3, 48% for RG2, 6% for ZAMG1 and <1% for Euler II). These ranges are tolerable especially as compared to other calibration scenarios from Tscheikner-Gratl et al. (2016a) [1]. This calibration point merges the water flow of major parts of the system while being not yet influenced by immediate overflow structures. Data collection in sewer pipes which fulfil this criteria is therefore an effective and economic approach, especially if the available resources are lacking. Figure 7. Deviation of CSO volume for design storm event Euler II, measured rainfall from ZAMG1 and own gauges RG1-3 for different assumed measurement campaigns and their according calibration steps Coming back again to scenario II, where both measurement points are located at the outlet to the WWTP, the big step of improvement can also be related the high increase of the connected discharging area. The second measurement point (south western outlet) is connected to 19.7ha of impervious area and the first (north western outlet) to only 5.3ha.

11 (a) Scenario I, step 1 (b) Scenario I, step 2 (c) Scenario I, step 3 (d) Scenario I, step 4 (e) Scenario I, step 5 (f) Scenario III (g) Scenario II, step 1 (h) Scenario II, step 2 Figure 8. Overall agreements for different calibration steps in scenarios I, III and II

12 3.2 Sensitivity analysis for scenario II To further investigate the loss of information, which stems from a less detailed distribution of measurement stations, all calibration parameters of scenario II were varied simultaneously within defined ranges. The aim is to show possible variations in model prediction with similar performance during calibration. For this sensitivity analysis, the two calibration points from scenario II are considered. They are situated directly at the outlets to the WWTP. The analysis was carried out for three different rainfall inputs. We considered a design storm Euler II with a return period of 10 years, another with a return period of 5 years and also a single measured rain event (rain event RE03). For each rainfall input, the model parameter were varied 300 times randomly within defined boundaries. To expel models with gross deviations, only models which resulted in a NSE of at least 0.80 at both calibration points are considered for the evaluations (57 and 88 models for rp=5y and 10y, resp.). Figures 9 and 10 show only results for the simulations with both design storms. The measured rainfall input evoked only very little flooding volumes between 0 and 10m³ and is neglected for further evaluations. Figure 9. Absolute values and density plots of flooding volume and CSO volume for design storm events Euler II (rp=5y and 10y) and random calibration parameter variations (scenario II) All but one of the random models with a good agreement at the calibration points overestimated the flooding volume while underestimating the CSO volume. Still, the deviations in the CSO volume with values between -21% and -3.8% are significantly smaller than for the flooding volume with a range from -8.9% to 79.5%. Especially for rainfalls with higher intensities (i.e. higher annualities) and consequently higher volumes, relative errors are smaller. A better model performance for conditions of high than of low flow could also be found out in Ahmed (2012) [28].

13 The absolute occurring range of CSO volumes is about the same for both rainfall series (1022m³ for rp=5y and 1262m³ for rp=10y). Also if we neglect the flooding volume s outlier(s), the occurring range for the absolute flooding volume is in the same order of magnitude (with 505m³ for rp=5y and 645m³ for rp=10y). The impact of coarse-grained sensor distributions shows up to be minor when regarding the CSO volume and higher for the resulting flooding volume. The resulting ranges for both volumes are remarkable smaller than their absolute differences to the values from the reference scenario. Consequently, a dense distribution of models with similar deviations can be seen. This indicates a possible systematic error in calibration, but the outlier proof that the given boundary conditions for parameter variations did not restrict to hit the reference scenario. Figure 10. Deviation from to the reference scenario and density plots of flooding volume and CSO volume for design storm events Euler II (rp=5y and 10y) and random calibration parameter variations (scenario II) With increasing precipitation, high flow peaks are generally more likely to be discharged through manholes, i.e. flooding, than to stay within the drainage system until they pass the next CSO structure. CSO s are usually built at downstream ends of sewer systems, the function of the upstream system behaviour can be compared to the function of an inherent throttling valve. At a certain tipping point, runoff is rather transformed into flooding than into overflow [14, 29]. Hence, calibration points after CSO structures are not able to represent processes like flooding, as detailled information is generalized with both flooding and overflow events. Specific reason for higher deviations in flooding compared to CSO volumes can be found in the original planning criteria for pipe dimensioning. It can be assumed, that the current drainage system

14 has been planned based on classifying the catchment as a residential if not even a rural area by then. Dimensioning according to Austrian guidelines [24] is based on risk approaches. Depending on the used method for dimensioning pipe diameters, sewers have to comply with different requirements regarding their frequency of occurring flooding events. Assuming the case of an initial classification as a residential area while using a rational method for pipe dimensioning, planning engineers would have had to proof a flooding frequency of maximum once every 2 years (i.e. 50% probability for a flooding event per year). This would explain the obvious transformation of the here evaluated design storms into flooding to a higher extent than into CSO, as they have return periods higher than 2 years. A second specific reason could be found in the population growth. The population in the surrounding area of conurbations like Innsbruck increased significantly in the last century [30]. From the late 1950 s, in Telfs it almost tripled until now (from 4786 inhabitants in 1951 to 5438 in 1961 and in 2017) [31]. Pipe dimensions might not have been adapted correspondingly. Thus, their capacity reserves decreased accordingly and flooding becomes more likely. 4 Conclusion The evaluation of different measurement campaigns showed significant improvements of model performance with an increasing number of calibration points. Especially uncertainties regarding an unfavourable choice of input data can be lessened by a favourable sampling of output data. A model based approach was used to evaluate different calibration scenarios with 1, 2 and 5 calibration points. They were validated subsequently with different rainfall sets, including measured rainfall series from available surrounding measurement stations and artificially created design storm events with different return periods. Then they were compared to scenarios processed in Tscheikner-Gratl et al. (2016) [1]. Even for different rainfall inputs and the neglection of spatial distributions of occurring intensities, their validations resulted in very good agreements and low deviations from the used reference scenario. Models calibrated to more than one calibration point can reproduce the occurring CSO and flooding volumes with almost neglectable deviations. Often, economic or practical reasons restrict the execution of extended measurement campaigns. This study shows that with a careful selection of input data also one well chosen calibration point can express and predict the system behaviour of a small case study to a satisfiable extend for engineering purposes. Still, the remaining error in the agreement with the used measurement data can be lowered substantially with an appropriate expansion of the in-sewer measurement campaign. The performed sensitivity analysis exemplified a possible error range due to the loss of information. For design storms with return periods of 5 and 10 years, it shows higher possible deviation ranges for flooding than for CSO volumes. Especially for rain events with higher return periods, the prediction of flooding volumes gets more and more uncertain. Simulated CSO volumes seem to converge to a certain limit. They should be considered more for the reflection of the system behaviour under moderate rainfalls. This analysis ushers the necessity of distributed measurements to get a better understanding of the overall system behaviour. It further emphasizes the importance of conferring resources to the calibration process. These resources are meant in terms of ensuring the availability of different kinds of measurement data as well as the time used to enable a well considered planning of data collection.

15 5 References [1] F. Tscheikner-Gratl, P. Zeisl, C. Kinzel, J. Leimgruber, T. Ertl, W. Rauch, and M. Kleidorfer, "Lost in calibration: why people still don't calibrate their models, and why they still should a case study from urban drainage modelling," Water Science & Technology, vol. 74, no. 10, pp , [2] D. Muschalla, M. Schutze, K. Schroeder, M. Bach, F. Blumensaat, G. Gruber, K. Klepiszewski, M. Pabst, A. Pressl, N. Schindler, A. M. Solvi, and J. Wiese, "The HSG procedure for modelling integrated urban wastewater systems," Water Science & Technology, vol. 60, no. 8, pp , [3] F. di Pierro, S. Djordjević, Z. Kapelan, S. T. Khu, D. A. Savic, and G. A. Walters, "Automatic calibration of urban drainage model using a novel multi-objective genetic algorithm," Water Science & Technology, vol. 52, no. 5, pp , [4] M. Kleidorfer, A. Deletic, T. D. Fletcher, and W. Rauch, "Impact of input data uncertainties on urban stormwater model parameters," Water Science & Technology, vol. 60, no. 6, pp , [5] A. Deletic, C. B. S. Dotto, D. T. McCarthy, M. Kleidorfer, G. Freni, G. Mannina, M. Uhl, T. D. Fletcher, W. Rauch, J. L. Bertrand-Krajewski, and S. Tait, "Defining Uncertainties in Modelling of Urban Drainage Systems," presented at the 8th International Conference on Urban Drainage Modelling and 2nd International Conference on Rainwater Harvesting and Management, 7th - 12th September, 2009, Tokyo, Japan, London, [6] G. Freni, G. Mannina, and G. Viviani, "Assessment of data availability influence on integrated urban drainage modelling uncertainty," Environmental Modelling & Software, vol. 24, no. 10, pp , [7] V. Notaro, C. M. Fontanazza, G. Freni, and V. Puleo, "Impact of rainfall data resolution in time and space on the urban flooding evaluation," Water Science & Technology, vol. 68, no. 9, pp , [8] G. Leonhard, Development and application of software sensors and reverse models for urban drainage systems: Model-based approaches for gaining more information from measurement data (Forum Umwelttechnik und Wasserbau). Innsbruck: Innsbruck Univ. Press, [9] A. Deletic, C. B. S. Dotto, D. T. McCarthy, M. Kleidorfer, G. Freni, G. Mannina, M. Uhl, M. Henrichs, T. D. Fletcher, W. Rauch, J. L. Bertrand-Krajewski, and S. Tait, "Assessing uncertainties in urban drainage models," Physics and Chemistry of the Earth, Parts A/B/C, vol , p. 3 10, [10] M. Kleidorfer, M. Moederl, S. Fach, and W. Rauch, "Optimization of measurement campaigns for calibration of a conceptual sewer model," Water Science & Technology, vol. 59, no. 8, pp , [11] T. Vonach, F. Tscheikner-Gratl, W. Rauch, and M. Kleidorfer, "Sensor placement for hydrodynamic model calibration," in 14th International Conference on Urban Drainage (ICUD 2017), Prague (Czech Republic), 2017, pp : Den Haag: International Water Association. [12] M. Kleidorfer, U. Tschiesche, F. Tscheikner-Gratl, R. Sitzenfrei, F. Kretschmer, D. Muschalla, T. Ertl, and W. Rauch, "Von den Daten zum Modell: Anforderungen an hydraulische Entwässerungsmodelle in kleinen und mittleren Gemeinden," in Wiener Mitteilungen, vol. 231, T. Ertl, Ed., 2014, pp. J1-J14.

16 [13] D. Muschalla, M. Sulzbacher, J. Leimgruber, R. Maier, T. Ertl, R. Neunteufel, F. Kretschmer, M. Kleidorfer, and F. Tscheikner-Gratl, "An efficient way from data to model (DATMOD) - renovation and adaptation planning for small and medium size sewer networks," [14] F. Tscheikner-Gratl, P. Zeisl, C. Kinzel, J. Leimgruber, T. Ertl, W. Rauch, and M. Kleidorfer, "Uncertainties in hydrodynamic modelling regarding rainfall measurements and calibration," in Computing and Control for the Water Industry Conference (CCWI 2016), Amsterdam, The Netherlands, [15] J. Gironás, L. A. Roesner, L. A. Rossman, and J. Davis, "A new applications manual for the Storm Water Management Model (SWMM)," Environmental Modelling & Software, vol. 25, no. 6, pp , [16] G. Burger, R. Sitzenfrei, M. Kleidorfer, and W. Rauch, "Parallel flow routing in SWMM 5," Environmental Modelling & Software, vol. 53, pp , [17] J. Yazdi, "Rehabilitation of Urban Drainage Systems Using a Resilience-Based Approach," Water Resources Management, [18] Y. Gong, X. Li, D. Zhai, D. Yin, R. Song, J. Li, X. Fang, and D. Yuan, "Influence of Rainfall, Model Parameters and Routing Methods on Stormwater Modelling," Water Resources Management, [19] DWA-A 118, Hydraulische Bemessung und Nachweis von Entwässerungssystemen [Hydraulic design of urban drainage systems], Deutsche Vereinigung für Wasserwirtschaft, Abwasser und Abfall ev, Magdeburg, Germany, [20] J. E. Nash and J. V. Sutcliffe, "River flow forecasting through conceptual models part I A discussion of principles," Journal of Hydrology, vol. 10, no. 3, p , [21] R. H. McCuen, Z. Knight, and A. G. Cutter, "Evaluation of the Nash Sutcliffe Efficiency Index," Journal of Hydrologic Engineering, vol. 11, no. 6, p , [22] J. A. Nelder and R. Mead, "A Simplex Method for Function Minimization," The Computer Journal, vol. 7, no. 4, pp , [23] Q. Duan, S. Sorooshian, and V. K. Gupta, "Optimal use of the SCE-UA global optimization method for calibrating watershed models," Journal of Hydrology, vol. 158, no. 3, pp , 1994/06/15/ [24] ÖWAV RB11, "Regelblatt 11 - Richtlinien für die abwassertechnische Berechnung und Dimensionierung von Abwasserkanälen [Guidelines for the calculation, dimensioning and design of sewers]," Österreichischer Wasser- und Abfallwirtschaftsverband, Vienna, Austria, 2009, p. 96. [25] C. Mikovits, W. Rauch, and M. Kleidorfer, "Importance of scenario analysis in urban development for urban water infrastructure planning and management," Computers, Environment and Urban Systems, [26] S. De Toffol, M. Kleidorfer, and W. Rauch, "Vergleich hydrodynamischer und hydrologischer Simulationsmodelle bei der Berechnung der Emissionen von Mischwasserbehandlungsanlagen," in Kanalmanagement 2006, R. Haberl, Ed. (Wiener Mitteilungen, Vienna: University of Natural Resources and Life Sciences (BOKU), 2006, pp. H1-H20. [27] V. Weilguni. (2015). ehyd Bemessungsniederschlag [Design rainfall events in Austria]. [28] F. Ahmed, "A Hydrologic Model of Kemptville Basin Calibration and Extended Validation," Water Resources Management, journal article vol. 26, no. 9, pp , July

17 [29] R. Sitzenfrei, C. Urich, M. Möderl, and W. Rauch, "Assessing the efficiency of different CSO positions based on network graph characteristics," Water Science & Technology, vol. 67, no. 7, pp , [30] C. Mikovits, F. Tscheikner-Gratl, W. Rauch, and M. Kleidorfer, "Integrierte Betrachtung von Anpassungsmaßnahmen und Rehabilitierung," presented at the Sanierung und Anpassung von Entwässerungsmaßnahmen - Alternde Infrastrukur, Landnutzungsänderungen und Klimawandel, Wien, [31] Statisik Austria. Bevölkerung von Bezirken: STATcube - Statistische Datenbank von STATISTIK AUSTRIA + Regionale Gliederung / Gemeinden. Acknowledgements This work is submitted and under review for the journal WARM. A previous shorter version of the paper has been presented in the 10th World Congress of EWRA Panta Rei Athens, Greece, 5-9 July This work was funded by the Austrian Climate and Energy Fund in the project CONQUAD (9th Call of the Austrian Climate Research Program project number KR16AC0K13143). Franz Tscheikner-Gratl is financed by the Marie Skłodowska Curie Initial Training Network QUICS. The QUICS project has received funding from the European Unions Seventh Framework Programme for research, technological development and demonstration under grant agreement no We would also like to acknowledge the input and good cooperation with the Gemeindewerke Telfs, the operating company of the case study.