Understanding the urban drainage system of Fehraltorf. Enhancing the reliability of Fehraltorf’s SWMM model through calibration
The hydraulic behavior and the capacity of sewer systems are assessed through hydrodynamic modeling. They are widely used in drainage planning, although their parameters are mostly estimated and the models are only checked for general plausibility. The Master‘s Thesis assessed the effect of calibration on the model reliability on the basis of a case study, namely Fehraltorf.
Fehraltorf is located in the agglomeration of Zurich and Winterthur and is inhabited by 6’292 people. Most of the village of Russikon and Fehraltorf are connected to the WWTP of Fehraltorf. The DWF amounts up to 66 l/s, whereas the residual flow Q347 of the main receiving water (Luppmen) is 72 l/s. The infiltration of the groundwater is 30.1 l/s, as a significant share of the drainage system was built below the average groundwater table.
During the thesis the conversion of the Mike Urban model to SWMM was revised and further adjustments were implemented. The data for input and calibration was chosen depending on its reliability and further processed if necessary. Rain events were selected and split for calibration and validation. Parameters known to be changed in the calibrations were selected and in the sensitivity analysis the most sensitive, and thus, important parameters were determined. Then, the calibration was conducted manually as well as automatically with simulated annealing. The effect of calibration was assessed for the calibration and validation events visually and with various objective functions, viz. Nash-Sutcliffe Efficiency, total volume, peak flow and time to peak. The hydraulic performance of the sewer system was tested by simulating a 10-year design rainfall event in the uncalibrated and the calibrated model.
For the simulations, the data of the rain gauge on the WWTP had to be discarded because it showed high deviations to the ones installed by EAWAG. The rainfall data and flow data (of the FloDars) collected by EAWAG were assessed to be reliable, but as the measurements were installed recently, only four rain events were available for calibration and validation. The objective functions could be raised to a satisfying level in the manual calibration and were further improved in the automatic calibration. The validation confirmed that the reliability of the model could be increased with calibration. Furthermore, the assessment of the sewer system revealed that there are less capacity limits in the calibrated model then in the uncalibrated one, but further adjustments are needed due to numerical instabilities.
Even though the validation suggests that manual as well as automatic calibration enhance the reliability of the model because of improved goodness of fit assessed by different objective functions, further input data and reliable calibration data is needed to perform the calibration once more.
After all, calibration is strongly recommended by the author of this thesis to not only improve parameter values which were only estimated in the uncalibrated models but also to detect systematic errors in the data and the model structure.