Hydrological models, like the Stormwater Management Model (SWMM), play a crucial role in water resource management. However, their effectiveness often depends on calibration based on historical rainfall-runoff events. Typically, calibration of urban drainage models is based on a select few observed rainfall-runoff events, which can be chosen from a broader dataset using various selection methods. This study explores the interaction between event selection criteria and objective functions in hydrological modelling. Four event selection approaches (Rainfall Depth (RD), Maximum Rainfall Intensity over a Five-Minute Interval (H5), Mean Intensity (MI), and Hyetograph Centre of Mass (HCM)) and six objective functions, were tested using high-resolution data from 42 rainfall-runoff events in an urban catchment area. Additionally, this research conducts a Global Sensitivity Analysis (GSA) of the drainage model parameters using the Morris method through an innovative integration of Mat-SWMM and the Sensitivity Analysis for Everybody (SAFE) Toolbox, creating a robust framework for assessing how various uncertain input parameters affect the hydrological model outputs. This integration facilitates the GSA process and allows researchers to conduct sensitivity studies on SWMM models using widely available tools. It provides a reproducible, scalable, and computationally efficient framework suitable for handling hydrological datasets of various sizes. Results reveal that sensitivity to model parameters varies significantly depending on the characteristics of the rainfall events. Only five influential parameters, out of the ten tested parameters, were identified and subsequently optimised using a genetic algorithm. The study highlights the strengths of different event selection criteria for specific outputs: the RD criterion provided accurate estimates for total runoff volume, whereas the HCM and H5 criteria performed better in estimating peak flow rates. The results show that traditional objective functions, such as the squared difference, are particularly sensitive to the choice of the calibration event selection approach.
Optimizing hydrological modelling on real urban catchment: impact of calibration data selection and objective function
Assaf, Mohammed N.;Salis, Nicolò;Manenti, Sauro;Tamellini, Lorenzo;Creaco, Enrico;Todeschini, Sara
2025-01-01
Abstract
Hydrological models, like the Stormwater Management Model (SWMM), play a crucial role in water resource management. However, their effectiveness often depends on calibration based on historical rainfall-runoff events. Typically, calibration of urban drainage models is based on a select few observed rainfall-runoff events, which can be chosen from a broader dataset using various selection methods. This study explores the interaction between event selection criteria and objective functions in hydrological modelling. Four event selection approaches (Rainfall Depth (RD), Maximum Rainfall Intensity over a Five-Minute Interval (H5), Mean Intensity (MI), and Hyetograph Centre of Mass (HCM)) and six objective functions, were tested using high-resolution data from 42 rainfall-runoff events in an urban catchment area. Additionally, this research conducts a Global Sensitivity Analysis (GSA) of the drainage model parameters using the Morris method through an innovative integration of Mat-SWMM and the Sensitivity Analysis for Everybody (SAFE) Toolbox, creating a robust framework for assessing how various uncertain input parameters affect the hydrological model outputs. This integration facilitates the GSA process and allows researchers to conduct sensitivity studies on SWMM models using widely available tools. It provides a reproducible, scalable, and computationally efficient framework suitable for handling hydrological datasets of various sizes. Results reveal that sensitivity to model parameters varies significantly depending on the characteristics of the rainfall events. Only five influential parameters, out of the ten tested parameters, were identified and subsequently optimised using a genetic algorithm. The study highlights the strengths of different event selection criteria for specific outputs: the RD criterion provided accurate estimates for total runoff volume, whereas the HCM and H5 criteria performed better in estimating peak flow rates. The results show that traditional objective functions, such as the squared difference, are particularly sensitive to the choice of the calibration event selection approach.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
