Sensitivity analysis of a generic urban flow model : 2D modelling with empirical hyetographs and CDS rain

Sammanfattning: Severe flooding events in recent years have underlined the importance of accurate hydrological modelling in urban areas. There are many important parameters relating to both the rainfall distribution and properties of the land on which the rain falls that controls the impacts of the rain event. While the importance of input parameters such as initial water content, topography and extent of hardened surfaces is widely known, their impact on hydrological response in urban areas is not thoroughly understood. In 2017, scientists from the Swedish Meteorological and Hydrological Institute, SMHI, presented five empirical hyetographs based on long-term Swedish municipal rain data (Olsson, et al., 2017). The hyetographs vary in location of peak and distribution of intensity. Olsson (2019) evaluated the hydrological response in terms of water depth of the hyetographs in relation to Chicago Design Storm (CDS), a common design storm used in Swedish hydrological modelling, and found that CDS nearly always overestimates the hydrological response in comparison to the empirical hyetographs, meaning historical rainfall intensity distributions, developed by SMHI.  The aim of this thesis is to analyse the robustness of empirical hyetographs by conducting a sensitivity analysis of a generic urban model with a variation of input parameters. A statistical analysis of data on hardened surfaces, topography and initial water content was conducted to find the median, 5th and 95th percentile respectively of the range of values in Swedish conditions. These values were applied to a synthetic urban model and run together with the five empirical hyetographs and CDS in MIKE 21. The results indicate that the empirical hyetographs are not very robust relating to variations in infiltration capacity (initial water content and amount of hardened areas), while they are more robust when varying the topography. The variation of topography also resulted in large variations in water depth, time to peak and extent of flooded area, while variation of initial water content and amount of hardened surfaces had smaller, although still clear, effects. Furthermore, the results show that hyetographs with a late peak are more sensitive to variations in initial water content and hardened surfaces than hyetographs with an early peak. On average, CDS estimates the response in comparison to the empirical hyetographs accurately, with an average overestimation of 0.5%, but does not capture the range and complexity of the empirical hyetographs. Including the frequency of the different hyetographs, CDS gave an overestimation of 5% compared to the empirical hyetographs. The highlight the uncertainty in using CDS as input for urban cloudburst modelling, but limitations in form of catchment properties in a study area and the importance of the input values limits the generability of the study.