Developing a Rainfall-Runoff Routing Model using  Spatially Distributed Travel Times : Modelling a Cloudburst Event in an Urban Catchment

Abstract: The future holds challenges for urban areas when it comes to handling pluvial floodings, occurring when the rainfall intensity exceeds both the man-made and natural infiltration and drainage capacity. To gain understanding of the effects and needed measures, tools for modelling the urban response to events such as cloudbursts are needed. The aim of this project was to build a model using the Spatially Distributed Travel Time (SDTT) approach to model the rainfall-runoff response of an urban watershed. The model was developed in ArcGIS Pro using a built-in module ArcPy allowing for the use of a Python script to ensure fast calculations and simulations on grid cell basis. In total six smaller watersheds within the larger catchment were modelled with a variety in size and degree of urbanisation. Unlike fully distributed models solving for both the continuity equation and momentum equation, the models save time by applying kinematic wave approximation solving the steady state, uniform continuity equation and the Manning’s equation. The study uses only one calibration parameter representing the upstream area contributing to runoff, used for adjusting the travel times to ensure they are not too slow which could generate a delay and underestimation of the peak discharge. The model was parameterized for a cloudburst event that occurred in the city of Gävle, in the year of 2021, and was validated against a fully distributed model (MIKE 21) simulating the same event. The generated response from the SDTT model successfully returns similar hydrographs to that of a fully distributed model in most cases. It performed very well in high urbanised areas with an even spatial distribution of the two land cover classes used, impervious and pervious surfaces, and small volumes of depressions. In areas with lesser degree of urbanisation and larger depression volumes collecting runoff, the simplified model struggled to capture the draining dynamics of these. However, the model managed to match the time to peak reasonably well in the struggling areas as well. To increase the applicability of the model the upstream area contributing to runoff should be based on physical characteristics and not calibration. Further, the model should be applied to other areas preferably using other rainfall event data or design storms, as well as investigate the performance using more than two land cover classes. Finally, a sensitivity analysis could be performed for parameters that were now set to fixed values, done so to reduce the calibration.