A parameter parsimonious approach for catchment scale urban hydrology – Which processes are important?

Increased urbanization and increased observed precipitation intensity and -frequency due to climate change call for urban hydrological models capable of describing urban flow dynamics in data-scarce areas. A parameter parsimonious rainfall-runoff model, DDDUrban, forced by precipitation and temperature in which most model parameters are estimated from a detailed digital elevation model using GIS or taken from the literature is presented. Snowmelt and evapotranspiration are calculated using an energy balance approach, with proxy models for the energy balance elements driven by temperature and precipitation. The model focusses on subsurface and surface flow processes using an analysis of travel time distributions which indicates that the shape of the urban hydrograph is largely independent of the comparatively very rapid process of water transport in conduits. The model uses an estimate of the distribution of subsurface velocities as a function of saturation. The study shows that the calibrated mean of this distribution agrees with the saturated hydraulic conductivity estimated from infiltration measurements. The model has been calibrated and validated on observed runoff data at a 10 min temporal resolution for two Norwegian catchments in Oslo and Trondheim with acceptable validation results measured by the Kling-Gupta Efficiency criterion (KGE = 0.56–0.69). Simulations show that precipitation infiltrated on permeable areas contributes, on average, to the total flow at a fraction corresponding to the areal fraction of permeable areas. In addition, simulations show that for saturated conditions, a significant part (~30–60%) of the flood peak is derived from saturation excess overland flow. Simulation of snowmelt indicates that a more detailed model for the spatial distribution of snow accounting for snow removal, is needed. The catchment-scale effects of Low Impact Developments in the form of 10 m² raingardens are simulated. In a residential area with 500 houses, 60 raingardens can reduce the flood peaks about 10%. A higher number of raingardens further reduce the flood peaks, but raingardens of too low capacity may increase secondary flood peaks for episodes of multiple heavy precipitation events.