On the sensitivity of urban hydrodynamic modelling to rainfall spatial and temporal resolution

摘要

Cities are increasingly vulnerable to floods generated by intense rainfall,because of urbanisation of flood-prone areas and ongoing urbandensification. Accurate information of convective storm characteristics athigh spatial and temporal resolution is a crucial input for urbanhydrological models to be able to simulate fast runoff processes and enhanceflood prediction in cities. In this paper, a detailed study of thesensitivity of urban hydrodynamic response to high resolution radar rainfallwas conducted. Rainfall rates derived from X-band dual polarimetric weatherradar were used as input into a detailed hydrodynamic sewer model for anurban catchment in the city of Rotterdam, the Netherlands. The aim was tocharacterise how the effect of space and time aggregation on rainfallstructure affects hydrodynamic modelling of urban catchments, forresolutions ranging from 100 to 2000 m and from 1 to 10 min.Dimensionless parameters were derived to compare results between differentstorm conditions and to describe the effect of rainfall spatial resolutionin relation to storm characteristics and hydrodynamic model properties:rainfall sampling number (rainfall resolution vs. storm size), catchmentsampling number (rainfall resolution vs. catchment size), runoff and sewersampling number (rainfall resolution vs. runoff and sewer model resolutionrespectively).Results show that for rainfall resolution lower than half the catchmentsize, rainfall volumes mean and standard deviations decrease as a result ofsmoothing of rainfall gradients. Moreover, deviations in maximum waterdepths, from 10 to 30% depending on the storm, occurred for rainfallresolution close to storm size, as a result of rainfall aggregation. Modelresults also showed that modelled runoff peaks are more sensitive torainfall resolution than maximum in-sewer water depths as flow routing has adamping effect on in-sewer water level variations. Temporal resolutionaggregation of rainfall inputs led to increase in de-correlation lengthsand resulted in time shift in modelled flow peaks by several minutes.Sensitivity to temporal resolution of rainfall inputs was low compared tospatial resolution, for the storms analysed in this study.