Results of modelling and prediction of urban pluvial flooding

Making use of the urban storm water drainage models that were implemented during the project, as well as of the high resolution rainfall estimates that were collected and post-processed, a number of tests were conducted mainly around two topics:

  1. Evaluation of urban storm water modelling approaches, including comparison of fully-distributed vs. semi-distributed models, detailed analysis of model resolution and evaluation of model nesting approaches, amongst others.
  2. Analysis of the impact of rainfall input resolution on urban drainage modelling outputs and consequent identification of rainfall resolution requirements for urban hydrological applications.  

These tests involved several pilot locations, making the results robust and enabling drawing general conclusions and useful recommendations for urban hydrologists. The results of these tests have been summarised in a number of reports and scientific publications; the main reports/outputs are listed below. For a full list of publications, please visit our publications webpage.

Some of the main lessons learned from these tests were the following:

  • One size does not fit all! The type of urban drainage (flood) model to be used depends on:
    • Purpose (e.g. CSO reduction? flood visualisation? Real-time / off-line applications?)
    • Data availability: surface data, sewer data & rainfall data
    • Available computer power
  • There is a strong interaction between the temporal and spatial resolution of rainfall inputs and urban drainage models. Disparity between these (e.g. very fine spatial resolution with relatively coarse temporal resolution; very high resolution urban drainage model forced with relatively coarse resolution rainfall data) may lead to significant loss of information and to ill-posed models.
  • Fully-distributed models are generally desirable (over semi-distributed models), as they provide a more realistic representation of urban runoff processes and allow better visualisation of results. These models are particularly desirable in areas in which runoff ponding (before it reaches the sewer system) is a relevant flooding mechanism. Current tendency is in fact towards the use of fully distributed models. However:
  • Runtimes of fully-distributed models are still problematic. Options for overcoming this include use of nested and hybrid models, and making best use of available computational power.
  • Fully distributed models require far more detailed data for their implementation and forcing. Such data are not always available and are often hard to process. When detailed data are not available for model building, semi-distributed models may be a better option.

The temporal resolution of rainfall inputs showed to have a stronger impact on urban drainage modelling outputs, as compared to the spatial resolution. In terms of urban drainage model sensitivity, results suggest that rainfall inputs with temporal resolutions below 5 min are required for urban hydrology, whereas spatial resolution of ~500 m – 1 km may suffice (for drainage areas > 1 ha). However, from an atmospheric / meteorological perspective, measuring rainfall at high temporal and spatial resolution can lead to improved accuracy and better understanding of the microphysics of the atmosphere, which in turn can translate into improved rainfall forecasts.