Quantifying changes in future precipitation and uncertainty remains a challenge.
Hydrological impacts of the changes are not fully understood either. These concerns form
the research aim, which is to understand changes in the spatial-temporal structure of
rainfall in a future climate so as to investigate its hydrological impact, especially in the
context of catchment floods. To achieve this aim, high-resolution future precipitation is
generated through a rainfall simulation tool developed in this thesis, and the hydrological
impact of rainfall changes on an urban catchment is then assessed.
The first part of the thesis focuses on selecting the suitable stochastic structure used in
the simulation tool. Two spatial-temporal stochastic generators are compared, not only
in terms of rainfall statistics but also of their behaviour in yielding a realistic catchment
runoff.
The second part concerns the evaluation of a 12-member 20-year-long convectionpermitting
ensemble of hourly precipitation for the British Isles, for both the current and
future climates. The extensive evaluation for the current period demonstrates its added
value in impact studies and its potential use for downscaling purposes. The investigation
for the future climate scenario reveals future rainfall changes in more detail, in particular
for high-impact events.
Building upon the identification of a realistic rainfall generator and the examination of
climate model ability to simulate realistic precipitation, the third part focuses on
developing a simulation tool which can well combine both sources of information. This
tool is designed to provide rainfall information with a resolution suitable for urban
climate impact studies. After the calibration, an extensive validation is carried out and the
performance of the simulation tool for both current and future periods is evaluated.
The final part uses this simulation tool to assess the impact of rainfall changes upon the
estimated future urban flood risk. Three simplifications of the rainfall input, which are
usually made in engineering applications, are selected here. The effects of each
simplification on the accuracy of the estimated sewer flood risk are quantified. Results
highlight the estimation bias in the current risk assessment approach caused by these
simplifications, and also provides valuable information for decision makers in terms of
effectively reducing potential errors in sewer network designs caused by the rainfall
input.