Geoscience Reference
In-Depth Information
in near future forecasts of rainfalls from either projections of radar images into the future or numerical
weather prediction, which at present are not very reliable.
•
Flood forecasting is best carried out using adaptive models that can take account of prediction errors
if discharge or stage data is available in real time by telemetry. This is most easily done with relatively
simple models, such as the Lambert ISO model for small catchments, or transfer function type models.
•
Rainfall-runoff modelling is also being increasingly used for prediction of flood frequencies in com-
bination with either long rainfall records or a stochastic rainstorm model as an alternative to a purely
statistical estimation of flood frequency. This approach has the potential to take account of the changing
nature of the hydrological response of a catchment with changing antecedent conditions and rainstorm
volume or intensities, but at the present time is associated with significant uncertainties.
•
There are strong realisation effects in flood frequency estimation, both between different observation
periods and in realisations of stochastic rainfall models. The observed period is but one realisation of
many possible sequences of floods in a length of record and this needs to be taken account of in model
calibration.
•
An important use of rainfall-runoff models is to assess the impact of future land use and climate change
on catchment hydrology, particularly extreme events and water yields for water resource evaluation.
Such scenario projections should be associated with an estimation of uncertainty to provide a reality
check on the prediction of future changes.
•
With some exceptions of dramatic change in small catchments, the natural variability of hydrological
systems in space and time, and generally short periods of available observations, make it very difficult
to detect, study and understand the impacts of past change as evidenced in hydrological observations.
•
The spatial complexities of land use change, for example in urban areas and in agricultural field
drainage systems make it particularly difficult to predict the impacts of change.
•
The impacts of climate change have, to date, been assessed using scenario simulations that are condi-
tional both on the rainfall-runoff model being used and on the particular climatic scenario being used
as forcing data. Such projections should be associated with an estimate of uncertainty or risk, but there
have been very few studies in which any attempt has been made to assess uncertainty of the impacts
of change. The uncertainties are such that it becomes a political issue as to how far to protect against
the impacts of future change.
