Geoscience Reference
In-Depth Information
Predicting the longer term hydrological effects of change of both climate and land use is certainly
a currently fashionable indoor sport. Concerns about the impacts of global warming, deforestation and
other changes have been enhanced by public and political awareness of the greenhouse effect, the ozone
hole over Antarctica, the El Ni no effect, satellite pictures of burned and burning forests in Amazonia
and the increasing demands of a growing global population on the world's freshwater resources. Hydro-
logical modelling contributes only a small part of the global atmosphere-ocean-land modelling systems
that are being used to predict changes into the next century. It is not, however, an insignificant part.
Global circulation models (GCMs) are known to be sensitive to how the hydrology of the land surface
is represented. There is also no doubt that both global and local changes will have feedback effects on
hydrological processes in the next century, but the problem is how best to predict whether those impacts
will be significant when hydrological systems are subject to so much natural variability.
It follows that, as with models of present-day hydrological responses, we should be qualifying our
predictions by estimating uncertainties of the impacts of change or, put in another way, the risk of seeing
a certain degree of impact on flood peaks, minimum flows or the usable water resource in the future.
Translating uncertainty into a future risk can provide a valuable contribution to the decision-making
process. Risk is the more acceptable face of uncertainty. The decision maker is then faced, rather than a
prediction surrounded by a fuzzy cloud of uncertainty, with the more manageable problem of assessing an
acceptable risk. The information provided by the simulations is essentially the same (see Section 7.15).
8.2 Short-Term Future Risk: Flood Forecasting
Managing risk in the short term is one of the most important applications of rainfall-runoff modelling,
particularly for flood forecasting which requires decisions to be made as to whether flood warnings should
be issued on the basis of the data coming in from raingauges, radar rainfall images, stream gauges and
the model predictions as the event happens in “real time”. This is a risk management problem because of
the uncertainties inherent in the modelling process. It is a case where taking account of the uncertainties
might make a difference to the decision that is made (Krzysztofowicz, 2001a; Todini, 2004).
There are a number of simple principles for real-time operational forecasting that can be expressed
as follows:
•
The event of greatest interest is the next event when a warning might (or might not) need to be issued.
•
The next event is likely to be different from all previous events in some of the details of rainfall
pattern, volumes and observation uncertainty; rainfall forecast uncertainty; antecedent condition and
runoff generation uncertainty; and rating curve uncertainty where discharges for overbank flows are
of interest.
•
Allowing for uncertainty means being right more often in terms of bracketing when warning thresholds
are crossed. It also gives a more realistic representation of forecasting capability in communicating
with professional partners.
•
Estimating the uncertainty associated with any forecast should not be the end point of the analysis.
There are still issues about communicating the assumptions of the analysis to those who have to act
on predictions, of managing uncertainty by post-event analysis, and of trying to reduce the uncertainty
by more effective measurements (and longer term improvements in the underlying science).
The requirement is for the forecasts and warnings for an area at risk of flooding to be made as accurately
as possible and as far ahead as possible or with the greatest possible
lead time
. We can use a model for
these predictions that has been calibrated on historical data sets but, because every event has its own
peculiarities, we should expect that during a flood event the model predictions of river stage or discharge
will start to deviate from those values being received online at a flood forecasting office from the telemetry
