INTRODUCTION - Modelling Uncertainty in Flood Forecasting Systems - page 8

Krzysztofowicz, 1993; Krzysztofowicz et al., 1994). Proper evaluation of uncertainty is
the basis for such systems.
Hiding the uncertainty associated with forecasts creates an illusion of certainty, the
consequences of which may be very serious. An example of this is the flood event during
the Spring of 1997 on the Red River in the Grand Forks, North Dakota, USA. The
estimated 49 ft flood crest led city officials and residents to prepare as if this estimate
were a perfect forecast (Krzysztofowicz, 2001a). The forecast was seriously
underestimated as the actual flood crest of 54 ft overtopped the dikes, inundated 80% of
the city and forced almost the total evacuation of its citizens. After the event, Eliot
Glassheim, City Council Member in Grand Forks, North Dakota, observed (Glassheim,
1997):
“…the National Weather Service continued to predict that the river's crest at
Grand Forks would be 49 ft… If someone had told us that these estimates were
not an exact science,…we may have been better prepared.”
1.3.2 Benefits of uncertainty quantification in flood forecasting
Krzysztofowicz (2001a) presented various benefits of probabilistic forecasts in
hydrology. In particular for flood forecasting, uncertainty levels attached to forecasts
have the following benefits:
• They are scientifically more honest as they allow the forecaster to admit the
uncertainty and to express the degree of certitude.
• They enable an authority to set risk-based criteria for flood warnings and emergency
response.
• They provide information necessary for making rational decisions enabling the
user to take risk explicitly into account.
• They offer the potential for additional economic benefits as a result of the fore-
casts to every rational decision maker and thereby to society as a whole.
• In the absence of rational decision making procedures (for example risk-
based warnings), the information regarding uncertainty (for example the
confidence intervals, the probability of exceedence of certain levels, etc.) helps
decisionmakers to use their own judgment more appropriately for decision making.
1.3.3 Aim and objectives of the present research
The aim of this research is to develop a framework and tools and techniques for
uncertainty modelling in flood forecasting systems. The objectives of the research are:
• To identify the need for uncertainty assessment and to distinguish types and
sources of uncertainty in flood forecasting;
• To review the theories of uncertainty representation and customary methods
of uncertainty modelling;
• To develop a suitable framework and methodology, and propose improvements to
existing method(s) for the better treatment of uncertainties, particularly the
propagation of uncertainty through models, in flood forecasting;