INTRODUCTION - Modelling Uncertainty in Flood Forecasting Systems - page 7

1.3 Subject of the present research: uncertainty in flood
forecasting and

warning

Broadly, the subject of this study is about uncertainty in flood forecasting and warning
systems. Management of uncertainty in the context of flood forecasting and warning
implies a series of actions to be undertaken for the proper quantification of uncertainty in
the forecasts and to make use of the uncertain forecasts in decision making (including
flood warning and emergency planning). The actions should be focused to achieve a
better consequence and normally involve (Maskey and Price, 2003a): (i) identification of
the sources of uncertainty, (ii) reduction of uncertainty, (iii) quantification of uncertainty
in inputs, (iv) propagation of uncertainty and the quantification of uncertainty in the
output, and (v) use of the uncertain information in flood warning decision-making. A
complete coverage of these steps for uncertainty management is beyond the scope of the
present study. The focus here is on methods and methodologies for the propagation of
various sources of uncertainty through a flood forecasting model for the quantification of
uncertainty in flood forecasts. The frequent appearance of “warning” in this thesis is
mainly to relate the usefulness of uncertainty quantification in practice.
The issue of uncertainty in flood forecasting and warning is further described in
Subsection 1.3.1, and the potential benefits of uncertainty quantification are presented in
Subsection 1.3.2. The aim and objectives of this research are presented in Subsection
1.3.3, and the two application examples that are part of this thesis are introduced in
Subsection 1.3.4.

1.3.1 Uncertainty in flood forecasting and warning

In almost all circumstances, and at all times, we all find ourselves in a state of uncertainty
(De Finetti, 1974). Uncertainty is therefore a common experience in everyday life and
flooding is no exception. Indeed, all natural hazards are inherently uncertain in essence.
Imperfect knowledge about the procedures and data generates uncertainty in the forecasts
of floods. Various sources of uncertainty in flood forecasting are presented in Subsection
2.2.2. Errors in the forecast of the flood stage or of the time of arrival of flood conditions
may lead (Smith and Ward, 1998) either

• to under-preparation and therefore to otherwise avoidable damage (if the forecast
stage is too low and/or the forecast timing of inundation is late), or
• to over-preparation, unnecessary expense and anxiety, and to a subsequent loss of
credibility (if the forecast stage is too high and/or the forecast timing of inundation is
premature).

The fact that there exists uncertainty in flood forecasting raises issues about the reliability
and credibility of flood warnings. A scientific solution to this problem is the
implementation of uncertainty and risk-based flood warning systems. The risk-based
design of civil engineering structures is increasingly common (see Tung and Mays, 1981;
Van Gelder, 2000; Voortman, 2002; Vrijling 1993; Vrijling et al, 1998). Well formulated
mathematical methods are available for risk-based flood warnings (Krzysztofowicz,
1993). The risk-based warnings are more rational, offer economical benefits and their
needs are increasingly being realised (see Kelly and Krzysztofowicz, 1994;