CONCLUSIONS AND RECOMMENDATIONS - Modelling Uncertainty in Flood Forecasting Systems - page 138

the anticipated flood and the consequence is the damages (also known as disutilities) of the
flood. The uncertainty in the forecasted flood is commonly expressed probabilistically.
The estimation of damages due to flooding is a rather complicated issue and generally involves
the lack of data for the probabilistic quantification of uncertainty. In this case the hybrid
concept, called fuzzy-probabilistic risk may be used whereby the uncertainty in the estimated
damage is characterised by a fuzzy membership function based on limited available information
Maskey (2004). The author anticipates growing applications of such an approach and believes
that enough attention should be given to further explore such concepts in future research.
7.2.4 Towards uncertainty and risk-based flood forecasting and
warning systems
Risk-based design of civil engineering structures is becoming 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
economic benefits and their needs are being increasingly realised (Kelly and
Krzysztofowicz, 1994; Krzysztofowicz, 1993; Krzysztofowicz et al., 1994). But their
widespread implementation in practice is yet to be witnessed. Therefore the
implementation of such a system should be encouraged whenever practicable to achieve
additional economic benefits from flood forecasting and warning systems.
7.2.5 Uncertainty assessment in data-driven modelling
Some of the data-driven techniques, such as the fuzzy rule-based systems and the fuzzy
regression, work with imprecise data and implicitly incorporate the uncertainty concept in
modelling. These models however do not have the flexibility of using uncertainty analy-
sis methods based on other theories (e.g. the most popular probability theory). In the case
of an ANN-based model (most popular so far of the data-driven techniques), model per-
formance is, by and large, based on the difference between the observed and model pre-
dicted results using the measures like Root Mean Square Error (RMSE). Methods for the
application of more commonly used uncertainty techniques, such as based on probability
and fuzzy set theories should be explored for the ANN-based and other data-driven models
so that these models can also be used in decision making based on uncertainty and risk.