A fundamental source of model uncertainty is the definition of model structure, including models of both flood occurrence and of the consequences of a given flood magnitude). Most flood risk estimation problems (particularly those involving rainfall-runoff modelling) allow a choice between a number of different perceptual models of the processes involved. Different implementations of these perceptual models as equations and code will produce a larger number of conceptual and procedural models. This plethora of different models and approaches proves that the transformation process from a ‘mind model’ to a material model is not straightforward. Therefore, it is not surprising that models may have slightly different structures and that these different structures may give different results (see e.g. Smart et al. 1999). In recent years an increasing amount of computer power has resulted in an increase in complexity in models from simple lumped models to highly distributed ones. The aim of increasing complexity has been to increase the realism of the resulting models, but the same dilemma noted above arises. An increase in complexity will introduce more model data and parameter requirements, where both data and parameters may be uncertain.

The model structure has to be chosen according to the circumstances, the task and the data available. For example, it may well be that many rivers can be approximated by a transfer function for certain tasks and do not require more complicated solutions e.g. the simpler solution may perform equally well in terms of model data in comparison to a more complicated one (Chow et al., 1988; Moussaa and Bocquillonb, 1996; Rashid et al., 1995; Rutschmanna and Hagera, 1996; Young, 2003), or a 3 dimensional representation may not be necessary for relatively shallow flow (Sellin and Willets, 1999). However, a physically based model representation may be chosen over a simpler structure in order to resolve the effects of different flood risk management options. The problem is complicated by the fact that it might be sometimes difficult to determine if a certain model structure has failed or is in error (Christensen, 2003). This is currently seen as more problematic for rainfall-runoff models than for flood routing models, because flow dynamics and subsurface boundary conditions are much more difficult to specify from physical considerations.

References

Chow, V.T., Maidment, D.R. and Mays, C.W., 1988. Applied Hydrology. McGraw?-Hill, New York.

Christensen, S., 2003. A synthetic groundwater modelling study of the accuracy of GLUE uncertainty intervals. Nordic Hydrology, 35(1): 45-59.

Moussaa, R. and Bocquillonb, C., 1996. Criteria for the choice of flood-routing methods in natural channels. Journal of Hydrology, 186: 1-30.

Rashid, R.S., Mizanur, M. and Chaudhry, M.H., 1995. Flood routing in channels with flood plains. Journal of Hydrology, 171(1-2): 75-91.

Rutschmanna, P. and Hagera, W.H., 1996. Diffusion of floodwaves. Journal of Hydrology, 178: 19-32.

Sellin, R. and Willets, B., 1999. Three-Dimensional Structures, Memory and Energy Dissipation in Meandering Compound Channel Flow. In: M.G. Anderson, D.E. Walling and P.D. Bates (Editors), Floodplain Processes. John Wiley, New York.

Smart, G.M., 1999. Turbulent velocity profiles and boundary shear in gravel-bed rivers. Journal of Hydraulic Engineering (ASCE), 125(2): 106-116.

Young, P., 2003. Top-down and data-based mechanistic modelling of rainfall-flow dynamics at catchment scale. Hydrological Processes, 17: 2195-2217.

Go to

Risk and Uncertainty (Description and Definition)