Flood simulation models may have different requirements, depending on their objective. Criteria for the selection of the appropriate tool are often based on: engineering staff time needed for model development, overall consultancy time for product delivery, speed of computation, completion time for a simulation, accuracy level of results, data requirements, numerical robustness, user-friendliness of the software and possibly others, depending on the objective of the model. These objectives may be related to flood risk analysis, flood forecasting, flood control and be based upon a variety of causes, such as storms, dam or dike breaks, hurricanes, typhoons or similar low atmospheric pressure phenomena. Recently (2004) attention has also been drawn once again to the devastating effects of tsunamis. All these application areas of numerical models have their own requirements, as will be discussed briefly in the sequel.
In many countries, insurance companies are using flood risk maps, sometimes based upon relatively simple and quick estimates obtained via simple rules in GIS. In these cases it is assumed that the value of insured property does not justify the more complex laws defining the detailed flow of water.
If the economic interest is greater, a first improvement is found by applying 1D (one-dimensional) steady flow models with GIS post processing to develop topography based flood frequency contour lines (e.g. FEMA procedures: www.floodmaps.fema.gov/fhm/). However, there is a tendency now to base flood risk analysis on more detailed combined 1D and 2D unsteady flow models for flood prone areas with valuable assets and a complex infrastructure. Federal and local governments have also become more aware of the potential of using such models for evacuation planning. In The Netherlands, for example, more than 60 % of the country is subject to flood risk and for most of these areas integrated 1D and 2D hydrodynamic models have been developed to study the effects of potential dike breaks and to provide guide lines to authorities in setting up evacuation plans. Besides producing flood depths, these models have to be capable of providing accurate estimates of flood wave propagation celerities over dry beds.
Flood forecasting sets quite different requirements. Speed of producing a forecast is one of the most important criteria, especially in areas where flash floods occur. For this reason, numerical models behind a river catchment flood forecasting system are usually 1D hydrodynamic models, gradually replacing the simpler hydrological routing techniques. There is a tendency to include partly 2D hydrodynamic models, which is already common practice in flood forecasting systems for coastal areas and seas. Numerical models for flood forecasting are usually embedded in a flood forecasting platform, such as the Delft FEWS system (Werner et al., 2004), which has recently been installed in the UK to provide flood forecasts for nearly all river basins in the country.
Important criteria for numerical models supporting flood control are accuracy, flexible schematization options, numerical robustness and consultancy time for model development and use. Currently, state-of-the-art for flood control is the use of combined 1D and 2D models (e.g. Hesselink, 2003). The former use of flood cells has been replaced by complete 2D flow descriptions, whereas sub-grid channel flow is still better described in 1D. Flood control models should be based upon reliable physical descriptions and schematizations, as part of their use is in extrapolation of calibrated models to extreme situations which have never occurred. One of the reasons to build models for flood control is the study of downstream impacts, especially cross border effects. Downstream impacts of flood control are changed flood wave celerity and changed flood peak attenuation. Higher flood wave celerities result from deepening of the river and the construction of embankments. This, in turn, leads to increased peak floods downstream. The construction of flood retention areas has opposite impacts and may be used to compensate the negative impacts. Model selection criteria then follow from the detail in which potential economic, environmental and social impacts have to be studied.
The analysis of floods caused by dam- and dike breaks requires extremely robust numerical methods, especially for the description of flooding of dry areas and the correct propagation of the wave front. Moreover, model accuracy, partly based upon the ability to describe the full hydrodynamic equations, is important, as will be discussed in the section on software and model validation. As dam- and dike break simulations are nearly always made for the prediction of their potential effects, data
The quality of the model fully depends on its descriptive capabilities of the physical system in terms of topographic and roughness data, the representativeness of the equations and the numerical methods applied. However, it has to be kept in mind that the overall model accuracy also follows from the quality of the description of the dam failure mechanism and the assumptions made here.
Floods generated by the passage of low atmospheric pressure zones such as hurricanes, typhoons and the geologically induced tsunamis require the modelling of 2D flow in coastal zones, seas and oceans and may set requirements such as the description of Coriolis forces, the use of spherical coordinates and curvilinear grids, the specification of moving atmospheric pressure fields, special ways of handling initial data etc. A possible integrated use of 1D, 2D and 3D models may provide advantages here.