ABSTRACT
In recent years, floods have become a matter of increasing concern in many countries. Flood risk is assessed by hydrologists using rainfall-runoff models, which are deterministic models of the processes by which rainfall input over a particular catchment is converted into surface runoff. Traditionally, the spatial and temporal structure of the rainfall input has been highly simplified and the flood risk assessment has focused on selected individual rainfall events. However, such an approach ignores the antecedent conditions which may be an important determinant of risk. For example, a moderate event following an extended wet spell may pose a higher risk than a severe event at a time when the soil is relatively dry. The risk of extreme flood events can therefore best be assessed by running a rainfall-runoff model continuously over extended time periods, so that the full range of antecedent conditions is represented (Wheater, 2002). Such an exercise requires long records of input rainfall data at a relatively fine temporal resolution (e.g., hourly) and incorporating spatial variation, which can be hydrologically significant even for small catchments (Wheater et al., 2000a).
