Estimating flood frequency by continuous simulation
This thesis explores several important hydrological modelling topics surrounding the use of continuous rainfall-runoff simulation for flood frequency estimation. A continuous simulation methodology suitable for flood frequency estimation is developed. The methodology features a rainfall-runoff model (TOPMODEL, e.g. Beven, 1997), a new profile-based stochastic rainfall model (developed in this thesis), and an uncertainty estimation procedure (Generalised Likelihood Uncertainty Estimation, or GLUE e.g. Beven and Binley, 1992). By explicitly accounting for a catchment's soil moisture conditions, allowing the direct simulation of long return period flood events (via the coupling of TOPMODEL with the stochastic rainfall model), and quantifying the uncertainty associated with the simulated flood estimates, this methodology is an attractive alternative to the more traditional statistical and event-based techniques available for flood frequency estimation. It is tested successfully using data obtained from five, gauged, UK catchments. In addition to exploring the possible consistency between flood peak and continuous flow rainfall-runoff model parameterisations, the methodology is used to examine the potential impacts of climatic change upon flood frequency. Two further issues are also addressed. These are: the choice of stochastic rainfall model (for use within continuous simulation studies), and the modification of a pulse-based stochastic rainfall model for enhanced extreme rainfall simulation.