Coastal flooding is a natural hazard causing devastation to many regions throughout the world, induced by the coincidence of high spring tides, large storm surges and waves. To reduce the risk posed by coastal inundation, warning systems have been developed to enable preparations to an expected threat. Although current operational predictions provide invaluable warnings, uncertainty in model formulations and input datasets, can lead to errors in forecasts. In order to provide coastal managers with the best possible information with which to make decisions, recent research has begun to focus on the movement from deterministic to probabilistic forecasting, which aims to explicitly account for uncertainty in the system. This research described the implementation of a regional tide-surge-wave model for the Solent-Southampton Water estuarine system, a region that is likely to experience increased risk of coastal flooding in the coming century. The accuracy of the model predictions were examined relative to in-situ measurements and those obtained from independent systems. Using the model, sources of error were examined and their effects upon the model predictions quantified, with particular reference made to the spatial variability throughout the region. In light of recent research, a probabilistic modelling approach, utilising a Monte Carlo technique used to provide a forecast capable of representing the uncertainty in the system, within a suitable time-frame for real-time flood forecasting that included an hourly Kalman filter data assimilation update. The findings presented in this thesis will be of interest to coastal modellers working in complex estuarine environments where the influences of tide-surge-wave interactions upon model predictions are uncertain. Furthermore, the application of a computationally efficient model, presented here, will provide a useful comparison with traditional physically-based systems to those wishing to quantify uncertainty in regions where computational resources are low