Forecasting of floods using combined nowcasting, mesoscale NWP and hydrological models

Studies of high rainfall events in mountainous terrain in New Zealand have been conducted. Although mesoscale Numerical Weather Prediction (NWP) models have undergone drastic improvements in recent years notably in increases in spatial resolution and more sophisticated parameterisations of important processes that have increased forecast accuracy. It is therefore reasonable to attempt to couple mesoscale model outputs to hydrological catchment models and thus attempt flood forecasting. However, it is clear that Quantitative Precipitation Forecasting is a difficult problem. This is because precipitation is not one of the primary variables of the model and is diagnosed indirectly from the the primary parameters. To the extent that NWP is an initial value problem, further improvements should be made by more accurately specifying the initial atmospheric state at higher resolution. To this end assimilation of high resolution cloud water content and rainfall data from satellites and rain radar should result in improvements in prediction of particularly rain and severe weather events. Some examples of severe storms interacting with topographic features are presented with cautionary results. It is suggested that for some time into the future it will be necessary to combine a nowcasting approach with mesoscale models to achieve satisfactory results