Use of radar data in short-term quantitative precipitation forecasting

The use of radar data for short term (up to three hours) quantitative precipitation forecasting is investigated. A large variety of procedures have been developed and used over the years for radar-based short term quantitative precipitation forecasting. However, they were not comprehensively analyzed, and their advantages and limitations were not accurately characterized. To overcome this deficiency, a general statistical procedure is formulated and tested on a large radar data base. The procedure treats the horizontal advection of reflectivity fields and models their intensity changes statistically. Several measures, widely used in hydrology and meteorology, are applied to evaluate the performance. The performance analysis reveals that there is predictability in rain not properly exploited by the statistical procedure.; This justifies the investigation of cloud models as a means of extending the rain predictability. A variational assimilation framework is developed and applied. The radar data assimilation in cloud models is mathematically complex and computer-intensive. This requires its thorough investigation before practical application. Experiments with simulated data are conducted to facilitate such investigation. A two-dimensional cloud model is used to study the influence of assimilation formulation, data type and errors on performance. The experiments show good results qualitatively, but poorer quantitatively. It is noticed that the error in finding the optimal solution dominates the one caused by the data inaccuracy. A one-dimensional cloud model is used to investigate the effect of model uncertainty on assimilation. A Monte Carlo simulation is performed with this model. Results show fairly good performance. This indicates that a compromise should be made between the cloud model complexity and its ability to describe the rain dynamics.; A real life case study is conducted to analyze the variational assimilation of radar data in a one-dimensional cloud model for short-term rainfall forecasting. Results indicate that the cloud model performs better than the statistical approach. However, limitations caused by the assumption of one-dimensionality are obvious. This strengthens the conclusion relative to the compromise between complexity and accuracy that is to be achieved for the benefit of short-term rain forecasting.