Simulation of a mid-latitude convective storm initialized with bistatic Doppler radar data

An analysis is developed to initialize a cloud resolving model from an explicit structure of a precipitating convective system derived from multiple-Doppler radar observations. The different fields of the model prognostic variables that compose the initial state of the simulation are estimated or retrieved using a 4Dvar assimilation method in which the model is used as a week constraint using two time level observations. This permits to obtain physical fields consistent with the observations and the equations of the model. This method is applied on a mid-latitude summer storm sampled by the McGill bistatic Doppler radar network that occurred on the 2nd of august 1997. During the 30 min forward simulation, the model succeeds in representing the observed features of the three main cells that compose the storm in terms of precipitation distribution and evolution of the convective activity. After this period, the model produces less stratiform precipitations. Comparisons with a Lagrangian persistency prediction are performed and show an notable improvement in the short term forecast.