| Summary: | The quality of numerical precipitation prediction depends on the accuracy of the model reproducing the true initial state of the atmosphere prior to the forecast. Typically a numerical model needs a spin-up time of several hours until its hydrological cycle is established. Assimilation of precipitation data can reduce the spin-up time significantly and consequently opens the possibility of nowcasting with Numerical Weather Prediction (NWP) models. We further enhanced the physical initialisation scheme (PIB, Physical Initialisation Bonn) by Haase (2002) in order to improve quantitative precipitation nowcasting with a high-resolution NWP model. The assimilation scheme takes as an input a radar based precipitation product and a cloud top height field retrieved from satellite observations. During the assimilation window, PIB adjusts the vertical wind, humidity, cloud water, and cloud ice to force the model state towards the measurements. The most distinctive feature of the algorithm is the adjustment of the vertical wind profile in the framework of a simple precipitation generation scheme. In this paper, we present an identical twin experiment, which reveals how the model variables are adjusted during the assimilation window, and which demonstrates the consistency of PIB with the physics of the NWP model. Three case studies with real measurements demonstrate that the scheme improves the forecast of the precipitation patterns, as well as the dynamics of the events. These improvements are found both during the assimilation window and for the first hours of the free forecast.
|
|---|
