Initialization of mesoscale models using radar data

Simulation of precipitation with numerical weather prediction models strongly depends on the accurate representation of the initial state of the atmosphere. A model consistent assimilation of precipitation reduces the spin-up time of the hydrological cycle significantly and improves nowcasting of surface pressure, precipitation, and dynamics.

A physical-dynamic initialization (PI) scheme for the non-hydrostatic limited-area model LM (Lokal-Modell) of the German Weather Service (DWD) is currently under development at the University of Bonn. It uses two-dimensional reflectivity measurements of the DWD radar network (C-band) as input - experiments with three-dimensional data are in preparation. Within the PI scheme the LM profiles of vertical wind, specific water vapor, and cloud water content are modified, so that the radar derived precipitation is produced by the model. Cloud top and base heights are derived from radar observations and LM analysis. The vertical wind inside the cloud is calculated assuming a simplified precipitation mechanism.

The LM precipitation forecast with the PI scheme is verified using the radar simulation model (RSM) developed at the University of Bonn and successfully implemented for operational use at the DWD (for the LM) and soon at the Finnish Meteorological Institute (for the High Resolution Limited Area Model, HIRLAM). The RSM can be seen as a first step in four-dimensional variational assimilation of radar reflectivities.

We will explain the PI technique and present a forecast verification for a convective weather situation.