Extended Abstract (260K)JP1J.15
Physical Initialization to incorporate radar precipitation data into a Numerical Weather Prediction Model (Lokal Model)
Marco Milan, Univ. of Bonn, Bonn, Germany; and F. Amen, V. Venema, A. Battaglia, and C. Simmer
We implemented the PI (Physical Initialization) method in the non hydrostatic limited-area model LM (Lokal Model) of the DWD (German Weather Service). The goal is the improvement of quantitative rain nowcasting with a high resolution NWP model. Input radar data is a DWD product: the national radar composite for 16 radars with a spatial resolution of one kilometer and a time resolution of 5 min. The conversion from reflectivity to rain rate is already made by DWD. This data is interpolated on the LM grid (2.8x2.8 km resolution) in order to calculate the analysed precipitation rate which depends on the observed precipitation and the model precipitation. The analysed precipitation rate is the starting point of PI. Our PI algorithm assumes that the model analysis is perfect and connects observation space directly with model space. It takes into account two cloud processes: condensation and evaporation. PI for every time step assimilates precipitation data, calculates the LCL (Lifting Condensation Level) and adjusts the LM profiles of vertical wind, specific water vapour and specific cloud water content taking into account whether the grid point is above, inside or beneath the cloud. Currently the cloud top is set to a constant value. In the near future we will use satellite data in order to calculate the cloud top. The model is tested in convective cases over Germany. In the tests we have already performed we met the problem that after the assimilation window there is a short time period where the model produces several strong convective cells for about an hour, which are not present in the radar data, a phenomenon we currently investigate. The LM precipitation forecast is evaluated using the radar observations. We have calculated several objective skill scores like hit rate, false alarm rate and threat score, together with spatial autocorrelation both for the precipitation forecast and for the radar data. Comparisons beetwen the results of PI and the LHN (Latent Heat Nudging) method implemented at the DWD will be performed as well.
Joint Poster Session 1J, Assimilation of Radar Data in NWP Models (Joint with 32Radar and 11Mesoscale)
Monday, 24 October 2005, 1:15 PM-3:00 PM, Alvarado F and Atria
Previous paper Next paper