Original paper

Radar data assimilation experiments using the IPM WRF Rapid Update Cycle

Schwitalla, Thomas; Wulfmeyer, Volker

Meteorologische Zeitschrift Vol. 23 No. 1 (2014), p. 79 - 102

85 references

published: Jun 1, 2014
published online: Mar 26, 2014
manuscript accepted: Nov 15, 2013
manuscript revision received: Nov 15, 2013
manuscript received: Jun 6, 2013

DOI: 10.1127/0941-2948/2014/0513

BibTeX file


Open Access (paper can be downloaded for free)

Download paper for free


The impact of assimilating radar radial velocities and reflectivities from the French and German radar network on nowcasting to short-range quantitative precipitation forecasting (SRQPF) was investigated during COPS IOP10. For this purpose, the Weather Research and Forecasting (WRF) model was applied in a convection permitting configuration and a resolution of 3.6 km covering Central Europe. Four different experiments were carried out to study the influence of assimilating 3-dimensional radar data by employing the first developed convection permitting WRF rapid update cycle (RUC) with a 3-h cycle interval over Central Europe. For the control experiment, only conventional observations were selected for assimilation. For the other experiments, radar radial velocities and reflectivities were added. Additionally, the reflectivity operator was slightly modified to study the influence of a smaller lower boundary for the rain water mixing ratio. The results show a positive impact on SRQPF over Eastern France when applying radar radial velocities in addition to conventional observations and Global Positioning System Zenith Total Delay (GPS-ZTD). Radial velocities reduce the overestimation of 3-h precipitation compared to the control experiment without radar data. The bias and RMSE show an improvement of about 10 % respectively. When applying also radar reflectivities in addition, the 3-h precipitation bias is reduced by about 50 % compared to the assimilation of radial velocities only. The frequency bias for different precipitation thresholds now shows values < 1$<1$. This is possibly related to the different microphysics schemes in the forecast model and the applied moisture partitioning scheme in the 3DVAR due to its simplicity. The experiment with a modified reflectivity operator shows no improvement compared to the original version. The results demonstrate the great potential of assimilating radar data using variational techniques on the convection-permitting scale.


3DVARWRFRadar dataConvection permittingCOPSnowcasting