Additional Ensemble Perturbations to Correct the Atmospheric Field through Assimilation of Radar Reflectivity

In the data assimilation using ensemble Kalman filter, variables are analyzed with observational data and correlations between predicted variables calculated by ensemble forecasts. However, the correlation with predicted radar reflectivity becomes zero if rainfall is not forecasted at the observation point in all ensemble members. Therefore, assimilation of radar reflectivity causes no impact at such points.

In the present study, we propose a method to correct the atmospheric field accurately through assimilation of radar reflectivity even if rainfall is not forecasted at the observation point. Namely, we added artificial ensemble perturbations at points where rainfall was not forecasted in half of ensemble members before the analysis. The perturbations, which were obtained in each vertical layer and analysis time, were based on the ensemble-based sensitivity of rainfall to the atmospheric field within the calculation domain. The atmospheric field can be corrected by adding the perturbations because the perturbations were produced from the relationship between rainfall and the atmospheric field.

To confirm the advantage of this method, we conducted assimilation experiments with a nested local ensemble transform Kalman filter (LETKF) system. Using this system, we assimilated dense surface observational data and polarimetric radar data (radial wind and reflectivity after rain attenuation correction by differential phase) in the cases of tornadic supercells in 6 May 2012 and 2 September 2013. The number of ensemble members was 50, and the horizontal grid interval was 1 km. The comparison between experiments with and without additional ensemble perturbations showed that the assimilation with the perturbations increased the amount of midlevel water vapor and rainfall at points where no rainfall was forecasted. Therefore, this method has a possibility to improve short-term precipitation forecast.