A comparison of convective-scale ensemble predictions using initial conditions produced by ensemble data assimilation and by downscaling from global ensembles of the Unified Model in the Korea Meteorological Administration

SeHyun Kim* and Hyun Mee Kim
Yonsei University

In numerical weather prediction, forecast uncertainties are caused by initial condition error, boundary condition error, and model error. In the high-resolution limited area ensemble prediction system, its boundary conditions are provided by downscaling global ensemble predictions and the model error is considered by using a scheme such as the Random Parameters (RP). The initial conditions can be provided by either downscaling the global ensemble predictions or by other ensemble perturbation generation method. In the downscaling method, for the early forecast hours, it can resolve mostly synoptic or meso-scale structures and may not resolve small-scale structures that can be represented in a convective-scale forecast model.

The Local ensemble Transform Kalman Filter (LETKF) was developed for the convective-scale operational Unified Model (UM) in the Korea Meteorological Administration (KMA), and the forecast performances of the initial perturbations generated by the LETKF were compared with those of the initial perturbations generated by downscaling the global ensemble predictions of the KMA. The prediction results of two systems were evaluated for two periods in February and August 2014 by using surface observation data. In terms of power spectrum analysis, the LETKF resolved the small-scale structures better than the downscaling method for the early forecast times. For 10 m wind speed and 1.5 m temperature, the LETKF showed comparable root mean square error (RMSE) with the downscaling method even though the spread of the LETKF is smaller than that of the downscaling method. For the precipitation forecast, both systems showed similar equitable threat score (ETS). More detailed analysis will be presented in the meeting.

*email: kshyunn@yonsei.ac.kr
*Preference: Poster