A study of the Impact of Multiscale Initial Condition Perturbations Based on a Nested Ensemble Kalman Filter on Convection-Allowing Ensemble Forecasts

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Wednesday, 5 February 2014: 9:45 AM
Room C202 (The Georgia World Congress Center )
Aaron T. Johnson, CAPS/Univ. of Oklahoma, Norman, OK; and X. Wang

This study addresses the question of how to optimally generate the initial/lateral boundary condition (IC/LBC) perturbations for large-domain storm scale ensembles. It is necessary for an optimally designed ensemble to include perturbations that sample all sources of forecast error, including both model and IC/LBC error. In contrast to convection-parameterizing regional and global scale ensembles, the most effective methods of sampling such errors in convection-allowing storm scale ensembles over large domains are still unknown.

A separately-submitted paper showing the importance of near grid-scale IC perturbations relative to larger scale IC and physics perturbations suggests the hypothesis that a method of generating IC perturbations that are flow-dependent on all scales will improve precipitation forecast skill, compared to a common practice of using relatively coarse resolution IC perturbations.

In this study a data assimilation configuration is designed to assimilate both meso/regional-scale conventional observations as well as convective-scale radar observations using the Ensemble Square Root Filter (EnSRF). A real data case of upscale growing convection is used to demonstrate the ability of the multiscale EnSRF configuration to retrieve the regional-, meso-, and convective-scale features needed for a skillful precipitation forecast. The inner nest of convective-scale data assimilation improves the forecast skill throughout the life of the developing MCS, compared to mesoscale data assimilation only. The EnSRF configuration results in improved forecast skill at most lead times, compared to a similar 3DVar configuration, when radar radial velocity observations are assimilated. Further skill improvements are attained at many lead times by also assimilating radar reflectivity. The EnSRF provides an analysis ensemble with multiscale, flow-dependent perturbations consistent with the analysis uncertainty. The hypothesis that such perturbations can improve the sampling of IC errors in a storm scale ensemble is then tested with an OSSE to isolate the impact of IC/LBC errors from the other sources of forecast error.