Impact of Assimilating High-Density GOES-Derived Wind Vectors on Numerical Analyses of Severe Storm Environments on the U.S. Central Plains

The forecast accuracy of severe storms partly is limited by uncertainties of mesoscale shear variability in near-storm environments. Ongoing research by the coauthors and collaborators explores potential improvement in mesoscale details of shear surrounding severe convection by assimilating dense GOES-derived atmospheric motion vectors (AMVs; i.e., cloud drift winds) into convection-permitting WRF ensembles (4-km horizontal grid spacing) using the Ensemble Kalman filter. The primary goal of this is research is to understand the impact of AMVs on the numerical analysis of convective environments over the U.S. plains relative to the next most abundant conventional sources of upper air observations (e.g., routine operational radiosondes and commercial aircraft data). Prior to the operational status of the new GOES-R series, which has the potential to make the highest quality AMV data sets of all previous GOES, our current work seeks to understand the impact of the rapid-scan GOES-11-derived AMVs collected on 5 June 2009. This date was selected because GOES-11 super-rapid-scan operations were performed during the period of severe storms that formed that afternoon-evening, including the highly-documented VORTEX2 Goshen County, Wyoming, tornadic supercell event. Several valuable VORTEX2 data sets were available for verification (e.g., several proximity environment soundings).

On average, AMV observations have the greatest impact on wind analyses at altitudes above 500 mb. AMVs have a moderate impact below approximately 700 mb, and very little impact between and 700-500 mb. Such impact is highly correlated with the vertical distribution of available AMVs (the most are retrieved at upper levels and the least are retrieved in the mid-troposphere). The most positive overall impact in reducing wind bias (verified using unassimilated radiosonde wind profiles) occurs in the 500-300 mb layer, where biases in the u- and v-components of the wind are both reduced by assimilation of AMVs. Above this layer and near the ground, reduction of ensemble bias of the v-component of the wind tended to increase bias in the u-component, possibly suggesting overall directional error in the retrieved AMV direction.

Despite mild-moderate degradation of the wind analyses in the lower troposphere (an important shear layer for predicting severity of thunderstorms storms), nested storm-scale (1.33 km horizontal grid-spacing) ensemble forecast experiments using initial conditions from the parent mesoscale domains with or without GOES AMVs assimilated demonstrate a relatively small sensitivity to forecasted storm-scale structure and severe weather shear metrics. Mesoscale details of shear analyses with and without the assimilation of AMVs will shown.