Using near-storm soundings from the Mesoscale Predictability Experiment to analyze convection permitting model accuracy: 28 May 2013

The Mesoscale Predictability Experiment (MPEX) observed several convective events in the Great Plains during the spring of 2013, including a convective line on 28 May. This event began in the Oklahoma Panhandle as a discrete supercell developed along the dryline, followed by further convective development along the expanding outflow boundary and upscale development into a convective line that moved northeastward into southern Kansas. Upsonde teams from the National Severe Storms Laboratory, Purdue University, Colorado State University, Texas A&M University, and SUNY Oswego sampled the pre-convective environment and near-storm environments after supercell initiation. Soundings launched underneath the anvil region of the convective line captured a classic V-shaped thermodynamic structure characteristic of downburst producing storms, and was accompanied by very strong surface winds. The upsonde teams captured not only the anvil region but the cloud-free region downwind of the anvil, thereby allowing for an assessment of how well a numerical model can correctly reproduce the environment surrounding a convective line.

A 36-member nested domain ensemble is created using the Weather Research and Forecasting (WRF) model to study this event. The outer domain (CONUS) has 15 km horizontal grid spacing while the inner domain encompassing Kansas, Oklahoma, and the Texas Panhandle has 3 km horizontal grid spacing. All available conventional observations, including WSR-88D radial velocity and reflectivity observations, are assimilated using an ensemble Kalman filter (EnKF) from the Data Assimilation Research Testbed (DART). The experiment begins assimilating conventional observations at 0100 UTC 28 May with hourly cycles until 1900 UTC. Five minutes cycles begin at 1900 UTC with the assimilation of radar data along with MADIS mesonet and Oklahoma Mesonet observations on the inner domain. The resulting 5-minute analyses replicate the observed convective evolution quite well. The question is how well the numerical model reproduces the unobserved features, with a focus on comparing the model-produced soundings with the special MPEX upsonde observations both within and nearby the convective area. This type of comparison is needed to improve our understanding of radar data assimilation within convection permitting numerical models.