Insights into Mesoscale and Storm-Scale Predictability Gained through Ensemble Sensitivity Analysis

A number of techniques exist to examine how uncertainties in the initial conditions of a numerical simulation impact the growth of forecast error. Ensemble Sensitivity Analysis (ESA) is a fairly recent approach, where a response metric is regressed against the state space across the ensemble of forecasts. Therefore, specific regions in space and time can be identified where the uncertainty of the state has the greatest impact on a specific forecast outcome. Most of the work involving ESA has traditionally focused on synoptic scale application. We explore in this presentation the potential utility of ESA on the mesoscale and storm-scale.

We first consider a suite of high-resolution WRF simulations of the Alabama portion of the 27 April 2011 tornado outbreak, with specific focus on how heterogeneities in the inflow and storm-scale state influence the generation of low-level vertical vorticity in the simulated Tuscaloosa-Birmingham supercell storm. ESA reveals that the buoyancy in the base-state inflow and flanking downdrafts are the most influential in this regard. The influence of ancillary thunderstorms upstream of the target storm will also be discussed using the same framework.

ESA has also been applied to the 15 April 2017 case from the VORTEX-SE project, where tornado development was largely overforecast for the VORTEX-SE domain in northeast Alabama. This analysis will reveal mesoscale details that dictated the underproduction of low-level updraft rotation, specifically the role of antecedent nocturnal/morning warm advection precipitation over adjacent regions of central Alabama.

A summary of results for the specific mesoscale and storm-scale applications will be provided, along with thoughts about the general utility of ESA for these scales. Thoughts will also be offered about best practices for targeted observation in future field campaigns.