Mesoscale Ensemble Sensitivity and Observation Targeting of Dryline Convection

Prediction challenges still exist to correctly model and forecast severe convection along the dryline. Deterministic mesoscale models have sufficient grid resolution, along with sophisticated data assimilation systems, to model severe storms, but the timing, location, and severity of these storms remains a challenge to reproduce in part due to initial condition errors. Ensembles can be used to characterize these errors, which may yield critical information to forecasters about the predictability of convective initiation (CI). Additional predictability information is gathered by utilizing the statistics of the ensemble to assess sensitivity of dryline CI to environmental influences and produce meaningful observation target locations to improve prediction.

For this presentation, we have simulated two cases of dryline CI, 3 April 2012 and 15 May 2013, from a 50-member WRF-DART EAKF ensemble. Ensemble sensitivity analysis is applied to convective response functions to analyze dynamical sensitivities at the surface and aloft. Data denial experiments are carried out to address the impact of surface and upper-air observations on CI predictability. Analysis reveals that convection is sensitive to surface thermodynamic fields no more than 12 hours prior to CI in advective regimes over higher terrain. This implies the importance of observing temperatures that may advect into the response region aloft and inhibit subsequent convection. Surface moisture advected from the western Gulf of Mexico is also shown to be a primary sensitive feature for convection. Aloft, sensitivities to geopotential height, moisture, and temperature exist 24 hours prior to CI and exhibit unique positional and magnitude sensitivities. Dryline convection is seen to be considerably sensitive to the instability of air columns advected into the vicinity of the response region. Furthermore, sensitivity to the large-scale trough, shortwave, and jet positioning is evident and indicates the impact of forcing aloft for storm development. Data denial experiments suggest that non-linearity dominates the convective forecast response to additionally assimilated observations at longer lead times, and the subsequent utility of the linear sensitivity-based targeting technique will be discussed. The application of this observation targeting technique in real-time with the Texas Tech University Ensemble System will also be presented.