An Investigation of the Conditional Practical Predictability of the 31 May 2013 Heavy-Rain-Producing Mesoscale Convective System

During the local evening hours on 31 May 2013, a supercell thunderstorm responsible for a deadly tornado in El Reno, OK grew upscale into an intense mesoscale convective system (MCS). This MCS exhibited two distinct phases to its existence: an initial quasi-stationary phase, responsible for significant flooding and multiple fatalities in the Oklahoma City, OK metropolitan area, and a subsequent backbuilding phase as the nocturnal low-level jet intersected the cold-pool–reinforced, east-west–oriented stationary boundary across central Oklahoma along which the MCS formed. A previous study of this event showed that high-resolution deterministic forecasts of this MCS had large sensitivity to convection initiation (CI) associated with variations in horizontal grid spacing and planetary boundary layer parameterization.

MCSs can significantly impact society with heavy rain and damaging winds, yet the extent to which they are predictable is somewhat limited. Further, most MCS predictability research either prescribes CI or implicitly assumes that it will occur, likely overstating MCS predictability because CI is associated with its own limited predictability. Thus, this research seeks to quantify MCS predictability when CI is not prescribed or assumed – or the conditional predictability – for the 31 May 2013 central Oklahoma MCS. Specific focus is given to documenting forecast sensitivity to initial and lateral boundary condition uncertainty for CI and initial upscale growth into an MCS as well as the MCS’s subsequent backbuilding phase.

A fifty-member ensemble adjustment Kalman filter-based cycled data assimilation and numerical simulation forecast system is used to facilitate this study. The resulting ensemble initial conditions are used to initialize two-way-nested 15-/3-km numerical simulations that extend forward 36 h from 1200 UTC 31 May, or approximately 10 h prior to CI for the observed event, to 0000 UTC 2 June 2013. Ensemble sensitivity and composite analysis, conditioned on 6-h fractions skill score-based verification over Oklahoma, is used to illuminate forecast sensitivities for this case.

Two key findings are obtained from the analysis. First, in the most skillful CI and initial upscale growth ensemble member forecasts, the lower tropospheric baroclinic zone along which CI occurs and an upstream middle tropospheric trough that provides large-scale forcing for ascent in proximity to the baroclinic zone are located further northwest, with greater intensities, at and prior to CI. Accurate CI forecasts are found to be a prerequisite for skillful forecasts of the initial upscale growth phase of the event. Second, a dichotomy in forecast performance exists between the CI to initial upscale growth and backbuilding phases: ensemble members that are most skillful during the initial phases are generally less skillful during the backbuilding phase, and vice versa. This implies that accurate CI and upscale growth forecasts are not necessarily a prerequisite for accurate backbuilding phase forecasts. The presentation will document these findings, including hypotheses as to the associated physics and dynamics, and implication to MCS predictability.