High-Resolution NMMB Simulations of the 29 June 2012 Derecho

The 29 June 2012 derecho was one of the most destructive mesoscale convective complexes in the United States killing 22 people and resulting in millions of people without power. The large uncertainty in the timing and strength of the derecho was due in part to a lack of run to run consistency as was the case with NCEP's 4-km North American Mesoscale Forecast System (NAM), also known as the Non-hydrostatic Multiscale Model on the B-grid (NMMB). In this study, a version of the NMMB was run using initial and lateral boundary conditions provided from various operational NCEP models, as well as from developmental systems such as version 2 of the NOAA Earth System Research Laboratory (ESRL) Rapid Refresh (RAPv2) and the NCEP NAM Rapid Refresh (NAMRR). The NMMB runs of the derecho were evaluated with and without the use of the Betts-Miller-Janjic (BMJ) convective parameterization, as well as using several different microphysics schemes. The NMMB runs using the RAPv2 initial and lateral boundary conditions provided the best forecast from which physics sensitivities could be assessed.

The remainder of the study focused on assessing the impact of the microphysics schemes on storm evolution and structure. Runs were made using the Weather Research and Forecasting Model (WRF) Single Moment 6-class (WSM-6), as well as experimental configurations of the NAM/Ferrier microphysics, which included advecting the 3D rime factor (RF) array, reducing rain evaporation (by modifying the assumed temperature of the rain drops), reducing the rimed ice fall speeds, and modifying the number concentrations of large ice. The RF takes into account the relative growth of the ice particles through liquid water accretion, which is assumed to increase ice-particle densities, versus the growth of ice through vapor deposition, gradually adjusting ice densities and fall speeds of ice from low-density snow to higher-density graupel. Vertical cross sections revealed the need to advect the RF rather than treating it as a diagnostic quantity in order to have a more spatially coherent graupel field. Graupel was identified as areas of ice with RF>5, where ice growth was dominated by liquid water accretion. The result of the RF advection was less ice aloft and a more vertically oriented convective system. The motivation for reducing the rain evaporation was to increase the 1-km AGL reflectivity, and although this occurred, the system was too weak and its motion was too slow. Reducing the rimed ice fall speeds increased the ice mixing ratios aloft, caused almost all of the ice to melt in a narrower layer closer to the freezing level, which resulted in cooling that peaked farther to the rear of the system and at higher levels instead of closer to the surface. These changes produced a stronger system that moved faster, which was closer to what was observed. Introducing a variable upper limit for the number concentrations of large ice particles (NLImax) allowed for larger, faster falling ice in the convective region and smaller, slower falling ice in the stratiform and anvil regions. By setting NLImax to very small values, it was possible to simulate a system containing large hail, with simulated reflectivities exceeding 75 dBZ. All of these physics sensitivity tests were aimed not only at improving the forecast strength and timing of the derecho, but also improving the vertical structure of simulated radar reflectivity. These microphysics changes are currently being evaluated for inclusion in the next set of NCEP regional model upgrades.