15.5 NUMERICAL INVESTIGATION OF THE ROLE OF MID-LEVEL DRYNESS ON TROPICAL MINI-SUPERCELL BEHAVIOR

Glen Romine* and Robert Wilhelmson University of Illinois at Urbana-Champaign, Urbana, Illinois

1. INTRODUCTION cyclone mini-supercells by McCaul and Weisman (1996, 2001) significantly Tropical cyclones that approach land furthered the understanding of the behavior are known to bring with them many hazards, of these storms, yet left several issues to including high winds, torrential rainfall and address. In this study, the impact of mid- destructive storm surge. Further, some tropospheric dry air on tropical cyclone mini- systems are also accompanied by a supercell convection is examined in closer tornadic threat from tropical cyclone mini- detail. supercells. Tropical cyclone variants of the The case examined for this study is supercell are spatially smaller in extent than Hurricane Opal (1995), which generated those commonly found on the Great Plains, twenty-two tornadoes reported during the yet are still capable of generating significant landfall. Mean tornado reports for a Gulf tornadoes (McCaul 1991). Coast landfall are only 5.7 (McCaul 1991), Favorable vertical shear of the making the Opal case unique. Radar horizontal winds and buoyancy capable of investigations revealed many of Opal’s supporting severe thunderstorms is often tornado events were associated with mini- found in the right-front quadrant of the supercell type convection embedded within hurricane (e.g., Bogner 2000; McCaul 1991; outer hurricane rainbands. Novlan and Gray 1974; Hill 1966). Despite A high resolution simulation of Hurricane this common feature, not all hurricane Opal (1995) was conducted by Romine and landfalls result in significant tornado Wilhelmson using the MM5 (Romine and outbreaks. Within the broad favorable Wilhelmson 2002). The simulation by region, numerous rainbands may be present Romine and Wilhelmson (2002) well and the threat associated with any particular captured the observed evolution of Opal, rainband within this quadrant is uncertain. including the track, intensity and A challenge for forecasters is the precipitation distribution. Within the anticipation and identification of rainbands simulation, a favorable convective posing a public risk. These forecasters environment existed in the right front could better warn the public of potential quadrant of the hurricane supportive of hazards associated with rainband tornadic supercell-type convection using the convection given a better understanding of mesoscale model output criterion of the behavior and characteristics of outer Strensrud, Cortinas and Brooks (1997). rainband convective elements and the The innermost nested grid spacing of environment in which they occur. 1.1 km modestly resolved the convective Cloud-scale simulations of tropical elements within the simulated Opal outer rainbands. These convective elements displayed persistent, rotating updrafts, which is consistent with tropical cyclone * Corresponding author address: Glen Romine, mini-supercells observed during the actual Dept. of Atmospheric Sciences, University of event. Upon closer examination, it was Illinois Urbana-Champaign, 105 S. Gregory St., evident that the simulation also replicated Urbana, IL 61801; email:[email protected] several dry air entrainment episodes along Additional information is available on the web at: http://pampa.ncsa.uiuc.edu/~romine/opal.html the eastern flank of the hurricane. Mid-tropospheric dry intrusions roughly Soundings were extracted from the MM5 coincided with the position and orientation simulation, and used for inputs to the cloud of simulated outer hurricane rainbands, and model simulations. The soundings selected hence may play a role in the convective were in close temporal and spatial proximity evolution within these bands. Satellite to areas where convection within an outer observations also were indicative of a band hurricane rainband developed in the MM5 of dry air entraining along the eastern flank simulation. Recalling the questionable role of the hurricane several hours prior to that dry air intrusions may play, the landfall. soundings were extracted along the edge of Storms forming within a moist a midtropospheric dryness band. To test the environment have cold pools that are moist hypothesis, the same sounding was typically less intense than those that form in also modified by moistening the profile the presence of mid-tropospheric dry air. throughout the cloud bearing layer. Both Entrainment of dry air into convective cells soundings were then tested in the leads to evaporational cooling, which experimental WRF model at higher hastens the formation and strength of horizontal resolution. negatively buoyant air feeding the low-level WRF model idealized simulations are cold pool. This in turn could lead to carried out at 1 km and 500 m horizontal increased baroclinic generation of low-level resolutions on a 91x91x60 xyz grid with a vorticity that could subsequently influence 17.5 km model top. Convection is initiated tornado formation. using standard thermal bubble techniques, Gilmore and Wicker (1998) examined following the methodology utilized by the significant influence which mid- McCaul and Weisman (2001). While tropospheric dryness plays in supercell sensitivities between Lin ice and Kessler evolution. Notably, they found a direct microphysics schemes appear negligible, relationship between midtropospheric Lin ice is hereafter employed for greater dryness and cold pool intensity. Idealized consistency with the Reisner mixed phase simulations of supercell structure performed scheme within the MM5 simulation. Fig. 1 by McCaul and Weisman (1996, 2001) used illustrates the sounding input to the WRF a thermodynamic environment with fixed simulation described hereafter. 50 mb high relative humidity (90%) above the LCL. mixed surface layer CAPE was near 1000 McCaul and Weisman (2001) recognized JKg-1, concentrated in the lower portion of the need for a deeper examination of the the profile. potential role of midtropospheric dry air, noting that the magnitude of low-level vorticity was related to the cold pool 3. RESULTS intensity in their mini-supercell simulations. Idealized simulations were conducted using similar horizontal resolution to the 2. METHODOLOGY MM5 finest grid for assessment. Qualitative comparisons between the convective The grid spacing for the MM5 simulation elements generated within the MM5 could not practically be reduced any further simulation and those generated within the without violating model assumptions. As a coarse WRF simulations were in good result, the convective elements within the general agreement. In particular, vertical outer hurricane rainbands, which were velocities were slightly more robust within spatially small, were not fully resolved. the WRF simulations, but vertical and Therefore, to allow for a more detailed horizontal scales, precipitation patterns and analysis of the mini-supercell convection, low-level vorticity magnitudes were all quite cloud model simulations are employed to uniform in scale and magnitude. Cells further determine the convective structure. remained steady beyond two hours. With horizontal resolution improved to generally were less extreme than that 500 m, considerable increases in updraft shown by Fig. 2. Differences in cell vertical velocities and surface vorticity were orientation are likely related to differences in noted, roughly doubling in magnitude. azimuth between observed and simulated Convective behavior also seemed quite cell locations relative to the hurricane core. different, with increased tendency toward Preliminary results from these and splitting cells, as demonstrated by the higher resolution simulations and how they rainwater field shown in Fig. 2. Surface relate to the MM5 simulated outer rainband vorticity fields rapidly reached large values, convection will be presented. exceeding .02 s-1 in under 25 minutes from storm initiation. In contrast to the anticipated results, 4. REFERENCES initial simulations at 500 m resolution suggest the magnitude of surface vertical Gilmore, M. S., L. J. Wicker, 1998: The vorticity was only weakly moderated by influence of midtropospheric dryness on increased tropospheric humidity. These supercell morphology and evolution. differences became greater as the Monthly Weather Review: 126, 943– simulations progressed, with the dry profile 958. surface vertical vorticity 13% larger in McCaul Jr, E. W., Weisman M. L., 1996: magnitude, as shown in Fig. 3. Simulations of shallow supercell storms Much greater sensitivity was found in in landfalling hurricane environments. the choice of diffusion and microphysical Monthly Weather Review, 124, 408– parameterizations, which in turn appear to 429. have some impact on the degree of ______, ______, sensitivity to the humidity profile. The 2001: The sensitivity of simulated microphysical parameterization sensitivity is supercell structure and intensity to negligible early, but substantial after the first variations in the shapes of hour of simulation for the Lin ice and environmental buoyancy and shear Kessler microphysics options. While this profiles. Monthly Weather Review, 129, option had little impact at 1 km resolution, 664–687. Lin ice showed a 60% increase in surface Romine, G. S., Wilhelmson, R. B., 2002: vorticity magnitude by the end of the second Investigating convective elements in a hour. high resolution simulation of Hurricane A more complete vorticity budget Opal (1995). Preprints 25th Conf. On calculation likely is needed to determine the Hurricanes and Tropical Meteorology, attribution levels from individual generation San Diego, Amer. Meteor. Soc. terms for the tropical mini-supercell Stensrud, D. J., John V. Cortinas Jr., Harold convective environment. Sensitivity to E. Brooks, 1997: Discriminating variations in the height and depth of the dry between tornadic and nontornadic layer also requires further investigation. thunderstorms using mesoscale model Further testing is also clearly needed to output. Weather and Forecasting, 12, quantify the role of model physics options 613–632. for this case. Precipitation distributions of the cells generated in the higher resolution (500 m) 5. ACKNOWLEGEMENTS WRF simulations appeared to have exaggerated hook-echo shaped protrusions. The authors gratefully acknowledge the While the observed cells, shown in Fig. 4, support of NSF 99-86672, the use of NCSA did have more prominent hook-echo shaped computer resources, and the advising and features than was captured in the 1 km assistance of Brian Jewett. WRF and 1.1 km MM5 simulations, they

Fig. 3. Time trace of the surface vertical vorticity magnitude for the WRF 500 m Fig.1. Extracted sounding from the MM5 resolution simulations, with the original simulation used to initialize the WRF model (solid) and moistened (dashed) soundings. idealized simulations.

Fig. 4. Base reflectivity image of three mini- -1 Fig. 2. Surface rainwater and .01s vertical supercells within a rainband of Hurricane vorticity contour at 45 minutes from the 500 Opal. Precipitation footprints are roughly meter simulation. Primary convective cell is half-moon shaped, comparable with the near the domain center. shapes noted in the simulation (Fig. 2).