To date, there has been little success in explicitly simulating severe weather over the United States, using high resolution models. Since we know that the large scale flow combined with local topography creates triggering points for convection, we should in theory be able to accurately simulate the genesis and evolution of particular tornadic supercells. However, simulating the correct chain of events, requires several generations of storms that lead to THE supercell that produces THE tornado in question. The fact that this has been a challenging task, leads to many questions: Do we have sufficient information in the synoptic scale analyses? Is the local environment changing correctly as the storms pass? Are high resolution models behaving accurately, statistically? Is the large scale flow perhaps TOO non-linear? Or are there other factors that go into convective initiation that are not captured by the analyses (for example, does watering ones crops have any impact on where convection initiates?).
My hypothesis is that we DO have enough information in the synoptic scale analyses 12 hours in advance to explicitly simulate the genesis and evolution of a particular thunderstorm. To verify this hypothesis, simulations will be done for the Oakfield, WI tornado of July 18, 1996, using the University of Wisconsin Non-Hydrostatic Modeling System (UW-NMS: Tripoli, 1992). The simulations will be initiated with the 32 km North American Regional Reanalysis (NARR) data. The reason for choosing the Oakfield, WI event is that it was statistically very well predicted. This leads me to believe that the large scale control was great enough that forcing the real scenario should be possible by a predictable synoptic scale.
Once an accurate simulation has been achieved, I hope to determine the factors that lead to predictability of several generations of storms. A nested grid system is being used in this simulation, with four grids currently running (Figure 1). A fifth grid will be added once a satisfactory mesocyclone is achieved on the fourth grid. As each subsequent grid achieves the desired results another grid will be added, thus increasing the resolution until a tornado vortex is achieved. The properties of the grids thus far are shown in the table below (Table 1).
The results thus far, have shown that vertical resolution and microphysics play a large role in the genesis and evolution of convective storms. Coarse vertical resolution of 500m smoothes out the necessary capping inversion, allowing convection to occur too easily. Fine vertical resolution of 100m caps the environment so strongly that convection does not initiate on time. However, when convection did initiate with the fine resolution, it seemed that gravity wave resonance with the boundary layer flow played a key role in giving it a focus point (shown by cursor in Figure 2). Whether or not this is a real or numerical feature has yet to be determined.
Figure 1: Nested grid configuration with 4 grids: outer grids on the left, inner grids on the right.
Grid Number / Horizontal Points / Vertical Points / Horizontal Resolution (m) / Horizontal Size (km) / Vertical Resolution (m)1 / 117x117 / 48 / 30000 / 3450x3450 / 100
2 / 78x78 / 48 / 15000 / 1140x1140 / 100
3 / 128x128 / 48 / 5000 / 630x630 / 100
4 / 352x352 / 48 / 1000 / 350x350 / 100
Table 1: UW-NMS grid properties used in simulating the Oakfield, WI Tornado.
Figure 2: Horizontal Roll focusing convective initiation in one of the 100m vertical resolution runs. The horizontal resolution is 1km.