Climatology of Tornadic and Non-Tornadic Cell Motions in Tropical Cyclones

It is well accepted that the total motion of a midlatitude tornadic supercell is a combination of advection by the mean wind and a motion resulting fron the interaction of convective updraft with the vertically sheared environment. Several studies have developed methods to estimate midlatitude supercell motions from proximity soundings in the nearby environments. While such methods perform well in midlatitudes severe weather environments, several studies, such as Molinari and Vollaro in 2008 and Eastin and Link in 2009, have used them to estimate cell motions in tropical cyclones without confirming their applicability. Interestingly, Eastin and Link showed that the midlatitude methods may not be appropriate for the variety of supercells observed in tropical cyclones (such variety tends to be smaller in diameter, depth, and longevity and thus are often called miniature supercells). Thus a detailed evaluation of midlatitude supercell motion methods should be conducted for tropical cyclone cells (tornadic and non-tornadic). If the midlatitude methods are not very accurate (hypothesized), a more optimal method for tropical cyclones should be developed.

A database of “general proximity” soundings (~200) associated with tornadic cells has been acquired to complete the evaluation and development of a more optimal method to estimate cell motions. Radar reflectivity and base velocity data from the closest NEXRAD site to the tornado has been obtained from the NCDC website to observe the cell motions associated with the proximity sounding. Using GR2Analyst software, tornadic and non-tornadic cell tracks were plotted in 5 min increments beginning roughly 15-min before to 15-min after the tornado. The “convective mode” of these cells was determined using the definitions provided by Edwards et al. (2010).

After compiling the proximity sounding and corresponding observed cell motion datasets, a development of several C++ programs has been created to read individual wind profiles associated with the soundings, compute the various motion estimates from the wind profile, and compare the estimates to the observed cell motions. The cell motions will be computed using the following methods:

1. Pressure-weighted mean wind in the 0-6km layer (McCaul 1991) 2. M76 method (Bunkers et al 2000) 3. DJ93 (Ramsay and Doswell 2005) 4. D98 (Ramsay and Doswell 2005) 5. RB98 method (Rasmussen and Blanchard 1998) 6. Original Bunkers method (Bunkers et al 2000) 7. Modified Bunkers (Ramsay and Doswell 2005) 8. Modified RB98 (Ramsay and Doswell 2005)

Given the diminutive structure of most tropical cyclone convective cells (e.g., miniature supercells) and the strong dynamic constraints on cell propagation within the rotational environment of a tropical cyclone, it is expected that the motion estimate developed for midlatitude supercells will not perform well in tropical cyclones. Thus, we will use our dataset to develop an optimal motion estimate for tornadic and non-tornadic cells in tropical cyclones. We will follow methods similar to those outlined in Bunkers et al. (2000) and Ramsey and Doswell (2005). In particular, we will adopt the basic Galilean invariant scheme of Bunkers and determine the optimal combination/definitions of Vmean, Vshear, and D for our database of soundings/cells, but use winds in the cylindrical coordinate system (VR and VT) to account for azimuthal variations between the observed cell and the corresponding proximity sounding. The statistical methods used to test this optimization will follow those outlined in both papers.

For each possible variation of Vmean, Vshear, and D, estimates of storm motion will be computed for all proximity soundings. Errors (the vector difference between the observed and estimated motion) will be computed for each proximity sounding, and then the mean absolute vector error magnitude and median absolute vector error magnitude will be determined for variation of the Vmean, Vshear, and D parameters. The combination with the lowest MEA and MDVE will be the optimal method for estimating cell motions in tropical cyclones.

Finally, stratifications of the database by convective cell mode, tornadic versus nontornadic, radius from the tropical cyclone center, and radial distance between the proximity sounding and observed cell can be performed to distinguish any significant variations to this optimal method. For example, supercells and/or tornadic cells may deviate from the mean wind more than non-tornadic cells or linear convective clusters. Likewise, cells observed far from their corresponding sounding or within 200 km of the storm center may produce large errors due to a strong background gradient in wind speeds.