RUC-2 SUPERCELL PROXIMITY SOUNDINGS, PART I: AN EXAMINATION OF STORM-RELATIVE WINDS NORMALIZED TO SUPERCELL DEPTH
Richard L. Thompson and Roger Edwards
Previously published studies have examined storm-relative winds in supercell environments in terms of fixed pressure or height levels (e.g., Thompson 1998), and the goals of some of these studies was to discriminate between tornadic and nontornadic supercells. We have hypothesized that more reliable discrimination between tornadic and nontornadic supercells may be achieved through scaling storm-relative wind parameters to a measure of storm depth. In this part of our examination, we analyzed Rapid Update Cycle 2 (RUC-2) model soundings for various storm-relative winds, using layers whose depth and height were scaled to the height of the equilibrium level. For example, the mid levels of a so-called “low-topped” supercell, or an exceptionally tall storm, may not be well-approximated by the 500 mb level. Scaling the SR wind layers to the depth of the storm may allow more direct comparison of SR wind profiles through the entire spectrum of supercells. In addition, our results are compared to other published proximity sounding studies that examined storm-relative winds.
2. DATA AND METHODS
RUC-2 model soundings were collected for a nationwide sample of supercells identified during 1999. Each supercell was identified via WSR-88D imagery, based on the presence of persistent hook echoes/inflow notches, and at least 20 ms-1of cyclonic shear in 0.5 or 1.5o velocity imagery. One sounding was obtained for each of 188 radar identified supercells, with the nearest available RUC grid point data in the inflow sector of the supercell serving as an effective proximity sounding (Thompson and Edwards 2000a). The motion of each supercell, derived from animated WSR-88D imagery, was used to define the storm-relative framework. No other kinematic modifications were made because of software limitations. Analysis of other thermodynamic and supercell/tornado forecast parameters is presented in part II (Edwards and Thompson 2000), where temperatures and dew points were adjusted based on surface observations and the findings of Thompson and Edwards (2000a).
For this preliminary study, RUC-2 analysis soundings associated with 188 supercells from 1999 have been analyzed, the locations of which are shown in Fig. 1. The supercells occurred primarily in the central and eastern United States; and the most concentrated number of cases extended from North Dakota southward to Texas. Of the 188 supercells, 92 were nontornadic (NCDC 1999), 62 produced weak tornadoes with F0-F1damage, and 34 spawned significant tornadoes (F2-F5 damage, after Grazulis 1993). Data were analyzed for these three subsets of supercells.
Figure 1. Geographic distribution of al RUC-2 supercell proximity soundings analyzed from 1999 (through September 9).
The RUC-2 soundings were normalized to the equilibrium level (EL) height by dividing each sounding into ten equal height layers from the surface to the EL. We refer to each layer as a “storm depth slice”, based on the top of the layer. For example, the “30% storm depth slice” refers to the layer from 20-30% of the EL height, and the calculated storm-relative (SR) wind speeds were the mean values for each slice.
3. PRELIMINARY RESULTS
The distribution of EL heights for each supercell subset is presented in Fig. 2. The EL heights for each subset were quite similar, and covered nearly the same range of values. The similarity of EL heights amongst the subsets is confirmed by the mean values in Table 1. The mean 10% storm depth slices represented nearly the same vertical depth for each subset; thus, differences in mean SR wind profiles between the supercell subsets were not attributable to systematic variations in storm depth from one subset to another.
Figure 2. Box and whiskers plot of EL height for each supercell subset. The top and bottom of each vertical bar denotes the maximum and minimum values in each subset, while the top and bottom of the shaded boxes mark the 75th and 25th percentiles, respectively. The median value is marked by a horizontal line across each box.
Table 1. Mean EL height (m) for the separate groups of nontornadic (NonTor), weak tornadic (nonsigTor), and significant tornadic supercells (SigTor). The mean 10% layer depth is also shown for each subset.
Composite vertical profiles of storm-relative wind speed are shown in Fig. 3. The vertical profiles of SR winds were remarkably similar in the mean for the three classes of supercells, especially in the lower and upper troposphere (e.g., the 10-20% and 50-90% depths in Fig. 3). However, some difference was noted in the lower and mid levels of the storms, namely the 30% and 40% of storm depth slices. In these slices, the mean SR wind speeds were approximately 2 ms-1 stronger for the significant tornadic supercells than both the nontornadic and weakly tornadic supercells. This difference is consistent with the findings of Thompson (1998; hereafter T98) in that tornadic supercells were associated with stronger SR wind speeds in the mid troposphere than nontornadic supercells. In the mean, the weakest SR wind speeds also tended to occur about 1.2-2.4 km higher above the surface in the nontornadic supercells compared to the significant tornadic supercells. Given the similarity between the nontornadic and weakly tornadic supercells, only the nontornadic and significant tornadic supercells will be discussed hereafter.
Figure 3. Graph of mean SR flow for each storm depth relative layer for nontornadic supercells (medium gray), weak tornadic supercells (light gray) and the subset of supercells producing strong or violent tornadoes (black).
A closer examination of the distribution of SR speeds for the nontornadic and significant tornadic supercells is presented in Fig. 4 for the 30% storm depth level. The significant tornadic supercells tended to have substantial SR wind speeds in the mid troposphere - 75% of the cases were associated with SR wind speeds greater than 8.8 ms-1. A similar distribution of SR wind speeds was noted in the 40% storm depth slice (not shown), where 75% of the significant tornadic supercells had SR wind speeds greater than 8.2 ms-1. T98 found that tornadic supercells were associated with 500 mb SR wind speeds greater than 7.9 ms-1, which generally agrees with our findings. Also, the significant tornadic supercell SR wind speeds were offset by about one quartile in the middle 50% of the distribution (e.g., the median value for the significant tornadic supercells in Fig. 4 was equal to the 75th percentile of nontornadic supercells). This offset infers a significant difference between the two subsets in the 30% slice.
Figure 4. Same as Fig. 2, except for SR wind speeds in the 30% storm depth slice for nontornadic and significant tornadic supercells
There were some noteworthy differences between the mid level SR winds in this sample (188 supercells), and the somewhat smaller sample (132 supercells) from T98. While the tornadic supercells were associated with substantial SR wind speeds in the mid levels, our sample showed stronger SR wind speeds for the nontornadic supercells. The significant tornadic supercells occupied the same part of the parameter space in Fig. 5 as the nontornadic supercells. The 50% slice, which represented the 4.7-5.9 km layer in the mean, was more appropriate for direct comparison to T98’s 500 mb level. Much like Fig. 5, and in contrast to T98 , there was little distinction between the supercell subsets in the 50% slice. The same range of SR wind speeds were associated with both significant tornadic and nontornadic supercells, which suggested little difference between the two samples.
Figure 5. Scatter diagram of SR wind speed for the 30% and 10% slices. The significant tornadic supercells are denoted by solid squares, and nontornadic supercells by open circles.
Another concern is the relatively small mean difference of roughly 2 ms-1 in Fig. 3 between the significant tornadic and nontornadic supercells. Thompson and Edwards (2000a) found a mean RUC-2 analysis sounding error of only -0.2 ms-1 for SR wind speeds in the 4-6 km layer, but a mean absolute error of 1.9 ms-1. The magnitude of the mean absolute error is nearly the same as the mean difference between the significant tornadic and nontornadic supercells in Fig. 3. Thus, it may be difficult to consistently discriminate between tornadic and nontornadic supercells using mid level SR winds alone.
Figure 6. Scatter diagram of SR wind speed for the 50% and 10% slices. Conventions are the same as Fig. 5.
Distributions of SR wind speeds were also examined in the lower and upper portions of the storms. As shown in Figs. 5 and 6, there was some tendency for the significant tornadic supercells to be associated with SR wind speeds of 10 ms-1 or greater in the 10% slice. However, as suggested by the scatter diagrams and Fig. 7, the distinction between significant tornadic and nontornadic in the 10% slice was not clear for SR wind speeds greater than 10 ms-1. There was considerable overlap between the clusters of SR wind speeds, which suggested little overall difference between these samples.
Figure 7. Box and whiskers plot of SR wind speed in the 10% slice. Conventions are the same as Fig. 4.
The SR wind speeds for both subsets of supercells in the 90% slice (roughly 9.5-10.6 km) were consistent with the 18-20 ms-1 SR wind speeds calculated by Kerr and Darkow (1996) in the 10-11 km layer for a sample of tornadic storms, and with the larger data base of Rasmussen and Blanchard (1998). However, very little difference existed between significant tornadic and nontornadic in the 90% slice (Fig. 8). Both supercell samples spanned nearly the same inter-quartile ranges (e.g., the 25th, 50th, and 75th percentile values were nearly the same for each), with no apparent discrimination between tornadic and nontornadic supercells.
Figure 8. Same as Fig. 7, except for the 90% slice.
4. CONCLUSIONS AND FUTURE WORK
The environments of 188 supercells from the central and eastern United States were examined with RUC-2 analysis soundings. The RUC-2 soundings were scaled to storm depth by dividing the height from the surface to the EL into ten equal slices, with the intent of improving upon SR wind parameters at fixed height or pressure levels. The results of this preliminary study suggest that the difference between tornadic and nontornadic supercells may not be quite as pronounced as found by T98 at 500 mb. However, there was a tendency for significant tornadic supercells to be associated with stronger SR wind speeds in the 30-40% slices, which represented the layer from roughly 2.4 to 4.7 km (e.g., the 700-600 mb layer).
The results of this study suggest that SR wind speeds alone do not discriminate well between tornadic and nontornadic supercell environments. More consistent discrimination between tornadic and nontornadic supercells may be achieved through various SR winds in combination with other parameters. An example of such a combination is shown in Fig. 9, which is similar to the multiple parameter comparisons presented in part II of this study.
Figure 9. Scatter diagram of SR wind speed in the 30% slice and surface parcel LCL height. Conventions are the same as Figs. 5 and 6.
As of late June 2000, our supercell sounding data base had grown to 350 cases. We expect to gather and analyze additional RUC-2 supercell soundings through early 2001, and we anticipate a total sample size in excess of 500 soundings. In addition, we may normalize to maximum parcel height. A larger sample should allow more robust comparisons between tornadic and nontornadic supercells.
We thank the computer support staff of SPC for their aid in making the data and images available. We appreciate the keen discussions and feedback from our colleagues at SPC and NSSL. Special thanks goes to Bob Johns, SPC SOO, for his encouragement and constructive comments throughout this project, and to John Hart for his numerous modifications of the N-SHARP sounding analysis software.
Edwards, R., and R.L. Thompson, 2000: RUC-2 supercell proximity soundings, part II: An independent assessment of supercell forecast parameters. Preprints, 20th Conf. Severe Local Storms, Orlando, FL, Amer. Meteor. Soc. (this volume).
Grazulis, T.P., 1993: Significant Tornadoes: 1680-1991. Environmental Films, St. Johnsbury, VT, 1326 pp.
Kerr, B. W., and G. L. Darkow, 1996: Storm-relative winds and helicity in the tornadic thunderstorm environment. Wea. Forecasting, 11, 489-505.
NCDC, 1999: Storm Data, 41, 2 through 9.
Rasmussen, E. N., and D. O. Blanchard, 1998: A baseline climatology of sounding-derived supercell and tornado forecast parameters. Wea. Forecasting, 13, 1148-1164.
Thompson, R. L., 1998: Eta model storm-relative winds associated with tornadic and non-tornadic supercells. Wea. Forecasting, 13, 125-137.
_____. and R. Edwards, 2000a: A comparison of Rapid Update Cycle (RUC-2) model soundings with observed soundings in supercell environments. Preprints, 20th Conf. on Severe Local Storms, Orlando, FL, Amer. Meteor. Soc. (this volume).