RADAR OBSERVATIONS OF NORTHEASTERN UNITED STATES TORNADOES

Kenneth D. LaPenta*, George J. Maglaras, John S. Quinlan and Hugh W. Johnson

National Weather Service Forecast Office, Albany, NY

Lance F. Bosart and Thomas J. Galarneau

The University at Albany/SUNY

1. Introduction

During recent years, there have been a number of major tornado outbreaks in the northeastern¹ United States (Table 1). The introduction of the WSR-88D radar produced significant improvement in the National Weather Service's ability to warn the public of tornadoes during these events. While advances in the warning process have been impressive, there remains room for improvement. Modern data archival capabilities provide a unique opportunity to re-examine tornado events. Radar analyses of a large number of northeastern tornadoes will provide a better understanding of tornadic thunderstorm morphology and the radar characteristics that best identify impending tornadogenesis. This will lead to additional improvement in warning accuracy and lead time.

Storm Data (U.S. Department of Commerce 1993-1998) was used to identify tornado cases to be studied, and WSR-88D archive level III and IV data (Crum et al. 1993) were collected from the nearest radar site. Based on radar data for the 86 tornado cases studied, structural characteristics of each tornadic storm were identified. Rotational velocity characteristics and maximum gate-to-gate shear values were calculated in order to evaluate their utility in tornado detection.

*Corresponding author's address: Kenneth D. LaPenta, National Weather Service Forecast Office, CESTM, 251 Fuller Road Suite B300, Albany, NY 12203-3640; email: Kenneth.LaPenta@nooa.gov.
¹ In this study the northeastern U.S. was considered to be New England, New York, New Jersey as well as central and eastern Pennsylvania.

2. Storm Types

While supercell thunderstorms (those with significant and persistent mesocyclonic circulations) produced half of the 86 tornadoes examined, a variety of storm structures were associated with the tornadic thunderstorms (Table 2). Nearly 30 percent of the tornadoes formed along lines of thunderstorms. About 40 percent of the tornadoes developed with bowing lines or cells. Two thirds (67 percent) of the tornadoes associated with bowing lines and cells formed on the apex of the bow. Tornadic thunderstorms occasionally exhibited several different structural characteristics as they evolved through their life cycle. Supercells were observed in lines of thunderstorms, or bowing line segments evolved into supercells. Appendages or hooks were observed in less than half the tornadoes. Radar detectable boundary interactions appeared to have played a role in about one fifth of the tornadic storms. Horizontal vorticity generated along the boundaries may be an important vorticity source for low-level mesocyclones through tilting and stretching (Markowski et al. 1998). Use of additional data sources (satellite imagery and surface data) may have identified additional cases where boundaries played a role in tornado formation. Gate-to-gate rotating velocity couplets of varying intensities were identified with a majority of the tornadoes. However, more than a third of the tornadoes (mostly F0 and F1) were associated with non-rotating wind maxima and may have been indicative of gustnadoes. A number of tornadoes produced no discernable velocity signature due to time and spatial sampling limitations of the radar (see section 4).

Table 1. Recent major northeastern United States tornado outbreaks. The number of tornadoes isbased on individual listings in Storm Data (U.S.Department of Commerce 1993-1998). Tornadoes that affected multiple counties arecounted separately.

Date	Number of Tornadoes	Maximum Intensity
29 May 95	8	F3
8 Nov. 96	8	F2
3 July 97	11	F2
31 May 98	38	F3
2 june 98	16	F3

Table 2. Frequency of occurrence of various tornadic thunderstorm characteristics. Some tornadoes exhibited multiple characteristics.

Characteristics	Percent of the 86 tornadic storms exhibiting a given characteristic
Supercell	49 %
Cell on line	29 %
Bowing cell or line	34 %
Percent of bow cases with	67 %
Tornado associated with wind max	29 %
Hook or appendage present	41 %
Boundary interaction	19 %

3. Tornado Detection

For each of the 86 cases, radar data have been examined for available elevation slices and for volume scans prior to and after the tornado touchdown time. In order to provide a set of non-tornadic cases, 34 thunderstorms that produced mesocyclones, but not tornadoes, were also studied. A large number of parameters were extracted from the radar data for each tornado case and analysis of this data is in progress. Preliminary analysis showed that the WSR-88D calculated maximum observed gate-to-gate velocity shear below 3 km may be useful in identifying tornadic storms. The shear (S) in units of s^-1 is defined as:

S = V_r / (D * 1800) (1)

where V_r is the rotational velocity (kt) calculated across adjacent pixels, and D is the distance (n mi) over which the shear calculation is made. The effects of beam spreading with range in both the actual shear calculation and the ability of the radar to resolve storm features required that the observed shear values be adjusted for range. At 30 n mi from the radar site, the smallest resolvable distance (using archive III and IV data) between adjacent pixels is 0.5 n mi. It increases to 1.0 n mi at 60 n mi from the radar. We can observe how shear varies with range for a V_r of 60 kt. At 30 n mi (D=.5), the shear for a 60 kt V_r is .067 s ^-1. At a range of 60 n mi (D=1.0), the shear for the same 60 kt V_r decreases to .033 s^-1. Close to the radar (at about 15 n mi, D=0.3), the shear would increase to .111 s^-1. As D in Eq. (1) shrinks further, the shear increases rapidly. In order to account for the variations of D with distance, and for the decreasing resolution of the radar, several adjustments were made. First, for ranges less than 30 n mi, V_r and S were determined over a distance of 0.5 n mi. At distances beyond 30 n mi, shear values were normalized by setting D equal to D/0.5 as a first approximation. Further research will evaluate other means of better normalizing S for varying distances from the radar. Where a well defined rotational couplet was not detected, S was calculated at the location of the maximum gate-to-gate change in inbound and outbound velocities.

Brooks et al. (1994a) and Brooks et al. (1994b) discuss the fact that many supercells do not produce tornadoes. Their conceptual model of the development and maintenance of low-level mesocyclones associated with tornadogenesis suggests that different physical mechanisms are responsible for mid-level mesocyclones associated with supercells, and with low-level mesocyclones associated with tornadogenesis. Where the radar beam elevation and resolution allow, S in Eq. 1 provides an estimate of low-level mesocyclone strength. In order to assess the strength of the mid-level mesocyclone, the maximum velocity differential (V_m) across the entire thunderstorm was determined. As a first approximation, V_m was adjusted for range from the radar based on the variation in mesocyclone strength with distance in Mesocylcone Recognition Guidelines (Andra et. al 1994). Figure 1 plots tornadic strength as a function S and V_m (adjusted for range from the radar). In area I (upper right) there were 38 events, 37 (97 percent) of these were tornadoes, and 14 (37 percent) of which were F2 or greater intensity. The one nontornadic event occurred in a remote area of the Catskill Mountains. Area I was selected in order to minimize the inclusion of non-tornadic storms and to maximize the inclusion of F2 or greater intensity tornadoes. In area II (middle), 65 percent of events were tornadic, with 11 percent F2 intensity or greater. Area II was selected in order to include the remaining F2 intensity or greater tornadoes and to maximize the percentage of events that were tornadic when compared to area III. In area III (lower left) 50 percent of the events were tornadic, with no F2 intensity or greater tornadoes.

4. Data Limitations

There are two primary sources of errors in the data used in the study; radar sampling limitations and errors in the data used in verifying events. As the radar beam moves away from the radar the beamwidth increases, and under normal atmospheric refraction, the elevation of the beam increases. At 60 n mi from the radar the beamwidth increases to 1 n mi with a beamwidth of 2 n mi at 120 n mi. Under standard atmospheric conditions, the beam reaches a height of 1 n mi at a range of about 47 n mi. An actual tornado signature is rare and can only occur very close to the radar where the tornadic circulation is several beamwidths wide (Brown 1998). As range increases, it becomes more and more difficult to resolve features associated with a tornadic thunderstorm, and the radar beam is likely to overshoot low-level features. The availability of archived radar data and the quality of the data also is a problem. Initially, over 200 tornadoes were identified for potential study, but radar data was available for less than half of them. Even when radar data was available, there were times when the data was incomplete due to missing volume scans or missing elevation slices. Storm Data (U.S. Department of Commerce 1993-1998) was used to identify tornadoes used in this study. Radar data suggest there may be errors in times assigned to several tornadoes and there were inconsistencies in a few reports. Also, it is possible that tornadoes in remote areas were never identified.

5. Discussion

This study examined tornadic (and nontornadic) thunderstorms in the northeastern portion of the U.S. Low-level gate-to-gate shear values (S), which provide evidence of the strength of small scale (1 km or less), low-level mesocyclones, were compared to the larger scale (1 km to 10 km), midlevel mesocyclone strength (V_m) in order to assess a storm's tornadic potential. Both S and V_m were adjusted for range from the radar. Large values of S and V_m were highly correlated to tornado occurrence. However, for smaller values of S and V_m, the correlation was not as good and it was more difficult to distinguish between tornadic and nontornadic storms. Future research will attempt to improve the results. Different schemes for adjusting S and V_m for range will be investigated. In addition, temporal variations in S and V_m, as well as other WSR-88D derived parameters will also be studied in order to statistically determine the most useful radar parameters for tornado detection. This research is part of a larger project designed to employ all available data in the tornado warning decision making process. Companion studies are currently evaluating cloud to ground lightning and satellite data to assess their usefulness.

In this study, half the tornadoes examined were produced by supercells, with bowing cells or lines responsible for a third of the events. In some cases, there was little or no evidence of a tornado in the radar data. This reflects the difficulty of the tornado warning process. The warning decision must be based on more than just radar signatures. Warning meteorologists must be aware of synoptic scale environmental conditions and the likelihood that they could support tornadoes. In addition, other factors such as the presence of boundaries, spotter observations and storm history must be included in warning decisions.

Acknowledgments. Support for this research was provided by a COMET Cooperative Grant UCAR-09915806. The authors would like to thank the National Weather Service Offices at Binghamton (NY), Mount Holly (NJ), Gray (ME) and State College (PA) for providing the radar data used in this study. They would also like to thank Julie Gaddy of the National Weather Service's Eastern Region Headquarters for her constructive review of this paper.

References

Andra, D., V. Preston, E. Quetone, D. Sharp, and P. Spoden, 1994: An operational guide to configuring WSR-88D principal user processor, Operations Training Branch, Operational Support Facility, Norman OK.

Brooks, H. E., C.A. Doswell III and R. B. Wilhelmson, 1994: On the role of midtropospheric winds in the evolution and maintenance of low-level mesocylcones. Mon. Wea. Rev., 122, 126-136.

_____, C.A. Doswell III, and J. Cooper, 1994. On the environments of tornadic and nontornadic mesocyclones. Wea. Forecasting, 12, 606-618.

Brown, R. A., 1998. Nomogram for aiding the interpretation of tornadic vortex signatures measured by Doppler radar. Wea. Forecasting, 13, 05-512.

Crum, T.D., R.L. Alberty and D. W. Burgess, 1993. Recording, archiving and using WSR-88D data. Bull. Amer. Meteor. Soc., 74, 645-653.

Markowski, P. M., E. N. Rasmussen and J. M. Straka, 1998: The occurrence of tornadoes in supercells interacting with boundaries during VORTEX-95. Wea. Forecasting, 13, 852-859.

U.S. Department of Commerce, 1993-1998: Storm Data, 35-40.