The Littlefork Tornado: A Post-Event Analysis
National Weather Service Office
At 2:38 a.m. CDT on August 9, 1993 a tornado touched down on a trailer home, killing the two people inside, two miles east of Littlefork, Minnesota (Figure 1). The intent of this article is to focus on the conditions that may have led to the development of the tornadic thunderstorm, although the line that it was part of was moving into slightly more stable air. It was determined that moderately unstable air, as well as favorable wind shear, existed over Minnesota on the early morning of August 9.
Figure 1. Map of northern Minnesota, indicating several of the locations mentioned in the study. Littlefork is approximately 16 miles southwest of International Falls, Minnesota (INL).
2. SYNOPTIC ENVIRONMENT
At 7:00 p.m. CDT Sunday, August 8, a sea-level cyclone and associated occluded front were located in central North Dakota (Figure 2). This north-south oriented front was moving quite rapidly to the east, and by 4:00 a.m. CDT Monday, August 9, it extended roughly from Roseau to Thief River Falls (TVF), and then southwestward to Fargo, North Dakota (Figure 3). The lift from this frontal system, along with PIVA at 500 mb, and low level jet convergence at the apex of this occluded front helped to produce a large area of showers and thunderstorms across the eastern Dakotas. In fact, the National Severe Storms Forecast Center (NSSFC) in Kansas City, Missouri, issued Tornado Watch #773 for a large portion of the eastern Dakotas at 3:00 p.m. CDT, which expired at 11:00 p.m. Fargo radar reported several tops to 55,000 feet during the evening, and they issued Severe Thunderstorm Warnings as far east in their County Warning Area as Hubbard County in Minnesota (1:43 a.m.). According to the 7:00 p.m. CDT August 8 rawinsonde observation from International Falls (INL), the tropopause was at 39,800 feet, and the tropopause at Bismarck, North Dakota (BIS) was 39,000 feet. Based on these two observations, it can be deduced that the storms were overshooting the tropopause by perhaps as much as 15,000 feet.
Figure 2. NMC sea-level analysis for 7:00 p.m., August 8, 1993 with isallobars (mb/3hr) shown in solid lines for net falls and dashed lines for net rises.
Figure 3. NMC sea-level analysis for 4:00 a.m., August 9, 1993 with isallobars (mb/3hr) shown in solid lines for net falls and dashed lines for net rises.
By 11:35 p.m. CDT on August 8, the showers and thunderstorms had maintained their intensity as they moved into north central Minnesota. At 12:55 a.m. CDT, a public report of a tornado in extreme eastern Polk County near the town of Gully was
received via NAWAS. WSO INL issued a Tornado Warning at 1:03 a.m. for northern Clearwater County. There were no reports of touchdowns across northern Clearwater County as of 1:30 a.m., and the warning was allowed to expire. A Severe Thunderstorm Warning was issued for Beltrami County at 1:22 a.m. Again, no reports of severe weather were received across Beltrami County. The warning was allowed to expire at 2:30 a.m.
The radar operators at Fargo and Duluth, and the staff of WSO INL did not see anything indicating a severe, not to mention tornadic, thunderstorm over this area. In fact, Duluth radar indicated mainly D/VIP level 3's and a few 4's, vertically stacked, with very low cores and no hail spikes moving through Koochiching County. The maximum top reported from WSO Fargo, North Dakota, at 2:35 a.m. CDT on August 9, was 42,000 feet, but was located about 110 miles southwest of Littlefork in Hubbard County. Duluth radar at this time indicated tops generally in the 30-35,000 foot range along this line of thunderstorms. All indications were that the line of thunderstorms had decreased too below severe thunderstorm criteria.
3. A CLOSER LOOK AT THE PRE-STORM ENVIRONMENT
The troposphere across northern Minnesota was moderately unstable during the evening of August 8, and the morning of August 9, according to the soundings across the upper Midwest (Figures 4-6). Lifted Indices (LI) at 7:00 p.m. CDT on August 8 were quite low, indicative of significant instability, where the thunderstorms initially developed across the eastern Dakotas. Utilizing the SHARP sounding analysis software package (Hart and Korotky, 1991), lifted indices from the 7:00 p.m., August 8 rawinsonde data from BIS and INL were computed to be -7°C and -4°C, respectively. The K-Index at BIS was 37, indicating the potential for numerous thunderstorms. The K-Index at INL was a less threatening 20. Showalter Indices for BIS and INL were -4°C and -2°C, respectively. However, the SWEAT Index was 387 for INL and 312 for BIS. According to the SHARP Workstation User's Manual (Hart and Korotky 1991), "SWEAT values over 300 indicate a potential for severe thunderstorm development; values over 400 favor tornadic storms, providing a trigger exists for releasing the potential instability". The hodographs shown in this article indicated favorable wind shear for INL and St. Cloud (STC). The mean wind between 0 and 3 km for INL was 193 degrees at 18 knots at 7:00 p.m. August 8. On the other hand, STC had a SWEAT index of 420 at this time, but a mid-level cap around 700 mb seemed to suppress convection across central Minnesota (Figure 6). From the SHARP sounding analyses, the height of the capping inversion at INL was slightly higher and it was weaker than at STC. The weaker cap at INL may have been broken by upward vertical motion associated with at least three meteorological entities. One such feature was the isallobaric fall center that moved across northern Minnesota (Figures 2 and 3). Another such feature was the passage of an 850-mb trough across northern Minnesota (Figure 7). Yet another such feature was the low level convergence associated with speed convergence along the low level jet, over northern Minnesota. According to Storm Data, no severe weather was reported across central or southern Minnesota on August 8 or 9.
Based on the above information, it can be seen that the troposphere was quite unstable across Minnesota at this time, and possibly more unstable over eastern North Dakota. Figure 8 shows' the Nested Grid Model (NGM) 4-layer LI initial conditions, as well as the forecast LI for the period following the occurrence of the "Littlefork Tornado." The 4-layer LI is generated utilizing the NGM, which assimilates observational data to make an analysis by using the Optimum Interpolation method. These data are then used to initialize the NGM, which uses the sigma coordinates system. According to the Forecaster's Handbook No. 1 (DOC 1993), "the first four layers of the NGM extend upward about 1500 m above the ground. Calculating the lifted index for each of these four layers ensures that surface-based inversions do not obscure the estimate of the lower troposphere static stability." In addition, the computation also uses the forecast 500-mb temperature. At 7:00 p.m. CDT Sunday, August 8, a "pocket" of extremely unstable air ( -8°C) was located just east of the frontal complex across northeast North Dakota. By 7:00 a.m. CDT, lifted indices of -4°C or less was common across all but the Arrowhead of Minnesota. These values were justified, since afternoon high temperatures were in the upper 70s and lower 80s on August 8, with dew points into the lower 70s across the eastern Dakotas and much of Minnesota, satisfying sufficient lower tropospheric instability and moisture.
We can surmise that although the thunderstorms were moving into slightly more stable air, the vertical shear profile was more favorable for tornadic development over northern Minnesota. The greater wind shear over northern Minnesota may have compensated for the "weaker" instability, and created an environment that was more favorable for tornadic thunderstorm development.
Davies-Jones (1984) notes that "most mesocyclones form when storm-relative, low-level, environmental winds are strong (10 m/s or more - roughly 20 knots) and veer markedly with height (90 degrees or more in the lowest 3 km)." To refresh your memory on the significance of using the storm-relative (SR) winds, Davies-Jones et al. (1990) state that "the ambient absolute vertical vorticity is insignificant in comparison with the horizontal vorticity associated with the vertical shear." Klemp (1987) further adds that convection may begin in an environment with no ambient vertical vorticity. However, in this case, the production of vertical vorticity must result entirely from the tilting of horizontal vorticity (right hand side of the vertical vorticity equation) contained in the ambient wind shear. As the parcel enters the updraft of the thunderstorm, the axis of spin is tilted upward in the rising air, and horizontal vorticity is converted to cyclonic rotation (Klemp 1987). "Since it's the updraft that tilts the environmental horizontal vorticity into vertical vorticity, it's the storm relative winds, not the ground-relative ones, that are important" (Davies-Jones et al. 1990).
In a weakly sheared environment, the outflow spreading beneath the storm is able to cut off the supply of warm and moist air necessary to maintain the updraft for a tornadic thunderstorm. However, in a strongly sheared environment, the thunderstorm can maintain itself for a much longer period than if it developed in a weakly sheared environment.
Klemp (1987) states two reasons for thunderstorm maintenance in the presence of strong westerly vertical wind shear. The first is that in such a highly sheared environment, strong storm-relative low-level easterly inflow prevents the cold outflow from cutting off the supply of moist air. Research by Wilhelmson and Klemp (1978), and that by Thorpe and Miller (1978), support this notion. Secondly, for an updraft that can maintain itself, lifting pressure gradients act to reinforce new updraft growth on the southern and northern portions of the central updraft. Research by Schlesinger (1980), and by Rotunno and Klemp (1982) are supportive of the latter.
A look at the stability parameters for INL at 7:00 p.m. on August 8 indicates that the "Littlefork Tornado" likely developed in an environment which favored multicellular convection, even though the stability could have dramatically changed by 2:00 a.m. on August 9. Indeed, WSO Duluth radar indicated multicellular convection in the Littlefork area, propagating along an outflow boundary from previous severe convection in the eastern Dakotas. In addition, this outflow boundary that moved across northern Minnesota may have provided an added source of cyclonic vorticity for the tornadic thunderstorm that moved through Littlefork. To add credence to the belief that this thunderstorm was multicellular, the Bulk Richardson Number (BRN), often perceived as a reasonable predictor of general storm type, was 64 at INL (Hart and Korotky, 1991). The BRN is roughly a measure of instability (CAPE) divided by a general value of wind shear. Weisman and Klemp (1982) found that a BRN of 45 had a high correlation to supercell thunderstorm development, while larger values tended to suggest multicell thunderstorms. However, it should be noted that the BRN is a poor predictor of storm rotation, since the BRN shear is a "bulk" measure. Lazarus and Droegemeier (1990) comment that it doesn't take specific effects of directional and speed shear components into account. According to the BRN, apparently the storm environment across northern Minnesota did not support supercell thunderstorm development, although Moller et al. (1990) found that multicellular structures may contain individual elements that have supercell characteristics.
A better quantitative measure of low-level storm-relative (SR) inflow and rotation are helicity (or streamwise vorticity). Again, Davies-Jones et al. (1990) assert that a general threshold for mesocyclone formation is that storm-relative winds should be at least 10 m/s with veering of at least 90 degrees in the lowest 3 km. From this it was calculated that the "general" storm-relative helicity threshold in the lowest 3-km necessary for mesocyclone formation is 150 (m/s)². In the Littlefork Tornado case, the 7:00 p.m. CDT August 8 SR Helicity was 125 (m/s)² at INL, but with an increasing low-level jet (LLJ), as shown on the 7:00 p.m. STC hodograph, it is quite possible that this value may have become greater after dark. The SR Helicity at STC at this time was 162 (m/s)², which surpasses the previously calculated threshold. This LLJ also shows up quite nicely on Figure 7, which is the 850 mb NMC analysis from 7:00 p.m. August 8. Doswell (1991) indicates that the LLJ is "one ingredient that could be added to the wind profile to yield an appropriate turning of the hodograph." Davies-Jones et al. (1990) state that "rapid increases in helicity over 1-2 hours have been observed in some cases" associated with LLJs. The initial value of helicity at the time of any sounding can change greatly with time, including INL and STC for this case. Actually, the storm relative shear at 7:00 p.m. from 0-3 km was 118 degrees, but the SR inflow from 0-3 km was marginal at 16 knots from 156 degrees. Again, with increasing wind speed from the southerly LLJ, vertical wind shear would have increased, and may have contributed too substantially higher helicity values later in the night.
From the information presented in this study, we can reasonably conjecture that the multicell thunderstorm that produced the "Littlefork Tornado" developed due to a combination of moderately unstable air and relatively strong vertical wind shear. This combination may have created a more favorable environment for tornadic development than what is present over the eastern Dakotas. After all, tornadic thunderstorms can develop in an environment of strong vertical wind shear and marginal instability. This is the case with the "Littlefork Tornado", since all indications were that it was moving into more stable air, when in fact, the thunderstorms appeared to head directly into an area with favorable storm-relative wind shear, especially considering the intrusion of the LLJ. In the realm of tornadic thunderstorms, multicellular structures produce tornadoes on a much less frequent basis than supercells. However, as Moller et al. (1990) have shown, these multicellular structures may contain elements having supercell characteristics.
As the National Weather Service modernizes its services in the next few years, tools such as the wind profiler network and the VAD wind profiles available through WSR-88D will be integral in helping operational forecasters modify existing conditions with upstream changes in both wind direction and speed. The SHARP program will continue to be a useful program for analyzing the pre-storm environment, but more emphasis in the future should be placed on monitoring changing environmental conditions. Unfortunately, it remains cost-prohibitive to expand the upper-air network across the country, so it will be increasingly important to use the new tools that will be more accessible to operational forecasters in the coming years.
The author is extremely grateful to WSO La Crosse, Wisconsin MIC, Glenn Lussky, for his numerous suggestions and comments on the original manuscript, and to WSFO Minneapolis, Minnesota Science and Operations Officer (SOO), Richard Naistat, for his much appreciated assistance with the latter stages of the review process. Other thanks to Edward Berry, Central Region, SSD for additional scientific suggestions and comments to help me formally publish this study, and to Brad Bramer for providing some of the maps used in this study.
Davies-Jones, R., 1984: Streamwise vorticity: The origin of updraft rotation in supercell storms. J. Atmos. Sci., 41, 2991-3006.
____________, D.W. Burgess, and M. Foster, 1990: Test of helicity as a tornado forecast parameter. Preprints, 16th Conf. Severe Local Storms, Kananaskis Park, Alberta. AMS (Boston), 588-592.
DOC/NOAA/NWS, 1993 : National Weather Service forecasting handbook No. 1: Centrally produced guidance and analysis products-Volume 1-AFOS, NGM. R. Meiggs, Ed., GSA printer, II-1-1 to II-1-36.
Doswell, C.A. III, 1991: A review for forecasters on the application of hodographs to forecasting severe thunderstorm, Mon. Wea. Rev., 16, 1-16.
Hart, J.A., and Korotky, J., 1991: SHARP Workstation v1.50 user's manual, NWS Forecast Office, Charleston, WV, from NWSTC PC Applications Program Guide, IV-1-IV-14.
Klemp, J.B., 1987: Dynamics of a tornadic thunderstorm, Ann. Rev. Fluid Mech., 19, 369-402.
Lazarus, S.M., and K.K. Droegemeier, 1990: The influence of helicity on the stability and morphology of numerically simulated storms. Preprints, 16th Conf. Severe Local Storms, Kananaskis Park, Alberta. AMS (Boston), 269-274.
Moller, A.R., C.A. Doswell III and R. Przybylinski, 1990: High precipitation supercells: A conceptual model and documentation. Preprints, 16th Conf. Severe Local Storms, Kananaskis Park, Alberta, AMS (Boston), 52-57.
Rotunno, R., J.B. Klemp, 1982: The influence of the shear-induced pressure gradient on thunderstorm motion. Mon. Weather Rev., 110, 136-151.
Schlesinger, R.E., 1980: A three-dimensional numerical model of an isolated deep thunderstorm. Part II: Dynamics of updraft splitting and mesovortex couplet evolution. J. Atmos. Sci., 37, 395-420.
Thorpe, A.J., and M.J. Miller, 1978. Numerical simulations showing the role of the down drought in cumulonimbus motion and splitting. Quart. J. Roy. Meteor. Soc. 104: 873-93.
Weisman, M.L., and J.B. Klemp, 1982: The dependence of numerically simulated convective storms on vertical wind shear and buoyancy. Mon. Wea. Rev., 110, 504-520.
Wilhelmson, R.B., and J.B. Klemp, 1978: A three-dimensional numerical simulation of splitting that leads to long-lived storms. J. Atmos. Sci. 35 : 1037-63.