Devastating Tornado Outbreak of March 20, 1998
A review of the weather factors that came together to produce an outbreak of tornadoes in north-central North Carolina and Southside Virginia 10 years ago, including one of the strongest tornadoes recorded in the region: the deadly F3 Stoneville tornado.
Prepared by Steve Keighton, Science and Operations Officer
Summary of the March 20, 1998 tornadoes and associated damage
Tornadoes are relatively rare in the western Piedmont of North Carolina and Virginia compared to eastern parts of these states, and strong tornadoes are especially rare. In fact, since 1950, only a few F3 or stronger tornadoes have been recorded in the region around the Triad area of northwest North Carolina (Figure 1), and one of them was the deadly tornado of March 20, 1998 that tracked through the town of Stoneville. The F3 tornadoes shown in Figure 1 near Winston-Salem occurred in 1985, 1989, and a brief one in May of 1998. The F3 tornado that passed through Stoneville, NC on March 20, 1998 that is shown in Figure 1 was not the only tornado to touch down in this region on that day. Figure 2 shows that on March 20, 1998 there were actually three tornadoes, all spawned from the same thunderstorm, which touched down across northern NC and into southern VA. The two other tornadoes produced damage in the F1 category on the Fujita scale. [Note: The Fujita scale (tornado damage from F0 to F5 with associated estimated wind speeds) was modified in 2007 to account for more recent research on a large variety of damage, and is now known as the Enhanced Fujita or EF scale. F1 (now EF1) damage is thought to be associated with wind speeds of between 86-110 mph, while F3 (now EF3) damage is associated with wind speeds of between 136-165 mph.] Figure 2 shows there was also some straight-line wind damage in the break in the line between the two tornadoes near the NC/VA state line (blue plus signs), while there were quite a few large hail reports (green dots) in the region as well (some were as large as a golf ball).
Figure 1. Historical tracks of F3 and stronger tornadoes from 1950 - 2006. The March 20, 1998 tornado that tracked through Stoneville was one of these very rare tornadoes in the northwestern Piedmont of NC.
Figure 2. Tornadoes (red tracks with F-scale ratings plotted next to them), severe wind reports (blue plus signs), and severe hail reports (green dots) for the day March 20, 1998 and overnight that night.
Zooming in closer to region where the significant tornado damage occurred (Figure 3) shows that the first tornado briefly touched down near the small town of Pine Hall in eastern Stokes County and was only on the ground for about a mile and half and for about 3 minutes, but still produced quite a bit of damage. This was at about 3:15 pm local time. The second, and strongest tornado of the day, first touched down near Mayodan at 3:25pm, and traveled for about 12 miles to just northwest of Eden, produced a path width ranging from 100 yards to 800 yards, and was on the ground for 24 minutes doing extensive damage along the way. Sadly, two people lost their lives as the F3 tornado passed through Stoneville, and at least 27 others suffered injuries. The damage total from these two tornadoes was approximately $34 million. The same thunderstorm then produced a third tornado after it crossed into southern VA, touching down near the town of Sandy Level just before 400pm. This tornado was also on the ground for about 12 miles, lasted for about 25 minutes, produced F1 rated damage, and caused about $1 million in damage to homes and vehicles. Fortunately, there were no reported injuries from this tornado in VA.
Figure 3. Specific tracks of tornado damage (in green) across northern NC and southern VA. Times are in universal time (UTC), which is 5 hours later than local time (EST).
A few photos of the tornado as it approached Stoneville were taken by various people, and are shown below.
Figure 4. Photo taken by Matt Alberts and Keith
Smith while driving north on Hwy 220 with a friend
near the Mayodan exit. They were looking north and
the tornado crossed from left to right.
Figure 5. Taken in the vicinity of Stoneville by Drew Kohler.
Photo courtesy of the Greensboro News-Record.
Figure 6. Photo by Ray Priddy as tornado approached
Stoneville. Courtesy of the Town of Stoneville.
A video of the tornado and some of the damage can be viewed here.
Some photos of the damage can be found on the Town of Stoneville web site, under the Tornado selection.
Atmospheric recipe for tornadic storms
Strong tornadoes, such as the F3 Stoneville tornado, are almost always spawned by a particularly dangerous, long-lived, and rare kind of thunderstorm known as a supercell. The atmosphere not only has to somehow provide all the right ingredients to produce supercells, but other factors typically have to be present for a supercell to produce a tornado, especially a strong one. A very rare set of circumstances happened to come together during the afternoon of March 20, 1998 to produce not only several supercell thunderstorms, but these long-lived strong tornadoes as well.
First, the spring season in this region, and indeed across most of the United States, is a common time for many of the ingredients to begin to come together for tornadoes. The climatology of the Piedmont region suggests that by far the most strong tornadoes (F2 or stronger) occur in the months of March through May, with a secondary but smaller peak in September through November. These are the transition seasons, when cold dry air from Canada and the northern Plains is clashing with warm moist air from the Gulf and Atlantic ocean. Tropical storms moving inland in the fall also likely play a role in this secondary peak in strong tornadoes, although most tornadoes associated with the remnants of tropical storms tend to be weaker. Warm, moist unstable air is one key factor in producing the kinds of strong thunderstorms capable of spawning tornadoes, and these air masses are becoming more common again during the spring months. However, there are other important ingredients that become less common as we get into the summer months even though the air masses tend to get even more unstable by then. These other factors which are still common during the spring months include strong upward forcing ahead of the air mass boundary (such as a cold front), and strong energy in the upper levels of the atmosphere to help enhance the lift (often strong fronts and strong upper level energy are part of the same package in a larger storm system). A very critical ingredient that this strong upper level energy can help produce is known as vertical wind shear (winds increasing in speed and changing in direction quickly with altitude). This vertical wind shear, combined with upward forcing from the unstable atmosphere and additional forcing from a potent larger scale storm system, are the ingredients necessary for these supercell thunderstorms, which will be described a little more later.
On March 20, 1998, all of these ingredients were coming together by the afternoon over the NC and VA Piedmont areas. Figure 7 is a 4-panel image showing a number of important features that were approaching the region by early afternoon that day (1800 UTC or 100pm). In the upper left panel, a strong trough of low pressure in the upper levels (centered over TN) was providing abundant upper level energy including strong winds speeds aloft, but also bringing cooler, drier air aloft toward the region (orange and white colors in the lower left panel represents the drier air coming in from the southwest). At lower levels (the right hand panels), a surface low pressure center over eastern KY was dragging a strong cold front (seen more clearly in Figure 8) across GA and into the western Carolinas, while moist air was in place ahead of the front (green shading in the lower right panel of Figure 7).
Figure 7. 1800 UTC 4-panel analysis images showing upper level energy in the upper left (pink colors), with black lines closer together indicative of stronger winds; the lower-left image shows moist air (green) and drier air (orange) at mid levels with winds also stronger where the black lines are closer together; the upper-right image shows surface pressure fields; the lower-right image is lower level moisture and again some indication of strength of winds by the proximity of black lines. The images are courtesy of Penn State University.
Figure 8. Surface chart valid at 1200 UTC (or "Z"), which is 700am EST. A radar echo depiction, contours of constant pressure (thin blue) and surface frontal positions (red, blue, magenta) as well as locations of low and high pressure (red L's and blue H's) are also shown. Image courtesy of Unisys.
An upper air radiosonde instrument is launched by balloon every morning at the Triad International Airport near Greensboro to measure the vertical profile of temperature, moisture, and winds. This plotted profile was then modified with the surface conditions of the early afternoon to reflect the warming air mass and changing winds ahead of the approaching cold front (shown in Figure 9), and indicated the atmosphere was just becoming unstable enough for potentially strong thunderstorms by early afternoon in this region. It also showed that the wind shear was very favorable for supercell thunderstorms. As a result, the Storm Prediction Center in Norman, OK, who overnight had issued an outlook for a risk of severe thunderstorms during the afternoon, then issued a tornado watch for western North Carolina in the morning, followed by another for the Piedmont areas of NC and extreme southern VA at 1:00pm and lasting into the early evening (see Figure 10).
Figure 9. A thermodynamic diagram (know as a "Skew-T") showing a vertical profile of temperature (red), dew point temperature (green), a lifted parcel of air becoming buoyant, which could represent a thunderstorm updraft, and changing wind speed and directions with height (right-hand side of diagram in white). This was from an 1800 UTC instrument launch, but surface conditions were modified based on 2000 UTC conditions near Winston-Salem and Greensboro.
Figure 10. Tornado watch box issued by the Storm Prediction Center starting at 1:00 pm local time.
Supercell thunderstorms are dangerous enough because of their strong updrafts, long life cycle, and rotating nature (characteristics they develop because of the environmental wind shear), and this rotation, which actually can help strengthen the updraft, can occasionally result in the development of much smaller scale rotation near the ground, that being the tornado. However, tornadoes don't develop within every supercell storm, and in fact, most of the time they do not, however these storms almost always produce large hail and/or damaging straight line winds, and often frequent lightning. So there are some other special ingredients which are thought to be favorable for tornado production within a supercell, even though the complexity of how these other factors come together in just the right way to bring the damaging circulation to the ground is still not completely understood. Two such factors that meteorologists can watch for are low-level wind shear along an air mass boundary, which can help cause spin about a vertical axis that is then stretched vertically by a strong updraft that moves overhead, and another is just the right touch of cool, moist air near the storm which can lower cloud bases and also add to shear about a horizontal axis which can then be tilted into the vertical by the storm's updraft. [For more explanation of supercells and tornadogenesis, there are a couple of good sites: National Severe Storms Laboratory Primer; and the NWS "Jetstream" online school for weather.]
A surface warm frontal boundary can aid in creating both of these additional ingredients discussed above, and such a boundary happened to be in the vicinity of northern North Carolina during the afternoon of March 20, 1998. Figure 8 above shows that the position of the warm front early in the morning was still near Charlotte in southern NC, but Figure 11 below shows that by 1800 UTC the warm front had lifted northward to near the NC/VA border. This change in air mass brought the warm, moist air which contributed to the unstable conditions in the northwestern NC Piedmont, but with the boundary still nearby, any developing storms that approached it could have a better chance of producing tornadoes. Warm fronts don't have to be present for tornadoes to develop from within supercells; sometimes cool moist outflow from other nearby storms, or from within the supercell storm itself can provide this same kind of boundary. These features are often hard to identify in observational data depending on where they are compared to the weather instruments (including radars), and even harder to predict where they will be a few hours ahead of time. Even the exact position of a warm front can be tricky to pin-point sometimes.
Figure 11. Surface observations and frontal positions at 1800 UTC (100 pm), with the warm front indicated by the red line.
Tracking the supercell storms and determining
the tornado threat using satellite and Doppler radar
(animation links below requires Quicktime...please be patient as it may take more time to load)
Imagery from NOAA geostationary weather satellites (GOES) orbiting the earth at 22,000 miles in space is useful for monitoring the progress of the large scale storm system (jet steam energy, cold front position, areas of moisture vs. dry air, etc.), when additional sunshine may lead to a more unstable atmosphere, and then also the evolution and movement of the developing thunderstorm clouds themselves. You can see these thunderstorms bubbling up across western NC and then moving into central NC and southern VA in this rapid scan GOES animation (new image taken every 5 minutes). You might also be able to see areas of low-level solid clouds across northern and eastern NC, and while the boundary between these solid clouds and scattered clouds holds generally in place (this is essentially the warm front), the boiling thunderstorms eventually shift across where this boundary appears to be in north-central NC. The faster you make the loop go, the easier it might be to see.
Figure 12 below shows a single image of Doppler radar reflectivity from the KFCX radar in Floyd County VA, close to the time the second tornado touched down near Mayodan. Reflectivity indicates the intensity of the energy returned to the radar, which is a measure of how many and how large the rain drops are (and possible hail stones for the bright red or purple colors in this particular image). The northern-most supercell near the VA state line produced all three tornadoes in the Blacksburg County warning area on this day as it moved into southern VA, while another supercell immediately to it's south followed a track just to the south of the first storm but did not produce any tornadoes.
Figure 12. KFCX WSR-88D reflectivity image at 2027 UTC, March 20, 1998, near the time the Stoneville tornado touched down near Mayodan.
Click the link below (or here) labeled "Larger View Radar Reflectivity" for an animation showing how several storms tracked through the area during the afternoon. Note the two strong storms that pass just to the west of Reidsville NC. The first and northern-most one is the tornadic supercell that produced the F3 tornado in Stoneville, and the one right behind it did not produce any tornadoes (only a few hail reports). Other supercells to the south of the first two (moving east and south of Greensboro) produced mainly large hail during the time of the loop, but later in the evening and overnight more tornadoes formed from supercells in the Granville County area as well as around Raleigh (see Figure 2 again), but none were as strong or as long-track as the tornadoes from the Stoneville supercell storm.
The second link below (or here) labeled "Zoomed Radar Reflectivity" shows a close-up view of the two storms that tracked from near Winston-Salem, across Rockingham County NC, and then just west of Danville VA. Note the similarity in shape, and sharp gradient in the reflectivity values on the southeastern flanks, including the occasional appearance of a notch in the reflectivity on the southeast side. These two long-track supercells do not necessarily have the classic "hook echo" shape that some larger supercells in the Midwest and Plains often do, but the sharp gradient and notch still indicate a very intense updraft and a storm structure suggestive of the same organization and danger as more "textbook" supercells. The location of the notch and updraft are coincident with where the rotation in the storms can be expected, and it is in this southeastern flank of the supercell (assuming it is traveling in a typical direction from southwest to northeast) that tornadoes are most likely to form. Finally, it is interesting to note that the second supercell appears to be more intense in terms of the reflectivity values in the storm, and this storm did produce more hail then the tornadic storm (including a couple of golf ball size hail reports). A couple of theories will be presented in the summary as to why the second storm may have had a stronger updraft yet did not produce any tornadoes.
The third link below (or here) labeled "Zoomed Radar Velocity" shows the Doppler wind velocities associated with these two supercell storms, at least in the lower levels of the storm. [Note: NWS forecasters have the ability (then and now) to view multiple layers of the storm in order to examine three-dimensional structure to the reflectivity and wind patterns within the storm. For simplicity we are just showing the lowest slice of the radar data here in this write-up, but we always consider the full structure of these storms for important features and evolutions in the upper levels of the storms that may influence what happens next closer to the ground. We also now have more efficient technology for quickly viewing all slices through the storm and comparing reflectivity and velocity images in the same window. Resolution of these data have also improved since 1998.] Interpreting the velocity images in this animation requires knowledge of where the radar is located, which is near the upper-left corner of the image. Cool colors (greens and blues) represent winds blowing generally toward the radar, while warm colors (oranges, yellows, reds) represent winds blowing away from the radar. Notice in the vicinity of both storms (you may have to refer back to the reflectivity animation) that there are small areas where cool and warm colors are nearly side-by-side when considering the radar's line of sight. This indicates an area of rotation. These signatures are known as "mesocyclones" (which are rotating updrafts), and are not actually resolving the tornadic scale circulation, but rather the slower rotation of the entire storm. Within these broader circulations, under the right circumstances, a tighter circulation can form and a tornado can reach the ground. If these mesocyclone signatures persist long enough, grow intense enough, and deepen upward or downward with height, or if we believe the environment around the storm is prime for tornado formation, then a decision will be made to issue a tornado warning for the imminent threat of a tornado. In addition, if trained spotters or law enforcement spot a lowering funnel cloud or tornado already on the ground, the warning is likely to be issued as well.
In this case, the evolution of the circulation strengthened to a point where forecasters thought a tornado warning was prudent about the time the storm crossed into Rockingham County. We discovered later that this was actually a couple of minutes before the first brief tornado touched down near Pine Hall. However, the warning was issued within a minute or two of the second tornado touching down near Mayodan, and over 10 minutes before the devastation in Stoneville. Tornado warnings continued with this storm into southern VA. The circulation associated with the second storm just to the south of the tornadic storm at times was nearly as strong as the first one, but overall was a little less. Figure 13 shows a comparison of the low-level rotational strength of the two mesocyclones; the red line represents the northern storm's strength, while the yellow line represents the southern storm's strength. Nevertheless the characteristics of this second circulation at multiple layers, the overall structure and strength of this storm, and the environment around it prompted the issuance of a second tornado warning for it as well.
Larger View Radar Reflectivity
Zoomed Radar Reflectivity
Zoomed Radar Velocity
Figure 13. Chart showing strength of rotational signature of the two mesocyclones in terms of Doppler velocities in the lowest layer of the storm (about 5,000 ft above the ground). The red line represents the northern-most tornadic mesocyclone, and the yellow line represents the southern non-tornadic mesocyclone. The times of the two longer track tornadoes are indicated by black lines at the top.
The F3 tornado that devastated Stoneville NC and surrounding areas on March 20, 1998, was a very rare event for the region yet climatology shows it was not the only one of that intensity in the northwest NC Piedmont area. Supercell thunderstorms and sometimes relatively weak tornadoes might occur in this region once every year or two, but the kind that track for several miles producing extensive damage and an extreme threat to human life might occur only once a generation in this general area. As discussed above, the specific factors that must be present to produce a tornado can be subtle to detect, and really don't all come together very often at all. Even 10 years later, we still aren't certain why the northern supercell produced such an intense tornado while the southern supercell (which appeared to be stronger at times), only produced large hail. The exact position of the surface warm frontal boundary likely played the most important role, with the first storm getting close enough to feed off of pre-existing wind shear along the boundary as well as the mix of cool, moist air. The southern storm might not have ever encountered this boundary as the boundary if it began to lift north in that area ahead of the storm. Smaller scale outflow boundaries, not detectable by the radar at that range or by surface observations in the area, may have also helped to generate the tornado with the northern storm, or created a locally stable environment for the second storm, but we think this is not as likely.
The NWS uses all the information we have available to make the tough decision to warn or not to warn for a tornado, hopefully before it has actually touched down. However, given that much of the tornado formation process is still not well understood, these decisions will not always be perfect. We are likely to err on the side of caution, and so some tornado warnings may not verify with an actual tornado (although the storm is still likely to have dangerous wind or large hail with it). Some tornadoes (usually weaker ones) may have very subtle signatures with them and make touch down briefly without a prior warning.
The unusual circumstances that came together on March 20, 1998, and the rarity of such an intense tornado in that area doesn't mean, however, that folks should not be prepared for this to happen again at any time, especially during the spring months when the ingredients that produced this significant event are more likely to be present. People should stay alert to weather forecasts and outlooks for potential hazardous weather, and if tornado watches are eventually issued make sure you know what you will do if a tornado were to approach, whether you are in your home, your office, school, or in you vehicle. Keep a television or radio on if a watch has been issued, frequent the our web site, keep an eye to the sky, and if you have a NOAA All Hazards Weather Radio, that might be the best source for updates and possible warnings. If a tornado warning is then issued for an imminent threat in your specific location, you will then be ready to take immediate action to protect yourself and those around you from such a dangerous situation.