The August 23, 1998 Severe Weather Outbreak Across Central and Northeast Wisconsin
Jim Skowronski
National Weather Service
Green Bay, WI
Part I - Overview
Three supercell thunderstorms affected the NWS Green Bay county warning area during the late afternoon and early evening of Sunday, August 23, 1998. The first two storms developed quickly over central Wisconsin and then tracked east-southeast through the Fox Cities. These two storms primarily produced large hail (1 to 2 inches in diameter) and some wind damage. A brief tornado touched down with the first storm near Menasha (Winnebago county).
The third significant storm developed farther north. This storm evolved more slowly, but went through the classic life-cycle of a tornadic supercell and spawned the Door county tornado. The tornado began as a waterspout over the Bay of Green Bay, about 5 to 10 miles from the Door county shoreline. The tornado moved ashore and remained on the ground for about another 5 miles (Figure 1). The multiple vortex tornado was over a half mile wide at times. Most of the damage over Door county rated F2 on the Fujita scale, though some F3 damage was identified.
A. Synoptic Overview
The synoptic scale regime in which these storms developed featured a fairly zonal flow across the country, with relatively fast westerly winds for late August. Jet streaks of up to 100 knots (at 300 mb) were propagating across the northern United States (Figure 2). As is common with zonal flow regimes, weak short-wave troughs were moving quickly eastward through the flow, and the numerical models were doing a poor job of handling the individual features.
B. Meso-Scale Overview
Some precipitation occurred across the area during the previous night as a warm front pushed north across the region. By late morning, the surface pressure pattern suggested that the front was well north of the area. However, due to the weak nature of the low-level flow it is possible that the front simply weakened to the point where it was no longer discernable. In any case, a moist and rather unstable airmass was found across the state. Surface dew points were above 65 F statewide, winds were fairly uniform from the west-southwest, and speeds were quite light--generally 10 knots or less. This environment made it very difficult to determine where (and even if) severe weather would occur. The Storm Prediction Center's initial Day 1 severe weather outlook placed all but extreme northwest Wisconsin under a moderate risk of severe weather. Perhaps underscoring the difficulty in assessing the severe threat that day, the risk was downgraded to a slight risk at 1100 UTC despite the fact that no new model data or upper air data were received between the two issuances.
Part II - The Severe Weather Event
By late morning, water vapor imagery showed a well-defined short-wave trough pushing east across central Upper Michigan. An area of rain and thunderstorms had developed over Upper Michigan and northern Wisconsin in response to this feature, including one severe storm that caused a fatality in Upper Michigan when a tree fell on a camper. Although it had yet to become evident on the 1600 UTC surface analysis, a boundary was beginning to develop across northern Wisconsin. It is possible that the boundary was the remnants of the warm front that pushed into northern Wisconsin the day before, or it may have simply been generated by the thunderstorm complex across Michigan. In any case, the boundary was becoming evident on 1 km visible satellite imagery (Figure 3), and it was strengthening as the rain-cooled air associated with the Michigan storms moved south. The boundary extended from near Rice Lake toward Marquette. The airmass south of this boundary was beginning to destabilize as cloud debris from the overnight rain was rapidly dissipating. Farther west, somewhat drier air was working eastward across Minnesota. The leading edge of the drier air was ill-defined, and did not appear to play a role in the development of the convection later in the day across Wisconsin.
By 1800 UTC, the short-wave trough and thunderstorm complex had moved farther east, but the boundary over northern Wisconsin continued to strengthen and had become much better defined in surface data and on satellite imagery. Whatever the origins of this boundary, it now appeared in both satellite and surface data (Figure 4) to be an outflow boundary separating rain-cooled air to the north from the warm and increasingly unstable air to the south. The boundary extended from near Rhinelander to central Marinette county to the northern bay of Green Bay. In addition, another reinforcing outflow boundary had developed and was moving east across Upper Michigan. A thunderstorm developed at the intersection of these two boundaries and moved east just south of Iron Mountain, Michigan. This appeared to drive the original boundary farther south, and it became visible on the KGRB WSR-88D base reflectivity product.
Meanwhile, the atmosphere over central Wisconsin continued to destabilized significantly with CAPES approaching 4300 j/kg and surface based lifted indicies to -10. A weak pressure trough was apparent on the 1800 UTC surface analysis and a developing cumulus field could be seen on the visible satellite imagery to be concentrated in this area. Temperatures in this area had risen into the lower to middle 80s and dew points had risen into the middle 70s. Modifying the GRB sounding from later in the day would show that these values would just about remove the cap and allow convection to fire.
By 1845 UTC (Figure 5) some of the cumulus in central Wisconsin were showing considerable vertical development, and they first became visible on the KGRB WSR-88D at 1918 UTC (Figure 6). The high CAPE environment allowed the developing thunderstorms to grow rapidly, and they quickly acquired supercell characteristics. Although the GRB sounding from later that afternoon would show a classic straight-line hodograph (Figure 7), the storms turned to the right and were moving from about 295 degrees at 32 knots. This motion produced a 0-3 km helicity of 89 m^2/s^2. Given the high degree of instability, this was apparently sufficient to create a rotating updraft. The first severe thunderstorm warning of the afternoon was issued at 1957 UTC for these storms, and the WSR-88D had already detected a mesocyclone with this storm.
As the storms continued to grow, the WSR- 88D continued to show a sharp low-level reflectivity gradient on the inflow side of the storms, and would occasionally show at least an appendage at the back edge of the storms and occasionally even an ill-defined hook. Storm Relative Velocity (SRM) products showed moderate mid-level rotation, but the storms had trouble building this circulation down into the low-levels.
By 2100 UTC, the WSR-88D was showing two well-developed supercells headed for the Fox Cities (Figure 8). The WSR-88D indicated that the first storm had developed a flanking line of cumulus. Also at this time, the WSR-88D began to detect development on the meso-scale boundary over northeast Wisconsin. The development was occurring at the point where Forest, Marinette, and Oconto counties meet, and would eventually become the supercell that spawned the Door county tornado.
There were several differences between the northern (Door county) storm and the two southern storms, the most obvious of which was the presence of the mesoscale boundary. In addition, temperatures and dew points over northeast Wisconsin were a little lower than in central portions of the state. As a result, the modified sounding showed that there was likely still a cap that needed to be overcome before the storm could develop. It is likely that mesoscale convergence along the boundary was needed to break the cap. This probably limited the number of storms that could develop, but allowed the tornadic supercell that did develop time to go completely through its life cycle before its inflow got cut off by surrounding storms. This fact is important because the northern storm developed much more slowly that the two storms to the south. It maintained an easterly motion for about an hour, and throughout that time VILs remained rather low. Just before 2200 UTC, the storm started to intensify rapidly, cell based VILs suddenly exceeded those of the southern storms, and the storm made an abrupt turn to the right (Figure 9). A severe thunderstorm warning was issued for Marinette county at 2215 UTC when it became apparent that the storm would not push east into Upper Michigan as its previous track suggested.
By 2230 UTC, the southern storms had begun to weaken and were no longer a major severe weather threat. The northern storm, however, continued to intensify rapidly as it turned southeast along the Wisconsin/Michigan border. The storm developed a well defined "V-notch" on the east side of the storm, and a very tight reflectivity gradient on the inflow side. Much like the southern storms, the storm had acquired moderate mid-level rotation, but the rotation was slow to build downward. The storm motion was a bit slower than that of the southern storms, which yielded a slightly higher 0-3km storm relative helicity of 107 m^2/s^2.
By 2245 UTC, a well defined hook had developed in the southwest part of the storm (Figure 10), and reflectivity cross sections of the storm showed a very well defined Bounded Weak Echo Region (BWER). By 2300 UTC, WSR-88D storm relative velocity information began to show the circulation building down through the storm. In addition, the storm had turned even more to the right and was now moving from about 305 degrees at 25 knots. This motion produced a 0-3 km helicity of 135 m^2/s^2.
Although a report of a funnel cloud was received from around Menominee, Michigan before the storm moved off shore, the first confirmed tornado touchdown (as a waterspout on Green Bay) occurred around 2310 UTC. The lowest elevation WSR-88D velocity pattern was still very convergent at this time (Figure 11), but the circulation above was quite intense. The tornado likely remained on the bay for nearly 20 minutes (although there are no eye-witnesses that it was on the water for the entire time after it formed) and then moved on shore at 2330 UTC. At this time, the WSR-88D indicated very strong cyclonic rotation from 0.5 degrees up through at least 3.4 degrees. The strongest low-level rotation occurred during the next 2 volume scans, at the time the widespread F2 and occasional F3 damage occurred. The .27 nm base velocity product detected a gate-to-gate shear in excess of 63 knots (Figure 12). It is interesting to note that although a major tornado was on the ground and the low-level rotation was intense, the rotation above 1.5 degrees was rapidly weakening. The tornado probably dissipated around 2344 UTC, after moving about two-thirds of the way across Door county. Only a much weaker circulation remained by the time the storm crossed the shoreline and moved over Lake Michigan.
Overall, the tornadic supercell which produced the Door county tornado underwent a classic evolution process. The storm initially developed along a pre-existing boundary, in an environment where some forcing was necessary to break a weak cap. It developed rather slowly at first as it moved with the mean wind flow. The storm eventually turned to the right of the mean wind flow and began to intensify more rapidly. Shortly thereafter, the storm developed a sharp V-notch along the leading edge, and a very tight reflectivity gradient along the inflow region. A persistent mid-level circulation was noted, but this persisted for some time before the circulation built down into the low-levels of the storm. The storm developed a classic hook-echo pattern in the low-level reflectivity pattern and a well defined BWER. Large hail occurred at this time. By the time the tornado first touched down, the circulation had built down to low levels, but was strongly convergent. The tornado remained on the ground while the low-level circulation intensified and the mid-level circulation started to weaken. Finally, the low level circulation rapidly weakened as the tornado dissipated. What made this storm so unusual was that this type of storm is rarely seen in Wisconsin.
Conversion to HTML and image editing by Jeff Last
