Figure 1. Severe Weather Reports from 1200 - 2300 UTC 2 July 1997.
This case illustrates the use of real time vertical wind data from WSR-88D VAD wind profiles and special rawinsondes to update earlier model output and standard soundings in order to improve the prediction of convective storm mode.
Introduction:
Severe storm outbreaks are often associated with well-defined baroclinic disturbances that are typically resolved by the standard observing network. Forecasters have been successful in predicting these severe weather episodes, called "synoptically evident events" (Doswell et al. 1993). These outbreaks sometimes produce primarily supercells and tornadoes, whereas other outbreaks may consist mostly of bow echoes and damaging wind gusts (e.g., Johns and Hart 1993). Anticipating the most likely mode of severe convection remains a challenging task for the forecaster, since proper prediction is dependent on the accurate assessment of the four-dimensional evolution of vertical wind shear and instability. This in itself can be a complex process, since the environmental shear and instability can exhibit substantial mesoscale variability (Brooks et al. 1996).
A number of physically-based parameters are available to aid in the determination of a preferred mode of convection (e.g., supercell, multicell, linear), including the Bulk Richardson Number (BRN) (Weisman and Klemp 1984) and Storm-Relative Environmental Helicity (SREH) (Davies-Jones et al. 1990). Furthermore, recent studies have focused on ways to discriminate between supercells that produce tornadoes and those that do not (Brooks et al. 1994, Stensrud et al. 1997, Thompson 1998). In these studies, the character of the mid-level storm relative wind has been found to play an important role in the formation of low-level mesocyclones. However, operational application of these concepts is not always clear-cut from the forecaster perspective. First, forecasters must initially determine where and when convection may develop, and convective initiation itself remains a difficult scientific problem (e.g., Crook 1996) that is not well handled by operational numerical models (Weiss and Jungbluth 1994). Second, it may be difficult to identify a dominant mode of convection since that can vary at different times during a convective outbreak, or multiple modes may be present simultaneously during the event. This makes the a priori assessment of a preferred mode most difficult. Finally, the high resolution data needed to routinely evaluate the vertical wind structure has traditionally not been available to forecasters on a day-to-day basis. Now, use of wind profilers and VWP (Velocity Azimuth Display Wind Profile) data provides an opportunity to observe the local vertical wind profiles in real time that can be critical in anticipating the most likely mode(s) of convection.
Description of case:
During the afternoon and evening of 2 July 1997, a series of severe thunderstorms swept across lower Michigan, Indiana, Ohio, eastern Kentucky, western West Virginia, southwestern Virginia, and northwestern North Carolina (Fig. 1). The storms produced a number of damaging tornadoes up to F3 intensity with additional wind damage caused by bow echoes and downbursts producing winds up to 100 mph across the region.
The outbreak resulted in 7 fatalities and more than 100 injuries over southeast lower Michigan alone, where property damage exceeded 135 million dollars. Because this was one of the most significant convective weather episodes in Michigan history, the case will focus on the lower Michigan part of the outbreak.
An intense upper low located over Minnesota at 1200 UTC on 2 July 1997 (Fig. 2) moved eastward across the upper Great Lakes during the subsequent 24 hours. The system was unusually strong for the time of year with a 75 to 80 kt jet streak at 500 mb over the southern quadrant of the circulation. A deep surface low pressure center (996 mb) nearly coincident with the upper low was also moving eastward (Fig. 3). A strong cold front trailed south from the low; the cold front surged eastward across the Lakes in association with the jet streak aloft, while a warm front lifted northward into lower Michigan. The warm sector air mass was very moist in the low levels with surface dew points of 70 to 75 deg F. Dry air aloft and steep low/mid-level lapse rates (~7.5C/km) were evident in the 1200 UTC soundings at Davenport, IA (DVN) and Lincoln, IL (ILX) (Fig. 4). The steep lapse rates and moist axis were forecast by the 0000 UTC 2 July ETA model run to move into Indiana and lower Michigan during the afternoon, and into western Ohio by evening.
The very unstable air mass present over the mid-Mississippi Valley region did not initially extend northward into lower Michigan. Observed 1200 UTC SHARP soundings at Gaylord, MI (APX) (Fig. 5) and Detroit, MI (DTX) (Fig. 6) revealed that the air mass over lower Michigan was not especially favorable for deep convection. CAPE computed by lifting the most unstable parcel was approximately 700 J/kg, and wind profiles were clearly unfavorable for supercell formation since SREH was less than 100 m²/s² (Fig. 7). (Note: SREH based on assumed storm motion of 20 degrees to the right of and 85% of mean wind - 20R85)
Forecast Evolution:
Model forecast fields from the 0000 UTC 2 July 1997 ETA model indicated that the environment over lower Michigan would become increasingly favorable for severe convection during the afternoon. During this period difluent mid/upper-level flow just ahead of the jet streak was forecast to move across Lake Michigan while the low level moisture and instability axis (model CAPE of 3000-5000 J/kg) was expected to spread into Indiana and lower Michigan. Strong mid-level drying in association with the jet streak was forecast (and observed in water vapor satellite imagery) to spread across lower Michigan and much of the upper Ohio Valley region. A composite chart of representative low and mid-level forecast wind maxima valid at 0000 UTC 3 July 1997 is shown in Fig. 8. The synoptic pattern on the forecast chart is similar to the dynamic outbreak pattern conceptual model of Johns (1993). In fact, Fig. 8 most resembles the pattern conducive for the development of squall lines with bow echo-induced damaging winds, since the low-level jet is more parallel to the mid-level jet axis compared to the "classic" tornado outbreak pattern (e.g., Newton 1967). The possibility of bow echoes is enhanced by the forecast of extreme instability, which is common with warm season derecho events (Johns and Hirt 1987).
Click for evolution of ETA 0000 UTC 2 July 1997 forecast of 500 mb heights/vorticity (85 KB)
smaller version (29 KB)
Click for evolution of ETA 0000 UTC 2 July forecast of mean sea-level pressure (MSLP) (58 KB)
smaller version (18 KB)
An additional aspect of this case involves the presence of Convective INhibition (CIN) and its impact on convective development. Warm temperatures above the boundary layer over Iowa limited the southern extent of thunderstorms the previous day, and the ETA model forecast indicated that CIN could possibly preclude thunderstorm formation across Indiana and Ohio during the afternoon, with model generated convective precipitation developing only over the northern tip of lower Michigan by 0000 UTC 3 July 1997.(Figure 9a)
Although pattern recognition techniques can be very useful in forecasting severe thunderstorms, they cannot be used exclusively to infer the expected storm type since sub-synoptic conditions and processes favorable for mesocyclones may also be present within the dynamic outbreak pattern. As noted by Johns (1987, 1993) and Przybylinski (1995), some dynamic pattern episodes produce both derechos and tornadoes, with tornadoes often occurring in the northern portion of the bow echo (associated with the cyclonically rotating comma head). Thus, it is necessary to examine specific parameters from observational data and model forecasts that can be used to determine possible modes of convection. ETA model forecasts from the 0000 UTC (2 July 1997) run are used since operational decisions related to late morning zone forecast updates, NOWcasts for initial thunderstorm development, initial coordination with emergency preparedness officials, and local office staffing requirements would need to be made before output from the 1200 UTC (2 July 1997) ETA was available for guidance.
Specific forecast parameters related to convective storm type, and grid point soundings from the 0000 UTC ETA model were examined. Forecasts of BRN showed values larger than 50 moving into most of lower Michigan and Indiana by 1800 UTC and remaining over the warm sector into the evening hours, suggesting that multicell storms were most likely to occur. However, for very large values of CAPE (> 4000 J/kg), the BRN is dominated by the CAPE such that BRN is large regardless of the values of BRN shear (Stensrud et al. 1997). Because of this, it is especially useful to also examine SREH, although this requires a "forecast" of expected storm motion. ETA model SREH values were predicted to increase to 150-200 m²/s² over portions of lower MI by 0000 UTC 3 July 1997; indicating the possibility of mid-level mesocyclone formation (Figs. 10a, 10b, and 10c)
It is important to remember that the ETA model SREH computation is based upon an assumed
storm motion that is to the right of and slower than the deep layer (850-300 mb) mean wind. [For
mean wind speeds >30 kts, the assumed deviant motion is 20 degrees to the right of and 85% of
the mean wind (20R85)]. For storm motion vectors that are closer to the hodograph, the SREH
would diminish to less than 100 m²/s², implying that mesocyclone development was unlikely.
For a wide range of reasonable storm motions, low-level storm relative winds were forecast to
strengthen to more than 20 kts over lower Michigan during the afternoon, a necessary (but not
sufficient) factor when considering mesocyclone development. Thus, SREH concepts applied to
model data suggested a possibility for supercell formation, but only for cells moving substantially
to the right of the mean wind.
To explore the likelihood of tornadic supercells to develop, mid-level storm relative winds are examined. Recently, Thompson (1998) applied basic findings from cloud modeling work by Brooks et al. (1994) that indicated a balance between the mid-level storm relative winds and thunderstorm outflow is needed in order to enhance persistent low-level mesocyclone formation. Thompson found that supercells were more likely to produce tornadoes when the mid-level storm relative winds were >15 kts. ETA forecasts based on an assumed 20R85 storm motion resulted in mid-level storm relative winds around the threshold value of 15 kts, however storm motions closer to the hodograph produced mid-level storm relative winds unfavorable for sustained tornado production. As a further test of tornadic supercell potential, values of BRN shear were computed. Stensrud et al. (1997) advocate the use of BRN shear as a surrogate for storm relative mid-level flow, citing the advantage that BRN shear is independent of storm motion. Utilizing MM4 mesoscale model output in their study of selected severe weather cases, they found that BRN shear of 40-100 m²/s² indicated a likelihood of tornadoes for supercell storms, whereas values less than 40 m²/s² were associated with storms dominated by outflow (e.g., bow echoes). For this case, ETA model forecasts of BRN shear tended to be in the favorable tornado range, suggesting an increased tornado potential for any supercells that might develop. Note, however, that the favored range of values found by Stensrud et al. is based on computations from the MM4, and they caution that these preferred values may not directly apply to observed soundings or wind fields from other numerical models.
Finally, ETA grid point wind forecasts (Fig. 11) through 1800 UTC suggested very little directional shear would be present above 700 mb over lower Michigan. Although a small amount of clockwise turning was forecast in the lowest 3 kilometers after 1800 UTC, the shape of these forecast hodographs (not shown) was similar to those shown by Johns and Hart (1993), which were associated with widespread bow echo episodes. The combination of large instability and moderate-to-strong low-level shear is also consistent with the findings of Weisman (1993a, 1993b), who found these conditions associated with the production of numerically simulated long-lived severe wind events. (Note: relatively long-lived bow echoes can also occur in areas of modest instability, especially during the cooler months when strong dynamic forcing is present)
Loop of 0000 UTC 2 July 1997 ETA forecast sounding for DTX at 00 hr, 06 hr, 12 hr, 18 hr, and 24 hr (valid at 0000 UTC 3 July).
The above loop shows a combination of steep lapse rates (leading to strong instability) and moderate wind shear (especially below 3 km) forecast between 1800 - 0000 UTC.
In summary, application of pattern recognition techniques and physically-based parameters related to convective storm type using ETA model forecast data suggested potential for bow echo formation producing widespread wind damage. This appeared to be the most likely convective evolution, especially for storms moving with the mean wind. However, for right-moving storms the potential for tornadic supercell development could be expected to increase, especially for cell movements well off the hodograph.
Observed wind profiles and continuous assessment of storm type potential:
Vertical wind profiles from area WSR-88D radars at KGRR and KDTX and an 1800 UTC special sounding at APX are used to update the shear profiles as the day progresses. In addition, surface data should be utilized to determine subtle, but key, changes in the low-level wind fields (and instability) during the intervals between model forecast times. Since the environment was forecast by the ETA to become substantially more favorable for intense convection than was indicated by the 1200 UTC soundings, it is critical to monitor real-time information in order to determine changes in convective potential, and to assess the accuracy of the model forecast trends. Visual inspection of VWP time sections from KGRR and KDTX (Fig.12) shows the low-level winds strengthened and displayed considerable veering with height below 850 mb prior to 1800 UTC as the upper low began to move across the western Great Lakes. In addition, animation of the regional surface plot from 1200 - 2300 UTC shows the winds strengthening with time (as well as backing a little) and surface dew points increasing substantially over portions of lower Michigan as the intense surface low tracked eastward across the Great Lakes.
Construction of hodographs based on VWP data can be accomplished using the SHARP workstation, including the computation of kinematic parameters. Figure 13 shows a modified hodograph from DTX at 1800 UTC based on VWP data through 3 km. Note the SREH is now approximately 400 m²/s² (based on an assumed storm motion of 270 deg/23 kts) and the mean low-level storm relative (inflow) vector from 0-2 km is from 150 deg at 25 kts. A time series of SREH at DTX (Fig. 14) shows a significant increase in mesocyclone potential during the afternoon hours, especially for right-moving storms. By 1700 UTC, SREH at DTX had increased from less than 100 m²/s² to around 250 m²/s² and continued into the 350-400 m²/s² range prior to the storms moving through southeast lower Michigan.
The strengthening of the wind fields and subsequent increase in SREH was also evident at APX, with SREH increasing from less than 100 m²/s² to more than 300 m²/s² from 1200 to 1500 UTC. The 1800 UTC special sounding from APX (Fig. 15) confirmed much stronger low-level winds had moved into lower Michigan during the morning with 50 kt flow above 700 mb. In fact, the observed wind profiles were stronger than predicted by the ETA model winds, resulting in much larger SREH values than the 0000 2 July 1997 UTC ETA output had indicated.
The real-time wind data showed that supercell potential was increasing not only for right-moving storms but for storms moving along the mean wind direction with 1800 UTC SREH values at APX of 354 m²/s² (right-moving) and 250 m²/s² (no deviation) (Figure 16). Although the BRN (87) and BRN shear (28 m²/s²) at APX appear to favor multicellular convection and outflow dominated storms, the notable increase in SREH and the continued favorable storm-relative low-level flow also means that forecasters should be alert for supercell formation, especially for any persistent right-moving storms (which there were).
Conclusions:
Severe thunderstorms moved across lower Michigan during the afternoon of 2 July 1997 producing widespread wind damage, 7 fatalities, and more than 100 injuries. The storms exhibited bow echo and supercell characteristics ( see radar loops below) , which complicate the warning decision-making process. Prior anticipation of the likely mode(s) of convection can improve the warning-decision system, since forecasters are more likely to correctly recognize and assess the severe potential of storms when they understand the mesoscale environment and its influence on storm type and evolution. Pattern recognition can be helpful as a first start in shaping one's general forecast of convective mode, areal extent and timing. Model forecast data are typically used to provide this initial assessment of the convective potential prior to thunderstorm development. However, operational models cannot be expected to routinely resolve mesoscale features that impact thermodynamic instability and the vertical wind structure. When possible, it is important to update model forecasts by incorporating real-time observational data, such as VWP data and special soundings, into the forecast methodology. As illustrated by this case, these types of real-time information can be used to more confidently determine the likely modes of convective storms, especially during rapidly changing dynamic situations when the latest soundings are not representative of conditions that develop just a few hours later. Furthermore, it is important to remember to focus on the real atmosphere during short-term thunderstorm forecasting, including the need to synthesize information from a variety of observational sources (such as surface data, radiosonde, radar, satellite, lightning network, wind profilers) in order to develop a four-dimensional picture of the atmosphere. Model data is often helpful in formulating an expected picture, but that image needs to be updated by observational data whenever possible.
Radar Data:
Radar data is intended to show the general evolution of the convective storm event during the afternoon of 2 July 1997 in lower Michigan.
Note: Download of radar imagery (due to the large file size) may take a considerable amount of time for slower modems.
Click here for 14 frame animation of KGRR Base Reflectivity product (246 KB) (0.5 deg. ) from 1800 UTC to 1915 UTC. Note the transition from initial multicell to supercell structures along the convective line in lower Michigan toward the end of the loop.
Click here for 36 frame animation of KDTX Base Reflectivity product (808 KB) (0.5 deg. ) from 1848 UTC to 2222 UTC. Note the evolution of isolated supercell to bow echo (with embedded supercell) structures along the convective line in southeastern lower Michigan toward the end of the loop.
Satellite Data:
Satellite data is presented to show the general evolution of the convective storm event during the afternoon of 2 July 1997 in lower Michigan.
Note: Download of satellite imagery (due to the large file size) may take a considerable amount of time for slower modems.
Click here for animation of 4 km water vapor satellite imagery (751 KB) from 1945 - 2345 UTC 2 July 1997. Examination of the water vapor satellite imagery showed very dry air (reddish orange enhancement) in the mid and upper-levels of the atmosphere across much of lower Michigan and the upper Ohio Valley during this severe weather episode. The distinction between this drier mid and upper level air and the higher level moisture circulating around the upper low over western Lake Superior appears to delineate the northern extent of explosive convective development over lower Michigan, eastern Indiana and Ohio.
Click here for animation of 2 km visible satellite imagery (2.7 MB) from 1545 - 2245 UTC 2 July 1997.
Storm Summary Information:
For additional information on this case including a complete summary and damage survey of all the severe weather in the DTX county warning area for this day, go to this NWS Central Region DTX homepage (http://www.crh.noaa.gov/dtx/july2.htm).
References:
Brooks, H.E., C.A. Doswell III, M.T. Carr, and J.E. Ruthford, 1996: Preliminary analysis of
soundings from VORTEX-95. Preprints, 18th Conf. On Severe Local Storms, San Francisco,
CA, Amer. Meteor. Soc., 133-136.
Brooks, H.E., C.A. Doswell III, and J. Cooper, 1994: On the environments of tornadic and
nontornadic mesocyclones. Wea. Forecasting, 9, 606-618.
Crook, N.A., 1996: Sensitivity of moist convection forced by boundary layer processes to low-level thermodynamic fields. Mon. Wea. Rev., 124, 1767-1785.
Davies-Jones, R.P., D. Burgess, and M. Foster, 1990: Test of helicity as a tornado forecast
parameter. Preprints, 16th Conf. On Severe Local Storms, Kananaskis Park, Alta., Amer.
Meteor. Soc., 588-592.
Doswell, C.A. III, S.J. Weiss, and R.H. Johns, 1993: Tornado Forecasting: A Review. In The
Tornado: Its Structure, Dynamics, Prediction, and Hazards, C. Church, D. Burgess, C. Doswell,
and R. Davies-Jones, Eds.. Geophysical Monograph 79, Amer. Geophysical Union, Washington,
DC. 557-572.
Johns, R.H., and J.A. Hart, 1993: Differentiating between different types of severe thunderstorm
outbreak: A preliminary investigation. Preprints, 17th Conf. On Severe Local Storms, St. Louis,
MO, Amer. Meteor. Soc., 46-50.
Johns, R.H. and W.D. Hirt, 1987: Derechos: widespread convectively induced windstorms. Wea.
Forecasting, 2, 32-49.
Johns, R.H., 1993: Meteorological conditions associated with bow echo development in
convective storms. Wea. Forecasting, 8, 294-299.
Newton, C.W., 1967: Severe Convective Storms. In Advances in Geophysics, Vol. 12,
Academic Press, 257-303.
Przybylinski, R.W., 1995: The bow echo: observations, numerical simulations, and severe
weather detection methods. Wea. Forecasting, 10, 203-218.
Stensrud, D.J., J.V. Cortinas, Jr., and H.E. Brooks, 1997: Discriminating between tornadic and
nontornadic thunderstorms using mesoscale model output. Wea. Forecasting, 12, 613-632.
Thompson, R.L., 1998: ETA model storm-relative winds associated with tornadic and
nontornadic supercells. Accepted for publication in March 1998 Wea. Forecasting.
Weisman, M.L., and J.B. Klemp, 1984: The structure and classification of numerically simulated
convective storms in directionally varying wind shear. Mon. Wea. Rev., 112, 2479-2498.
Weisman, M.L., 1993a: The genesis of severe, long-lived bow echoes. J. Atmos. Sci., 50, 645-670.
Weisman, M.L., 1993b: The influence of the coriolis force and directional vertical wind shear on
the evolution of simulated long-lived bow echoes. Preprints, 17th Conf. On Severe Local
Storms, St. Louis, MO, Amer. Meteor. Soc., 562-566.
Weiss, S.J, and K.A. Jungbluth, 1994: Evaluation of ETA model guidance for the prediction of
severe local storms. Preprints, 10th Conf. On Numerical Wea. Prediction, Portland, OR, Amer.
Meteor. Soc., 383-385.
Acknowledgments:
The producer of this instructional component would like to thank Steve Weiss (SPC) for content collaboration and analysis of data. In addition, thanks go to Dick Wagenmaker (DTX), Bruce Smith (APX), and Gary Gavet (GRR) for supplying much needed data. Thanks also to the WSR-88D OSF/OTB for graphics support and editing reviews.