The Greensburg Tornado Environment  on 4 May 2007  (a forecast and environment analysis by Jon Davies) 
  (c) Dick McGowan & Darin Brunin

The Greensburg tornado was the first tornado rated EF-5 using the new Enhanced Fujita scale, and the most intense tornado in the U.S. since 1999.   A deadly and tragic event that involved several nighttime tornadoes as wide as 2 miles at times, the setting certainly merits examination from a forecast and environment analysis standpoint.

Morning forecasts showed a broad trough over the western U.S., with southwesterly midlevel winds of 45-50 kts punching into southwest Kansas (see NAM/WRF forecast below), and divergence in the 500 mb wind field coming out into the central plains on the evening of 4 May:

The NAM/WRF forecast of CAPE for the evening of 4 May was impressive (3500-4500 J/kg) over western Oklahoma pointing into southwest Kansas ahead of a surface dryline (see below).  In addition, the NAM/WRF relative humidity forecast at 700 mb hinted at convective initiation (see arrow on 2nd map below) over the eastern Oklahoma panhandle area around 7 p.m. CDT at the edge of the CAPE axis, with temperatures of 8 C at 700 mb roughly approximating the edge of a "cap" in lower midlevels.
 

Precipitation forecasts (for 3-hr or 6-hr periods, see below) from both the RUC and the NAM/WRF also suggested that storms would develop through the evening in the TX/OK panhandle or southwest KS area along the dryline (see arrows below):
   
While great care is advised when reading and interpreting model precipitation forecasts (for example, the RUC is usually less reliable and often breaks out too much precipitation in areas that are capped), note that both the RUC and the NAM/WRF models agreed in general that storms would develop at the west edge of the dryline in the eastern OK panhandle/southwest KS area (see first 2 maps above). Moreover, the NAM/WRF precipitation forecast valid during the 6 hrs between 7 p.m. and 1 a.m. CDT (3rd map above) suggested that convection would move northeastward through western sections of south-central Kansas into the nighttime hours, and that areas further south would remain largely capped.

With convection looking like a reasonable bet along the dryline in southwest Kansas during the evening, the forecast environment could next be assessed in the area where storms were expected.  With winds forecast to approach 50 kts at 500 mb (see the NAM/WRF 500 mb forecast earlier), plentiful deep-layer shear (0-6 km shear values of 40-50 kts, not shown) would support supercells in southwest KS .  Low-level shear and storm-relative helicity (SRH) in the lowest 1 km (close to the ground) was also forecast by the NAM/WRF to be large in a SE to NW axis, maximizing near a warm front in central KS (see circled area on 1st map below). Combining this SRH with CAPE (shown earlier) via the 0-1 km energy-helicity index (EHI), an area of optimized SRH and CAPE was forecast over central and south-central KS (see circled area on 2nd map below) at early evening, suggesting potential for low-level mesocyclones and possible tornadoes with storms that formed near the dryline that could ingest this air mass from the east. 
     
This NAM/WRF forecast was similar to RUC forecasts of SRH and EHI (not shown), and quite impressive given that, from experience, the NAM/WRF often under-forecasts SRH and SRH-CAPE combinations. Furthermore, SRH in the plains ahead of strong midlevel troughs usually increases near and after dark due to diurnal backing of winds east of the dryline. This backing would be expected to increase SRH and EHI values even more through the evening.

After looking at deep-layer shear and SRH-CAPE combinations, an additional assessment factor involves whether supercells will be strongly "surface-based", realizing instability from air parcels originating near the ground (significant tornadic supercells are strongly "surface-based") rather than from air parcels that originate 1-2 km or higher above ground (making for "elevated" supercells, which do not produce significant tornadoes). One useful parameter in this regard is CAPE below 3 km.  Larger amounts of 0-3 km CAPE indicate that a storm is significantly surface-based and not occurring above a relatively stable layer near the ground.

Both the RUC and NAM/WRF forecasts for the evening of 4 May showed significant 0-3 km CAPE (below, in brighter colors) over the area where SRH-CAPE combinations were forecast to be strong (circled below) and where storms were forecast to form as shown earlier.
   
Although 0-3 km CAPE by itself means little regarding tornadoes (notice the broad areas of low-level CAPE depicted on all the maps above), areas that are forecast to be co-located with significant SRH and CAPE combinations (circled above) are important to note.  In addition, notice that the 3rd map above (NAM/WRF forecast valid at 06 UTC or 1 a.m. CDT) continued to show low-level CAPE present over central KS at a very late hour, well after dark when surface temperatures cool and storms are usually more "elevated" rather than "surface-based".   When an area like this is seen co-located with significant SRH and total CAPE at such a late hour, this definitely suggests potential for tornadoes well after dark and into the night, something worth noting in this case.

As those who followed this event know, a supercell did develop in the northeast TX panhandle at late afternoon and moved into northwest OK (not shown), producing a tornado near Woodward. This storm then appeared to get "snuffed out" by the warm temperatures and capping aloft, while new thunderstorm development occurred farther north along the KS/OK border near dark as winds backed diurnally, and the dryline retreated to the west.  Although not able to forecast the specific details, the morning NAM/WRF forecast was correct in its general scenario of showing storms building into southwest KS and western south-central KS during the evening.

A "south-end" storm organized and took hold near Protection KS by 8:30 p.m. CDT (see 1st radar image below).  Unobstructed by other storms to the south and east, this storm ingested the strong shear-CAPE surface-based environment that was forecast by the models (discussed earlier), and rapidly began producing supercell tornadoes.  As everyone knows, a very large and violent tornado struck Greensburg around 9:45 p.m. CDT (see 2nd radar image below), essentially destroying the entire town and killing at least 9 people. Subsequent tornadoes from the same supercell killed 2 more people in rural areas. Excellent warnings from NWS and media certainly kept the death toll from being larger.
 

Surprisingly, the SPC mesoanalysis page (based on RUC data) did not properly reflect the Greensburg tornado environment in real-time.  Look at the two SPC mesoanalysis images at 9 p.m. CDT (02 UTC) below.  The 1st map was inaccurate in showing Greensburg to be located within relatively dry high-LCL air (MLLCL > 2000 m) less than an hour before the tornado, with the mixed-layer moisture axis depicted too far to the east.  The axis of strong significant tornado parameter values (STP > 2.0, effective layer version using mized-layer parcels, 2nd map below) was also depicted too far to the east, with inaccurately large CIN values (darker blue > 100 J/kg of inhibition) indicated over the Greensburg area.
<SPC MLLCL 02 UTC  <SPC STP 02 UTC (effective layer, w/ MLCIN)

The reason for these poor analyses is the performance of the RUC itself, which incorrectly showed dry surface air mixing eastward past Pratt (PTT) and Medicine Lodge (P28), well to the east of Greensburg.  Look at the RUC analysis for P28 at 9 p.m. CDT (02 UTC), which is far too dry in low-levels:

Although the hodograph and wind environment on this estimated sounding was very favorable for tornadoes (0-1 km SRH > 200 m2/s2, sharp clockwise hodograph kink near 1 km, 0-6 km shear > 45 kts), mixed-layer computations from this flawed profile resulted in thermodynamic characteristics that were probably not representative of the true environment.  MLCAPE was depicted as only around 1600 J/kg, CIN was considerable (130 J/kg), LCL heights and resulting cloud bases were depicted too high, and little if any CAPE was shown to be present below 3 km.

Observationally, the early evening surface map (see 1st map below) showed the dryline to be much further west than indicated by the RUC, with observed dew points at PTT and P28 in the mid-upper 60s F.  In sharp contrast to the RUC, the NAM/WRF 3-hr forecast profile at P28 valid at 10 p.m. CDT (03 UTC) showed the deep low-level moisture quite well.  This profile (shown as 2nd image below) had more than twice the total CAPE seen on the RUC, much smaller CIN, and lower LCL/LFC heights.   (Unfortunately, the Dodge City observed sounding at 00 UTC was not useful in this analysis, as it was located just west of the deeper low-level moisture.)
<0143 UTC sfc map 

Based on this information, a correction to the 02 UTC raw RUC profile that deepens the low-level moisture (see below) as suggested by the surface map and NAM/WRF profile above, indicates that the local environment immediately southeast of the Greensburg supercell was much more impressive than shown by the raw RUC analysis and the SPC mesoanalysis:

With this correction, the mixed-layer CAPE more than doubles, jumping from 1600 J/kg to nearly 4000 J/kg, CIN drops from 130 J/kg to around 40 J/kg, and 0-3 km CAPE increases from negligible amounts to more than 100 J/kg!  As a result, 0-1 km EHI also jumps from 2.0 to nearly 5.0, a huge increase, and the environment is shown to be much more surface-based. The dry bias in low-levels with some RUC profiles in cases such as this one is something to be very much aware of, particularly since the RUC model drives the SPC mesoanalysis and has been the basis of several environment studies of supercell tornadoes.

In addition to the above thermodynamic corrections to the RUC analysis profile, low-level SRH in the local Greensburg area was probably larger than around 200 m2/s2, as suggested by the Haviland profiler data (data somewhat incomplete, not shown) and increasing 0-1 km SRH values on other area RUC profiles after 03 UTC (as high as 400-500 m2/s2 in the Hutchinson area at 05 UTC or midnight CDT, not shown).   As noted earlier, SRH east of a dryline typically increases after dark due to diurnal backing of winds.

In my personal database of RUC analysis soundings that I've used in several papers (including two published in Weather and Forecasting), I have 26 profiles that are reasonable estimates of violent (F4) tornado environments during 2001-2006. Below are median values of estimated environment parameters for those tornadoes compared to the Greensburg tornado environment based on the corrected RUC profile above:

Caution should be exercised in using parameter values to forecast and assess storm environments, because nature does not recognize "numerical values" and "thresholds", and at best we can only estimate the characteristics of a given storm environment.  Nevertheless, it is still useful to compare parameter value estimates of a given environment with those from a carefully developed database of similar events.  Notice that the Greensburg environment values are very comparable to those from the RUC database for violent tornadoes.  One noticeable factor that may have helped contribute to the intensity of the Greensburg tornado (it was rated EF-5) was the total MLCAPE, which approached 4000 J/kg, compared to a median of 3000 J/kg in the database.  It is a little unusual to have that much CAPE associated with a large shear/large SRH setting.  The F5 Moore OK tornado in 1999 was likewise associated with around 4000 J/kg of MLCAPE in a similar large shear/large SRH environment.

After reviewing the above forecasts and analysis of the Greensburg tornado environment, it is not surprising on 4 May 2007 that the storm forming near the KS/OK border and moving northeastward over Greensburg after dark produced a deadly tornado.  Also, given the earlier forecasts of large SRH and total CAPE, and large low-level CAPE with a surface-based environment extending well into the night over central KS, it is not entirely surprising that large tornadoes with the Greensburg storm continued well after midnight in central Kansas, rather unusual for the central plains.

Hopefully the above analysis is helpful in comparing this tragic but well-warned event to other violent and deadly tornado settings. 

Jon Davies  5/12/07

(graphics used in this study are from the UCAR RAP site, Earl Barker's site, the FSL RUC site, and the ARL archive site)

return to main page