Wednesday, April 1, 2020

The Jonesboro, Arkansas tornado on March 28, 2020: A subtle and difficult forecast setting.


As Covid-19 keeps expanding throughout the U.S. (we all need to keep following social distancing guidelines!), thoughts and well wishes go to people like fellow storm chaser Dr. Bill Hark in Virginia who has been sick with the virus.  We hope you recover soon, Bill.

With this going on and more than 4500 Covid-19 deaths now in the U.S., tornadoes may seem like "small potatoes".  But spring is upon us, along with the threat of tornadoes, no fooling intended on this April Fools Day.

Saturday's EF3 tornado that struck Jonesboro, Arkansas (AR) around 5:00 pm CDT (2200 UTC) is a reminder of that (see photos above).  With a tornado watch out a couple hours in advance and good warnings, no one died with 22 injuries reported, which is good news.

Most meteorologists on March 28 (including me) were focused on the potential for tornadoes over Iowa and Illinois, but the local environment really ramped up quickly in northeast AR during the afternoon, and was more subtle than one might expect prior to a large tornado.  So, I spent some time taking a closer look at this event.

Morning model forecasts did not suggest much low-level shear over AR compared to farther north near a warm front (see 0-1 storm-relative helicity / SRH forecast for mid-afternoon below):

But as the southern end of a large upper system approached (not shown), a low-level jet at around 5000 ft MSL (not shown) increased from 30 kt to 45 kt over northeast AR during the afternoon, helping to generate enough low-level shear to support tornadoes, which we'll discuss in a bit.  What might also be easy to miss is that a subtle boundary also appeared to play a role.

Below is a 3-panel composite radar image from late evening on March 27 through midday on March 28 showing this northeast-southwest boundary (see white arrows) drifting eastward across Arkansas during the 18 hours before the tornado:


Although it's not clear what originated this boundary (NWS-analyzed surface maps showed this as a cold front, which I don't think it was), it appeared to stop near Jonesboro and even back up a bit to the west during the afternoon.  Here's the surface map I analyzed at 2100 UTC (4:00 pm CDT) about 45 minutes before the tornado struck Jonesboro (notice the backed southeasterly surface wind at Jonesboro). This boundary is shown as a thick red-blue dashed line over AR:
And here's a composite radar image at 2125 UTC (4:25 pm CDT) with the boundary from the 4:00 pm CDT surface map above superimposed as a white dashed line:

Notice that the Jonesboro tornadic supercell (white arrow, producing its first tornado near Amagon AR at this time, southwest of Jonesboro) appeared to be moving northeast right along this boundary.  It's possible this boundary provided increased low-level wind shear and convergence to help support tornadoes.

Regarding storm environment parameters over northeast AR, the SPC mesoanalysis showed an area of enhanced low-level MLCAPE (0-3 km above ground, 1st panel below) and, although not particularly impressive, an area of somewhat enhanced 0-1 km SRH (around 150 m2/s2, 2nd panel below) near Jonesboro:


This "overlap" may have set the stage for enhanced tilting and stretching of low-level SRH with the supercell storm moving along the aforementioned boundary.  The 2100 UTC SRH over northeast AR (larger than forecast from the morning model runs) appeared to be the result of the low-level jet increase during the afternoon that I mentioned earlier.

An additional graphic I put together (1st panel below) shows areas of overlap of 0-3 km MLCAPE > 75 J/kg and 0-1 km SRH > 150 m2/s2 in purple from the 2100 UTC SPC graphics above:


Notice that the genesis region of the Jonesboro tornadic supercell was in the purple area over northeast AR, again suggesting localized potential for enhanced low-level tilting and stretching of SRH along the boundary discussed above.  The 2100 UTC effective-layer significant tornado parameter (STP, 2nd panel above) also suggested support for supercell tornadoes over northeast AR.

Here's the 1-hour forecast sounding from the RAP model for Jonesboro valid at 5:00 pm CDT (2200 UTC) at the time of the tornado, suggesting the localized support for supercell tornadoes near the boundary:


Although both STP and the energy-helicity index (EHI) weren't especially impressive (both parameters around 2.0) for a large EF3 tornado, they do indicate an environment supportive of tornadoes, one that may have been given a "boost" by the presence of the boundary in addition to the sounding environment shown above.

Here's a larger view of the forecast afternoon setting from the morning NAM model run, showing the large 500 mb trough moving through the central U.S. with the typical spreading jet stream pattern (thick white curved arrows) ahead of it, providing dynamic forcing and lift:


The insets on the graphic above show EHI and fixed-layer STP forecasts for mid-afternoon (again, not very impressive for northeast AR compared to areas farther north), and SPC tornado reports during the afternoon.

This case is a good illustration of how important it is to monitor features and parameters in real time, as the mid-afternoon SPC mesoanalysis and surface map indicated increasing low-level shear and support for supercell tornadoes near the boundary over AR, which was not forecast well by the morning model runs. 

Away from Arkansas, notice that over Iowa this was a "cold-core" tornado event  with a closed 500 mb low nearby and a boundary intersection (warm front and Pacific cold front) over central Iowa (see the surface map earlier above).  As is typical with cold-core settings, several tornadoes occurred near this boundary intersection over Iowa as it evolved northeastward during the afternoon, including this tornado northeast of Des Moines near Rhodes, Iowa after 4:00 pm CDT:

Back to Arkansas, the Jonesboro tornado setting and environment was more subtle than those accompanying recent large tornadoes such as the nighttime Nashville tornado back on March 3 and yesterday morning's tornado near Eufaula, Alabama (March 31).  In both those cases, the tornadic supercells were near a warm front or stationary front, and low-level shear was larger and more evident (0-1 km SRH 300-450 m2/s2) than near Jonesboro on March 28 (0-1 km SRH 150-190 m2/s2).

I hope everyone reading this navigates the Covid-19 outbreak carefully and in reasonable health over the next couple months!

- Jon Davies  4/1/20

Tuesday, March 3, 2020

March 3, 2020: Double-digit death toll from nighttime tornadoes in Tennessee



The first truly major tornado episode of 2020 took place last night in Tennessee where at least 24 people were killed by a tornadic supercell moving straight east through central Tennessee.  A large tornado moved through Nashville near the downtown (see above) after midnight, killing 5 people in Nashville and the area east of the city.

The most deaths (at least 18) occurred in and near Cookeville, 60-70 miles east of downtown Nashville, with a tornado from the same supercell.

The upper air setting associated with this tornado episode was a little unusual for March.  Most such events occur with a large midlevel (500 mb) trough moving east through the southern states of the U.S. and prominent southwest flow aloft over the Tennessee area.  But in this case, midlevel flow was west to east over Tennessee (see the 18-hour NAM model forecast below), with a positive tilt trough (thick dashed red line) approaching from the west-northwest (rather than the southwest) within a northern branch of the jet stream:


Another midlevel trough and closed low were way back to the west in northwest Mexico within a separate southern branch of the jet stream.  But the flow between these two jet stream branches was "spreading out" over the Tennessee area (see large white arrows in the graphic above), providing strong lift within an area where significant combinations of instability and low-level shear as indicated by the energy-helicity index (EHI) were located (see inset and the yellow oval shown above).

At midnight CST (0600 UTC, about 35 minutes before the tornado hit Nashville), the effective- layer significant tornado parameter (STP) from the SPC mesoanalysis was sizable (around 3.0) over central Tennesssee, indicating combinations of instability and wind shear that were favorable to support supercell tornadoes:

At around the same time, the surface map showed a warm frontal segment  moving east through central Tennessee, just ahead of the favorable STP environment shown above:

Areas near warm fronts where instability and wind shear are increasing (as indicated by the earlier STP graphic) are good tornado producers when the warm, moist air moves north and east, and this case was no exception.

The High-Resolution Rapid Refresh (HRRR) model forecast from mid-morning on March 2nd did a reasonable job suggesting the tornado potential for the coming nighttime hours going into March 3rd.  The model forecast supercell storms over western into central Tennessee (note the black rotation tracks on the radar forecast below), along with forecast 6-hour updraft helicity swaths from 0200 UTC to 0800 UTC (also below) suggested storm rotation and certainly possible tornadoes given the instability and shear environment discussed earlier.

Just prior to the tornado in Nashville, the RAP model analysis sounding at 0600 UTC showed a setting with large low-level and deep-layer shear (0-1 km SRH > 400 m2/s2, and 0-6 km shear > 60 kt) and adequate instability (MLCAPE around 800 J/kg):



Above, I've highlighted in yellow the most important parameters regarding support for supercell tornadoes.  The STP and EHI values, while supportive of tornadoes (around 2.0 to 3.0), weren't unusually impressive for a killer tornado with double-digit deaths, probably due to this being an early season event with large shear but relatively small CAPE, typical of cool season tornadoes in the southeast U.S.  But, notice the large amount of low-level CAPE below 3 km (125 J/kg of "3CAPE" at lower left on the graphic above).  Put that with the large low-level shear/helicity and deep-layer shear, and that probably facilitated strong low-level tilting and stretching of vorticity within the supercell storm updraft to generate a tornado stronger than one might expect with less than 1000 J/kg.

Here's another image of the tornado moving near downtown Nashville, as indicated by the power flash left of the skyline:


As I write this, preliminary survey information suggests at least EF3 intensity with the tornado in areas just east of downtown Nashville.  Given that this tornado occurred in a metro area after midnight, it is fortunate that the death toll in Nashville wasn't higher, possibly indicating some effectiveness of warnings there. Sadly, death tolls were larger farther east as the storm tracked toward east-central Tennessee in the middle of the night, a difficult time to make people aware of tornado warnings.

Jon Davies  - 3/3/20

***********  UPDATE 3/5/20 *************
The tornado that hit Putnam County and Cookeville around 70 miles east of Nashville has been rated EF4 (violent in intensity) with 18 deaths and 88 injuries. Please consider helping the tornado victims in Tennessee... here are a couple sites you can visit regarding donations:
CNN - How to help Tennessee storm victims
KIRO - How to help Tennessee tornado victims

Wednesday, January 22, 2020

ChaserCon 2020 is almost here... Tornado Forecasting class on Friday, January 31!



Hi everyone!

ChaserCon 2020 is a little over a week away.  I'm doing a Tornado Forecasting class on Friday evening January 31 from 7:00 to 9:30 pm, with a break in the middle.  You can sign up at:

              http://chasercon.com/node/18

It's $30.00 per person... Just scroll down to the bottom of the page at the link above to register for the class.

I'll do more examples and interactive exercises covering supercell tornadoes than last year at Wichita, including a cold-core tornado case.  So don't miss it if you're interested in forecasting.

There's room for around 100-110 people, and we're up to 83.  Sign up soon!

I'll also be doing a talk on Saturday afternoon (February 1) about the crazy and confusing tornadic HP supercell in northeast Kansas last May 28, as part of the conference.

This will be the last ChaserCon in Denver hosted by Roger and Caryn Hill, so come if you can.  Hope to see you there!

(Thanks to Bill Hark for the photo at top from my class last year in Wichita.)

- Jon Davies  1/22/20

Thursday, December 19, 2019

Southern tornado outbreak leaves 3 dead on December 16, 2019


A strong storm system moving across Louisiana (LA), Mississippi (MS), and Alabama (AL) on Monday, December 16, 2019  caused the first tornado deaths in the U.S. since May 2019.  A large long-track EF3 tornado above (at top) in west-central and central LA left one person dead near DeRidder, and was on the ground for over 60 miles, striking Alexandria around noon on Monday.

Another tornado (EF2) at late afternoon from a supercell embedded within a squall line killed a married couple in Lawrence County in northwest AL.  The bottom photo above shows an EF2 tornado north of Tupelo MS earlier from the same embedded supercell.  In all, around 30 tornadoes were reported in LA, MS, AL, and Georgia on December 16 through December 17.

The 2-panel base reflectivity radar composite image below shows the deadly tornadic supercell at midday over west-central LA (1st panel), and the deadly tornadic supercell within a bowing line segment at late afternoon over northwest AL (2nd panel).  Prolific tornadic supercells over southern MS are also indicated:


This storm system was a good example of a large "positive" tilt trough in mid-levels (see thick red dashed line on 500 mb forecast graphic from the NAM model below) causing a significant amount of severe weather.  As noted on the NWS "Jet Stream" educational site, positive tilt upper troughs (where the southern end of the trough is much farther west than the northern portion) tend to produce the least amount of severe weather, but that certainly wasn't the case on December 16:






A broad pattern of spreading jet branches (thick white lines and arrows on graphic above) ahead of this trough was causing upward forcing over the southern states.  With plenty of wind shear and enough instability present at midday over LA and MS to support supercell tornadoes (see inset showing energy-helicity index or EHI above, combining instability and low-level shear), the outbreak was in progress from late morning on.  Note that the southern jet branch above tended to define the southern end of the severe outbreak area, while the northern extent of instability defined the northern end.

Here's a U.S. surface map at midday showing the cold front and surface low pressure associated with the midlevel trough moving east across the western Southern states:


A RAP model sounding at midday at Alexandria LA shows how primed the atmosphere was for tornadoes over this area:


There was close to 2000 J/kg of MLCAPE, around 250 m2/s2 of 0-1 km storm-relative helicity (SRH), and over 50 kt of deep-layer wind shear.  Add to that a lack of convective inhibition (MLCIN) and plenty of CAPE in low levels below 3 km above ground (100-150 J/kg) to promote low-level stretching of updraft parcels, and this was an environment very supportive of supercell tornadoes with discrete or semi-discrete cells.

In recent years, the introduction and use of updraft helicity in computer model forecasts can help forecasters predict where rotating supercell storms (and, by association, possible tornadoes) may develop and track.  A HRRR model forecast from the morning of December 16 (below) suggested that long-track supercells would develop from central LA into southwest MS between 1500 UTC (9 a.m. CST) and 2100 UTC (3 p.m.CST), which was a good model forecast:


By mid to late afternoon at 2200 UTC (4 p.m. CST), the focus for supercells and tornadoes had shifted east to Mississippi and northwest Alabama, as seen on the SPC mesoanalysis graphic below:


The best combinations of low-level wind shear (0-1 km SRH) and low-level instability (0-3 km MLCAPE, see black oval-enclosed area above) were over Mississippi into northwest Alabama, where low-level stretching and tilting of wind shear within stronger storms would be enhanced, supporting low-level rotation and tornadoes.  The cells marked "A", "B", and "C" in the graphic above were consistent long-track tornado producers.

The supercells marked "B" and "C" on the SPC graphic above were particularly effective tornado producers.   From these cells, images in sequence below show a large mid-afternoon tornado northwest of McComb MS, a large tornado shortly before 5 p.m. CST near Columbia MS, and the EF3 tornado that struck Laurel MS at dark.  Power flashes are quite visible in the latter two images:



As is so often the case these days, excellent NWS warnings and media coverage probably saved many lives in this outbreak.

- Jon Davies  12/19/19

Tuesday, October 22, 2019

October 20, 2019 Dallas, Texas tornadoes after dark - no deaths or injuries!


Wow... it has been an active several days for tornadoes this past weekend in October (see my recent post about the EF2 tornado in Florida from T.S. Nestor).

As anyone watching news the last couple days knows, Sunday evening (10/20/19) saw several tornadoes in the Dallas, Texas area after dark, including one EF3 that struck the North Dallas-Richardson corridor around 9 pm CDT (top image above).  Another of EF1 intensity hit the Rowlett, Texas area a little later from the same supercell storm (middle and bottom images above, note the very impressive power flash!) .

**** Update 10/24/19:  Additional tornadoes have been surveyed by NWS Dallas/Ft. Worth in recent days, including a short-track EF2 tornado in Garland just before the Rowlett tornado mentioned above, and an EF1 in Rockwall after the Rowlett tornado. These tornadoes were from the same supercell that produced the North Dallas and Rowlett tornadoes. ****

No one wants to see damaging tornadoes in a metro area, especially after dark, but it is great news that there were no injuries or deaths reported in the Dallas area, largely due to a tornado watch and good warnings by NWS.

I've had several people ask me, "Isn't that odd for this time of year?"  Not really.  Based on  statistics over the past 25 years, Texas sees an average of around 9 tornadoes in October each year, and 58 tornadoes occur on average nationally in October.  So it does happen with the right meteorological settings.

Sunday evening's surface map (below, 7:00 pm CDT) showed a dryline west of Dallas with moist air (dew points upper 60's to near 70 deg F) that had moved back into north Texas on south winds during the day, with the deepest moisture from the Red River southward:

Tornado parameters from the SPC mesoanalysis at 7:00 and 8:00 pm CDT (below, enhanced energy-helicity index or EEHI, and the effective-layer significant tornado parameter or STP) suggested that combinations of instability and wind shear were quite supportive of tornadoes over the Dallas area as storms were developing rapidly and becoming supercells to the west:

The northernmost supercell on the 7:00 and 8:00 pm SPC images above produced the damaging EF3 tornado in north Dallas around 9:00 pm CDT, followed by the Rowlett tornado just after 9:30 pm CDT.  In fact, the warm sector setting east of the dryline over north Texas was so supportive of tornadoes that no boundaries were needed to help produce tornadoes, unlike last Friday evening in Florida.

A forecast of instability and wind shear from the RAP model sounding at Dallas a couple hours before the tornado shows excellent combinations of CAPE (instability), low-level wind shear (0-1 km storm-relative helicity or SRH), and deep-layer wind shear (0-6 km shear) were in place, with not much convective inhibition (CIN) in the environment.  These factors were all supportive of significant supercell tornadoes if discrete storms developed:


It is worth noting that the High-Resolution Rapid Refresh (HRRR) model from earlier on 10/20/19 (see below) was not successful in forecasting convective storms and supercells over the Dallas area that evening, although it did forecast storms in Oklahoma, and over west central Texas:


So, even with our often impressive automated model guidance of convective storms these days, the human forecast element is still greatly needed!

In my post this past weekend about the Lakeland, Florida tornado, I pointed out from radar images how that supercell behaved somewhat like a typical Plains tornadic supercell with one tornadic mesocyclone occluding and dissipating, while a new one formed to its east-southeast.  That same evolution was seen Sunday evening with the North Dallas and Rowlett tornadic mesocyclones, as is evident on the reflectivity and storm-relative velocity images below:

Looking at the larger synoptic picture, the NAM model 500 mb forecast for that evening showed a very large and strong midlevel trough (dashed red line below) moving through the Central Plains, with a typical "branching" jet pattern ahead of the trough.  This area of dynamic forcing overspreading the returning low-level moisture through the Plains was where the bulk of severe weather occurred Sunday evening and Sunday night:


One final note... The north Dallas tornado touched down 15-20 miles east-northeast of AT&T Stadium where the Dallas Cowboys were playing at the time of the tornado.  Although the soon-to-be tornadic North Dallas supercell stayed well north of the stadium, it is nevertheless very fortunate that the EF3 tornado did not directly impact the thousands of people at the game!

- Jon Davies  10/22/19

Saturday, October 19, 2019

October 18, 2019: A strong Florida supercell tornado with Tropical Storm Nestor


It's been awhile since I posted about a recent tornado case, so I pulled together some graphics about the meteorological setting with last evening's EF2 tornado after dark (see images above) near Lakeland in west-central Florida east of Tampa.  This tornado was associated with a tropical storm (T.S. Nestor, centered well out in the Gulf of Mexico at the time).  But instead of the supercell being embedded within an outer band of storms as with many tropical systems, it was discrete and occurred near an east-west stationary front, behaving in some ways more like a Plains supercell storm.

The stationary front is shown on the surface map below at about 0300 UTC (11:00 pm EDT), about the time of tornado development near Lakeland:


There was not a tornado watch in effect at the time, probably because the environment most supportive of significant tornadoes appeared to be out in the Gulf of Mexico closer to the center of Nestor, as indicated on the 0300 UTC (11:00 pm EDT) SPC mesoanalysis graphic below using the effective-layer significant tornado parameter (STP):

However, low-level wind shear (0-1 km storm-relative helicity or SRH) and low-level CAPE (0-3 km MLCAPE) on the SPC mesoanalysis at 11:00 pm EDT (below) were notably co-located together over west-central Florida near the stationary front:



This was in the same area where a supercell storm (shown in later graphics down below) was moving north-northeast near and across the stationary front.  The combination of these two low-level parameters near the ground probably facilitated low-level stretching and tilting of environmental vorticity into the storm's updraft to produce strong low-level rotation, even though total instability and numerical shear/instbility combinations did not appear especially large.  The stationary front likely provided additional low-level shear to add to the background environment as the supercall moved across the front.

A 3-hr forecast sounding from the RAP model at Lakeland valid at 0300 UTC (11:00 pm EDT) also suggests that the environment was supportive of supercell tornadoes:


Notice above that the red CAPE area extended only up to around 30,000 ft MSL (300 mb), with the "fattest" area of CAPE located near 10,000 ft MSL (700 mb), rather low in the sounding relative to the ground.  This is similar to many "cold-core" low-topped supercell  environments that support tornadoes in the Plains, suggesting significant upward air acceleration in low-levels, resulting in strong vertical stretching of low-level vorticity near the ground, even though total CAPE does not appear unusually large.  

So, the vertical distribution of CAPE near the stationary front was probably important in this case.  Also, deep-layer shear (0-6 km shear) near 30 kts on the sounding above, while not overly impressive, was just enough to support supercells and tornadoes.

This discrete supercell associated with a tropical system also behaved more like a Plains supercell on radar.  The Tampa radar reflectivity and storm-relative velocity images zoomed in below show the original tornadic mesocyclone northwest of Lakeland occluding and dissipating (drifting toward the back side of the storm while wrapping in rain-cooled air), while a new mesocyclone formed to its east and southeast, similar to the evolution of many Plains supercells:

Thankfully, even though the EF2 tornado was on the ground after dark for 9 miles and was nearly a third of a mile wide at times, there were no injuries due to timely warnings from NWS Tampa.

- Jon Davies  10/19/19

Saturday, August 17, 2019

Tornadoes in Kansas in mid-August! 8/15/19 in northeast Kansas


Supercell tornadoes undoubtedly can occur in Kansas in early to mid-August, but they don't happen that often, and even as a native Kansan, I've never seen one.  That changed on Thursday, August 15 when a well-established supercell (see Shawna's photo above) in northwest flow produced tornadoes southeast of Manhattan and southwest of Topeka.

Here's a map showing the approximate location of three tornadoes that occurred with the supercell pictured above in the 7:00-9:00 pm CDT time frame:
(A fourth brief tornado occurred with a separate trailing supercell after 9:00 pm CDT.)

I was chasing with James Skivers and my wife, following the southernmost supercell that developed northwest of Manhattan and moved sharply southeast across Wabaunsee County, Kansas (KS).  We missed the first brief tornado (EF0) from this supercell near the western Wabaunsee County line south of I-70 around 8:12 pm CDT when we were driving south to stay ahead of the storm through a low area with trees.  Here's Brian Miner's cool photo of the full supercell base and this first tornado, all in the same shot:


The second tornado (rated EF1 by Topeka NWS) was longer-tracked near dark, southwest of Alma, KS and northeast of Alta Vista.  The image of this tornado below is from Shawna's video, backlit by lightning looking east-northeast, with the rear-flank downdraft (RFD) and gust front visible:

A third brief tornado (EF0, below, looking north and also backlit by lightning) occurred near the time the second tornado was ending, back to its northwest under an occluded mesocyclone behind the one that generated the EF1 tornado:

A separate supercell back farther to the northwest produced a brief EF0 tornado after 9:00 pm CDT east of Alta Vista (not shown).

The satellite image below, with relevant features superimposed, shows the setting just before 7:00 pm CDT (0000 UTC 8/16/19):

Note the stationary front and outflow boundary (from earlier storms that moved into northern Missouri) close together over northeast Kansas near Manhattan and Topeka. Winds from the low-level jet (LLJ, around 850 mb or 5000 ft MSL) were impinging on and overrunning these boundaries, helping to initiate storms northeast of a "cap" that was inhibiting convection to the southwest.

Three different supercells are visible in the image above (black arrows show their southeastward motion), but only the southernmost cell near Manhattan was able to access the most unstable air along and south of the boundaries by virtue of its location and motion, probably a big factor in its ability to produce tornadoes about 90 minutes after the time of the satellite image.  The low-level jet was also increasing near dark, as is typical, enhancing low-level shear supportive of tornadoes.

The SPC mesoanalysis depiction of the effective-layer significant tornado parameter (STP) at 8:00 pm CDT (shortly before the tornadoes) showed large values supportive of tornadoes along and near the aforementioned boundaries in northeast Kansas (the enhanced energy-helicity index, an experimental parameter, is also shown):


Notice that, even though these tornado forecasting parameters appeared "favorable" over a rather large area extending north of the boundaries, the tornadoes were limited to the air mass south of and near the boundaries.  This suggests the importance of following the location and evolution of relevant boundaries in forecasting/nowcasting tornado potential, rather than just relying on parameter "bulls-eyes".

Also, here's the NAM model 500 mb forecast (winds at roughly 18,000 ft MSL) showing the northwest flow driving into Kansas, a little unusual that far south in mid-August, and the reason the supercells moved southeastward to the right of the midlevel flow:


Two shortwave troughs are marked on the forecast above (thick dashed red lines) showing a lead shortwave associated with the morning and early afternoon storms that moved from northeast Kansas into northern Missouri, and a fairly strong trailing shortwave not far behind crossing Nebraska.  Often, morning and early afternoon storms leave subsidence (sinking motion) and a "worked-over" air mass in their wake, reducing the chances of severe storms behind them.  But that was not the case here, with another shortwave immediately behind generating upward motion and intensifying low-level winds converging on and overrunning the surface boundaries.

All in all, an extremely interesting case for mid-August in Kansas!

- Jon Davies  8-17-19