An evolving study
The “Surprise Tornado Outbreak” of August 24th, 2016 dealt over 40 tornado reports to the Midwest, despite not even a 2% risk of tornadoes outlined in the morning by the Storm Prediction Center. Why this happened has been scrutinized in detail; however, this is not the first time nor the last time it will happen again.
Mesoscale Convective Complexes are large, quasi-circular thunderstorm systems that roam the Great Plains and Midwest during the late spring and summer months, often reaching peak strength overnight. Fueled by the evapotranspirative powers of vast fields of crops, these dump copious amounts of rain on the Corn Belt, and are self-sustained by the latent heat released by massive amounts of cloud condensation. Once peak strength is achieved, low pressure forms in their centers which induces large-scale rotation especially evident in the mid-levels of the atmosphere. Once the MCC decays in the morning, thunderstorms may weaken or cease to exist altogether. However, the mid-level Mesoscale Convective Vortex remains intact and continues to drift on.
Unfortunately, numerical modeling is not advanced enough to capture the strength and scope of these MCVs, and our sparse network of upper-air observations also fails to do them justice. The result? Mid-level winds could be significantly underestimated and surface response substantially distorted locally by forecaster analyses, even with morning observations. In essence, a significant atmospheric disturbance has now been created where previously it was not forecasted, and the worst part is, it’s not easy to measure.
Thus, remnant MCV tornado setups often come as “surprise” scenarios where ingredients come together just right for a few or even several tornadoes in a very local area. The problem is, these ingredients don’t indicate severe weather, leaving forecasters hesitant to even mention the word “severe” right up until it happens.
Accurate forecasts of remnant MCV environments are challenging even right up to the point when tornadoes first touch down. Effective preparedness, at that point, is lost. However, there are indeed precursors that may hopefully help assess risk in the hours leading up to the event.
Though not necessarily conducive of severe weather in a classic sense (i.e. damaging wind and large hail), remnant MCV environments can be quite conducive for tornadoes. In other words, just as certain setups carry predominantly a damaging wind threat, or maybe a large hail threat, this setup may carry mostly a tornado threat.
Why the environment is not conducive to damaging wind and large hail:
- Weak lapse rates (mid-level warming from MCC)
- Moist mid-levels (mid-level moistening from MCC)
- Weak bulk shear (typically ridging conditions)
Why the environment is conducive to tornadoes:
- High low-level CAPE (very warm, moist surface)
- High low-level SRH (backed surface flow, enhanced mid-level winds from MCC)
- Low LCLs (high boundary-layer relative humidity)
- Pre-existing surface vorticity (near stationary front that MCC developed along in the first place)
This is where communication is key; of course it sounds absurd to scrap the tried-and-true warning, “primary threats will be damaging straight-line winds and large hail, but a few tornadoes are also possible” in favor of “primary threats will be a few tornadoes, with damaging straight-line winds and large hail also possible”, but would this not be more representative?
Of course, a threat of tornadoes is often not forecast when sustained supercell activity appears borderline at best. However, most tornadic storms forming within this environment are short-lived mini-supercells; these are deceivingly small with much of their rotational strength confined to the lowest levels out of sight of the radar network, and tornadoes may even be brief enough to slip between radar scans. This only adds headaches to the warning process.
I am examining over a decade of Storm Prediction Center tornado reports in the Midwest (April 1st – September 30th, 2006-present), using archived upper air and visible satellite observations to determine the presence and impact of remnant MCVs.
Events tended to be accompanied by:
- Isolated/localized tornado reports
- Lack of severe wind/hail reports
Events tended to share a common synoptic pattern:
- Weak/split upper level flow
- Ridging (“Ridge Riders”)
- Significant severe convection upstream (most often Northern Plains)
Tornado Reports: 148
Remnant MCV tornado events appear to happen roughly 2-3 times per year in the Midwest.
Remnant MCV tornado events appear to be most common in the area from late spring into early summer, but may occur from late April through late September.
Though skewed rather heavily due to the August 24th, 2016 tornado outbreak, the remnant MCV events that average the most tornadoes per event seem to generally occur on both ends of the season, in late spring and late summer.
Fortunately, the majority of events only produce one or two reported tornadoes. However, about a quarter of events will produce more than 5 tornadoes.
Due to their placement within synoptic patterns with CAPE and shear environments that are highly unfavorable for strong, long-track tornadoes, remnant MCV tornadoes are generally weak, with only a small fraction rated as significant. The vast majority are not even rated.
For each of the cases, a visible satellite image was chosen at a time where MCV locations were most discernible, marked “X”. Their direction of motion is labeled with an arrow. Approximate locations of tornado reports are marked with upside-down triangles.
Most readily visible is observational proof of the theory that, as in hurricanes, tornadic activity is most likely in the right-front quadrant of the MCV following its motion. In almost all of these cases, the direction of motion of the MCV either pointed toward future tornadoes, or put all future tornadoes in its right-front quadrant.
Further corroborating this theory, the MCV-forward-motion-relative positions of all tornado reports are shown above, in relation to the analyzed position of the MCV (marked with a circle at (0,0) at the time of the first reported tornado. Looking down the direction of movement of the MCV, the vast majority of reported tornadoes occur in its right-front quadrant. A handful of left-front quadrant reports do exist, most in very close proximity to the MCV. In general, the reports appear to spread out and further to the right of the MCV motion with time (likely due to differences in MCV motion vs. tornadic supercell propagation or boundary orientation).
All tornado reports:
For the sum of all cases, a coherent smattering tornado reports exists in a corridor from central IL into central IN, down into western KY and up into western OH. Is there an “MCV Alley”?
Analyzing the paths of these MCVs from early morning through late afternoon, there appear to be trends as the season progresses.
- April MCVs take a southerly, zonal track and are generally fast-moving
- May MCVs originate from central Plains convective events and trek northeastward out of large-scale troughing
- June MCVs begin forming from the north/central and high Plains and drift eastward at a slightly slower pace
- July MCVs start taking advantage of northwest flow regimes and many originate in the north-central Plains
- August MCVs are mostly all embedded within northwest flow and originate almost exclusively in the northern Plains
- September MCVs return to a more zonal, southerly track
Despite the somewhat small per-month sample size, a seasonal oscillation in the direction of MCV propagation that was expected is readily apparent. After a brief period of mostly zonal motion, MCVs come from the SW out of western troughing early in the season, then from the W and eventually the NW as ridging builds in and northwest flow dominates, before returning again to a more zonal direction
Again considering the sample size limitation, the average MCV propagation speed undergoes a similar seasonal cycle as expected. Initially swift-moving, with some samples exceeding 40mph, the MCV’s average speeds decrease into the summertime as stronger jets migrate north and patterns become more stagnant (the period late July into early August, as well as the month of September is deemed unrepresentative due to too small a sample size).
MCV birth locations are marked as points. Each point was taken at 0300z, which seemed to be the average time that a mature convective complex began producing a significant mesoscale circulation. This circulation typically evolved from a bowing line segment with strong bookend vortex, which may or may not have evolved from a single significant supercell (as a side note, that a single rotating supercell is able to alter its environment enough to manifest in a broadening, strengthening wind field that remains intact for almost 18 hours and nearly one thousand miles and ultimately provides the support for a number of more tornadic supercells is breathtaking).
Early on in the MCV’s formation out of robust convective activity, radar analysis shows that the MCV may shift erratically, often pushing southward with a strong rear-inflow jet then abruptly jumping north into a region of stratiform rain (thus MCV paths above are often disconnected from formation points slightly). Starting around daybreak when convective activity dissipates, visible satellite analysis shows the MCV can then travel largely with the synoptic flow in a uniform direction (thus 12z-21z paths can be drawn as straight lines). At the end of this period, 21z marks the average time that the next-day tornadic convection occurs, and also closely coincides with the MCV ceasing to be discernible using radar or visible satellite.
The “Typical” Tornado-producing Remnant MCV
For the 12 cases, the typical tornado-producing remnant MCV is born in eastern Kansas the night before, strengthening overnight as it drifts through Missouri. The MCV crosses into Illinois just after noon, and the trajectory the steering flow takes it determines exactly where tornadoes may become a threat.
At midday, an analysis of average MCV location, motion and resultant tornado reports should look familiar. Almost all tornado reports are in its right-front quadrant!
At about the time of peak tornadic activity, an analysis of of averaged CAPE, 0-3km SRH, 700mb winds and surface winds reveal exactly what would be expected for a remnant MCV tornado event in their expected climatological region:
- Mid-level vort max (MCV) and associated CVA
- Peripheral enhanced mid-level flow
- CAPE gradient (stationary front)
- Backed surface winds (and resultant high SRH)
Why is this Important?
We are NOT good at this!!
Assessing 1200z Storm Prediction Center Convective Outlooks for the events, the vast majority of remnant MCV tornado reports occurred outside of areas with any delineated tornado risk or even severe thunderstorm risk.
That begs the question: how well, numerically, has the Storm Prediction Center performed for the events studied?
Not so well, unfortunately, with the 1200z morning convective outlook carrying a dismal detection rate of remnant-MCV tornadoes when a “miss” is defined as a report occurring outside of a “2%” tornado risk or within a “General Thunderstorm” risk area.
Now, AS A DISCLAIMER, this is not meant to criticize the efforts of the professionals who strive to anticipate all impactful severe weather events with the lowest false alarm rate. By nature, this specific scenario defies both of these goals; MCVs are quite difficult to anticipate even the morning of, and most MCVs drift by without one tornado ever being spotted. Furthermore, in recent years, the SPC and local NWS offices have taken huge strides in delivering continuous, ever-evolving predictions when complex scenarios unfold, and in communicating the uncertainty that plagues such scenarios. Now, however, we are faced with the task to fine-tune our forecast process and hone in on gaps in our understanding and skill. This is one such gap, but further research will only help close it, making next “surprise” tornado outbreak hopefully less of a surprise.