Saturday, 14 October 2017

An Analysis of the 2017 Tornadoes at Three Hills and Jenner, Alberta

Despite being a season that many chasers considered a dud (at least as far as tornado sightings are concerned), Alberta had three quite memorable tornadoes of what ended up being 10 confirmed events in 2017. June 2nd in Three Hills, June 9th in Mannville, and October 7th in Jenner were all events that local chasers would have liked to have observed through their viewfinders, that also caught many forecasters off guard (save for maybe Mannville). Images of a local mowing his lawn in front of a spectacular, tall tornado churning up brown earth, or of snow on the ground in front of a fully condensed funnel gave these events a quirky "Albertan" flavour, and propelled them to instant fame on social media. The sting of not seeing such photogenic spectacles was also real for those who make a habit of seeking them out. But two of these twisters also appeared to be special in another way, in that they likely originated from supercells in otherwise seemingly benign environments - which is likely the main factor in why they were somewhat of a surprise to forecasters and chasers alike who might otherwise be anticipating them. While supercell tornadoes occurring in June are not uncommon, those occurring in October over brown fields with still-melting snow drifts from a recent winter storm seem a little counter-intuitive.

The following is intended to be, as in previous posts, an informal discussion of the meteorological background environments of the Three Hills and Jenner tornadoes of 2017. I will attempt to make a case that these tornadoes originated from parent thunderstorms that were supercells - that is, those with rotating updrafts - which involve different mechanisms responsible for tornadogenesis than those of landspout tornadoes. Landspouts often arise from a different set of environmental conditions and are often found beneath rapidly growing updraft towers of non-rotating thunderstorms, which also commonly occur in Alberta. All data has been collected from the College of DuPage website, the SPC Mesoanalysis Archive, and Radarscope, and all annotations are mine. A lack of upper air sites and proximity soundings causes me to rely on prog soundings and derived values on mesoanalyses, when surface and upper air analyses are not available - so these may not be entirely representative of the environment. Any scholarly work will be directly linked and not cited as in a traditional, formal paper given the informal nature of this blog. A glossary of (undefined) acronyms will be included at the end of the post. The main goals of this post are to document these events for future reference, and to attempt to accurately diagnose the mechanisms responsible for them to help spotters and forecasters alike in better anticipating similar future occurrences.

Three Hills

On June 2, between approximately 4:40-5:10PM MDT (2240-2310Z), a tornado was observed tracking from WNW-ESE a few kilometres north of the town of Three Hills, Alberta. In the end the tornado was given a rating of EF1 due to structural damage assessed on a property next to Hwy 21 where a barn sustained heavy damage.

Meteorological Background Environment and Discussion

A 23Z 500mb RAP mesoanalysis reveals a mid-level trough over eastcentral Alberta into southwest Saskatchewan, and a speed max associated with an embedded vort max in WNW flow nosing into southern Alberta near tornado time. At the surface, a NW-SE trending wind shift axis/effective dryline was draped from the central foothills out across southcentral Alberta. Light, relatively moist NNE surface flow was found along the northern side of this boundary, and dry westerly flow was found to its south. A tight moisture gradient characteristic of strong drylines was present near Three Hills, with surface dewpoints of 9C observed along the northern side close to dewpoints of -3C just to the south.

23Z RAP 500mb Mesoanalysis. Courtesy COD.
23Z Surface obs. Note the sharp dewpoint gradient near the dryline. Annotations mine. Courtesy COD.

DCVA ahead of the approaching vort max would have served to steepen mid-level lapse rates and increase CAPE along the moist side of the boundary, which was characterized by SBCAPE values (per COD mesoanalyses) in excess of 1000J/kg in the pre-storm environment. Prog soundings indicated more modest MLCAPE values near 600J/kg as a result of the "skin layer" of shallow moisture characteristic of the time of year. MLLCL heights are thus also quite high (progged to anywhere from 1500 to over 2000m AGL). The mid-level speed maximum would have also grazed the area from the south - and when superimposed over veered, light northeasterly surface flow north of the boundary, more organized convection - including supercell structures - could not have been ruled out, due to a local increase in deep layer wind shear. Convergence along the low level boundary was also the main source of lift in initiating thunderstorms, which may have also contained embedded misovortices as a result of the breakdown of horizontal shearing instabilities along the wind shift line. And finally, very steep low level lapse rates in excess of 9C/km existed across much of southcentral Alberta.

18Z NAM prog sounding valid at 21Z representative of near-storm environment. Steep lapse rates are readily noted, as well as thin layer of near-surface moisture. Calculated shear values are lacking due in large part to progged WNW surface flow, when in reality observed flow was NNE. Courtesy COD.

23Z SBCAPE (derived). Courtesy COD.
22Z MLLCL heights. ~2000m near Three Hills, right on the northern edge of the image. Courtesy SPC Meso Archive.
22Z 0-3km (low level) lapse rates. 9.5C/km+ (nearly dry adiabatic) lapse rates depicted near Three Hills. Courtesy SPC Meso Archive.

With an environment characterized by marginal deep layer and low level shear, seeing a tornado such as Three Hills - with its classic "dust tube" appearance - may automatically be assumed to be a landspout without much further thought. Moreover, it occurred in an environment of high LCLs, steep low level lapse rates, seemingly weak shear, and along a wind shift axis, which are all conditions typically found in landspout environments. However, the presence of marginally ample deep layer shear (as mentioned above), as well as storm motions nearly parallel to the dryline (which would provide a ready source of low level streamwise vorticity to the thunderstorm updraft) may have resulted in the development of a mesocyclone in the parent tornadic thunderstorm.

The only real fly in the ointment that I personally see (for the tornado being driven by supercellular processes) are the high LCLs. Typically, high LCLs (roughly "cloud base height") reveal deep, dry, well-mixed boundary layers with strong evaporative cooling potential, which would result in cold and stable RFDs with air that is resistant to lifting beneath the main updraft - preventing the spin-up of near ground vertically-oriented vorticity to tornado strength. However, the ingestion of the strong baroclinic circulation associated with the tight moisture gradient would have resulted in a steady dose of streamwise vorticity for the thunderstorm, strengthening its low level mesocyclone, and the resulting dynamic "suction" effects beneath the storm. The steep low level lapse rates in the subcloud layer would have also provided little resistance to vertically accelerating parcels originating from near the ground, so these two effects combined may have compensated for the high LCLs and colder rear flank downdraft air. In fact, this excellent paper by Jon Davies notes that tornadoes can occur with higher LCLs in high plains environments (like Alberta) than would otherwise be expected elsewhere - perhaps due to the offsetting effect of steep low level lapse rates.

Radar Presentation and Visual Characteristics

Unfortunately, I only have one velocity scan saved from the lowest tilt (hindsight is always 20/20), at which level only marginal cyclonic shear appears to be present. A very well-defined pendant (hook) echo is present on reflectivity at tornado time, revealing perhaps a few larger hydrometeors amid the drier air of the RFD being wrapped into the occlusion. A very progressive rear flank gust front spills out to the south of the storm into the drier air south of the boundary, but doesn't cut off the storm's warm sector and inflow along the boundary right away.

Lowest level scan when tornado was ongoing. Beam would be intersecting the storm at near 6Kft, approximately 40 miles north of XSM (Strathmore Radar). Courtesy RadarScope.

The other telling feature in the following radar loop from YouTube is the deviant motion of the tornadic thunderstorm compared with all other convective cells moving along with the mean cloud wind, as it becomes anchored to the boundary and VPPGFs associated with a rotating updraft contribute to the storm's motion during the time it is rotating.

Moreover, visual clues pertaining to the storm's structure lend to the hypothesis that the parent storm possessed a rotating updraft. Many of the shots taken of the Three Hills tornado were taken from right in town, which was too close to get a sense of the overall structure. The following shot was taken from further south, which seems to reveal high-based clouds comprising the rear flank gust front arcing out away to the south of the storm, and a well defined RFD "cut" (which evaporates most cloud matter due to its being quite dry, and wraps into the updraft base). The blue streamline in the image is my depiction of the environmental streamwise vorticity attendant with the boundary being ingested, tilted vertically, and stretched by the thunderstorm, and contributing to the development of a low level mesocyclone - but not directly to tornadogenesis. It is unknown to what extent regular theorized effects of the RFD played in tornadogenesis, or in the potential of pre-existing low level vertically-oriented vorticity near the wind shift line.

Photo: High River Online. I rotated the original to level the horizon.

The overall structure and radar presentation of the Three Hills storm seems to reveal that it possessed updraft rotation. Landspout tornadoes often simply occur when strong updrafts (towering cumuli) develop over pre-existing "swirlies" near the ground, helping them to spin up in an environment of steep low level lapse rates - before any downdrafts occur. The presence of downdrafts are thought to be of crucial importance to supercellular tornadogenesis, while landspouts tend to occur before downdrafts become well-established, during the updraft phase - and can be seen pendant to or beneath dark, flat cloud bases in the absence of much other structure.A great example of this is the beautiful "dust tube" (landspout) that occurred near Mannville on June 9:

Photo: Troy Bader. This landspout is still considered a tornado, even if no condensation funnel is visually hanging from the cloud base. Here, we see the classic dust tube beneath a dark, flat updraft base.


On October 7th, between about 2:50-3:10PM MDT (2050-2110Z), a tornado was observed tracking from NW to SE, approximately 15km south and southwest of the village of Jenner, Alberta. Since it passed over open fields, the tornado was given a rating of EF0. Environment Canada determined that the tornado (at least at the time of writing) was a landspout.

Meteorological Background Environment and Discussion

The 00Z October 8th (6PM MDT on the 7th) 500mb analysis reveals a subtle shortwave trough axis from SW Alberta into central Saskatchewan, which is rotating through flow aloft about a larger upper circulation centered over northern SK and SE NWT. A cold pool of exceptionally cold temperatures at 500mb also exists over central Alberta, with temperatures to -30C indicated, edging into the Jenner region. At the surface, a SW-NE oriented wind shift axis is translating across southern Alberta, likely as a surface reflection of weak forcing associated with an embedded vort max in the mid-level trough (not shown). Winds were moderate southwesterly ahead of the convergence axis, with moderate northwesterly flow behind. A sharp moisture gradient also existed along the wind shift, with dewpoints near 3-4C being sampled ahead of it, and ranging from from -6C to -10C behind it - a similar gradient to the one observed near Three Hills on June 2. The convergence along the line likely helped to pool the moisture into the region, with some of it possibly being added to the environment with the ongoing snowmelt in SE Alberta after a recent winter storm.

00Z Oct 8 500mb analysis. Courtesy COD.

2030Z Oct 7 Surface obs and derived SBCAPE (RAP). RAP often underestimates moisture in the region which likely accounts for the lack of derived SBCAPE, near 250J/kg. NAM progged MLCAPE values near 500J/kg in a narrow corridor ahead of the boundary. Courtesy COD.

With such cold temperatures aloft, it is not surprising that at least some modest instability developed across southern Alberta - especially in the pool of relatively moist air ahead of the boundary. Indeed, steep low and mid-level lapse rates contributed to pockets of progged MLCAPE values near or in excess of 500J/kg near the best moisture - relatively high values for the time of year. Moreover, environmental deep layer shear was progged to be slightly greater than at Three Hills, with near 30 knots of effective shear forecast for the area. However, given the time of year and position of key synoptic features, a low tropopause would make for low-topped convection/low ELs. Thus, the majority of the buoyancy was confined to the lower and mid troposphere, including a share of low level (0-3km) CAPE values in excess of 200J/kg near Jenner. This, combined with relatively low LCLs (owing to smaller temperature-dewpoint spreads near ~9/4 in the pre-storm environment), and appreciable low level shear (where winds veered approximately 90 degrees from SW at the surface to NW a little ways aloft in the low levels) was likely crucial to the development of the tornado. Contrary to Three Hills, storm motion was orthogonal to the initiating boundary, so it appears the surrounding background environment alone played a large role in tornadogenesis to the exclusion of any other known mesoscale features.

18Z NAM prog sounding valid at 21Z near Jenner. Note the low EL, appreciable low level CAPE and MLLCLs, low ELs, and steep lapse rates. Surface winds were again progged to be westerly, but in reality winds were backed to southwesterly - which would serve to further increase low level turning. Courtesy COD.
18Z NAM progged MLCAPE valid for 21Z. A narrow corridor of MLCAPE values approaching 500J/kg exist where moisture has pooled ahead of the boundary. Courtesy COD.

Effective shear (EBWD) at 20Z reveals a single barb of 35 knots right over Jenner - which would be supportive of updraft rotation. Courtesy SPC Meso Archive.

MLLCLs at 20Z. Note the couplet of high to low MLLCLs between the dry and moist air across the boundary, with approximately 1000m AGL MLLCLs indicated near Jenner. Courtesy SPC Meso Archive.

After reviewing the data as well as photo and video evidence, it seems to me that the Jenner tornado almost certainly originated from a low-topped supercell. Low-topped supercells are essentially "miniature" versions of their traditional cousins, manifesting smaller versions of the same features - both visually observed, and on radar. Here is a good link that explores low-topped supercell features specifically as observed on radar. Of course, a supercell occurring in Alberta in October with single digit temperatures and dewpoints is quite rare indeed - let alone one that could produce a tornado. However, it turns out that the key ingredients were likely there, since radar returns and visual clues almost certainly reveal supercell tornadogenesis.

Radar Presentation and Visual Characteristics

Historical radar captures reveal that a mesocyclone likely developed between 230-240PM MDT, and dissipated between 310-320PM MDT. The tornado cycle likely occurred between 250-310PM MDT, based on ground reports. The storm was unable to recover from the occlusion near 310PM and began to dissipate thereafter. Several supercell features appear to be evident on reflectivity and velocity at various tilts, including a hook echo, inflow notch, V notch, BWER, and cyclonic rotational couplet. The moderately fast storm motion straight down the radial toward XBU (Schuler Radar) likely augmented some of the inbound velocities while masking some outbounds. A series of scans will be presented from the lowest tilt to see the evolution of the mesocyclone, and then an analysis of all 3 scans (on RadarScope) around the time the tornado developed (at 250PM MDT) follows. These images are likely best viewed in sequence (the blog gives the option to view images this way):

230PM MDT Base reflectivity and velocity. The following frames are all courtesy RadarScope via Mark Simpson. We begin by looking to the rear of the cell WSW of Iddesleigh for development of a mesocyclone.
240PM MDT. A hook and cyclonic shear couplet begin to rapidly take shape SW of Iddesleigh.
250PM. The tornado cycle begins. At this distance (50 miles from XBU), the beam intersects the storm at about 6500ft AGL.
300PM MDT.
310PM MDT. The tornado cycle has just ended, and we are already seeing signs of rapid dissipation.

A closer look at 250PM MDT reveals the following:
Tilt 1. A fairly classic supercell presentation, with all annotated features. The presence of a V notch also suggests the existence of a strong updraft, which often implies rotation.

Tilt 2. An echo overhang indicates a strong updraft (associated with supercells), with a subtle BWER being seen here. Also, a mid-level mesocyclone appears to be clearly sampled on velocity at this tilt.

Tilt 3. Directly above where we saw the BWER in tilt 2, we see the strongest reflectivity values at tilt 3. This is consistent with a strong updraft, which is likely rotating - though here in the upper levels of the storm, we are above the mesocyclone.

Visually, some impressive photographs and videos were taken of this event. These images also reveal textbook supercell features, and most notably the RFD "cut" - a feature that spotters look for as a clue that a tornado may be about to form, which appears here consistently throughout the tornado's lifetime. This implies that supercellular processes were likely responsible for this impressive October tornado.

Photo: Jeff Johanson. This shot was taken from highway 884 looking NW at 248PM. It is truly amazing to see a supercell funnel beginning to develop behind melting snowdrifts in the ditches and brown October fields in Alberta.

Photo: Jodi Aebly. This shot was taken at 308PM, looking north on 884 when the tornado was fully condensed and had tracked to the east of the highway, 20 minutes after the above photo. The RFD cut is still plainly visible.

Photo: Screen capture from a video taken by Jeff Johansen, at about 309PM. Here, he was looking SE toward the now dissipating tornado. Our RFD cut is once again visible.

Reflections and Conclusion

For me personally, as a chaser and amateur forecaster, both of these events were eye-opening experiences. I specifically recall the irony of arriving home in Calgary from Tornado Alley an hour before the Three Hills tornado, and the pain of having seen the parent thunderstorm while driving by and ruling it out as being "too high-based" to do anything interesting. When I should have been giving my wife my undivided attention after having been away for 3 weeks, I spent a little while cursing with deep regret that I missed seeing the "Canadian Wray" - a spectacle that I had just spent 3 weeks seeking out. Nonetheless, we live and learn as chasers and forecasters alike.

It was also easy to initially rule out a supercell tornado in Alberta in October when I had first seen the pictures. However, upon closer investigation, myself and several of my peers were intrigued to find that this had likely originated from a low-topped supercell. Exactly 6 months earlier (to the date), another surprise low-topped supercell went briefly severe warned near Buffalo (also very close to Jenner) on April 7th. So it has been a season rich in learning experiences in that we shouldn't rule out the possibility of such occurrences from the get go due to being in abnormal times of year. If the environment is there, spinny things are possible. Nonetheless, the above analysis seems to reveal that both tornadoes were born of supercell thunderstorms. However, I have been mistaken in my diagnoses before. I am hoping this discussion continues among the greater minds of the atmospheric science community.

What this all is teaching me is not to downplay those more benign-looking days - especially after a cold front scours out most of our moisture the day previous following a more classic, warm sector severe setup. These days often hold some goodies for the lucky few potato farmers or oilfield workers who happen to be out there - low risk, high reward days with no chasers to be found. It's easy to rule out severe weather when there appears to be marginal CAPE and moisture, shear, and when there is westerly surface flow in Alberta, which has a dry, downslope component. We need to keep an eye out on the following types of days (especially when all other necessary ingredients for thunderstorms exist):

- When there are pronounced moisture gradients, usually found along wind shift/convergence axes with minimal temperature differences across them;
- When enough moisture is present to yield even modest CAPE values;
- When there are steep low level lapse rates - especially near axes of convergence where even marginal instability exists;
- When there is still just enough wind shear to organize convection

Most of these ingredients often occur together during "cold core" days following the passage of an energetic trough, when most ingredients seem too marginal for severe weather within a cool, drier, and unsettled upper trough/low. It turns out that even seemingly innocuous days can have just enough residual moisture, shear, and instability for a little magic to happen - even outside of the classic summer season. Here is another good paper that addresses the fact that these kind of days (closed 500mb lows in particular) often catch forecasters off guard, which means many chasers will also be caught off guard as well. However, the more we can learn from these events, and become savvy in our attention to detail - resisting the psychological urge to downplay a given setup, the more we can anticipate them in the future.

A glossary or acronyms follows, in rough order of appearance throughout this post. If you're still unsure what these mean, it's a great opportunity to study! :)

MDT: Mountain Daylight Time

SPC: Storm Prediction Center

EF_: Enhanced Fujita (Scale)

RAP: Rapid Refresh Model

COD: College of DuPage

DCVA: Differential Cyclonic Vorticity Advection

CAPE: Convective Available Potential Energy

SB (prefix to CAPE): Surface-based

ML (prefix to CAPE): Mixed-layer

LCL: Lifting Condensation Level

ML (prefix to LCL): Mixed-layer

AGL: Above Ground Level

NAM: North American Mesoscale Model

EL: Equilibrium Level

RFD: Rear-flank Downdraft

VPPGF: Vertical Perturbation Pressure Gradient Force

BWER: Bounded Weak Echo Region

No comments:

Post a Comment