A monster supercell on May 22, 2014 produced giant, destructive hail and an EF-3 tornado just west of Albany, New York. In general, the event was poorly forecast.
The supercell formed outside of the “slight risk” area outlined by the Storm Prediction Center 1630 UTC forecast. Additionally the tornado probability was <2% and not contoured in the SPC forecast. The morning Area Forecast Discussion from Albany summed up the severe weather threat in the following way:
THEREFORE...WE FEEL THE THREAT OF SEVERE THUNDERSTORMS IS VERY LOW
(BUT NOT ZERO PERCENT). THE BETTER THREAT APPEARS TO BE LARGE HAIL
BUT AGAIN THAT IS A VERY LOW THREAT.
I don’t think anyone (myself included) expected a long-lived supercell producing a long path of significant severe weather. The storm gave us an opportunity to look at how this supercell looked with dual polarization radar – a relatively new tool in our severe weather arsenal.
The storm started out with a classic large hail signature in Fulton County, NY to the north of the New York State Thruway. The above video from the town of Broadalbin shows baseball size hail falling from the storm. While low level rotation was unimpressive the mid level mesocyclone was cranking.
From a dual pol perspective this was a classic large hail case. A large area of high reflectivity (near 70dbz in places) with ZDR near 0 and pockets of low CC. Because hail tends to tumble as it falls it appears spherical to the radar – hence a differential reflectivity near 0. Not shown is KDP which is fairly low over Broadalbin – only about 1º/km showing that liquid water was not a large contributor to the high reflectivity – the hail was! In addition, a three body scatter spike (note very low CC downraidial of hail core) was present from the lowest elevation slice (near 2,000 ft AGL) all the way up to 20,000 feet! Storm top divergence was also approximately 100 knots which is impressive for a northeastern US supercell.
The supercell went on to produce large hail in the town of Amsterdam, NY as well. The largest hail fell in an area with somewhat low reflectivity (55dbz over Amsterdam as opposed to 65-70dbz farther east). This is another example of why dual pol variables are better at detecting large hail than using reflectivity alone!
Picca and Ryzhkov (2012) showed giant hail detection is possible by locating areas of wet hail growth above the freezing level, and in particular, in the hail growth zone near -15C. They cite CC <0.9 and ZDR <0 as clues. This event showed that signal over Amsterdam when giant hail was indeed falling. You can see very large Z near Amsterdam at 9500ft coincident with sub-zero ZDR and CC <0.9. This signature is also present up into the hail growth zone above the -10C isotherm.
Rapid tornadogenesis took place following the large hail reports north of the Thruway. At 1928 UTC there is some sign of a reflectivity appendage at the lowest tilt. In addition the mid level mesocyclone is beginning to descend (not present at 0.5º or 0.9º tilts but is higher up at 1.3º). This was a tough event to warn for – as there weren’t many clues until the tornado was touching down.
By 1933 UTC tornadogenesis has occurred with a 56 knot gate-to-gate delta-V at 900 ft AGL. The tornado warning came out at 1939 UTC. By 1942 UTC a weak tornado debris signature is present and by 1951 UTC a clear tornado debris signature is present with very low CC, high Z, and near zero ZDR.
Given the strength of the tornado (EF-1 to EF-3), the proximity to the radar site, and the path of the tornado (through heavily wooded areas) I’m very surprised the TDS was not more dramatic. The “slam dunk” TDS only lasted 1 volume scan and only extended up to about 3500 ft AGL. Not particularly impressive.
One possible explanation for the relatively unimpressive debris signature was that there was a tremendous amount of rain and hail that was wrapped into the rear flank downdraft (KDP and Z quite high and ZDR near zero) that could have skewed the correlation coefficient higher than you’d typically expect with hydrometeors beings the dominant signal?
One other interesting feature is the presence of a ZDR arc shortly after tornadogenesis. The ZDR arc did not precede the tornado which frequently happens (no benefit for the radar operator) and was somewhat unusual because it appeared so rapidly.
You can see the ZDR arc here at 1947 UTC along the reflectivity gradient of the storm’s forward flank. ZDR is quite high in places (nearly 7db) and there is clear separation between high ZDR and high KDP. This is an example of hydrometeor size sorting in a highly sheared environment. The high ZDR is indicative of large drops and the smaller drops are getting pushed farther into the storm’s forward flank where KDP shows high water content. The ZDR arc appeared as the tornado reached its maximum intensity (EF-3) in Duanesburg.
The sudden appearance of a ZDR arc might indicate that the storm was moving into a more strongly sheared environment where hydometeor size sorting occurred quickly. Not surprisingly, this is where the tornado was the strongest. At the time of tornadogenesis there is a region of high ZDR displaced well east of the low level mesocyclone but it is coincident with high KDP indicating little separation from size-sorting. Prior to and during tornadogenesis there was no real signal from a ZDR arc to help forecasters with this storm.
The EF-3 tornado provided a unique opportunity to view the dual pol characteristics of this northeastern US supercell. Unfortunately, the storm’s rapid tornadogenesis was preceded with few, if any, clues on radar. When the tornado was on the ground the debris signal was surprisingly muted (many far weaker and farther from the radar tornadoes have produced much more dramatic TDS around here) but did eventually give forecasters confirmation of a tornado in the absence of any reports in real-time.