What happened? May 23, 2020
Several convection allowing models (CAMs), most notably the morning through midday runs of the HRRR, initiated deep convection near the Colorado/Kansas/Nebraska border area this afternoon. Although there was some weak convection for a couple of hours, in the 23-01z time frame, this convection struggled.
The Storm Prediction Center (SPC) even issued a Tornado Watch, but for most of the day, they remained fairly skeptical about deep convection in the area. Even their mesoscale discussion (MD) from mid-afternoon only highlighted a 20% probability that a severe weather watch would be needed.
It wasn’t just the HRRR. The image to the above-left shows robust simulated radar reflectivity swaths. Several 12z CAMs initiated convection over northeastern Colorado and sustained it for at least a few hours in a northeasterly direction. Based on model forecast simulations, convection would initiate near a dryline and eventually intensify as it encountered a more favorable environment to the east.
Note above that it was not just the HRRR (brown) that showed robust updraft helicity (UH) swaths across the area. The 12z NSSL, shown in red, similarly showed a stout updraft passing near the Colorado/Kansas/Nebraska border area. The 12z HRW NMMB (blue) did as well. While not perfect, the 12z 3km NAM seemed to have a better handle with convective evolution.
The 12z run showed CI over northeastern Colorado, which was wrong in placement, but the general idea was close to what actually occurred a bit farther to the southeast. Interestingly enough, the 18z run blew up convection up and down the dryline across a sizable portion of northeastern Colorado, which did not verify.
Cloud cover and thermodynamics
Mid and high level clouds were prevalent across much of eastern Colorado and northwestern Kansas through the day. As Chip Redmond discussed early this afternoon, it turns out that the clouds likely limited daytime heating. The forecast sounding below is an area averaged sounding for Sherman and Cheyenne counties in northwestern Kansas, from the 17z HRRR, valid at 21z.
Note the forecast average 2-meter temperature of 82 degrees F. St. Francis (KSYF) was only 76F at 21z and temperatures were generally in the mid to upper 70s in the area at the time, when looking at nearby METAR and Kansas Mesonet stations. Right off the bat, temperatures are approximately 5F cooler than forecast. While the sounding above shows near-surface heating eroding convective inhibition (CINH) as temperatures warmed into the lower 80s, more CINH was in place than was forecast.
Above is a graphical representation of the 17z HRRR forecast. Not only are temperatures roughly 5F cooler than forecast, but dew points were also lower than forecast. St. Francis’ dew point was only 52F at 21z, which was 4F cooler than forecast. While Colby (KCBK) was progged to have an 80/58 temperature/dew point, the observed conditions were 77/55.
Basically, not only was it cooler than forecast, but it was also drier than forecast. The end result is more CINH, on top of what was already forecast to be relatively large temperature/dew point spreads. This is not a conducive environment for sustained convection.
Initiating boundary and CI
There was a noteworthy dryline bulge in northeastern Colorado, near the Kansas border. Forcing in vicinity of the dryline (boundary) bulge over higher terrain was sufficient for CI. In northeastern Colorado, convective temperatures were in the mid to upper 70s.
As convection was moving into a relatively dry near-surface environment with sizable CINH at somewhat lower elevations, the convection struggled and ultimately collapsed. Note that convective temperatures were in the lower 80s in northwestern Kansas, but actual temperatures were only in the mid to upper 70s, creating an environment hostile for surface-based convection.
Mesoanalysis showed an area of substantial CINH in northwestern Kansas. Based on observations, this is where cloud cover seemed to most notably limit surface heating. In addition, drier low-level air than forecast also played a role.
This was not a result of a traditional capping inversion, which usually involves warm temperatures in the mid-levels. Instead, most of the CINH was due to dry low-level air and a lack of near-surface heating to erode CINH.
Summary
It is always a good idea to look at severe weather setups from an objective, quantitative standpoint. Instead of getting lured by a simulated radar forecast map or bright colored UH swaths, take a deeper dive into why a model is producing (or not producing) storms.
It is easy to “cherry pick” soundings, so try to use area-averaged forecast soundings to get a better idea of what the broader scale environment looks like. If area-averaged soundings are not available, pick and choose multiple point-and-click soundings over a broader area.
In today’s case, the first red flag should have been lingering cloudiness. While forecast models did show morning clouds, the broken/overcast cloud deck lingered into the afternoon across northwestern Kansas. As a result, instead of convective temperatures in the lower 80s being reached by 21z, actual temperatures lingered in the mid to upper 70s.
Additionally, 2-meter dew points were also at least a couple of degrees cooler than forecast. In an already borderline environment with drier air than forecast, it should have been evident that convection might struggle to sustain itself, if it could even initiate at all.
