Tuesday, April 29, 2008
I won't post as often as I've been doing recently, but yesterday's Virginia tornado event was quite newsworthy, and had a very interesting environment worth discussing.
Above at top is the Suffolk, VA RUC profile at 20z (4 pm EDT) in close proximity to the tornadic supercell approaching Suffolk at the time (see radar image farther down, above). I've modified it to properly show the low-level moisture. Notice how the CAPE is essentially "squeezed" down below 500 mb, with the fattest CAPE at 700mb, and lots of CAPE in low-levels. Compare that profile to a nighttime tornadic profile from north-central Kansas on 4/25/08 at 05z (also shown above). On that profile, the low-level shear was larger, but the fattest CAPE was much higher up in the profile (around 400mb) and the total CAPE was distributed through a deeper layer, with not as much low-level CAPE as the Suffolk modified profile. This is more typical of tornado profiles in the plains, and contrasting this with the Suffolk profile at top, it is interesting to see how variable tornado environment profiles can be.
Yesterday's Virginia environment appeared to have CAPE and low-level shear (storm-relative helicity, or SRH) co-located down close to the ground, which may have facillitated or optimized tilting and stretching of low-level shear (SRH) for tornado development. I'm just speculating, but the contrasting 4/25/08 profile in Kansas may have required more low-level shear (SRH) because of less CAPE down low in the environment, and therefore less vertical stretching.
Yesterday's case also points out some problems with RUC profiles that I've talked about in recent presentations at NWS offices. In some cases with rapid moisture return and strong warm advection, the RUC does not moisten the low-levels enough. An extreme example was the RUC's handling of a tornado outbreak setting in Illinois on 4/20/04. Regarding 4/28/08 in Virginia, examine the unmodified Suffolk RUC analysis profile at 20z that I've also shown above, near tornado time. (Remember, the profile at top was modified by me to be more representative of the low-level moisture). The unmodified RUC profile has a dry "bubble" near the ground that appears erroneous and unrepresentative of the true environment, particularly if you compare that profile to the WRF analysis profile 2 hours earlier at Suffolk (also shown above) that seems much more reasonable in its low-level moisture depiction. Using mixed-layer lifted parcels (as most SPC mesoanalysis graphics do), the unmodified Suffolk profile yields about 50% less total CAPE than the modified profile, as well as less low-level CAPE, not a good representation of the setting.
This is significant, because SPC graphics are driven by the RUC. Notice how the Significant Tornado Parameter (STP) at 20z (shown above) using mixed-layer parcels and effective shear/SRH (that computation is supposed to be the "best" one, according to SPC research) showed no STP values whatsoever in this damaging tornado situation (at least EF-3 intensity in the Suffolk area). The main reason for this and problems with other SPC graphics yesterday was the erroneous low-level RUC profiles in the southeast Virginia area, similar to the unmodified Suffolk profile discussed above.
This is an error that forecasters need to be keenly aware of, and also addressed by SPC researchers regarding the SPC mesoanalysis output. The problematic graphics feeding off the RUC errors may have masked yesterday's environment and had something to do with SPC's slowness to issue a tornado watch in Virginia after tornado warnings were already being issued. Additional research also needs to focus on better operational highlighting of environments where CAPE is squeezed low in the environment, and co-located in the vertical with low-level shear/SRH, a situation not that easy to pick out using operational graphics of commonly-used supercell tornado parameters.
Jon Davies 4/29/08
Sunday, April 27, 2008
A strong capping inversion with warm air aloft pre-empting thunderstorms over the southern 2/3rds of Kansas was a big issue last Thursday afternoon (4/24/08). Many storm chasers targeted south-central Kansas waiting for convection to develop in convergence along the dryline there, but no storms developed. Looking back at some forecast and analysis maps from last Wednesday and Thursday, there were some definite clues that most of Kansas would remain capped on Thursday.
It is often helpful to reference 700mb temperatures (roughly 10,000 ft MSL) as a crude ballpark guideline for where the cap is setting up, particularly if convection occurred on the previous day and can be used as a comparison guide. For example, above, examine the 700 mb forecast temperatures from the NAM/WRF model 12-hour forecast valid the day before (Wednesday evening) at 00z 4/24/08. Then compare that with the observed composite radar graphic next to it at the same time (00z). Notice that temperatures above 7 or 8 deg C (the 7.5 deg C isotherm is marked by a dotted line on both images) seemed to delineate an area south of which storms did not develop. As such, that might be a good first guess for locating the cap using the next day’s models.
The 12-hour and 18-hour NAM/WRF 700mb temperature forecasts on Thursday valid at 00z 4/25/08 and 06z 4/25/08 are also shown side by side above, with the 7.5 deg C isotherm marked similar to the previous graphic. Although there is a “dip” in the 7.5 deg C isotherm over central Kansas at 00z, notice that the general pattern suggests a broad “tongue” of warm air aloft over most of Kansas, and that by 06z, the 7.5 deg C isotherm has moved northward close to the Nebraska border. This suggested that the cap would build in strong over Kansas during the evening, limiting convection to areas near the Nebraska border as an upper disturbance moved across overnight. The 02z radar composite (also shown above) confirms that evening convection was indeed limited to northwest Kansas where a supercell was approaching Hill City (see my earlier blog post about the ensuing nighttime tornado environment).
Also shown above are mixed-layer CAPE and CIN at 20z (3 pm CDT) on Thursday from the SPC mesoanalysis, and 700 mb temperatures at the same time. This is a good example showing that CIN values alone should not be relied on to identify a cap. Notice that the 20z CIN graphic shows little or no CIN was present in the Alva-Medicine Lodge area and near Kinsley KS, incorrectly suggesting an absence of capping along the dryline. However, the SPC 700mb temperature graphic suggests that all of south-central Kansas was capped with temperatures 7 to 8 deg C or above over a broad area, when using the previous day’s values as guidance.
Shawna Helt and I ventured west to the Salina area on Thursday afternoon, but when we checked data around 3 pm CDT, it became clear from satellite and SPC graphics that daytime thunderstorms over central or north-central Kansas probably would not happen. The strength of the cap discussed above, and the slowness of the upper disturbance coming out of Colorado (not shown), sent us on back to Kansas City before dark due to a tight schedule.
It is also important to understand that ballpark values of 700 mb temperatures suggesting the location of an inhibiting cap in the plains vary by situation and time of year (for example, 4 or 5 deg C in March to as high as 13 or 14 deg C in July). So be careful and don't use them as a "concrete guideline". Also, though such convection typically is very high-based, it is important to remember that strong heating and convergence over elevated terrain (such as the high plains west of 100 W lon) can pop storms within what looks like a “capped” area at 700 mb. This is a very common occurrence in the summer months. So there are lots of caveats to the above where experience is the tool that really helps.
Hope this is helpful regarding recognizing some future potential cap “busts” :-).
Jon Davies 4/28/08
Nighttime tornadoes, particularly after midnight, are somewhat unusual in Kansas and the central plains. This is largely because nighttime cooling at the surface below warmer temperatures aloft in many plains severe weather episodes prevents storms from being surface-based where “parcels” of air near the ground can rise rapidly to aid in tornado development. An exception sometimes occurs when strong warmth and moisture moves north in the plains at night in response to the approach of a strong springtime storm system. The deadly nighttime Greensburg area tornadoes last year were an extreme example of this type of situation.
On 4/25/08 after midnight (around 12:26 am CDT), a damaging tornado (rated EF-2, 15 mile path) developed north and northeast of Beloit in north-central Kansas (see this link from NWS Hastings). The parent supercell had been traveling east for nearly 5 hours before producing this tornado (see composite radar images above). Why did this storm suddenly produce a moderately long-track tornado in the middle of the night in Kansas? There were several factors present over north-central Kansas to suggest why.
First, the storm environment was becoming more surface-based as the supercell moved east. Above, compare estimated 0-3 km CAPE (this low-level instability is one indicator of how surface-based a setting may be) on maps from the SPC mesoanalysis at 02z (9 pm CDT) and 05z (midnight CDT). Notice that earlier, when the tornado-warned supercell (circled “S”) had been near Hill City in northwest Kansas, there was no low-level CAPE (red), suggesting an environment that wasn’t really surface-based. However, by 05z, an area of low-level CAPE (red) was showing up over north-central Kansas, with the supercell moving into it. This was a more surface-based setting, a result of strong warm moist advection, with surface weather maps indicating a significant temperature and dew point rise at Salina (not shown) from 68/63oF to 70/64oF during 03z (10 pm CDT) to 05z (midnight CDT), at night.
A similar setup is seen with SPC maps of estimated total CAPE and CIN, also above. In red, CAPE (instability) was present across northwest and north-central Kansas throughout the evening. But large CIN in dark blue over northwest Kansas at 02z (indicating inhibition that would slow near-surface air parcels from rising rapidly beneath storm updrafts) was not a good environment for significant tornado development with the supercell located near Hill City. But later, at 05z, notice that the supercell was approaching an area with much less CIN (in light blue or white) over north-central Kansas, indicating a more surface-based environment, increasingly favorable for tornadoes if low-level wind shear was also present.
Also above, SPC maps of the 0-1 km energy-helicity index (EHI) at both 02z and 05z showed large combinations of low-level wind shear and instability supportive of supercell tornadoes (suggested by large EHI values) pointing northward into north-central Kansas. This is where the more surface-based environment at 05z was also located as discussed above, with the long-lived supercell moving eastward into this “more favorable” axis. These factors likely contributed to the unusual middle-of-night tornado near Beloit, when such events are relatively rare. As clusters of storms increased over southeast Nebraska and northeast Kansas after 06z, widespread organized outflow brought an end to the tornado threat.
Computer forecasts from the NAM/WRF model valid at 06z (1 am CDT) on 4/25/08 did a reasonable job hinting at potential for a nighttime supercell tornado environment as far out as 18 hours in advance. Note above the model forecast of strong EHI values (wind/instability combinations supportive of tornadoes, bright colors) at 06z, as well as the more surface-based environment suggested by forecast 0-3 km CAPE values at 06z (orange and yellow colors). These overlapping parameters from north-central and northeast Kansas into southeast Nebraska (where an EF-1 tornado occurred from another storm) suggested a model forecast “heads-up” to forecasters for nighttime tornado potential well in advance of this event.
Jon Davies 4/27/08