Tornado Forecast Busted 9-14-05 - Let's Discuss Why

I think yesterday's episode in western OK and northwest Tx will make a good study. Many thought this day was prime for a few tornadoes. Even SPC expected perhaps a few. What do you think went wrong? If someone can point us to some archived data and model runs that would be a great help.
I need to look at this some more, but my take on it is primarily it was the 500mb PVA (or more accurately) the lack thereof as shown on NAM model runs that contributed to subsidense and suppression of intense severe activity. This coupled with being on the right exit region of the 300mb jet. Jeff mentions in the REPORTS section that it was partly lack of a cap, but as I recall there was at least 10 degrees at 700mb - not exactly no cap; however given adiabatic heating did reach and surpass convective temp.

My take is even though storms got started (primarily in Tx) most were in slightly high LCL's and the subsidense preventing quality growth and strengthening over time. For some reason there were also two sets of storms some east of the others and with anvils hanging out almost east I believe. I'm thinking this may have screwed with the better storms (to the west) inflow air eventually.

Anyway that's just cursory. Any other observations or opinions here? I think it is important to know the reasons why tornadoes don't form as the corolary is knowing reasons why they do form.
The problem was mostly too much stuff went up at the same time... Which was a concern for me too, but I didn't think it would turn to crap nearly as quickly as it did. Very strong directional shear and moderate speed shear through the layer, extreme instability and defined moisture pooling along the east-west OFB in northwest TX/southwest OK made me think there would be a few strong supercells with a couple tornadoes to add into the mix. These storms were mostly defined supercells when they first initiated... But there was too much competition and a lot of the storms got cut off and it pulled a "Michigan-day" (LOL...) on everybody pretty much. That was the biggest factor: too much clutter as everything went up too quickly.
Let me throw out some ideas to ponder...

On May 3, 1999, there weren't "too many storms."

On April 26, 1991, there weren't "too many storms." In fact, there were supercells at about 40 mi. intervals producing F3 (Peabody), F5 (Andover), F4 (Cowley Co.) and F5 (Red Rock) tornadoes (from north to south) simultaneously.

My point is that we continue to miss something (perhaps unmeasurable with current technology) regarding tornadogenesis. What made May 3, with lots of supercells, some quite close together, such an efficient tornado environment?
I didn't end up chasing yesterday so I didn't pay real close attention to what was going on, but from what I saw, I would have to echo what Nick said. There were too many storms competing with each other in the area where the shear was most favorable for tornadoes. The one or two storms that did remain discrete for a while were further South where LCL's were higher and they didn't have E SE surface winds and a boundary to enhance tornadic potential. Also, storms that moved into SW Oklahoma were tracking nearly parallel to the front which meant that other storms fired and it became a cluster pretty quickly. IMO those were the main problems with yesterday. The potential was there, but it is hard to know how convection will evolve until it happens.
I think there were two main causes: cell motion nearly parallel to convergence boundary (front) and an imbalance between capping and forcing...

First off, model forecasts called for storm motions to the east or perhaps east-southeast. The storms in the initial cluster near Dickens, TX, moved northeastward for the most part (some ENE). So, given an area of sustained convergence, storms would move off to the NE, or ENE, only to have more storms develop to their southwest. The orientation of the storms meant that the precipitation from one storm fell into the inflow/updraft of the storm to its northeast.

Second, ample sunshine in the area meant that the convective temperature was likely reached over a relatively large area. Without any real cap, the strong convergence along the front and various OFBs (or along each particular storm's front-flank outflow).

IMO, upscale growth from supercell cluster to MCS occurred relatively rapidly mainly because of the two above-noted factors. Now, another question is why the strong low-level shear didn't yield a strong supercell, around which there should have been compensating subsidence. Since we're all chasers, we know it certainly isn't uncommon for initiation to yield a cluster of storms, with one or two updrafts dominating in time. Heck, this happened 5-29-04, among many other days. On those days, however, the shear vector was more normal to the convergence boundary (dryline on 5-29-04).

Looking at 0z data and model initializations last night, I almost cried. Assuming the model initializations were at least close to correct, the area along and immediately north of the stationary front was moderately unstable and moderately/highly sheared. Re really don't see the juxtaposition of strong shear (250-500 0-3km SRH, 50-90kts effective shear, etc) with strong instability (2500-3500j/kg SBCAPE) very often. As I noted in my REPORT post, it appears that prior model runs underforecast 850mb flow... Well, I guess it goes to keep all of us on our toes...

Biggest lesson learned: It doesn't really how much shear / instability is present if the storm mode is quasi-linear instead of discrete supercells. I strongly believe that, had we had discrete supercells near or north of the Red River yesterday, we would have seen tornadoes, and I still think there was the definite possibility of a strong tornado owing to the strong low-level shear. Instead, we saw a relatively rapid transition to MCS. There was a supercell that developed near or just norht of CDS that had decent rotation, but that was before the MCS-like convection south of Quanah moved northeastward and intercepted the warm, unstable inflow.

EDIT: I agree with Mike. Some of the biggest tornado days have seen many supercells in a relatively small area. As he noted, 5-3-99 had several supercells in very close proximity that still produced long-lived, significant tornadoes. I think this comes down to the orientation/position of the supercells relative to each other. Instead of one supercell precipitating into another's updraft/inflow (such as was the case yesterday I believe), the supercells on 5-3-99 were able to remain in close proximity, yet their position relative to each other didn't allow for "negative" storm interaction. I think such storm-scale things are largely unforecastable given what we currently know and can measure/forecast...
I thought we might be in trouble when the NAM began to seperate the instability from the shear. This turned out to be accurate and made tornadoes much more unlikely. The best CAPE values shifted southwest, toward CDS and points south, while the shear remained along and north of the boundary, particularly the very important low level turning we counted on to take advantage of low LCLs with storms crossing the OB/front.

The best storms fired off the dryline or the synoptic triple point and maintained some discretion and structure, like the storm I saw in northeast Dickens County around Dumont. This storm was somewhat elevated but maintained a vigorous updraft and strong rotation until other storms in the cluster undercut its efforts. As Nick mentioned, widespread convection made it hard for storms to stay organized, and even the more isolated cells--like the one we chased later from near Post to around Jayton-- rained into their own inflow. Or new storms east of them further contraminated source regions. Every storm I saw took on HP characteristics and outflow dominance rapidly. With other storms all around, any serious updrafts were quickly undercut.

Rather than subsidence inhibiting convection, we actually had too many storms too soon.

It is interesting to note that around 0300z or so, some of the storms actually intensified and showed modest rotation as they at last moved northeast enough to reach the boundary.

This was always a pretty marginal setup for tornadoes; we needed strong, isolated cells to take advantage of low LCLs and sufficient 1k SRH as they crossed the boundary. The storms that might have been close enough to the boundry to do this were in lower instabilities and a clustered environment, and the isolated storms further south were too far from the boundary when they did mature.
I think Amos and others covered this pretty well. I agree there was a significant disconnect between the favorable shear and the instability. The garbage junk on the low-level jet kept conditions near the outflow boundary around the Red River too cool to reach the convective temp, whereas further southwest this was less of a problem, but the forcing boundary extended out underneath the better upper-level flow, and well away from the favorable low-level shear environment closer to the Red River. The convective development upscale in that region was already well described, so why no storms further northeast? Again - there was a lack of subsidence in the wake of the morning convection to clear out the redevelopment until early afternoon, followed by limited heating, further limited by anvil overspreading the region of interest. A few small cells did try to develop leading the cluster, but there just wasn't enough juice to get them going. With the development occuring on the subsidence quadrant of the upper jet the area soundings at 00Z showed pretty lame lapse rates at midlevels as well. So, pick your poison, there were several working against the event.

As noted by Amos, I doubt many seriously saw the environment as great, it was marginal at best in my opinion as well, but certainly had the potential had a few things on the mesoscale worked out differently. I'd say the title of the thread is also unfair - it wasn't much of a tornado forecast - as I recall SPC only had 5% probs, and that seemed quite reasonable.

I don't think the quantity of storms has anything to do whether storms will produce tornadoes or not. Check out this radar loop from 8/18/2005 when Wisconsin had a 27 tornado outbreak. Cells were RIGHT next to each other.

Some good points all of you, and I tend to agree with pretty much all of them.

As for the title - I was just trying to get people's attention so they would comment. I realize it was never supposed to be a big tornado day and so it didn't really bust that much. But from discussions I read and from what I have heard about on other lists a lot thought it was expected to be a bigger day.

So in summary let me see if I can list the main problems with the day all together:

1) Convective temp reached over too large an area contributing to too many storms at once.
2) Anvils spread east or northeast and contaminated those areas with colder outflow and contributed to additional seeding.
3) Storm motion was mostly in a similar direction of the anvil so to some degree the cells eventually began to ride into rain cooled air.
4) Storm motion did not perhaps take advantage of orientation of the outflow or frontal boundaries to help increase SRH as much as it may have if it was oriented as Jeff mentions was forecast, though some brief strenthening was seen as the OFB passed through the cells.
5) 500mb negative buoyancy and 300mb right exit region subsidense made lapse rates lame so updrafts were not as intense as they might have been.
6) Storms formed in area that had reached convective temperature and was more unstable but also had a bit higher LCL's and less shear compared to the more favorable area in the direction of CDS, but storm attributes in current environment contributed to messing up the 'better' environment to the NE.
7) I'd have to verify but it seems I recall storms nearby earlier in the day which may have assisted in stabilizing more of the area along or north of the frontal boundary preventing cell formation earlier in these areas.
8) All of the above contributed to the quick formation of an MCS which further surpressed the ability for producing quality supercells and tornadoes.
I knew there was a decision to make when we left to go west to the shear or southwest to the instability. I'll take the instability every time, so off we went. I knew we were screwed when we saw three storms go up all at once right along each other. Some days aren't really a mystery as to what went wrong so much as days where you have to remind yourself not to get too excited. I wasn't really taken aback by the lack of tornadoes or even inflow-dominant sups yesterday, just really disappointed. But the storm we got south of Paducah was more than I expected upon initiation, so all in all a good chase.

I think setups this time of year are wishcatsed a bit more than the same type of day would be in May. The bar is lowered I believe for chase setups on the backside of the calendar.
Originally posted by Bill Tabor

5) 500mb negative buoyancy and 300mb right exit region subsidense made lapse rates lame so updrafts were not as intense as they might have been.

I'm not sure what you mean by 500mb negative buoyancy. I'd say the instability was sufficient along and immediately north of the front (in high low-level shear), but of course "sufficient" is subjective and, in this case, hypothetical, given that we didn't have a discrete supercell traverse the area to prove whether it was sufficient or not (though I don't believe the lack of supercells is related nearly as much to slight-moderate instability as it is to storm mode). In addition, if I'm not mistaken, the area was in the right entrance region of the 250mb jet streak, so transverse circulation about the jet streak should have resulted in upper-level diverse and resultant vertical motion. While vertical motion has an effect on lapse rates, large-scale vertical motion associated with transverse circulations or typical DPVA is on the order of centimeters per second (cm/s), which is several orders of magnitude less than the 10s of meters per second typical of deep moist convection across the plains. This means that synoptic-scale vertical motion does not directly/substantially impact individual convective updrafts, though the indirect impact (subsidence causing weak lapse rates due to low/mid level warming) certainly can lead to decreased instability and resultant vertical motion of updrafts.
Tom said...
"I don't think the quantity of storms has anything to do whether storms will produce tornadoes or not. Check out this radar loop from 8/18/2005 when Wisconsin had a 27 tornado outbreak. Cells were RIGHT next to each other."

I'm going to have to disagree with you on that one. Tornadoes can and do occur with storms imbedded in a squall line or MCS, but most of the strong tornadoes are produced by discrete supercells. I would say that most tornadoes(not just strong ones) occur with discrete storms, but the higher frequency of squall lines/MCS type convection means there are more opportunities for tornadoes from this form of convection. There are several reasons why isolated storms have a better chance of producing tornadoes, but the main reason is that it simply doesn't have to compete for energy with other storms around it.
Tim G.-that radar loop is hourly, so it doesn't give you a complete picture of the complete evolution of the storms. The storms that produced the strongest tornadoes (Viola and Stoughton) may not have been completely discrete, but they were in a position where they didn't have to compete with other storms for inflow. Also the storms were moving in such a way that they didn't rain into each other's updrafts.



Keeping on topic, I would have to agree with the asessment that there were too many storms on Wednesday-and none of them could get uncontaminated inflow.
I made the decision not to chase in the morning after noting two things which I have learned do not equal chasing success in the NW TX/Oklahoma areas.

Factor #1 was definitely the synoptic set-up with bad timing of the ejecting jet max from the southwest (too late in the day) and the fact that the forcing mechanism was a Cold Front -- I'll take a dryline over that anyday. The SW OK area recovered nicely after the morning elevated garbage, so the added CAPE and significant convergence along the front didn't help sustain the cap across a broad area.

Factor #2 was the presence of that early morning elevated convection. This shifted the extreme instability pretty far south of the Red River in TX, this area was disjointed from the best shear as others have noted. 500 mb winds were much weaker (~30 kts) and low-level flow much more uni-directional.

On second thought, I don't know of many scenarios that are good for chasing in my first 12 months in this state haha. Everyone's luck is bound to change around here -- in 2006 sometime! I love days like yesterday where the shear is off the charts, but something just didn't feel right. Another $50 in my pocket for when I totally bust my next chase...have a great weekend everyone!
In regards to instability... Below is the 00z 15 Sept WRF initialization of CAPE...


The NAM and RUC initializations looked similar as well (as the NAM should, since I think the WRF takes its initialization off of the NAM, though I could be wrong). In fact, the RUC (from SPC Mesoanalysis, which I've learned to take with a grain of salt) showed the max CAPE to be in southern OK, decreasing with southward extent into TX.
Thanks for that graphic Jeff, the best CAPE axis set-up just to the west of the morning elevated convection and it aligns well with where the storms probably had the best chance of moving off the boundary and become surface based. Another thing that hasn't been discussed is LCL heights were disappointingly in the 1200-1500 m range across this area. If Tds could've been a bit higher that would've helped matters.

I'd also like to give props to the GFS for this event. I gave weather briefing on Wed. Sept. 7th and all the signals were there in that 12z run that Tues/Wed. would be very active across the central and southern Plains (a full 6-7 days in advance!). It truly is amazing the advancement of our numerical weather prediction models on the synoptic scale. Errr mesoscale, yea thats what we need to work on.
I don't think it was a lack of instability or shear. Even with southward extent into Texas shear was sufficent for organized severe cells and in SW OK at least moderate instability was juxtaposed on 60-70 kts Effective Shear and 200-400 m2/s2 0-3km SRH. I think the lapse rates/subsidence and a large spatial area reaching convective tempature played a key role as well as the storm motion generally being parallel to the boundary.
Originally posted by Jeff Snyder
I'm not sure what you mean by 500mb negative buoyancy.

Right. I suppose that is a bad use of terms. I mean't suppressed by the 500mb NVA which was showing on earlier NAM model runs for that entire area and that I posted under FCSTS. There was supposedly a shortwave coming into west Tx past the Tucamcari profiler but I don't think it made it presence known to help assist during the daylight event. Maybe that forecast NVA never materialized completely - not sure. Anyway generally more of a subsidence issue.

So, you were thinking it was under the 250mb right entrance? Hmm, would have to check on that as it would be interesting, but I was pretty sure it was under the 300mb right exit.
I see several mentions of LCL here, but lately I've become of the mind that the big players, given what overall appears to be a relatively supportive synoptic environment, are down low, such as LFC heights, low level CAPE, and 0-1 km shear. Jon Davies has done fairly significant research into these topics, and we here at IND have spent a lot of time, especially over the last year to 16 months since the May 30, 2004 outbreak, discussing these parameters and watching to see how these parameters come into play and evolve during tornado events in our area.

LCL, while it has been shown to have some correlation with tornadic supercells (lower being preferable, of course), really only tells you something about the boundary layer relative humidity, and nothing about the quality of low level parcel acceleration that may be critical in promoting tornadogenesis. The parameters that do have something to say about the location and strength of this good low level parcel acceleration/stretching appear to perhaps hold value operationally in determining an environment conducive to tornadogenesis, with a low LFC generally implying little to no low level CIN, etc, and thus strong vertical motion in the low levels.

Check out Jon Davies' excellent website at:
Another problem--basically relating to the "instability and shear not coincident" problem noted by others--was the narrow nature of the favorable corridor for tornadoes immediately north of the front. I think the tight gradient of CINH (south to north) may have been a problem for storms crossing the boundary. If isolated updrafts could have moved parallel near/north of the boundary I think they could have utilized both boundary layer instability and the decent boundary layer SRH values in place and maybe popped out a tor or two. I don't think LFC heights were a problem, since moisture was deep and rich and 700mb temps do-able at 11C or so.

[This is ignoring the "mess of too many updrafts" problem, which of course was a biggie.]
Originally posted by afischer
Another problem--basically relating to the \"instability and shear not coincident\" problem noted by others--was the narrow nature of the favorable corridor for tornadoes immediately north of the front. I think the tight gradient of CINH (south to north) may have been a problem for storms crossing the boundary.

Absolutely. The best boundaries for efficient tornado production are those which have modified enough that there is only a weak instability gradient across the boundary itself (read: temps and dewpoints hardly drop), allowing storms to remain essentially surface-based while drawing on the significant increase in SRH that the backed winds north of the boundary provide.