Tornadogenesis question

[Mike Smith]

Dan and Pat,

A couple of comments are in order, I think.

First, I am in favor of "warn on forecast" (WoF) when it can be done on a consistently accurate basis.

However, when high-level officials say (in December, 2008) we are "five years away" (now 3 years and 9 months away) from initiating WoF we are kidding ourselves if we think we will be able to do it on a consistently accurate basis by then. I fear that, because these expectations have been publicly set, we will roll out something "not ready for prime time" that will damage the both the quality and perception of quality in the storm warning system.

Second, while I am glad to hear we can simulate the Greensburg tornado, there are several things that I think need to be stated:

• It is one thing to replicate an event and an entirely different thing to forecast an event.
• Greensburg was sampled by a DOW in its early stages. Unless we are going to roll out national CASA or teams of DOWs deployed across the U.S., that type of very high resolution data will not be available for tornado warnings except in Oklahoma.

Again, I wish you and the entire field luck with improving forecasts and warnings. I just am skeptical we are as far along at achieving accurate and reliable short term forecasts as some have stated.

Mike

 

[Patrick Marsh]

However, when high-level officials say (in December, 2008) we are "five years away" (now 3 years and 9 months away) from initiating WoF we are kidding ourselves if we think we will be able to do it on a consistently accurate basis by then.

Emphasis added above.


Mike,

Thank you for your comments and well wishes.

If you are referencing what I believe you are (regarding the December 2008 meeting), I was there and don't recall anyone saying that in 5 years we will be predicting tornadogenesis on a "consistently accurate basis." I do remember hearing that operational benefits of the WoF process are around five years away. I think you said it best, we are "five years away from initiating WoF". The key there is initiating. WoF is a process, not simply an end result. In the next 3 years and 9 months portions of the process will most likely be done in real-time or very-near-realtime. This summer we will be evaluating very-near-realtime model-generated analysis of convective environments (I believe 200km by 200km domain) at a 1-km resolution. What it can do is provide a much better picture as to thunderstorm updraft strength and ambient conditions than a simple radar analysis could ever do. This is part of WoF, and a first generation product could very easily be ready for use in the next 3 years and 9 months. Although, it is not "consistently accurate" forecasts of tornadoes, it could very well prove beneficial in hail forecasting.

Again, everyone needs to think of WoF as a process, something we are working toward, not just an end goal. In this respect, the aforementioned timeline is very reasonable.

 

[Mike Smith]

Patrick,

I agree with what you just wrote.

But that is not what was said. I am not going to name the official but it was explicitly said that we would begin issuing WoF warnings in five years. I was shocked when he said it (and he said it a second time in his prepared presentation).

I think we have to be very careful about setting expectations.

The words "consistent" and "accurate" are mine and I believe they are key to successful public warnings.

Please don't interpret this as not being in favor of WoF. I am in favor of the process (i.e., a scientific learning process leading to better warnings) as I will be of the end result provided it is accurate and consistent and not rushed because someone set an artificial deadline.

Mike

 

[Dan Dawson]

Dan and Pat,

A couple of comments are in order, I think.

First, I am in favor of "warn on forecast" (WoF) when it can be done on a consistently accurate basis.

However, when high-level officials say (in December, 2008) we are "five years away" (now 3 years and 9 months away) from initiating WoF we are kidding ourselves if we think we will be able to do it on a consistently accurate basis by then. I fear that, because these expectations have been publicly set, we will roll out something "not ready for prime time" that will damage the both the quality and perception of quality in the storm warning system.

Second, while I am glad to hear we can simulate the Greensburg tornado, there are several things that I think need to be stated:

• It is one thing to replicate an event and an entirely different thing to forecast an event.
• Greensburg was sampled by a DOW in its early stages. Unless we are going to roll out national CASA or teams of DOWs deployed across the U.S., that type of very high resolution data will not be available for tornado warnings except in Oklahoma.

Again, I wish you and the entire field luck with improving forecasts and warnings. I just am skeptical we are as far along at achieving accurate and reliable short term forecasts as some have stated.

Mike

Sorry for the delay in my reply.

Thanks also for the well-wishes. I want to clarify a couple things. First, in my Greensburg experiments, I do not claim that I have even replicated the tornado, let alone forecasted it, only the larger mesocyclone-scale circulations, due to the relatively coarse resolution of the model (1-2 km grid spacing). I performed assimilation cycles, and subsequent forecasts, using only KDDC data and no other data except for an environmental proximity sounding (by the way, it wasn't a DOW that sampled the early stages, it was the UMASS X-Pol radar -- part of Howie Bluestein's team). The idea is to assimilate the radar data for an hour or so before the storm becomes tornadic, and then cut loose with a free forecast, *as if* I was doing it in real-time, with no further information going into the model, to see what happens. For all intents and purposes this is a true "forecast", even though it was done after the fact. In such a case, I was able to achieve, even with just the KDDC data, a decent ensemble "envelope" of forecast ground circulations that can be used as a proxy for the tornado track, even though I'm not actually resolving the tornado. However, Greensburg was a rare case of a tornadic storm that was well-resolved by even an 88D, so it should not be taken as generally applicable to all tornadic supercell cases.

This is the sort of step-by-step groundwork I and others are trying to lay on the road to the whole WoF paradigm. In the future, experiments like the one I outlined previously will certainly be run in real-time. Right now we are just beginning to explore the playing field. Having not heard the fellow, I can't comment on what the high-level official was saying, but it sounds like he may have been able to choose his words more carefully, or there may have been some overly optimistic exaggeration of how quickly we can get there. It's tough to predict these things, however.

 

[Paul Knightley]

I sometimes wonder whether it's worth taking a step back and starting with the initial question, how do small vortices, such as those going down a plug hole in a sink, start? Well, we know it's down to the conservation of angular momentum as the vorticity is stretched down the plug hole by gravity. Now, apply that to the atmosphere, and upwards! You could then get a non-supercell tornado under a growing cumulus or cumulonimbus - again, stretched vorticity, in this case by a stout updraught and decent low-level instability.

So what about supercells? Well, surely it still boils down to the conservation of angular momentum? We just need to find where that AM comes from, and what's stretching it! Of course, that's the tricky bit!

We have had a number of non-supercell tornadoes here in the UK which *appear* to form in association with strong rear-inflow jets/downdraughts in line-squall/LEWP-type set-ups. Instability is fairly meagre, and the rear-inflow is rather cold - but tornadoes do appear to have developed. It may be a rather mechanical method to tornadogenesis - i.e. the surging downdraught causes vortices to develop at each end, which can then get stretched by a nearby updraught, rather than there being so much of a baroclinic nature to the development of vorticity.

These lines/LEWPs are very interesting, as they can often be shallow (<5km deep), but produce quite strong tornadoes (EF2). Perhaps they contain very shallow mesocyclones, and could even be classed as shallow supercells, but the fact that the rear-inflow seems to be tied into with descending areas of dry air from aloft means this inflow seems to be quite cold, and one would imagine not easy to be lifted back up by the updraught. So perhaps the research should examine the generation of very powerful updraughts close to the surface which could lift and stretch non-buoyant air.

 

[Mike Smith]

Dan,

Now, I am the one apologizing for the slowness of my reply.

Your information is very helpful in understanding what you are doing and I congratulate you on the "simulated forecast" approach.

Please let me suggest taking a look at the situations with hooks and Doppler rotation that bust like the tornado warning issued on this one, http://meteorologicalmusings.blogspot.com/2010/04/tv-stations-dilemma.html. It is my observation that, over the last 30 years, we often write papers about the Superoutbreaks, the May 3rds, and the Greensburgs and we don't focus enough on the non-obvious busts. I am reasonably happy where we are at warning of "significant" (using Grazulis' definition) tornadoes. Now, I would like to turn more of our attention to reducing the FAR.

The other suggestion I would make is to not release this product "before it is ready for prime time." The warning system works very well now. Before we start a WoF program (which will be an extremely difficult education process that I still have major concerns about) for the public, it needs to be thoroughly tested and validated in many parts of the nation.

What works in Oklahoma likely will not work in Florida due to differences in the nature of tornadoes and demographics.

Best wishes in your work and that of your colleagues.

Mike

 

[Joel Wright]

If indeed the warmth/moisture content of the RFD is so instrumental in determining whether a supercell will produce, I'd be interested in what causes RFDs to vary in temp/moisture content.

 

[Mike Heitz]

Really in the markowski paper's he says that the factors that seem to control the amount of negative buoyancy in the RFD seem to be tied to the lcl's and Relative humidity. He argues that the higher the cloud base and the lower the relative humidity the more evaporative cooling that occurs, causing more negative buoyancy (cold pool stronger).

 

[tom hanlon]

I think that the science of tornadogenesis is more than adequate.

I think the problem lies with the definition of a tornado and our focus on the minutiae. The rather focused definition of a tornado leads to a more or less identical storm having a chance of being "tornadic" while in most respects being a very similar to a storm that is not tornadic. We focus on this rather meaningless difference.


We can predict a storm that might produce, the storm will be a strong storm, the storm will be rotating, the storm will produce severe weather. We can do this rather well, with a rather decent geographic focus. Yet we insist on focusing on whether it will produce a "tornado" or not.

Look at some of the "tornadoes" you have seen.. you see a wall cloud, you see rotation, then for 5 minutes it produces rotation at the ground, perhaps a descent but brief funnel, some brief minor debris. So for 5 minutes we have a tornado producing storm, but outside of that 5 minutes has the storm really changed? it is still rotating, it is still severe, individuals near the storm still need to be aware. For five minutes we have a tornado, but around that time in all other respects the storm is the same.

The city in the path of the storm needs to be aware, not because there might be something attached to that storm that when it passes through town might meet the technical definition of a tornado. The town needs to be aware because tornadoes are attached to violent storms. We can and do predict violent storms and we do this pretty well. The technical presence of a tornado is meaningless.

Our point of focus , the tornado, is dependent upon a collection of relatively narrow symptoms of an otherwise severe storm. This focus is more or less meaningless, more damage could be done by 60 mph straight line winds then the little spinup that is technically a tornado, but we focus on the tornado.



My point is a rotating column of air/fluid is a rotating column of air/fluid. The focus might be narrow and strong, broad and weak, broad and strong, broad and weak. In addition the rotation can be at the ground, near the ground or significantly above the ground. In one case we might have a tornado, if a bunch of other events come together.

Imagine if we treated hail in a similar manner, we focused on the minutiae. Suppose we considered a storm that did not produce hail over 2.5 inches as different class of storm than a storm that only produce hail of 2.0 inches. Suppose we treated the hail like we treat the rotating inflow. Would the science of "hailogenesis" really be useful? Would we be any better off getting our predictive abilities beyond "this storm might produce big hail" Suppose hail of 2.5 was fuji-hail scale H0 and hail of 3 inches was fuji-hail scale of H5. Does this make the small hail meaningless?

The science and this question goes on because we all love that little photogenic detail the tornado. But predicting what storm is going to produce a tornado and exactly when , where and how strong is silly. An identical storm that does not produce is going to have very similar effects and outcomes to society as the storm that is also severe that does produce.

The tornado is a potential effect of a strong storm. Focusing on this effect is like focussing on the severity of a particular symptom of an illness rather than the illness itself.

 

[rdale]

An identical storm that does not produce is going to have very similar effects and outcomes to society as the storm that is also severe that does produce.

But as you mentioned - we already cover those types of storms. Unfortunately we're nowhere near perfect with them -- otherwise every SVR with "60mph winds are expected" would come true and that's far from the case.

 

[Paul Knightley]

...but not every rotating column is the same! Why, for example, did the 'Greensburg' storm produce several tornadoes well over 1 mile wide, which are extremely destructive, and other storms that evening did not? If we don't try to understand tornadogenesis more, we will never be able to answer these questions. The impacts are not the same!