"Our greatest responsibility is to be good ancestors."

-Jonas Salk

Wednesday, May 4, 2011

Black Birds

I find that the whole way this tornado discussion is going is disturbingly missing the point.

There is no compelling evidence that tornados are becoming more severe or more common, or for that matter, less severe or less common. There is no compelling modeling evidence that amounts to a meaningful prediction either way. So what should you say when you get hit by a tornado? You say "Dang! Shoulda moved to California when I had the chance." At least until the earthquake comes.

Does that mean that climate change had nothing to do with your tornado? Of course not.

This is a twenty-first century weather event. Everything is in a substantially changed and substantially changing context. Does it mean your bad fortune is "attributable" to climate change? Not in any reasonable sense of attribution. Unless, that is...

Unless your tornado happened to be part of an extraordinary subcontinental-scale tornado outbreak.

Now, this sort of outbreak event is not entirely unprecedented either, though it seems to be emerging as the single most severe instance in the satellite era.

But it's a, forgive me, black swan of a sort. One thing about really really severe events is that you can't do statistics on them. They are too rare to form a large enough collection to draw conclusions.

But you may have a flock of shouldn't-be-black stuff. Black doves. Black, um, pelicans. Black seagulls. Black other stuff. I'm not much on this bird business, but the thing is, although you know these things exist, and you know they are too rare to extract a trend, you shouldn't be seeing any of them very much.

Hot summers in Moscow. Year after year of flooding on the Mississippi. Huge tornado outbreaks. At what point do we get to look at the collection of weird events and say something is going on?

Let me admit, first of all, that there are all sorts of statistical warning signs associated with this question. Selection bias, observer bias, post hoc definitions. Our intuitions may well be misleading us.

On the other hand, there is the question of rolling a thirteen. The more we disturb the climate, the more excursions it will make into unfamiliar territory. As we perturb the climate, as it wobbles around more and more, it will more and more often hit these weird peaks.

Perhaps local events like tornados will show no trend, but tornado outbreaks, when they happen, will be more severe. The ocean, after all, is further from equilibrated with the atmosphere and with space. Air masses will encounter each other in unfamiliar ways. Perhaps strange things will happen. Perhaps (and I am not being Eli-style coy here, perhaps not) they already are happening.

The fact that the trend in tornados or strong tornados is too noisy to say very much doesn't change the fact that we just had an epic event, and that just about everywhere seems to be in a run of anomalous seasons. Are we enjoined from thinking about this for some reason? Maybe it's a misimpression, but maybe not.

Coupled GCM runs are expensive and not really mature. Atmosphere-only GCMs give us quasi-equilibrium responses and the atmosphere alone is strongly damped. But the bulk of the energy is in the ocean, which is a dynamic system with all sorts of bounce in it. What happens when you kick the ocean? You get wobbles on a decadal time scale. And what does this do to the atmosphere? Well, you're smarter than I am. You tell me.

The thing is, I see lots of people, some who ought to know better, buying into this sort of thinking, this instance from Revkin quoted by Kloor at the link above, emphasis added:
But that’s a meaningless assertion without asking whether there is evidence of a meaningful influence — meaning enough of a nudge to the atmosphere that the contribution from greenhouse gases is relevant to policy and personal choices, in this case in tornado zones.
No, no, no! The idea is not: radiative imbalance causes warming everywhere which nudges the climate a bit. So many people think like this but it is wrong. Radiative imbalance pushes the climate around so much that eventually it changes enough to restore the balance. Almost surely surface warming is part of that. But a lot else is part of it.

In my opinion, people have it backwards, though, as to which is the nudger and which the nudgee. Forced climate change causes global warming as a response, not the other way around. And a lot else changes too. Talking about "nudges" is so massively wrong. The global mean surface temperature is very stable. It takes a hell of a kick to move global temperature as much as we are moving it, and the climate system is responding directly to the kick, not to the temperature change.

So what I think we should expect is changing climate on the time scale of the memory of the upper ocean, roughly a decade. And this process will accelerate over the lifetimes of anyone now living. The whole concept of "climate" weakens under the circumstances. And the more we disrupt the system on a large scale, particularly the more carbon we spew, the more and the bigger the black birds we will see.

This is why I'm baffled by arguments that start from the slightly more energetic and moist air columns. That would make sense if all else were equal, in other words, on nudge-world. But there is no reason whatsoever that I can see to expect all else to be equal.


Nick Barnes said...

In other words, stationarity is dead?

"Turning and turning in the widening gyre
The falcon cannot hear the falconer;
Things fall apart; the centre cannot hold;
Mere anarchy is loosed upon the world,
The blood-dimmed tide is loosed, and everywhere
The ceremony of innocence is drowned;
The best lack all conviction, while the worst
Are full of passionate intensity."

Unknown said...

Wikipedia's list of North American tornado outbreaks shows a distinct trend in outbreaks (eyeball method). Why can not a statistical test be applied?

Jim Bouldin said...

No offense, but I'm having a helluva time following your argument(s) here. You're making this way too convoluted.

For possible trends up to the present, this can at least potentially be addressed with the 20th C Reanalysis (v2), looking at trends in favorable large-scale weather conditions, perhaps in combination with smaller scale atmospheric dynamics models that help predict actual tornado formation.

For trends moving into the future it's simply a matter of developing a better observing system, as it is with everything.

Adam said...

I admit to profound confusion, MT.

This sentence in particular makes me feel I'm walking a Möbius strip:

It takes a hell of a kick to move global temperature as much as we are moving it, and the climate system is responding to the directly to the kick, not to the temperature change.

Surely you are not saying the initiating factor of change is not the increased retention of energy (warming) in the atmosphere. Forgive me, but I am unable to see why the temperature change is not the kick. Perhaps you would be so kind as to put it another way.


Michael Tobis said...

Jim, a research program such as you suggest may well be possible, but I don;t think it would be easy. I think the mesoscale conditions that lead to tornados are understood, but how well understood are the large scale conditions that lead to such mesoscale conditions? Is that resolved in the reanalyses? Correctly so? Then, how do you draw a threshhold between potentially tornadic situations and otherwise?

Embedding mesoscale models into reanalyses is possible, but difficult and expensive. Which locations would you run?

I am not saying there is nothing there, but I don't feel compelled to sit down and start coding, either.

However, this is all beside my point. My point is precisely that while a positive result tells you something, a null result tells you very little. There could be a very strong anthropogenic component to the signal, perhaps even going back as far as the satellite records do, without that signal being monotonic. There is no guarantee that we will see a TREND in any particular phenomenon of interest.

An analogy: if you have a precise radar-based measure of the altitude of an ascending airplane, and conditions are laminar, you will be able to tell instantly that the plane is rising. If the plane enters turbulence, the instantaneous measure may show a decline about half the time even if the plane is ascending. You will have to watch for a long time to average out the wiggles.

In a perturbed climate, the wiggles are on the order of a decade long. So to extract a meaningful trend even from ordinary events you will have to watch for more than a decade. But in event classes that are intrinsically rare you will have an even harder time, and in events that are extraordinary you still will get a null result, because you don't have enough samples.

So Nick's summary is exactly right. The whole way we habitually think about climate is based on an approximation that it is near enough stationary that we can do statistics on it. But we already know this is sort of fake; that is why we often look back thirty years, long enough to get statistics but not long enough to get the big wobbles messing us up.

I am suggesting 1) as the climate is more strongly forced, the thirty year period gets to be too long and 2) if truly bizarre events like this April's tornados are considered, the thirty year period is already too short for obtaining a background repeat interval and 3) while the global climate is monotonically heated, local patterns may go up and down even more than in the holocene. All of this means null statistical results, at least in the usual tests, but it does NOT mean that the climate forcing is absent.

What we are doing intuitively, though, is adding up all the black swans in our minds and saying "uh-oh". This is a biased procedure, though. What I'm asking for is a way to turn this into something objective enough that we can get around the fact that our intuitions are making a paradoxical claim, i.e., that "extremely rare events are becoming more common". And the reason I'm asking for it is because I think that claim is actually very likely.

Michael Tobis said...

Adam, there is a difference between heating and warming; temperature is not energy. It is essentially the core claim of Spencer and Lindzen, the only naysayers with any clue about physics, that a small amount of extra energy simply goes into raising the albedo, and so there is a negative feedback and we live happily ever after.

There is a less benign scenario. Suppose the extra energy goes into melting ice so quickly that deep water formation is suppressed because the upper ocean becomes salt-depleted and thus lighter. This means that the ocean surface becomes colder, which in turn means that the atmosphere becomes colder. The radiative imbalance does not go away; indeed it becomes more severe. The whole system becomes more energetic (less ice, more water) but the atmosphere becomes colder. And you get a massive cooling glitch. This is still the dominant theory of the Younger Dryas. It is now believed that there is not enough ice left to allow such an event to recur, but there is some lingering question as to whether the models capture the small scale dynamics well enough to be sure of this.

Dol said...

Having just spent a good solid 5 minutes on the interwebs, I've concluded that MT needs to write us non-physicists a primer in physics-language use and how we should be using it to think about climate. It appears that heat is a rate of energy transfer - didn't know that! So when you're saying 'warming' you mean a temperature change, not a thermal energy transfer? (Temperature, it seems, only refers to systems that can be 'treated as' in equilibrium already, since otherwise there'd be heat transfer.)

Uuuurgh. I'll shut up now before I embarrass myself any more with further wikipedia-level understanding... But just to second Adam: confusing!

Michael Tobis said...

"heat" is or a form of "energy". "heatING" is a rate of change of energy, or a form of "power".

The above is undergraduate level physics and is clear.

Nevertheless, please note that I am sticking my neck out in this discussion. Don't take my argument for consensus. I am actually questioning consensus here.

zumone said...

The answer I believe is spatial not temporal, that is, not in their frequency but *where* the tornadoes are occurring. Look beyond tornado alley and the climate signal is clearly evident - here in New York State for instance there has been a distinct up-tick in the frequency, intensity of severe weather including tornadoes over the last 30 years. Certainly in the mid-west and south the climate forcing signal is lost in the noise, but look to the fringes and the evidence is clear.

Michael Tobis said...

(re my previous comment)

...or would be clear if I didn't have an extra "or" in there...

Jim Bouldin said...

Michael I have the vague feeling that we have some serious differences in assumptions.

Stationarity is dead yes, because well, it was never alive to begin with, which is to say, it's not relevant. The whole point of a statistical test is to determine the likelihood that your observations arose from different processes, and hence from two different distributions, which therefore, by definition, are non-stationary.

Yes, I of course agree that with any rare event you have to observe over a longer time period to get the same level of statistical confidence in possible changes, as you would with a more common event. Therefore, this problem has to be approached in a classical maximum likelihood (loosely speaking) approach, that is, by evaluating the relative likelihoods that different possible physical models would produce the observed results.

The only way I can imagine to do this, is to link various models together, whether they be statistical or mechanistic, that relate the relevant physical processes involved, at least from the scale of individual tornadoes up to the size of the frontal systems that give rise to them. This wouldn't necessarily have to be tied to increased GHG forcings at the global scale, although it would be best if it could be. Just documenting evidence for increased frequencies of potential "tornado forming" weather would at least give evidence of whether there is a trend in possibility or not. And the best source of data for having a shot at doing that, as far as 'm aware of, is the NOAA-CIRES 20th C Reanalysis, V2, because of it's relatively fine spatial scale, many vertical levels and numerous estimated meteorological parameters.

However, I don't claim to be an expert in this area either.

Michael Tobis said...

Jim, yes, but we don't have much indication of a trend in tornados.

I am postulating a trend in extreme tornado outbreaks. This is harder still; a single reanalysis would not have enough instances, and the pure models are not fine enough to capture the behavior.

The point is this: that the result is inconclusive is not a refutation. It is merely inconclusive.

But we perceive something happening at a higher level of abstraction: "weird weather". However, as scientists, we acknowledge that this perception is suspect. Is there some way of turning that perception into science? I think there might be.

Dol said...

I didn't do undergrad physics so it's unclear to me, sorry. Playing catch up. Having given this a read, I'm getting that the extra energy trapped by GHGs is heat energy, and that temperature doesn't need to change - but it pretty much immediately does, doesn't it? Extra photons leaving the ground, being absorbed, don't get much more than a few metres on average. The air/upper ocean isn't changing state: average kinetic energy is going up/temperature is increasing. So: "forced climate change causes global warming as a response, not the other way around" - I still don't get. Heat energy trapped = re-absorbed IR = extra-wigglin' atoms = almost all direct temperature change, very little state change = after the actual GHGs, the first causal link. So warming causes climate shifts: heat transfer in the system moving back towards equilibrium = what's gonna get us... What on Earth am I getting wrong? Is it just that the temperature change isn't the point, but the variability of that change and how re-equilibriating interacts with the climate system?

Perhaps this kind of article isn't aimed at non-physics folk like me, but I'm vexed I seem to be having so much trouble with this. Quite understand if you don't want to play physics tutor though.

Jim Bouldin said...

Then it seems that the point boils down to the need for multivariate, rather than univariate, statistical analysis to evaluate how unusual the various events over the last couple of years are.

Also, I think I'm implictly thinking more in terms of severe weather (e.g. severe thunderstorms and worse) than tornadoes per se, which may well make a big difference in how to approach the problem. And I think I'm doing so because of the issues that you raise regarding observer bias in the observational record, a problem which is obviated by resorting to reanalysis data and limiting oneself to whatever events such data can reasonably track.

Anna Haynes said...

Michael, who can talk about this stuff as well as you write about it?

Michael Tobis said...

Anna, aw shucks. I don't know, though, I'm kind of out on a limb with this one and can't imagine who I'd want to carry this particular ball for me.

I'll talk to an audience any time, though.

Anna Haynes said...

Thanks MT. But what heuristic could we provide to listeners, to tell them that you're not just some thinktank (or other) doofus chosen to lead them astray?

(the one thing they need, more than anything, being a good set of heuristics, IMO)

susan said...

As a complete non-physicist (with plenty of willing physicists to help me out and a good strong belief that science pretty much works and is all we've got) I struggle with all this. But having started looking at water vapor maps and watching them closely over the recent years, I'm seeing a repetitive and developing pattern which I'm not sure you're talking about (those narrow columns would be local?). I see some turbulence in the Pacific somewhere around Hawaii that travels in a beautiful and fascinating pattern, whorls, curlicues, etc. to the northeast. It comes in as the pineapple express, or crosses further to the south, and gets fierce in the midwest. It then explodes and we get floods and storms, eventually, on the east coast, after they're wreaked havoc in the midwest. In summer, of course, we have a whole other system coming off the coast of Africa.

This is where I get the idea that energy has increased and with more water vapor we're getting more extremes. I also get that contrast leads to things like Texas wildfires. But at this point most scientists would probably say I'm off the reservation. My only advantage is that as an artist interested in energy and patterns I have a good eye for the big picture). Even in the short time I've been watching the whole system has gotten more intense.

love it or hate it, it's beautiful:
(h/t Tenney Naumer)

Meanwhile, in the arena of complexity, vast swathes of different information are pouring in. Agulhas current. Methane. Fires in Scotland. What a grab-bag.

Michael Tobis said...

Susan, if you have perceived a specific change in a specific pattern it would probably be possible to verify it from the reanalyses. But you would need to be a bit more precise about what you see and what is changing about it.

When I say "column" I usually mean a vertical column from the surface to space at a particular latitude and longitude point, or in practice, above a particular area on the surface that is small compared to the general circulation.

Michael Tobis said...

Anna, a Ph.D. in atmospheric and oceanic sciences from Wisconsin is what, chopped liver?

Anna Haynes said...

well, yeah, there is that...

susan said...

dam' rejected me again and I didn't save. See if I can make it shorter, anyway. I notice if I have something big running in another window that's a no-no.

I'm talking about big repetitive water vapor patterns that when they get very white (I'm told this is height which can be several kilometers) are reliable predictors in my very small (couple years) observation span. They run from Pacific to Atlantic (Newfoundland or mid-Atlantic) and sometimes they loop south or north. Right now there's an obvious bit over Montana/Alberta/Saskatchewan you can see for yourself:

This is the bigger one I cited before:

Reanalysis won't work for me here; this is pretty hands-on but I ventured because I thought you might find it interesting.

Anna Haynes said...

> "...trends in favorable large-scale weather conditions..."

Changes in severe thunderstorm environment frequency during the 21st century caused by anthropogenically enhanced global radiative forcing. Robert J. Trapp et al, 2007, PNAS
("We use global climate models and a high-resolution regional climate model to examine the larger-scale...meteorological conditions that foster severe thunderstorm formation....[and] find a net increase during the late 21st century in the number of days in which these severe thunderstorm environmental conditions (NDSEV) occur.")

Jim Bouldin said...

exactly the approach I am advocating Anna, except that you now (hopefully) have an (hopefully) improved reanalysis data set to use in evaluation

Unknown said...

Dan Olner, I may or may not be able to clear anything up.

At this point you might appreciate the concept of internal energy. Any substance (say 1 gram of it at a certain temperature, in a certain state) requires a certain amount of energy to exist in that state at that temperature. The gram must have acquired that energy in its history. This is related to specific heat, but the gram may have changed state in its history, or something else like pressure may have changed its specific heat. The famous example is that the internal energy of graphite is different than that of diamond.

Global warming: around 93 % of the added energy is in the oceans. This comes back to us via changes of state (more water vapor in the atmosphere with its latent heat (internal energy)) that is released (becomes regular or sensible heat, or wind) in various circumstances, also in ice melting, and also in changes in ocean currents (a different sort of change of state), not to mention coral bleaching.

Michael, you must make the invisible visible. I read that somewhere ;)

Dol said...

Pete, thanks a lot for that. So actually the vast majority of climate change *is* state-change related, via the ocean - interesting. I knew most of the heat energy was going in the ocean - I didn't know most of the effects we need to worry about were so immediately coming back from there too. I think I get the way MT's pointing the causal arrow now, cheers. Phew.

susan said...

I see I didn't read properly. Will keep looking at those water vapor patterns and see if I can be more specific about the changes over time that I am seeing. It has been a while, but I think I'm getting the hang of it. Mostly it's just a question of being more powerful and storms being more frequent. Presumably you science guys have better ways of measuring and keeping track of this potent fact. I get distracted when deniers start calling me insane and a liar for pointing out what is all around us, and how obvious it is. This line of talk, I think, scares them more than the science, which they are better at obscuring since it requires so much skill and open-mindedness to understand.

Steve Scolnik said...

And the hits keep coming:
Second Mississippi April 27 Tornado Upgraded to EF-5
Amazing damage from this one, and a new intensity record for the state. (Hint: You don't need urban/suburban impacts to detect an EF-5.)

Dean said...

You say

"just about everywhere seems to be in a run of anomalous seasons."

So where does the Climate Extremes Index fit into this? Looks like it doesn't include tornadoes though.