It is time to stop quivering in our boots in pointless fear of the future and just roll up our sleeves and build it.
- Ray Pierrehumbert

Wednesday, May 12, 2010

Nashville Flood: Dubious Analysis by Nielsen-Gammon

I'm wiped out. Had a wonderful weekend as co-organizer of a somewhat scaled back Scientific Software Days, conferring with Steve about the imminent future of science communication, hanging around with him and other friends arguing the future of climate science, and even having the privilege of interacting with Bill Stein over dinner, which was a treat. Plus I'm seriously starting an exercise program which for a man of my sphericity is no small task, and trying to spin up some new projects, and, yeah, attend to the tedious details of the day job.

Today I was worthless; you can always tell those days because there are eighteen items on my Reader feed. That means I was in "I'll just poke around a bit and see if I wake up enough to work" mode all day. Sigh.

Consequently not much time/energy for blogging this week. Read my feed if you're bored!.

I'd like to draw your attention to this article by John Nielsen-Gammon. It makes me uncomfortable and I think it's wrong, but it's not obviously wrong.

My first response was "um, that can't be right".

My second response was "perhaps the question is ill-posed" but I think it isn't. There's a serious actuarial question here. To be sure, given that there is only one Earth rather than thousands, it's not one that is easy to answer.

The question is essentially "How much more likely is an event on the scale of the recent Nashville flooding as a consequence of anthropogenic climate change?" John says not much, but his answer changes when you drop the "anthropogenic" term. In other words, he's saying that changes in severe flood events to date have been real climate changes but not anthropogenic climate changes.

So I'm back to "um, that can't be right." I'm pretty sure there's something more or less non-obviously wrong there, but what is it?


Paul said...

Change is change whether anthropogenic or Milankovic induced. The current rate of change seems unprecedented in human terms. Adding energy to the climate system inducing unexpected and really unpleasant changes would seem an obvious connection. Why is it such heresy to link human induced change to unpleasant specific events?

I know that Stoat asserts this is the least certain of areas in climate science and is ever at the ready to call to account those over selling the link in the peer reviewed literature. RP Jr. does the same thing but usually because they conflict with or don't sufficiently acknowledge his contributions to the field.

I think this is the most critical area for climate science to advance its case for real action to reduce carbon emissions. Disasters with people on roofs tend to focus the mind of at least the fence sitting Democrats. With them on board, some meaningful legislation could be passed in the next few months.

With enough disasters, even the Republicans might get the picture but I'll be long gone by then.

Paul Middents

S said...

I think his reasoning overall makes quite a bit of sense, but I think there's a problem in the details: it seems that he's talking about attribution of climate change in a smallish region to anthropogenic cause, but I'm by no means convinced that a more general attribution over a large region can applied to effects on a smaller scale.

PS. Best of luck on your exercise regime. It may surprise you to learn that there are strangers out there who have been worried about your BMI.

Frank Bi said...

Nielsen-Gammon is comparing

(a) the current climate with

(b) the climate that'll have resulted in some artificial universe where climate remains static.

However, what we really want to do is to compare (a) with

(c) the climate that'll have resulted in the presence of only climate changes due to natural influences.

There are probably a number of ways to analyze the data to do such a comparison, but surely confusing (a), (b) and (c) and capitalizing on the resulting confusion isn't one of them.

Arthur said...

On the exercise - my advice is to start slow but make it a habit, which means not hurting yourself along the way so you have to take a break! I was getting a little spherical a few years back, but managed to get into a very steady 3-day per week 30-minute exercise habit that I've kept up ever since. It's made a big difference!

On Nelsen-Gammon's post - I think the distinction he's making is between "climate change" and "global warming", not so much between anthropogenic and natural causes. I.e. the increase in rain in TN was much more due to a change in precipitation patterns ("climate change") than to higher moisture levels in the air due to warmer ocean ("global warming").

Which is why climate change is a better description of what we're doing than global warming is - we're changing the atmosphere and oceans, not just warming things up.

t.dr.k.e.s. said...

I haven't read the article yet, so this may be senseless, could the physical system possibly tell where the change came from?

t.dr.k.e.s. said...

Yeah, disregard my last comment--I've read it now. I think one problem is visible in this paragraph:

"Keep in mind that we're talking about the two-day rainfall total for Franklin, TN, here. If there are 10,000 long-term precipitation stations in the United States (not a bad guess), you should expect one of them, somewhere, to receive a 10,000 to 1 rain event on average once every year. [Update: a better way to say that would be that you should expect there to be a 1 in 10,000 rain event in the US recorded about once per year.] And you might expect a similar frequency of occurrence of unusual 1-day rainfall events, or 6-hour rainfall events, etc. So this sort of extremely unusual rainfall event is actually quite common, just not at any particular location."

He seems to be treating those 10,000 weather stations as statistically independent. Whereas a particular weather system could hit thousands of them as it moved across the country.

Andrew said...

I'm not sure about his causal analysis, but he assumes (for simplicity's sake) that the events are randomly distributed in time and space, when they're likely to be clustered, depending on the climate. My understanding is that the rare events are beginning to cluster here and now.

Aaron said...

People are not good at understanding change. Thus, I like to look back at Dr. Deming’s work.

Plot everything in terms of standard deviations. If you accumulate 6 “standard deviations” without the plot crossing the mean, then the system is unstable, and you cannot expect a regression to the mean. That is 1 SD for each of 6 units of time or 6 SD for unit of time or 2SD for 3 units of time. (When I say “stable” or “unstable” I am referring an equilibrium state of a system with multiple positive and negative feedbacks. If you accumulate 6 SD in one direction, the system is “lost” and it is not going to go back to where it was. )

Given the extremes of in geologic history, recent natural climate has been very stable.

Plotted in terms of standard deviations human greenhouse gas emissions are at more than 6 SD over some previous baseline. Atmospheric greenhouse concentrations are at more than 6 SD over some previous baseline. Thus, we can expect 1 in more than a million weather events (rain, drought, wind, temperature) over that past weather baseline. Noting that in many cases, these rare events will be barely perceptible to human senses. That is, a 40F high temperature in January may not seem that extreme because we have seen 100F highs in July. However, it is still a one in a million event with respect to the climate baseline for that area on that date in January. My favorite rare weather events that are occurring with increasing frequency are rains on Greenland and the High Arctic.

We predicted more severe weather, and it is here. This is one of the great confirmations of a scientific theory in all of history. The recent drought in the SE and the recent rains become our baseline weather/climate for current levels of greenhouse gas/ latent heat in the atmosphere.

The corollary is, that if we continue to emit greenhouse gases, such that the atmospheric concentrations continue to rise another 6 SD, then we can expect future weather events to be as extreme in comparison to the current weather, as the current weather is to the past weather. The problem is, that another 6 SD is much less of an increase in greenhouse gas than what most climate models predict or assume. And, we have not even started to talk about carbon feedbacks.

And, it puts the whole concept of engineering at risk, because we do not have a statistical basis to estimate future storm events, because the amount of heat driving the weather is constantly increasing and the entire latent heat/ weather system is wildly non-linear.

JohnMashey said...

Exercise is good.
Don't wait until you have a heart attack.
From personal experience, neither that nor a quad bypass are really fun experiences.
You'll also have more energy.

David B. Benson said...

In climatology, of whatever form, do not assume statistical independence.

Leads to serious error.

Kooiti MASUDA said...

We (climate scientists) are confident in attribution of global climate change. We are confident in the theoretical outlook of global warming because the results of simple one-dimensional models and the global average of the results of comprehensive (though not perfect) agree. There is still uncertainty in aerosol forcing (both cooling and warming) and behavior of clouds (responding to temperature and to aerosols). But the uncertainties are not likely to be large enough to change the order-of-magnitude of the greenhouse effect.

We are not so confident in attribution of local climate change. We have more candidates of causes. There may be local climate forcing. Also, movement of energy within the climate system may cause local climate change without changing the global average.

We are less confident in attribution of particular weather events. There are factors which we (human beings) have to regard as random. We can make a forward statement that how global warming changes the probability of the particular type of weather events. And we can use the statement i an inverse manner somewhat like attribution. But attempts of explicit attribution will always be speculative.

Understanding of global climate change per se is normal science. But the interface between it and the society itself involves assessment of local extreme events, which is aptly called "post-normal science" or "trans-science" even without political contention.

n-g said...

Hmm...clearly I don't read this blog frequently enough. I wonder whether you or your readers would find my additional analysis including spatial correlation any more credible?

To summarize, taking spatial correlation into account (albeit crudely) makes the expected US frequency of a 1 in 10,000 rainfall event a good bit rarer. One should occur at some COOP site or another about once every 2-3 years rather than once every year.

EliRabett said...

Eli thinks you are being optimistic