"Our greatest responsibility is to be good ancestors."

-Jonas Salk

Tuesday, September 13, 2011

Nielsen-Gammon vs the New Normal

John Nielsen-Gammon, our Texas State climatologist came up with this scary image that most of you have seen, and that everyone in Texas ought to take a good long look at. I was one of the first to reproduce it, but I've seen in lots of places since, and with good reason.
Unfortunately, John has followed up on this contribution with what I consider a mistaken article, wherein he claims that

"Texas would probably have broken the all-time record for summer temperatures this year even without global warming."

Before we get into his argument and its drawbacks, let's note the obvious. We see that this years drought/heatwave is far outside the observed pattern distribution of events.

It's hard not to take note of the tremendous similarity of the situation here to Australia's a couple of years back, though Australia's droughts are on the other end of the El Nino seesaw. (Australia has, in fact, been extraordinarily wet of late.)

So is it a "New Normal"? Is Texas in perpetual drought now? Will we swing back and forth out of unheard-of droughts and unheard-of floods? Will Australia do the same, along with other parts of the formerly semi-arid subtropics? Certainly this is the intuitive impression that many of us come away with. Barry Brooks is no amateur, and he was at least willing to quote a colleague saying
“Given that this was the hottest day on record on top of the driest start to a year on record on top of the longest driest drought on record on top of the hottest drought on record the implications are clear...

It is clear to me that climate change is now becoming such a strong contributor to these hitherto unimaginable events that the language starts to change from one of “climate change increased the chances of an event” to “without climate change this event could not have occured”.
Clearly, we can say similar things in Texas this year. But should we? Nielsen-Gammon says we shouldn't.

Let me summarize his argument:
  • The temperature anomaly this summer is about 5.4 F
  • Global warming to date has led to a local warming of Texas summertime temperatures of 0.5 F, so the temperature anomaly can be divided into 0.5 F background warming + 4.9 F other warming.
  • There is a strong correlation between annual rainfall variation and annual temperature in the graph. N-G finds a second order curve that fits the data [about as well as the linear fit] (see comments), and figures that the low rainfall could account for most of the remaining 4.9 F
  • He sinks into the tea-leaf territory of the "AMO" and claims to pick up the balance
  • Leaving aside the odd idea of superposition of temperature anomalies and the very weak evidence for the AMO, clearly there is a plausible claim that the huge temperature anomaly is "mostly" "because of" the drought
  • There is no obvious trend in Texas toward drought, so climate change does not cause unheard-of droughts
  • Therefore this is a fluke and has nothing to do with climate, or that other fluke in Australia in '09, or all the other flukes we have been seeing lately
I don't buy this for a minute. I am, in fact, shocked by the seriousness with which this argument is being taken.

It is interesting that when I have run this by non-experts, they all think it is crazy. Are they right?

I've been struggling for an analogy, and have come up with nothing resembling a realistic real-world example that allows this fallacy, so allow me a parable instead.

There is a small, isolated urban country where the wild fauna have been eliminated, and the public is only familiar with pets: dogs, cats, hamsters, and a few horses used in ceremonial events. The entire population knows very little about other animals, and even the experts have acrimonious debates based on fossil records and old paintings, just as we are familiar with contemporary climate but have to extrapolate to ancient or future climates.

One day, there is an earthquake. Not only does a border fence fail, but that fence abuts on the neighboring country's great zoo. Many animals escape into our urban country, and it happens that one of them is an elephant, which they will perceive only as a large, bizarre animal.

However, our experts have been observing the zoo from a distance. They believe they have a good idea of the number of animals at the zoo from the number of feeding stations (visible from an observation tower), and a good idea of the total mass of the animals (calculated from the size of the food deliveries). They conclude that the average zoo animal is the size of a large dog. Therefore, the elephant is not an escapee from the zoo! It must be an extremely unusual dog or cat.

That's the best I can do. It makes no sense, does it? We have an invasion of phenomena which we have only weak characterization for. We have some idea of averages and trends because of physical constraints, but we know very little of the nature of outliers in the changing climate.

(This is to say nothing of anomalies due to transient climates for the present.)

Here is the thing. We are increasingly disturbing the climate. A truly bizarre season occurs in a particular place. Either these extraordinary events are connected, which is perhaps unlikely, or they are unconnected, which is extremely unlikely. That is, you are asking for a bizarre coincidence.

But now we add up the number of bizarre coincidences, for each of which John can make comparable arguments. The tornado outbreak this spring. The huge blocking event in Asia last summer which did so much damage in central Russia, Pakistan, and parts of China. The fires in Australia in 2009 and the floods this year. The floods in the midwest. Heat waves in Europe.

None of these are clearly part of local trends. None of these are particularly predicted in the literature, and as far as I know the GCMs don't indicate these things happening.

But, here's the thing. They are happening.

So when I look at John's plot, I see that there are only two possibilities. First, a bizarre coincidence as John suggests: a gigantic grey housecat with big teeth, floppy ears, enormous legs, and a strange nose. Second, an unexpected consequence of climate forcing. An elephant.

That is, what we have is not because of a change in the mean but because of a spreading, an expansion of the cloud of possibilities. From a dynamics perspective, that's not surprising in the least. We're passing, year by year, from one climate configuration to another at a very rapid pace, and we are used to thousands of years of unusual stability.

Does anyone actually expect "global weirding"? Well, I am not sure how we should specify an a priori metric for it, and without one we can't really formally detect it. And the models, well, we already know that the non-assimilating GCMs are very stingy with extreme events. Why? My theories on that are too vague for publication, but it's widely known to be true.

But when I see a graph like that one, I don't find myself saying, hmm, obviously not part of the trend, therefore natural.

Now my other analogy is emotionally fraught, and let me apologize if it offends anyone, but I have to say it. When I saw a couple of 100 story buildings falling down, I didn't say, there's no anthropogenic trend to date for buildings to fall down so they must have fallen down naturally.

I'm sorry, but I find that argument, ahem, less than compelling.

Let me offer a couple of simple propositions instead.
  • There's a first time for everything.
  • If you push something hard enough it will fall over.
I for one think the fan is no longer pristine.

I am working on a longer version of this article. It also ties into the less obvious but very similarly wrong arguments about rainfall anomalies and also to a case that the egregious Pat Michaels has been flogging.

I think it's time to take this bull by the horns. You can't apply small-signal arguments to large signals in nonlinear systems. So please stop it.

Update: Via Google Plus, Jonathan Abbey summarizes my argument nicely:
Climate characterizes the statistics of weather and the statistical bounds of weather. If we start seeing weather patterns change, that can indicate a change in climate.

The question is all about how likely it is that this weather would occur if the statistical parameters of the climate were held fixed as it has been since instrumental records began, say.

If weather like this is sufficiently unlikely under our previous understanding of regional climate, it may be (a piece of) evidence that the climate is itself experiencing a dislocation.

Which is sort of interesting.

44 comments:

Pangolin said...

So let us just say that we set aside the heat records. The consecutive months of "hottest average monthly" temperature. The extreme length and severity of the drought. Set all of that aside.

How do you explain the rise in overnight lows in a low moisture regime? That should be dry air above Texas and Oklahoma and nighttime temps should drop precipitously as they do in most deserts. Instead nighttime temps stayed very warm to set new records for highest low temperature.

Now, I could be wrong and the air could have been humid but I doubt it. Not for six months with little or no rain.

The rise in overnight low temps. should be like a banging gong telling us that this is Climate Change in action. The heat simply cannot escape like it has in the past.

Steve Scolnik said...

The non-linearity is key to the analysis. A more energetic system is by nature more variable as well as having a different mean.

dbostrom said...

I read Nielsen-Gammon's account with great interest. It's comprehensive, which is great, but I was struck a little by the number of assumptions it stacked up all pretty much in the direction of natural variation. I'm also struck by the fact that I'm not qualified to rate Dr. Nielsen-Gammon's judgement. :)

Presumably the correct weighting of the individual components involved each sit on a continuum. I'll hazard a guess that much deeper scrutiny will be needed to really tease out where each influence fits, along the lines of what's been done w/floods in England, with Dr. N-G's take intended as a first pass. A better considered answer might well be quite different.

dbostrom said...

Reading Thomas Friedman's latest column "Is It Weird Enough Yet?", I immediately thought to myself that Dr. Nielsen-Gammon's analysis will provide handy plugs for deniers to stopper their ears. Yet, consider that when and if they do so, they'll encounter a cognitive short-circuit acknowledging that some anthropogenic warming is making itself felt. After all, N-G does not exculpate AGW; the anthropogenic forcing signal is woven through his narrative.

William M. Connolley said...

Sometimes, unusual events just happen. It is very hard to know.

> Is Texas in perpetual drought now?

Dunno. Did you mean very much by that question - it sounds leading. If you mean "Texas is", then you expect a repeat next year, and should say so. What do you expect to happen next year?

Steve Bloom said...

Thanks for the cite to Fudd's First Law of Opposition. It's been a while. :)

Dol said...

"a gigantic grey housecat with big teeth, floppy ears, enormous legs, and a strange nose"...

I want one!

Bart Verheggen said...

Michael,

It sound like the main disagreement between John’s and your take is that you chose to look at this event in combination with other weather weirding at different places in the world, and John choses to look at this event in isolation. John looks at various indices of change in order to tray and attribute this one event to possible causes. Your argument sounds more like “this is too implausible to have occurred largely by chance”. Intuitively convincing perhaps, but not quite “hard” enough.

John’s argument appears pretty solid to me (except some nitpicks on his attribution of the warming signal into a natural and GHG component, see my latest comment at his post):
1. Both a physical and mathematical picture of the relation between drought and temperature.
2. absence of a forced trend in precipitation/drought
3. He asks the question “How much of the heat is accounted for by the drought? ”
And answers by attributing a fraction of the heat to the drought (which occurred by chance as per 2), based on the observed relation between the two.
4. From 3 he conludes that ¾ of the excess heat is due to the drought and ¼ is due to the warming trend, which happens to coincide nicely with the observed average warming in Texas.

The crucial steps 3 and 4 in this argument rest on the assumption that drought is the cause and temperature is the effect. Whereas he writes that they influence each other in both directions (positive feedback): “During summertime in Texas, temperature and precipitation are closely related, because dry soil (caused by previous lack of precipitation) leads to low dewpoints (and high temperatures), making it difficult for thunderstorms to form, thus leading to less precipitation, more dry soil, etc. (Myoung and Nielsen-Gammon 2010). ”
So the reverse question is also relevant: “How much of the drought is accounted for by the heat?” Since there is a clear trend in temperature this would presumably lead to a very different conclusion as to the question of natural variability vs anthropogenically forced. A more complete analysis ought to include that I’d say.

Anonymous said...

In that larger picture, even if John is 100% correct, what we've seen is that these small changes in temperature -- that GHG's add on --have devastating effects on current society. Therefore we should have a better understanding that we are very far from being able to deal with what we are in for. And isn't that the entire point of this whole issue? And won't this become more and more apparent with each new exchange of energy? Is there another argument that forces our hand on (at least) large -scale adaption and infrastructural changes. Then, when we realize how we f-ed it up so badly, maybe we can ask the Revkin's of the world why demanding future targets are so wrong-headed?

Brian Brademeyer said...

>>>> "When I saw a couple of 100 story buildings falling down, I didn't say, there's no anthropogenic trend to date for buildings to fall down so they must have fallen down naturally."

Wouldn't the more scientific proposition have been:

When I saw a couple of 100 story buildings falling down, I didn't say there's no trend to date for buildings to fall down naturally, so they must have been brought down anthropogenically (for which we have a wealth of previous evidence)?

n-g said...

Technical correction to "N-G finds a second order curve that fits the data about as well as the linear fit...":

The original graph was also a second-order fit. I simply used a different (and, I argued) better predictor (May-August precip rather than June-August precip). And, of course, a second-order fit will always be a closer fit than (or at worst, equally close as) a first-order fit. It won't necessarily be better, because it might be overfit, but it will be closer.

What, exactly is your argument? That unusual weather never happened before? That bizarre (let's say 1 in 500 year) events cannot happen unless the climate changes? That all bizarre events have become more likely with climate change? That the total number of bizarre events should increase with climate change?

I think the only defensible argument is encapsulated in the last sentence. I'll respond to it if/when you confirm that that is, in fact, your argument.

Michael Tobis said...

Of all the times for Blogger to eat a comment! Aargh.

n-g said...

I didn't know Blogger had a default comment logic checker! ;)

Michael Tobis said...

Short version: Tom Friedman and Al Gore believe, along with many if not most people, that weather is getting weirder and that anthropogenic forcing is responsible. Is their position scientifically defensible?

I think that it is quite plausible that it is defensible. I do not think John's approach is a good test. To the contrary, the weirder the event, the more likely it is to reject the hypothesis that the dominant change in the environment is the dominant factor. Perhaps that is why Pat Michaels uses it. But that is the main reason I think we ought to reject it.

If anything it is a sort of best-case analysis. It seems fundamentally flawed as a test.

I find several flaws with the details of the analysis but the core problem is that it's structured in such a way as to attribute as little as possible to anthropogenic climate change.

More to follow.

n-g said...

I agree that the weather is getting weirder, though I'd prefer a statement that's actually testable. It's way too fuzzy to be useful, except as a rhetorical technique that plays into people's tendency to assume that unusual events must have a cause that's different from ordinary events.

Surely some weird events will become more common. Others will become less common. Some will stay about the same. On purely statistical grounds, if the climate changes substantially the overall number of weird events will increase.

How do you decide whether a particular weird event has been made more or less likely because of global warming?

In my analysis, I couldn't find any justification for assuming that the lack of rainfall should be worse because of global warming. Temperatures are warmer because of global warming, so this particular event seems to have been made more likely. It's hard to say how much because the answer's sensitive to the apportionment of high temperatures to the lack of rainfall, but I'll proceed under the assumption that the rest of my analysis is correct.

We're a degree F warmer because of anthropogenic greenhouse gases. The standard deviation around the best-fit curve seems to be about a degree F. So an event which would have been close to the best-fit curve is one standard deviation off it. Given the lack of rainfall, a temperature which would have been expected to be attained about 16% of the time is now expected to be attained about 50% of the time.

So, this event (i.e., this particular combination of drought and heat) has been made three times as likely by anthropogenic greenhouse gases, with lots of assumptions built in. The least of which is what global warming is doing to our local PDF of precipitation in Texas, which could go either way.

Pangolin said...

To my layman's eyes the Nielsen-Gammon analysis appears to stop at the Texas border. As if global extreme weather events and their increased frequency should not be considered in the analysis.

Also I counted 19 uses of the word "average" in his explanation and only four uses (or variations) of the word "extreme." It seems though we are addressing an extreme weather event in a year of extreme weather events in a decade of extreme weather events.

It appears that Texas weather patterns for individual years don't fit closely to the average but rather show strong shifts between dry years and wet years. I don't see a lot of data points on or closely fit to the average lines on any of the graphs and without error bars or confidence interval shading it's hard for laymen to spot true outliers from normal variation.

Also note that where the years are noted on the originally published graph indicating the weighting of events toward the last decade they were omitted in the following graphs where different equations were used, apparently, to fit closer to the outliers.

It seems to be somewhat of a whitewash designed to downplay the impact of the original "scary" graph.

Aaron said...

As Jay Forrester and Ed Deming kept reminding us, people are not good at predicting the behavior of non-linear feed back systems. In particular weathermen and climate scientists study the one weather system, rather than the behavior of dynamic systems in general.

In his classes, Dr. Deming made his students look at the behavior of various systems as the systems went “out of control.” It was shocking how a dynamic system could be “in control” and apparently stable, then suffer some small chaotic event, “go out of control”, and exhibit violent behavior as the system moved toward an new equilibrium. We have been adding heat to the weather system, bumped it out-of -ontrol, and we can expect weather that we have never see before as the system seeks a new equilibrium.

John Nielsen-Gammon missed the point that he has a system that is out of control and that his system is violently seeking a new equilibrium. We can expect ongoing violent behavior until the weather system comes back into control. The studies that he cites all assume that the system is "in control" and that the old rules hold. However, those old rules do not apply to the new, “out-of-control” weather system.

Our assumption has been that the climate models would give us adequate warning of the effects and impacts of AGW. Both public and private policies are based on this assumption. At this point, I would say that we have reason to doubt the assumption. Moreover, climate models at their best do not include much in the way of carbon feedbacks (CH4) and ice sheet dynamics related to sea level rise. Such models are a weak foundation for policy.

chek said...

I wouldn't blame John N-G for showing caution in the circumstances.

As has been alluded to, step changes in the climate system may in their initial stages look like outliers. Only time and more data will show which if either they are.

In the current political climate, only when nature is ramming it full-on down our throats will we collectively even think of doing anything about it. Perhaps only a corresponding step change in evolution can save us from what seems an increasingly neanderthal mindset.

Out in the real world BAU certainly isn't a continuation of party time (except for an insulated elite), but is becoming an increasingly grim life for increasing numbers of people globally.

Michael Tobis said...

Bart, I think it matters little which is the cause and which the effect in the argument at hand.

The system can be modeled as stochastic for various purposes but an extreme season can only exist in a global dynamical context. That context in turn is increasingly dominated by rapid change and decreasingly by the late holocene stability.

Judging rare events in terms of quasi-stationary statistics hardly helps when our question is precisely whether the quasi-stationary regime still holds.

Michael Tobis said...

Steve, not necessarily.

A hot fry pan with hot oil in it is more energetic before I throw an ice cube in it, but less predictable and with a narrower range of extreme events.

But in the present case, the ocean is not in quasi-equilibrium with the atmosphere, and won't be until decades after we stop perturbing the system at least. So in practice I agree.

We will be seeing more and more unlikely events, I fear. At present I see things as going quite a bot worse than predicted. In which case the GCMs are not going to be especially helpful.

Michael Tobis said...

Belette, we are getting a second year in the La Nina phase, but we have had many La Ninas in the past without comparably extraordinary heat and drought.

(Actually I'm surprised John did not filter on Southern Oscillation Index somewhere in his analysis.)

So my intuitive bet would be some reversion to the mean along with some more modest continuation of drought conditions. If the second year continues as is, we will lose many more, perhaps most, of our trees.

In the long run, I don't think we have seen the last of flooding here either. Many subtropical regions have very large year-over-year variability, and Texas' peculiar geography (bordering both desert and ocean) makes us outstanding in that regard. Most places are dominated by inter-latitudinal gradients while Texas changes most dramatically from west to east.

This year, the whole state is getting west Texas weather. A couple of years ago we were getting East Texas weather. The models are quite indecisive about the future trajectory of the place, unlike states to our west, the true Southwest, where continuing desertification is pretty clearly indicated.

Michael Tobis said...

Brian, yes, exactly.

I see a year like this and say "how else is such an outlier possible if you ignore climate change"?

Michael Tobis said...

Pango: "As if global extreme weather events and their increased frequency should not be considered in the analysis."

A good point.

But I think John and I agree that we don't have a clear scientific statement that we can use for such analysis. Again there is a need for some sort of objective weirdness index.

Michael Tobis said...

Aaron: "Our assumption has been that the climate models would give us adequate warning of the effects and impacts of AGW. Both public and private policies are based on this assumption. At this point, I would say that we have reason to doubt the assumption."

With this, I agree.

"Moreover, climate models at their best do not include much in the way of carbon feedbacks"

Well, until the current generation, none at all. But I still contend that this coupling is ill-advised. We don't know enough about the carbon cycle to put it into a climate model. We want the climate models to get better, and adding a poorly understood dynamic (which is easily separable because it is slow and uniform at the time scale of weather dynamics) adds unnecessary complexity. In general, science makes progress by finding ways to reduce the complexity under consideration, not by piling on.

But I agree with your main point. It is hard to escape the suspicion that the last few years are telling us that supposedly "alarmist" science has been too complacent.

But it is also hard to demonstrate. That is important too.

Jeffrey Davis said...

The chart which shows CO2 and temps over the last 2 million years should hang above Nielsen-Gammon's bed. In it, temps just suddenly go vertical. Like an elevator. There's a stair step change in the system as it cools, but an abrupt transition as it warms. Granted these changes are in geologic spans, but they're also in response to the much weaker Milankovich forcings.

Alastair said...

I find the situation in Texas today to be similar the Arctic sea ice in September 2007. Then the ice melted to a state far from normal. At that time, I thought the ice might have passed a tipping point, but it has taken four years for the ice to get near to that record low extent again.

It could be that Texas has switched into the state of a desert, or it might take another four years until there is another drought as bad as this one. Who knows? The worry is that even if the latter situation is the case, it still means that Texas is doomed to be a desert. Just that it will take a little longer than next year. How long before the Arctic is free of sea ice?

Cheers, Alastair.

dbostrom said...

Dr. N-G mentions (concerning Texas, I believe) "We're a degree F warmer because of anthropogenic greenhouse gases."

I'd really like to know if that's strictly true of this summer in Texas, or a number that's the result of substantial smoothing of observational data and if so what's the appropriate span of such statistical treatment in a case like this.

It's my understanding that as we warm the planet, temperature won't rise homogeneously over time in a given location, but would more likely move in fits, starts, hesitations depending on the span of time one is looking at.

I ask because I can see that Dr. N-G's assignment of that degree contribution of warmth seems the most conservative (as in proper in the absence of enough information) approach, but on the other hand in terms of specific attribution for this particular summer it may actually not be descriptive, if indeed a lack of homogeneity is what we'd expect.

I suppose I'd better go beaver away to see if I can learn about that on my own.

n-g said...

Doug - That's a rough entirely model-based estimate based on ensembled simulations with anthropogenic forcing. The actual linear trend in Texas temperature has been near-zero century-scale, despite being rapidly up since the 1970s.

David B. Benson said...

Michael Tobis --- The prior question is in what sense has the climate of the past 10,000 years been quasi-stationary? For example
Variability of El Niño/Southern Oscillation activity at millennial timescales during the Holocene epoch
http://www.nature.com/nature/journal/v420/n6912/full/nature01194.html

dbostrom said...

Thank you, n-g, and sorry I asked; I did not read your account carefully enough the first time.

Anonymous said...

I remember back a few months we were all in a discussion revolving around trying to figure out how to ask that question about the possibility of a new normal in extreme weather that Michael has alluded to frequently. I came up with the idea to look at individual extreme event types (flooding, heat waves, etc) in correlation with NINO/NINA to see if there is a trend there over the last 50 years, but Micheal correctly pointed out that would be a curve fitting. Then there a brief discussion about jet streams and there movement during NINO/NINA. Did that conversation get anywhere? I can't remember.

Bart Verheggen said...

Michael,

"I think it matters little which is the cause and which the effect in the argument at hand."

That to me sounds strange. If we accept the premise that there is a human forced trend in temp but not in precip in Texas, than if a big portion of the high temps is due to the dryness, this big portion is natural.

If otoh a big portion of the dryness is due to the high temps, than a big portion of the dryness is man-made.

No?

Andy S said...

I agree with you on this. I think the feedback between drought and heat is stronger than Dr. Gammon does, but we don't have data as yet to show this. This year is just the first point on a new curve where the relationship between drought and temperature rises exponentially once the vegetation has exhausted its source of transpiration (soil water). We haven't recorded this occurrence before as evidenced by the forests (now dead and dying) which predate our record keeping and couldn't have existed under such conditions. I believe that future research will show this to be true. We'll see I guess.

A drought history derived from cypress tree rings from the Texas Hill Country and the upper San Bernard River by Dr. Cleaveland of the Univ. of Arkansas Fayetteville shows that since the 1400's, 1716 was the driest year in that region with 1976 coming in secondplace. I think this drought tops 1976 easily and thus that in 1716. This work was done under contract for the Texas water development board. It isn't fool proof research, but it is the first piece of work showing we're encountering weather not experienced before by Europeans in Texas.

Aaron said...

Michael,

If the models are known to under estimate effects and impacts, then the risk managers should compensate.

Risk managers should say this is a "known unknown", so we compensate by applying a risk factor and assuming that our total risk is X times larger than the model projected.

Whatever the quality of climate modeling, our risk management has been - much worse.

I assert, that AGW has changed global upper atmospheric circulation so that the jet stream has changed its patterns, and heat wave events such as 2010 Russia and 2011 Texas become very likely events rather than rare events. I see both events as continental scale atmospheric circulation patterns allowed by a weird jet stream.

Lets see, who would identify weather risks in say, Texas?

I look at the Attribution of the 2010 Russia Heat Wave, and I do not see the skill in DC to do a good job talking about heat and drought in Texas.

Ed Darrell said...

So, what do we call your parable? "The Elephant in the Weather Report?"

Enough of climate physics and chemistry for a moment. Has anyone paid much attention to the lepidopterist journals, the ornithology journals, and the farm reports? Seems to me we should pay attention to what the plants, animals and insects say about global warming and climate change -- they aren't influenced by their having watched Al Gore's movie, after all. We can trust them, can't we?

Alexander Ac said...

Barry Brook, not Barry Brooks, otherwise excellent :-)

EliRabett said...

John's statement that

"And, of course, a second-order fit will always be a closer fit than (or at worst, equally close as) a first-order fit."

Is only true in the limit that the second order fitting parameter goes to zero, and in noisy data that is a dangerous assumption.

n-g said...

Ed - For starters, there's http://www.sciencemag.org/content/333/6045/1024

"...we estimated that the distributions of species have recently shifted to higher elevations at a median rate of 11.0 meters per decade, and to higher latitudes at a median rate of 16.9 kilometers per decade. These rates are approximately two and three times faster than previously reported. The distances moved by species are greatest in studies showing the highest levels of warming, with average latitudinal shifts being generally sufficient to track temperature changes...."

Eli - To be precise, I was talking about polynomial curve fitting, for which my statement is true, although "closer" does not imply "better".

Michael Tobis said...

Eli, I don't think that makes sense. A least squares fit to data is a linear process; the N+1 parameter optimum always has a smaller residual than the N parameter one. That doesn't make it a better predictor, of course, as is well known. But it does make it a better "fit".

A five-parameter fit to five points has zero residual and carries absolutely no information about where you expect the sixth point to show up. But it's a perfect fit to the five points you have.

I don't see any reason that John's quadratic is justified, though.

Unknown said...

While it's virtually undeniable that anthropogenic climate change exists, there simply isn't enough evidence from a long enough period of time to determine whether this specific weather event is part of a pattern, or whether that pattern is the result of anthropogenic climate change. These conclusions - or the leading questions masquerading as conclusions - are unscientific speculation.

Michael Tobis said...

Whoa. Start by defining your terms.

What is the difference between scientific speculation and unscientific speculation?

Jim Bouldin said...

Ed, the problem is that the biota is one or more steps down the chain of causation, thereby introducing confounding effects, of which there are typically several possible. In other words, the data's not as informative, climatologically as the physical data is. However, it is more informative, ecologically.

Nevertheless there is good phenological evidence to be had, among others. I'm not sold on some of the range shift stuff. That's a particularly thorny metric. [And ecologically harder to evaluate the consequences of as well]. Some of it's fine though. There was a recent meta-analysis in Science claiming faster shifts than Parmesan et al had estimated a few years ago.

The other point is that biotic metrics used to evaluate climate change effects tend to be responsive to changes in climate central tendencies, not extremes. Climate extremes are very problematic in evaluating certain biotic changes--like range shifts. Essentially worthless I'd say.

Aaron said...

Ed Deming developed tests to check if a dynamic system was stable.

In 2002, those tests said the Arctic Sea Ice was out of control. My statements to that effect were called "alarmist".

I think now we can say that the Jet Stream is "out of control". (http://squall.sfsu.edu/gif/jetstream_norhem_00.gif)

Who wants to bet that the jet Stream will "revert to normal" and bring Texas its pre-2011 climate?

I do not think that is likely as long as Arctic Sea is less than its 1980 volume.

Frontiers of Faith and Science said...

Good luck with selling the idea that a drought in a state famous for droughts is evidence of a cliamte catastrophe.
I beleive that very soon- in a few more years- this perceptive bias that AGW believers engage in wil be seen as no different than those who see an image of Jesus or the Virgin Mary in some screen door or a piece of cookie dough.