The relationship between predicting the climate response due to forcings over a timescale of decades and predicting the climate over the period of a year, which is dominated by short-term variability, is tenuous at best. The mathematical statement is that those short-term climate predictions are an initial value problem whereas the long-term predictions are a boundary-value problem.
There are a few groups who are trying to bridge the gap (such as the recent Keenlyside paper), but this is very new and speculative work…and, as that RealClimate post makes clear, lots of scientists believe that this particular group is unlikely to have done it successfully.
As for the Hadley groups predictions of the yearly temperature average, you should note that all of the observed values have fallen within the error bars. Do I consider this a very impressive result? No, given the wide error bars relative to the relatively narrow range of variations over the past several years, it is not a particularly interesting test in my book. In particular, along the lines that brazil84 suggests, a prediction of 0.40 C ± 0.12 C would have included all of these data points too. However, it is worth noting that, contrary to the claim that scientists in the climate field are regularly making bold predictions with overconfidence, that is not the case here…They know that they can’t predict these climate fluctuations with very good accuracy as reflected in their errorbars.
I suggest you read the IPCC reports. Here are just a few examples of such predictions: James Hansen’s 1988 projections, the cooling due to Mt. Pinatubo’s eruption, and the moistening of the upper troposphere. The latter two are not only interesting predictions in their own right but are specifically tests of the water vapor feedback, which is important specifically to the issue of climate sensitivity…which is most of what this question all comes down to.
Lol. Only in climate science can you take data after the fact; tune a simulation to fit that data; and congratulate yourself on your wonderful “prediction.”
Actually that’s not true. The likes of Miss Cleo are also very good at making predictions after the fact.
Of course, James Hansen really did make some predictions back in 1988. Let’s see how he did, shall we?
First, let’s look at the actual, original 1988 projection. It apparently be found on the last page of this document
Note that the projection is made in terms of annual mean temperatures – not trends. One can read the anomaly for the last year of observations on the chart as approximately +0.3C. The anomaly for 2007 – 20 years later – is a little bit above +0.8C for scenario B – an increase of 0.5C
So how does that compare with reality? Let’s give Hansen the benefit of the doubt and use his own GISS temps, even though GISS has (apparently) been consistently higher in recent years than the other leading measures.
Could you please provide evidence from a peer-reviewed citation that the simulations were fit to that data? Thanks in advance.
Ah…Up to our old tricks again of trying to align graphs using the fluctuations in such a way that we cherrypick the result we want to get. (And, by the way, the difference in temp between the 1988 values and the 2007 value in Hansen’s Scenario B is actually barely over 0.4 C.) I will just repeat my link so that people can go look for themselves at how they align.
It is also worth noting that Hansen did use a climate model that had a sensitivity that is higher than the best estimate that we now have for it…although still within the realm of possible values. (So, even if we accepted your numbers that the trend in the data is only 3/5 of the trend in the projection, which I don’t, this would still imply a climate sensitivity well within the range of what the IPCC is projecting…e.g., it would be somewhere around 2.5 C or so.) And, note that many scientists at the time, and not just denialists, were skeptical of Hansen’s claim that we were yet seeing the effects of global warming…and thus skeptical that one could predict the general trend over the next 20 years period.
In fact, if you think Hansen did so poorly, I imagine you could find someone who made a projection at that time that turned out better than Hansen’s?
Sure. My evidence is the papers themselves. Take a look at the graphs. See the beautiful fit between simulation and data? If that’s not the result of tuning and tweaking, then it ought to be possible to run the simulation into the future and get a prediction that fits the data that beautifully.
Which the author has not done, as far as I can tell.
In any event, it should be emphasized that the “predictions” you are trumpeting were made after the fact.
Lol. If any lurkers are reading this, just look at the orginal graph. The difference between the value for 1987 and 2007 is CLEARLY approximately 0.5C.
Then look at the GISS temp graph. The difference between the value for 1987 and 2007 is CLEARLY approximately 0.3C.
Are you saying that the modelers surreptitiously tuned their models to fit these obscure bits of data in advance? So, then when someone else came along and checked, they fit perfectly? You clearly don’t know anything about how climate models are created and run, do you?
I can play the cherry-pick game too, my friend. If you look at 1985 vs. 2005, then the difference in the data is ~0.55 C, which is in very close agreement with the Hansen scenario B projection of about a 0.6 C difference between those two years.
Of course, the real way to do it right is not to go on individual year fluctuations but replot things on the same graph as shown in the RealClimate piece that I linked to. That shows that the prediction is doing pretty well…although clearly the fluctuations are large enough compared to the signal that one cannot make a conclusion to high precision. Given that the Hansen prediction used a climate model that has an equilibrium climate sensitivity close to the high end of the current IPCC best estimated range of 2 to 4.5 C, I’d say that is doing pretty damn good…worrisomely good. And, considering that there were probably a lot of people at the time who would have bet that the climate would just as likely been cooler twenty years from then warmer, I’d say Hansen has a much better track record than many in the field…particularly his detractors.
No, I’m not saying that. I’m saying what I said. Which is that the modeler took data from the past and tuned the model to fit the data.
I wasn’t picking cherries at all. I simply used the most recent year when Hansen made his prediction (1987) and the most recent year for which GISS data is available (2007). You are the one who chose the Hansen prediction, not me.
In any event, it hasn’t been until recently that Hansen’s prediction has been seriously diverging from reality. His reality, anyway.
In the PDF you linked to of Hansen’s 1988 prediction, the observed temperature difference for 1988 is listed as being at about 0.3[sup]o[/sup]C. His prediction for 2007 in Scenario B is for it to be at 0.7[sup]o[/sup]C. In the NASA graph you linked to of the observed temperatures, 1988 is listed at 0.3[sup]o[/sup]C and 2007 is listed as being 0.55[sup]o[/sup]C. So, indeed, instead of being 0.4 in difference it’s 0.25.
And?
If you compare the graphs like they do here and look at the blue and black lines. The blue line is a “prediction” of what would happen that year based on levels of gasses in the air and whatnot and the black line is what “did” happen that year. Compare 1973-1977 or 1984-1986 ranges of those two lines. Vastly different.
Trying to guess weather was one of the leading precursors to chaos theory. The issue is that. like in a fractal, even though there might be an overall “look” to something, like the Mandlebrot “Bug”, if you try to zoom in, you’ll get entirely different images with different inlets and flourishes with slight changes in what the variables you put in to begin with are. You can for certain that there will be a bug-shape in there, but what color and where everything will be specifically is all up in the air.
Quibbling over small periods of slight differences when the general flow of things is exactly the same is meaningless when you’re talking about climate and weather.
I’m not sure what your point is about Chaos Theory. The fact that weather is somewhat chaotic should make us more skeptical of long range predictions, not less. Hansen and company are trying to predict the consequences of the interactions of many different phenomena over a long period of time. Many of those phenomena are not very well understood.
Take for example a bean machine. There’s no predicting the fall of any single bean, so for each bean (or year) you add to the machine, there’s every chance that you’ll be way off in guessing which of the slots it will land in. Using the law of averages, you will guess that it will land in the place that will preserve the expected bell curve, but the bean really has no particular reason to actually end up in that spot. And yet, as time goes on you will always consistently get something that looks exactly like you expect.
Another example would be the economy. You might expect that as you insert more money into it, that the amount of commerce will rise by a certain percent, but there’s every possibility that there will be years where everyone holds back their money and years where everyone goes gung-ho. But if you look at it over the long scale, the amount of commerce will rise on average to the expected amount.
There’s any number of examples like this, and climate is accepted as being something which acts in this manner. But, the better a modelling system you can get, the finer the resolution will become. So even though we can’t say today what will happen on a year-to-year basis within a reasonable range of error, that isn’t a relevant argument against the fitness of the simulator. Eventually, possibly we will get it down to that resolution, but for the moment, no. What can be said though is that comparing the simulator over a 25 year period (for instance 1960 to 1985) seems to follow what is known, so the next 25 year period will likely continue to do so. But there’s no reason to expect the simulator to do better predicting the future than it did the past. It was far off during certain short periods of time and it can be expected to be far off for short periods of time in the future as well.
Addendum to my previous post (i.e. read it first):
Actually, the resolution might never become better than it is today. Like the economy, there’s some level at which there just is no way to know whether people are going to spend or save. In the case of the climate, even if we are able to model every nook and cranny of the planet, there’s still all the animals and buildings being built and destroyed and overall changing the course of the wind, not to mention the sun’s effect which varies based on sun spots which are entirely random so far as we know it, and so on and so forth.
At some point, there’s going to be a hard line where any global prediction of climate on that small a period of time becomes impossible to determine. The smaller a time frame you look at, the more likely this is.
But, like adding money to the economy, you do know what’s going to happen more and more finitely the longer a scale you look at.
It seems to me it is you who are wrong. Your hidden assumption is that uncertainties about the climate will cancel eachother out, resulting in a predictable pattern. While this may be true with a “bean machine,” there is no a priori reason to think that the climate will act this way over the next 100 years.
Actually that’s not totally true. There is in fact an a priori reason to think that the climate will stay somewhat stable: The fact that it has been somewhat stable for a long time. But that observation doesn’t really help the alarmists.
Anyway, please imagine a bean machine which is a little more complex: Each time a bean falls into a slot, it affects the probabilities of later beans falling left or right when they hit a pin. Different slots affect different pins in ways that you do not fully understand. Imagine that there are hundreds of millions of slots and pins.
In my opinion, that’s a better metaphor for the climate. And such a machine would be very difficult to predict. It may very well have trends that last a long time, before suddenly reversing themselves. A simple model might predict the machine’s behavior for quite a while before veering off course, hopelessly and forever.
It’s called “anthropogenic” for a reason. There’s no a priori reason to believe that it will stay stable with human intervention–or at least there’s no reason to believe that it would self-correct in a time-frame that does us any good.
There’s no reason to assume that a bean machine will fall into a clean bell curve directly under the dropper if you’ve got a hammer on the outside that swings to hit the side of the box periodically.
Adding in outside interference, there’s no reason to assume the status quo will remain nor that it will return to it’s former state on its own. No reason, unless you prove it one way or the other.
So if I wanted to see what would happen with the bean machine and hammer, assuming they were each 400 miles tall, I’d have to simulate it and the hammer in a computer, including everything I know about gravity and gravity fluctuations and air whorls within a confined space and so on, and then I’d be able to run it and determine the answer. Having done all that, I’d have no call to assume my result was particularly wrong unless other people came do a different conclusion with independently built simulators. And if all of the simulators said that the intervention of the outside hammer will cause a noticeable difference, well then that seems the best way to throw your vote so far as I would think it.
We have had the amount of atmospheric CO2 rise to the highest level ever seen in 650,000 years as best we can tell it, and we since we know how much CO2 humanity has released, and can attribute that 133% increase to ourselves, again there’s no reason whatsoever to assume that humanity can’t act as a force that can intervene on the state of the climate on a global scale, until you’ve proved otherwise.
So now, what is your a priori reason that everyone should assume that nature will continue its course with 133% more CO2 in the atmosphere introduced into than we have reason to believe has ever been in it before?
In science, how do you ever prove something with “a priori” assumptions?
That’s incorrect. The climate has taken huge hits in the form of asteroid impacts, mega-volcanos, and goodness knows what else. It’s gone through ice ages, warm periods, high CO2 levels, low CO2 levels, etc, without running away.
The natural assumption is that the climate is dominated by negative feedbacks.
But anyway, that’s a different argument from your argument that the climate on a 100-year time scale can be modeled and predicted analogously to a simple bean machine.
Yeah, and you’ve never heard the theory that the dinosaurs were wiped out by just such an event?
The thing we’re trying to answer is whether the planet is going to become less comfortable to live on and your own examples of asteroid impacts, mega-volcanos, and goodness knows are all things that prove that changes to the atmosphere can make the world quite uncomfortable for unreasonably long periods of time. You’re going to have to explain to me why I’m supposed to feel encouraged by the knowledge that the planet can indeed become an unreasonably inhospitable place (to humans) for insanely long periods of time. Yeah sure, it might rebound in 10 million years, but I just don’t find that knowledge to be reassuring.
So we’re now up to two questions:
Why should we assume “a priori” that humanity can’t act in the same manner as a large asteroid, rather than proving it to the best of our ability scientifically?
Why should I be encouraged by the likelihood that the planet will be able to recover over a period of millions of years?
The question is whether the climate is dominated by negative feedbacks or not. All of those events suggest that the climate is indeed dominated by negative feedbacks.
On the other hand, the climate models which are used to forecast CAGW all assume positive feedback. i.e. that increasing CO2 levels will increase temperatures, which will increase water vapor levels, which will increase temperatures more, and so on, and so on.
And feedback is the critical question here. As far as I know, nobody is claiming that the first-order effects of CO2 will cause catastrophic warming. The claim is that there will be interactions, which will cause more interactions, which will cause more interactions, and so on and so forth, increasing the temperature every step of the way.
When you are dealing with a chaotic system, it’s silly to think you can confidently predict that kind of response. Which is why your point about chaos is 100% wrong.
And it’s also why the Met Office forecast was so terrible. And why Hansen’s forecast is failing too.
Dominated? You must have a different definition of that word. A large asteroid collision takes a matter of seconds to occur. It took millions of years for your negative feedbacks to correct the system to its norm. How is that “dominated” by negative feedback?
If you introduce a sudden massive change into a system where the negative feedback is incredibly small in effect, there’s every reason to expect a domino effect period for all of the positive effect changes to ripple through the system and change it in all the ways that it will. For instance, one of the issues being considered is the delay in effect that global warming could have on the ocean. It could take decades for today’s air temperature change to percolate through to the total body of surface water, and that’s assuming that CO2 levels stay the same. Now, at some point, the last domino will fall and negative feedback will start to step in, but from what we can tell, the last domino isn’t going to be coming down until a century or two from now, and the negative feedback could take millenia or more to set things back as they were starting from that point.
It should also be noted that your contention that there aren’t negative feedback items in simulators is simply wrong. As much as is known about the climate is being modeled to the best of our ability and every time someone thinks of something that hasn’t been included, positive or negative, it’s added. That’s what scientists do. They try to out-think everyone else and catch everyone else for factors that weren’t included.
But being an iterative process, in general this means that things narrow down on an answer, not entirely reverse it. Larger factors are the first ones added, not the last, as they’re quite obviously the ones that are most apparent.
Self-correction: Apparently the lasting effect of things like the dinosaur extinction asteroid collision (if there was one), while having effects on the food chain that last millions of years (i.e. until larger creatures appear again), the actual effects on climate are more in the range of decades than millions of years.
But the same sort of simulators that say that negative feedback “reset” the climate in a period of a few decades assuming such a collision are predicting a change of centuries in regards to AGW. Take that for what you will.
And no, the Hansen prediction doesn’t visually appear to be widely disparate from the actual unless you look at it on a year by year basis, which I’m afraid just isn’t how it works (for reasons that have been explained and against which you have made no particular argument beyond that “it just isn’t so.”) If you want to know more about some of the various effects that make things more random the smaller the scale you look at, feel free to start a GQ question.
