Causal feedbacks in climate change
nature.com
nature.com
Sometimes, he wrote, we do get temperature increases in the geological record right before we see the CO2 go up. Historically, that's happened on a regular basis due to orbital variations. The fact that CO2 goes up afterwards is not reassuring, it's a positive feedback that drives the temperature even further.
That's why climate change is so worrisome; it's not just the direct heating effect of our emissions we have to worry about, but the additional heating from the extra CO2 released by the planet.
> In conclusion, our analysis provides direct confirmation that internal Earth system mechanisms rather than orbital forcing have controlled climate dynamics over the Pleistocene cycles. Moreover, they demonstrate the existence and importance of a feedback effect of temperature variability on GHGs in driving the dynamics (Fig. 1b). This confirms the existence of a positive feedback operating in climate change whereby warming itself may amplify a rise in GHG concentrations. As CCM infers causality directly from the time series, the consistency of our results with elaborate mechanistic analysis represents remarkable empirical confirmation and, moreover, provides a clock for the response times involved. We suggest that this new and powerful approach may also help to assess causality behind the numerous other time series we have for the Earth system.
They cite Sugihara, G. et al. Detecting causality in complex ecosystems. Science 338, 496–500 (2012) for this.
"Direct evidence for a positive feedback in climate change: Global warming itself will likely accelerate warming": http://phys.org/news/2015-03-evidence-positive-feedback-clim...
I think the point is that deviation from some 'normal' global temperature catalyzes or inhibits greenhouse gas proliferation. (At least in the positive direction, I haven't read it so I don't know if their proposal supports this.)
Another, much longer term factor (tens to hundreds of millions of years) has to do with the positions of the continents on earth, which affects how well ice can be stored. If there is considerable landmass near the poles and it's cold, then ice sheets can form, which increase the amount of reflected solar radiation (albedo) and decrease the amount of heat stored in the oceans (which have less volume due to ice sheet storage, and because sea level falls, the warm shallow seas that hold a lot of heat are much smaller). This can help cooling feedbacks. When the continents are arranged such that there is little polar landmass, these situations are reversed which leads to heating.
This was one of the ideas used in the "Day After Tomorrow" movie, although soundly criticized for presenting a very short time frame change into an ice age, that an overly warm earth can trigger a transition into an ice age is not unheard of. Lots of time questions about time frame and mechanism.
I lost the paper I read that was suggesting the mechanism was water vapor (a pretty important green house gas) which as the planet warmed the air carried more vapor which hits critical mass, turns into a nearly permanent covering layer of clouds in the upper atmosphere, which then causes water at lower altitudes to precipitate out as snow. Sort of a self induced nuclear winter type scenario.
The key conclusion of the paper is that the system responds in a non-linear way. That makes predicting its behavior with models like the one IPCC uses very challenging.
These are forcings because nothing in the Earth's climate causes them--they are essentially external inputs that change the fundamental limit of energy available the climate: insolation.
If the amount of solar energy available to the climate changes, then the temperature equilibrium of the climate must change. As things change, the changes can themselves cause further changes, producing nonlinear effects. That's what this paper is addressing: the idea that warming itself can cause additional warming.
But it all starts with a forcing. If the forcing is removed or reversed, then the equilbrium must shift back. Regardless of how strong a feedback might be, it can't overcome a forcing. (That's the essential distinction between the two.)
So: regardless of how much the temperature is run up by various feedbacks when the Earth gets closer to the sun, once the Earth moves away from the sun again, the temperature must go down again. And another ice age begins.
edit: article -> paper
http://www.uvm.edu/~cdanfort/csc-reading-group/sugihara-caus...
It irks me to see that term mixed into scientific discussions, because I am sure it will just be assumed that everyone knows what they mean by "casual" and accepts it as a useful scientific concept.
I hope that's not what you're saying, because it's absurd.
I don't know about never, but the "c-word" gets thrown around way too freely in my opinion. But yea, I am saying that the most successful science has had little to nothing to do with causality. You see the focus on causality is more common in the less successful endeavors such as the social sciences, perhaps it is time to consider that the causality concept is a red herring (at least when it comes to science).
To be sure, I am far from the first to entertain such thoughts: "Causality in physics has had bad press in philosophy at least since Russell’s famous 1913 remark: “The law of causality, I believe, like much that passes muster among philosophers, is a relic of a bygone age, surviving, like the monarchy, only because it is erroneously supposed to do no harm” (Russell 1913, p. 1). Recently Norton (2003 and 2006) has launched what would seem to be the definite burial of causality in physics. Norton argues that causation is merely a useful folk concept, and that it fails to hold for some simple systems even in the supposed paradigm case of a causal physical theory – namely Newtonian mechanics." http://philsci-archive.pitt.edu/4690/1/CausalFundam.pdf
Its why we have concepts like dependent and independent observations. We split observable dimensions into these groups so we can change the independent variable and find out their causal effect on the dependent variables. The whole reason we have "experiments" is so an experimenter can deliberately change the system in a focused way, outside of its normal dynamics, to reveal the causal structure of the system under observation. You can't do this by passively observing the system. This is why we have experiments. This is core to the scientific method.
"However short we make the interval t, something may happen during this interval which prevents the expected result. I put my penny in the slot, but before I can draw out my ticket there is an earthquake which upsets the machine and my calculations. In order to be sure of the expected effect, we must know that there is nothing in the environment to interfere with it. But this means that the supposed cause is not, by itself, adequate to insure the effect. And as soon as we include the environment, the probability of repetition is diminished, until at last, when the whole environment is included, the probability of repetition becomes almost nil.
In spite of these difficulties, it must, of course, be admitted that many fairly dependable regularities of sequence occur in daily life. It is these regularities that have suggested the supposed law of causality; where they are found to fail, it is thought that a better formulation could have been found which would have never failed. I am far from denying that there may be such sequences which in fact never do fail. Ic may be that there will never be an exception to the rule that when a stone of more than a certain mass, moving with more than a certain velocity, comes in contact with a pane of glass of less than a certain thickness, the glass breaks. I also do not deny that the observation of such regularities, even when they are not without exceptions, is useful in the infancy of a science: the observation that unsupported bodies in air usually fall was a stage on the way to the law of gravitation. What I deny is that science assumes he existence of invariable uniformity of sequence of this kind, or that it aims at discovering them." http://www.hist-analytic.com/Russellcause.pdf
Basically, if you do not have a notion of causality, you cannot specify a predictive model, which is a pretty fundamental part of science (although, of course, not all science deals with predictive models; sometimes even finding the correlation is an achievement).
Are you saying the above law isn't science and must have no predictive skill since there is no claim of causality involved?
R = PV / (nT)
P = nRT/V
etc.
The model would be something like:
experiment setup -> gas dynamics -> observed measurements
You would instantiate it like so:
(setup apparatus such that P=1 psi T = 289 n=5 mol R=12.2) -> (solve PV=nRT for V, V = ...) -> (V = ..., other values same as in setup)
But as I said, it's all discussed in Judea's book. Particularly, there is the epilogue, a reprint of this lecture: http://singapore.cs.ucla.edu/LECTURE/
Slide 26 is where he brings up Russell's critique of causality.
And this quote from slide 47 is Pearl's response: "Let us examine now how the surgery interpretation resolves Russell's enigma: concerning the clash between the directionality of causal relations and the symmetry of physical equations. The equations of physics are indeed symmetrical, but when we compare the phrases "A CAUSES B" vs. "B CAUSES A" we are not talking about a single set of equations. Rather, we are comparing two world models, represented by two different sets of equations; one in which the equation for A is surgically removed, the other where the equation for B is removed. Russell would probably stop us at this point and ask: "How can you talk about TWO world models, when in fact there is only one world model, given by all the equations of physics put together?" The answer is: YES. If you wish to include the entire universe in the model, causality disappears because interventions disappear - the manipulator and the manipulated loose their distinction. However, scientists rarely consider the entirety of the universe as an object of investigation. In most cases the scientist carves a piece from the universe and proclaims that piece: IN namely, the FOCUS of investigation. The rest of the universe is then considered OUT or BACKGROUND, and is summarized by what we call BOUNDARY CONDITIONS. This choice of INs and OUTs creates asymmetry in the way we look at things, and it is this asymmetry that permits us to talk about "outside intervention", hence, causality and cause-effect directionality. "
And if you can't read, then here he is presenting it in a recent lecture: https://youtu.be/9YMHqO6Z7AI?t=1731
Second, it sounds like Pearl is in agreement that cause and effect do not appear to be physical aspects of the universe (ie something to be independently measured), but are rather wholly dependent on the person/device performing the measurements. By dependent I don't mean in the sense that pressing a ruler up a against an object slightly changes its length, or the uncertainty principle, etc. I mean that the perspective of the observer looks like it is central to Pearl's conception of causality. If this is the case, it should be possible to do science without any use of the cause/effect idea at all. However, perhaps it is a useful heuristic in many cases.
Third, I am always wary about arguments regarding "scientists do x". Many people are passing themselves off as scientists these days who are not actually doing science according to the usual definitions. For an example, just look at the dearth of even attempted replications or proposed theories making precise predictions in fields like health and social science. Unsurprisingly, the vast majority of what gets claimed in those areas seems to be wrong when it is checked. The widespread misinterpretation of and games surrounding p-values occurring in these fields of research goes without mention... My point here is just because some people claiming to be scientists behave a certain way doesn't mean that is the most (or even a) productive way to act.
Also as Pearl also mentioned, this really only became a point of interest when computer science and artificial intelligence got going, because there non-narrative models seem to be successful as well, like neural networks, finite element methods, and so on.
A similar perspective applies to p-hacking; it's not so much that people deviate from the "one true way" of doing statistics, as that there are a hundred conflicting myths of what statistics should be and often they pick the one that is most convenient, rather than searching through to find the most applicable.
> Does pressure cause temperature or does temperature cause pressure?
This sentence is incredibly vague, which is why scientists use equations to do their work, instead of sentences.
If I hold n and V constant and change T, it tells me what change I caused to P by changing T.
If I hold n and P constant and change V, it tells me what change I caused to T by changing V.
And so on.
The equation is just an equation. But if I control some of the parameters, I can still cause stuff to happen, and the equation tells me what stuff I will cause.
"The goal of this account of causation in science has been to reconcile two apparently incompatible circumstances. On the one hand, causes play no fundamental role in our mature science. Those sciences are not manifestly about causation and they harbor no universally valid principle of causality. On the other, the actual practice of science is thoroughly permeated with causal talk: science is often glossed as the search for causes; and poor science or superstition is condemned because of its supposed failure to conform to a vaguely specified principle of causality. I have argued that we can have causes in the world of science in same way as we can retain the caloric. There is no caloric in the world; heat is not a material substance. However in many circumstances heat behaves just as if it were a material fluid and it can be very useful to think of heat this way. It is the same with cause and effect." http://philsci-archive.pitt.edu/1214/
All I'm saying is that once you start wondering a bit about this causality idea, it is really annoying to see it thrown around like something obvious.
Ironically, you're confusing cause and effect (or rather, observation and result) in how scientific progress is made: the observation of correlation leads to the formulation of cause (i.e. theory), and the manipulation of specific variables leads to the validation of effect (i.e. proof). Without either, fields of science do not mature at all.
Now here's the million dollar question: when it comes to complex models (like genetics, neurology or the climate), how do you isolate and manipulate specific variables to validate their effect?
Causality within the context of scientific research almost always means variable precedence and dependency. Did observation A precede observation B? If I change A, how and when will B change? Or vice versa?
That's what scientists are looking at when they study whether something is "causal." They're not trying to figure out if inductive reasoning can discover truth. They just provisionally assume it can and go from there.
1) it will be harder to undo the changes we caused or 2) greenhouse gas changes occurred as a result of temperature change
or some other combination of things.
i.e. positive feedback loop, but with drastically different time scales.