Guest blog by Chris Colose (e-mail: colose-at-wisc.edu)
UPDATE: This is Part 1 of two posts by Chris. Part 2 is here
RealClimate has recently featured a series of posts on the greenhouse effect and troposphere, articulating some of the more important physics of global warming from first principles. It is worthwhile reviewing these elements every so often with different slants just so the broad picture is not lost in the disagreement over details. This post extends on this theme to discuss one of the greatest sources of interest and uncertainty in the physical science of climate change: feedbacks. Feedbacks behave in interesting and often counter-intuitive ways, some of which can only be fully appreciated by mathematical demonstration. The previous posts at RC were criticized for being either too complex or too simple, so this post will feature two parts, with the second part providing some more of the technical details.
Feedbacks are components of the climate system that are constrained by the background climate itself; they don’t cause it to depart from its reference norm on their own, but rather may amplify or dampen some other initial push. These original “pushes” are forcings which are typically radiative in nature (such as adding CO2 to the air) and manifest themselves as a climate change when they are large enough or persistent enough to overcome the large heat capacity of the oceans, and thus change the annual mean radiative energy balance of the Earth. In a broad sense, a feedback means that some fraction of the output is fed back into the input, so the radiative perturbation gets an additional nudge (amplifying the forcing, a positive feedback or damping the forcing, a negative feedback). The major examples such as decline in ice extent in a warmer world, thereby reducing the reflected fraction of incident surface radiation are pretty well known at this point.
Another thought experiment can help to appreciate the implications. When we think about the terrestrial greenhouse effect, it is necessary to distinguish between those gases which condense and then precipitate from the air as solid or liquid rather rapidly (on Earth this substance is water) and those gases which can reside in the atmosphere for a very long time, and whose concentration is not so dependent on the temperature. Once you account for the spectral overlap between the various greenhouse gases and clouds in the sky, it is found that water vapor makes up roughly 50% of the modern greenhouse effect, clouds about 25%, CO2 is 20%, and the remaining gases (primarily methane, ozone, and nitrous oxide) make up the rest (Schmidt et al 2010, in press and a discussion of the paper here). This generally leads to popular claims like ‘water vapor is the most important greenhouse gas’ since it makes up the bulk of the infrared absorption in our atmosphere. This simple picture is incomplete however, since the total water vapor concentration is largely set by temperature and thus the non-condensable, long-lived greenhouse gases (chiefly, CO2) really provide the skeleton by which the greenhouse effect is maintained and what governs its capacity for change. In that sense, water vapor is in large part supported by the other gases and then amplifies their effect significantly.
If you could remove all of the CO2 from our atmosphere, aside from making the planet more efficient at losing its heat to space (thus cooling) you would do a couple of things. First, you’d lose much of the water vapor and cloud greenhouse effect since temperatures would be too cold for them to exist in appreciable amounts. Secondly, you would also get temperatures cold enough to the point where expanding ice cover greatly enhances the surface albedo of the planet and triggers a snowball Earth. This simple picture also holds true as the climate warms today, where it has long been noted that the increase in water vapor content of the upper troposphere should amplify the CO2 signal by a factor of about two.
CO2 concentration itself can act as a feedback to temperature on longer glacial-interglacial timescales in response to changes in ocean dynamics, temperature/salinity, as well as vegetation. It can also be a negative feedback to temperature on still longer timescales, where silicate weathering effects (determined by volcanic output and removal by precipitation) are thought to keep the climate in check over geologic timeframes. Often however, climatologists define some metric CO2 change (such as a doubling) which allows you to ignore the carbon-cycle feedbacks and focus on the ones which alter the radiative balance and temperature further. The principle feedbacks in this category are water vapor responses, surface albedo, cloud, and lapse rate effects.
Feedback behavior
The ultimate constraint on climate change is the Planck radiative feedback, which mandates that a warmer world will radiate more efficiently and therefore provide a cooling effect. For a blackbody, the emission goes like the fourth power of the temperature. So the question of how the other feedbacks behave is really of how they modify the Planck feedback. In order to decide whether a feedback is positive or negative, it is instructive to define a baseline sensitivity value that the climate system would have if no feedbacks operated at all. That is, if we perturb the climate with some forcing, what is the temperature change you would need to have to allow the planet’s energy balance to be satisfied. It can be shown that for every Watt per square meter radiative forcing the climate would warm by about 0.3°C without any other responses. To put this in perspective, it would take about five doublings of CO2 or a 7% increase in the total solar radiation hitting the Earth to produce the magnitude of climate change typical of glacial-to-interglacial transitions. Changes of this sort are well outside the bounds of what is characteristic of proxy records and observations, so this must mean that various feedbacks act to change the temperature much more than 0.3°C for a watt per square meter forcing. In other words, the aggregate effect of feedbacks is to be positive and enhance the so-called climate sensitivity relative to what it would otherwise be. Figure 1 below illustrates this.
The feedback factor is a value that is proportional to the no-feedback sensitivity value. It relates the fraction of the system output that is fed back into the original input, and takes on a value between 0 and 1 for positive feedbacks and less than 0 for negative feedbacks. It also means that the timescale it takes for Earth to reach a new equilibrium value is longer.
Feedback interactions
When multiple feedbacks operate, they can add together in rather odd ways. For instance, you might think that if you take a feedback that doubles the sensitivity to climate and another that halves it, they would cancel and bring you right back to the no-feedback sensitivity. You might also think that two feedbacks which each amplify the original forcing by 50% would add to double the no-feedback sensitivity. In fact, neither of these is the case, and the behavior emerges because multiple feedbacks interact with each other as well. One can imagine that if water vapor and the ice-albedo feedback are operating, the water vapor boost will mean more ice melt, which will mean further warming, more water vapor, still less ice, and so forth. Note that positive feedbacks do not inherently imply a runaway scenario; it just means that the final temperature change is higher than it would have been without the feedback being there.
Aside from just enhancing the temperature signal, the existence of feedbacks is really what allows for significant departures in planetary climate evolution from some reference state. It would not be possible, for example, to cover the whole planet with ice down to the tropics or to boil off Venus’ oceans without feedbacks kicking in and rearranging the climate system to be compatible with a completely new state. Although it is not feasible to trigger a runaway greenhouse like Venus even if we burned all the coal today, it should really be kept in mind that there’s nothing unique about our current climate except that we have adapted to it. It is very readily capable of changing fast and ending up in a completely new regime, and ruling such a scenario out cannot be done with high confidence.
Ice-Albedo feedback
The ice-albedo positive feedback arises because sea ice is less dense than its liquid form, it is more reflective, and the extent is highly sensitive to temperature. Water is somewhat unique in this regard since most solids tend to be denser than their liquid form, causing the ice to sink which would forbid an ice-albedo feedback. This means that if the planet warms the ratio of highly reflective ice to relatively high absorbing ocean and land surfaces will be altered. The key in the modern climate is the seasonality between absorption of solar radiation in the summer and the release of energy from the topmost part of the ocean into the lower atmosphere in the cooler months. Temperatures in the Arctic do not become substantially amplified in the summertime as you might expect in large part because a lot of energy is going into melt or evaporation. However, areas of open ocean water develop earlier in the melt season, raising the heat content of the ocean mixed layer and melting more ice. When the melt season is completed, there’s a lot of open water and heat in the mixed layer and large vertical heat transfer from the ocean to overlying air until the sea ice forms, resulting in a seasonal delay of warming from the radiation absorbed in the summer. This surface amplification of Arctic temperatures has emerged primarily in the autumn and winter and should progress into the spring and summer in the future (Serreze et al., 2009).
Lapse Rate Feedback
Why is the lapse rate (the temperature decline with height) important as a feedback? In the tropics, the temperature lapse rate is largely set by convection to stay near a moist adiabatic profile. In principle, this should decline in a warmer world resulting in the upper atmosphere warming more than the surface. This means that the bulk of the atmosphere radiates to space at a temperature warmer than it would have with no lapse rate change, and emission from warmer layers is more efficient than emission from cooler levels. This provides a negative feedback which partially compensates for the water vapor feedback. Interestingly, the two effects act in tango with each other and so the uncertainty in the water vapor+lapse rate feedback is smaller than the uncertainty in the individual terms.
In the context of anthropogenic global warming, all of these complex feedbacks and interactions end up boiling down to the question of how much warming you get from additional CO2 release into the atmosphere. The most recent IPCC AR4 assessment gives a range of about 2 to 4.5ºC at equilibrium. This is the so-called ‘Charney sensitivity’ which takes into account these fast feedbacks discussed above, as well as clouds which provide the greatest uncertainty in narrowing these estimates.
Estimates of this range have been based on not just GCM results, but constraints from observational data (the seasonal cycle, or volcanic eruptions) as well as the past climate record (Knutti and Hegerl (2008) provide a review). One problem is that high values of sensitivity are more difficult to rule out than low values, and some observations that are good for ruling out the low end do not constrain the high end very well. For example, volcanic eruptions display a non-linear relationship with the equilibrium sensitivity, so the peak in the probability distribution shifts only slightly for larger mean values of sensitivity.
Recently, some studies have expanded on this view to also include ‘slow feedbacks’ such as the response of ice sheets and vegetation that are important on hundreds of year timescales (Lunt et al 2010; Pagani et al 2010). These estimates show that the long-term warming should be even more than the Charney estimates, on the order of about 5°C.
In summary, the function of feedbacks is to modify how much you expect the climate to change for a given forcing. Part 2 will describes some of the basic mathematical relationships that are important for discussing feedbacks, as well as elaborate on the water vapor feedback to climate change. Water vapor provides the strongest feedback effect on Earth in terms of enhancing the sensitivity, and is also a key component in understanding the evolution of the terrestrial planets.
Well done — exactly the right level for interested laymen.
Given the uncertainties in estimating sensitivities, it could be useful to know more about the state of the scientific discussion, e.g.:
(1) Are there essentially two polar competing models for smallest vs. largest sensitivities, or a wide variety of models all over the spectrum, depending perhaps on which features are most emphasized?
(2) Do these preferences correlate with the background of the adherent? For example, do the people who study glaciers tend to reach a common conclusion which differs from that held by people who study stratospheric effects, or is there a correlation with the age of the researcher, or are the disagreements more idiosyncratic?
(3) Does one model (or agnosticism among them) predominate as a favorite, for whatever reasons?
(4) Is the uncertainty generally seen to be inherent, as in complexity theory, or does there seem to be a tractable precise answer that we converge towards as data accumulate and modeling improves? (E.g. is the uncertainty and the debate about it significantly more constrained now than 10 years ago?)
These are intended to be reasonably objective meta-level questions whose answers everyone could agree upon no matter what positions they hold individually with respect to the underlying science, and would give interested outsiders a general sense of the scientific trajectory.
Thanks again.
Thank you, Chris, for another technical but very understandable discussion of an important topic. I will be linking to this page in my Greenhouse Gases page.
Thanks. Since you mention overall climate sensitivity estimates, and their uncertainties, at the end of the post perhaps I could use the opportunity to ask a question perhaps a little off the topic of feedbacks per se.
As I understand it some of the possibility of high sensitivity comes from the possibility of strong aerosol cooling (which would ‘mask’ the effect of warming in observations up to present). This paper:
Gunnar Myhre, Consistency Between Satellite-Derived and Modeled Estimates of the Direct Aerosol Effect, Science 10 July 2009: Vol. 325. no. 5937, pp. 187 – 190
seems to reduce some of the uncertainty in aerosol cooling, putting the observations in the range used by the models. My question is, does this help significantly reduce the probability of sensitivities in the 4C+ range?
Talking to “casual” sceptics (as opposed to political fanatics), I often get the impression that they – paradoxically – refuse to believe in global warming because they think any positive feedback implies that temperatures will rise forever or spin out of control.
I wish there was some kind of toy applet/flash program that illustrated feedbacks in terms of something else than climate. Maybe water flowing in and out of a tank or something. If these causal sceptics could get the feeling of figuring it out on their own, rather than being told, I think they would be less likely to just retreat to another denial argument.
Oh dear, Schmidt et al (2010) is going to be Knorr (2009) all over again. Not for anyone with a clue, of course, but for those vast acres without a clue it will be used as “proof” that CO2 does not alter the greenhouse effect because even under a doubling of CO2, it still accounts for 20% of the greenhouse effect. Yes, this makes no sense whatsoever – but these same people argue that a constant airborne fraction is the same as constant CO2, or that it means anthropogenic CO2 isn’t increasing.
How can we make sense of these things for people who leave sense at home?
Well, one step in the right direction is for people with sense to have a really good understanding of the technical details of feedbacks, and this article will be a huge help in that respect.
Thanks for such a clear presentation.
As well as Charney sensitivity and feedbacks from land ice/albedo changes (“Earth System Sensitivity”), we also have to consider feedbacks from the carbon cycle. The issue is that CO2 levels tend to go up because the temperature has risen, causing further warming.
Unfortunately, these effects can interact with an Earth System Sensitivity of ~5 degrees to give some very large temperature changes.
I’ve discussed this before in Real Climate posts: see for instance here and
here and also some spreadsheets and tables here.
It would be good to see this covered in Part 2.
Chris, thanks for taking the time to do this. I took a look at your blog http://chriscolose.wordpress.com/ and there’s plenty of good material there too. It really illustrates to me how there is no excuse for ignorance and clinging to misinformed views.
Thanks, this is very helpful. Even better would be the ability to get one click PDF versions of these backgrounder posts. Then, I could easily use them for teaching purposes.
Regarding explaining the water/CO2 difference to people, it occurred to me to use an analogy of our bodies. Everybody knows that if you drink a lot of water, you really don’t gain much weight, except briefly. You get, well, human rain (downpour?). But if you eat even a modest increase in fat regularly, over time you gain weight. Replace that fat with CO2, and the analogy for a system that can flush excess water and not something else, and I think it makes it more understandable.
Additionally, people who gain a lot of weight have more water in their body contributing to the weight, but it isn’t because they drank a lot or it.
Nice summary, except that two items popped out at me at being at odds with other analyses.
1) A small, but positive cloud feedback. Many other publication show a negative feedback associated with clouds. All show a large uncertainty associated with clouds, possibly larger that the entire CO2-induced warming.
2) Saying that CO2 is long-lived is true went comparing to water vapor. However, compared to other atmospheric gases, CO2 is relatively short-lived, witness by the large seasonal variability.
I don’t understand why the moist adiabatic profile should decline in a warmer world. Could you elaborate? Thanks.
[Response: It’s because the Clasius-Claperyon equation is non-linear. The amount of latent heat released by condensation from a saturated parcel that moves up the atmospheric column is a function of the temperature. The warmer it is the faster the saturation goes up as temperature does. So, for warmer starting conditions, you get slightly faster latent heat release during ascent, and so the effect of the moisture (which is to warm the atmosphere relative to dry conditions) is enhanced. Thus the relative increase in atmospheric warming (in unstable conditions) compared to the surface. This increases OLR out to space, and so is a negative feedback to the initial radiative perturbation. – gavin]
Since you are discussing feedbacks, I would like some expert who believes in a positive global feedback to tell ignorants what he thinks about Roy Spencer’s belief in a negative feedback:
http://www.drroyspencer.com/research-articles/satellite-and-climate-model-evidence/
For instance, in his last Figure he compares the energy flux from the earth, given by a satellite, and the average temperature of the lower atmosphere. He finds parallel pieces of straight lines, with a slope which corresponds to a negative feedback. I do not find the flaw.
The clear definition here of positive and negative feedback helps avoid what seems to have been a lot of confusion, where the idea of ‘positive’ is often thought of as a good thing and ‘negative’ as a bad thing. At least we from the electronic control system world can be assured that what we are reading from real climate science is consistent with our terminology. I think the confusion is that many people say things about the climate without any idea at all of the meaning of these words. It is easy to get the idea this error comes from scientists, so it is good to have a clear reference.
We still have a problem with the terminology in the use of the phrase, ‘mixed layer’ in the ocean. It seems that scientific modeling imagination has misinterpreted terminology from the world of underwater sound, and adapted it as if there is a heat trapping effect in such a layer. There is not. In fact, when there exists a ‘mixed layer’ that condition means that the heat is absolutely not trapped, and instead, it is churned up to the extent that there is no temperature gradient throughout that mixed layer. Below that layer, the more stable lower temperature gradient takes over without a discontinuity in the temperature function. (The derivative of the gradient is discontinuous in simple acoustic models, but in reality, not so much.) On the other hand, when there is not a mixed layer there is a mild heat containing state where there is a temperature gradient. Since it is temperature gradient that determines refraction of sound waves, this is important in the world of submarines and anti-submarine warfare, where the existence of a ‘mixed layer’ means that a submarine can be detected by a simple sonar on a destroyer at medium ranges. If the submarine ‘goes below the layer’ it is relatively secure from being detected. And if the layer goes away, as it often does in summer conditions, it means the submarine can rather easily defeat the destroyer. I only say this about submarines to point out that this terminology that you borrow is not a trivial imagining in my mind, but something of great importance that has long been known and studied in another field.
As an aside, this has particular meaning in regard to the XBT story. The purpose of the XBTs used by destroyers is to measure the temperature gradient. Nobody involved in ocean acoustics cares much about the absolute temperature, thus the efforts to use the records of these as a reference are bound to be inexact since there was never much need to control the manufacture of these to give a temperature reading. Thus, there can be imagined calibration corrections, but they have little relation to reality. In fact using these to demonstrate heat content history of the oceans is highly questionable.
But back to the ‘mixed layer’ issue. I have been writing for some time about the way ocean heat content was being analyzed, and have been particularly perplexed by the way vertical mixing is handled. Since I now realize that there is a belief that there is a mixed layer trapping heat, I think I see that the vertical mixing model is based on this simplifying error, whereby it is not necessary to model vertical temperature structure at any significant depth. But I see this as an error that is being set in perpetuity by the frequent reference to this new way of thinking about ‘mixed layer.’
Dan, cloud is tricky for a couple of reasons. In this context, one problem is that we are specifically looking at a feedback from temperatures in general, not at the net contribution of cloud in total. They may not be the same, if there are other factors bearing upon cloud formation than temperature.
The diagram Chris provides suggests that cloud feedback ranges from small positive to significant positive. I am not aware of work in which cloud feedback is observed negative — except the “IRIS” hypothesis which was really only ever a speculation and turns out to be invalid when checked with empirical measurement. But I could be educated on this if you have a reference!
I do know that the tables of forcings in IPCC 4AR (Figure 2.21, page 205) have cloud shown as a definite negative forcing (though with large uncertainties). This is, as I understand it, related mainly to the effects of cloud formation from aerosols; not from temperature change. Hence this is not a temperature feedback.
The second point, on the lifetime of CO2: unlike most gases CO2 does not have an easily defined atmospheric lifetime, precisely because there is an active carbon cycle involving a number of processes and reservoirs. The lifetime in the atmosphere of a single molecule is quite short, because it can cycle through the carbon cycle comparative quickly (a few years) and this is mainly because of the seasonal cycle. But this is not really all that significant for the longer impact, because it IS a cycle, and most of what is taken in by vegetation in one season is returned in the next. So the lifetime in the more useful sense of how long it takes a pulse of carbon into the atmosphere to be removed again is much longer, and this is why CO2 is properly considered a long lived gas; certainly much longer than that lifetime of individual molecules due to the seasonal cycle.
This is a point of confusion for a number of folks and on a number of websites. A carbon cycle tutorial style post might be a useful thing to consider at R!
#4 (Harald) – Your question prompted me to try to come up with non-climate examples of non-runaway positive feedback. I finally came up with what I think is a good one, and it’s called “positive feedback”. When someone “likes” a web site, video, story, or entry, other people are more likely to visit. The stronger the positive feedback, the bigger the increase in visitors.
In rare cases, the feedback is so strongly positive that a runaway response is generated and the site becomes “viral”. But most of the time, the positive feedback simply increases the number of visitors a bit.
I expand on this analogy at http://blogs.chron.com/climateabyss/2010/09/positive_feedback.html
Good. thanks especially for this: Estimates of this range have been based on not just GCM results, but constraints from observational data (the seasonal cycle, or volcanic eruptions) as well as the past climate record (Knutti and Hegerl 2008 provide a review).
and this: Recently, some studies have expanded on this view to also include ‘slow feedbacks’ such as the response of ice sheets and vegetation that are important on hundreds of year timescales (Lunt et al 2010; Pagani et al 2010).
A nice clear article – thanks. It was good to read something about lapse rate, which I hadn’t understood before.
It seems to me that it’s also important to point out that feedbacks will vary depending on the state of the planet at the time, and therefore the climate sensitivity will also be different at different times. For example, there is no ice albedo effect at all on an ice-free world, and I’ve read that there is potentially a very large positive feedback from methane today (in frozen soil and seabed) which would have been much smaller in the geological past.
He lost me on the graphs also. However, something he said earlier might shed light on my earlier post. He mentioned that some scientists claim that warming reduces cloud cover, which results in more sunlight reaching the surface, hence a positive feedback. If this is the positive feedback associated with clouds, then shouldn’t it be called positive feedback due to cloud reduction. Saying clouds are a positive feedback implies that more clouds lead to more warming, as opposed to the opposite.
Dr. Spencer claimed the opposite – that reduced cloud cover led to the observed increased warming.
I take it you just aren’t bothering to discuss release of sequestered (i.e. frozen) carbon as permafrost and methane hydrates melt.
Sorry, I got that ReCaptcha problem where my first attempt didn’t have the recaptcha, and when I tried again it said “duplicate post detected” so I don’t know if either attempt got through… Here’s a fresh copy just in case – please eliminate any actual duplicate.
I’d like to recommend a book on this topic for those wanting to go deeper: Understanding Climate Change Feedbacks [2003: National Academies Press]
book link on NAP website paperback for $28.35 or free PDF download for personal use only.
On the issue of how to understand feedbacks in mathematical terms, start with their Box 1.1 on p. 19-20. I found this clarified the issue nicely for me. This also addresses why feedback factors are not simply additive, and why the presence of one positive feedback can make the system more sensitive to the addition of a second feedback.
The book was produced by an editorial panel of 19 leaders in the field, chaired by Dennis Hartmann of UWa in Seattle, and including many well-known experts such as Betts, Busalacchi, Manabe, Sarewitz, Wallace, Weaver (yay Canada!), Wofsy and Wood.
– Jim P.
http://stephenleahy.net/2010/09/23/arctic-ice-in-death-spiral-risks-climate-catastrophe/
For all these RECAPTCA posts:
1. Use a browser like Firefox (http://www.mozilla.com/).
It has a feature where it remembers the form data (the comment), so that you can click back and get your comment text.
2. If you are unsure of the recaptca text, click on ‘Try another challenge’ in order to get an easier one.
3. You may also try the Lazarus Firefox addon, https://addons.mozilla.org/en-US/firefox/addon/6984/ which has a feature to remember all form data.
Plants.
(my italics)
That was from the abstract here:
http://dge.stanford.edu/labs/caldeiralab/Caldeira_research/Cao_Bala2.html
Since this must be reconciled with the “10 day rainout” criterion discussed by Gavin here :
https://www.realclimate.org/index.php/archives/2005/04/water-vapour-feedback-or-forcing/
it must be discussing a large effect i.e. the total reduction of water vapour produced by this effect in 10 days must be significant. So we must check with the body of the paper here:
Is this generally accepted? (To Chris and or Gavin)
[Response: The oceans are an effectively infinite source for atmospheric water vapour in the large scale. That is what determines the global effect. Over land, there are all sorts of reasons for humidity to do different things and changes in plant physiology might be one of them. – gavin]
Chris Colose: It would be a good idea to use the readability tool if your word processing software has one. You need to get the level down to about third grade for most people, not including most readers of RC. Your sentences are too long. I can’t do something else while reading your post.
The idea of dividing your post into sections is great. Start at 3rd grade level then shift up a few grade levels at a time. The first mathematical symbol will be an absolute stop for most people. Please use 2 mathematical levels.
Re: Schmidt et al 2010
Very unlikely trace gases hold up the water vapor. Please use some imagination.
[Response: I prefer to study the real world. Thanks for the advice though. -Gavin]
To Jacques (#12) – “Since you are discussing feedbacks, I would like some expert who believes in a positive global feedback to tell ignorants what he thinks about Roy Spencer’s belief in a negative feedback:”
At the risk of being mistaken for someone claiming to be an expert, I’ll offer my perspective, which is that Spencer’s blog article doesn’t make much sense. First, some basic concepts. Climate sensitivity entails a temperature response at equilibrium, which is approached asymptotically, typically over multi-century timescales. With appropriate models, it is possible to utilize interim data to infer sensitivity, but intervals of a year or less are very treacherous. Early on, a climate perturbation (e.g., a CO2 forcing) will induce a temperature change, which in turn initiates a feedback. At this point, the imbalance caused by the feedback will be at, or moving toward, the higher end of its potential range, but as the temperature responds by moving toward its equilibrium value, the imbalances due to forcings and feedbacks will diminish. Unless one knows where we are under these changing conditions (which can also be affected in unpredictable ways by temporarily changing dynamics of energy exchange within elements of the climate system), the relationship between forcings/feedbacks and imbalances can’t be accurately interpreted. Multi-year, and when feasible, multi-decadal data are preferable for accurate assessment.
However, there is a more serious problem with the Spencer argument. He suggests that in the case of positive feedbacks, the TOA energy imbalance should change more with temperature than in the no-feedback case, and even more than with a negative feedback scenario. To see why that is not necessarily true, consider two hypothetical scenarios – one involving warming due to a CO2 increase and the other warming from an El Nino. In the first case, theory, models, and empirical data indicate that the CO2 increases will reduce OLR (outgoing longwave radiation), and that the consequent warming will cause water to evaporate, so that its greenhouse effect further reduces OLR, exacerbating the warming – a positive feedback. Now consider an El Nino that warms the sea surface and then the atmosphere. Here, the initial effect on OLR will be the opposite of the CO2 effect – instead of OLR diminishing, it will increase due to the extra heat emitted from the sea. As with CO2, the warming will also add water vapor to the atmosphere, with a consequent greenhouse effect. In this case, however, the water vapor effect will be to reduce OLR rather than further increase it. In other words, the water vapor will amplify warming, but tend to nullify the TOA energy imbalance.
Effects on solar (shortwave) radiation are also likely to differ between a CO2 forcing and an El Nino. As an example, CO2-mediated atmospheric warming will tend to reduce relative humidity (RH) until water evaporation compensates for, but does not outweigh, the change. In contrast, El Nino warming imposed on an unwarmed atmosphere will tend to increase RH. The effect on cloud formation is likely to differ, and there is evidence for increased low cloud cover associated with El Nino events, which would increase planetary albedo and induce a cooling effect due to increased reflection of solar radiation. Although the data are scanty, some evidence suggests that in contrast, cloud cover in the lower latitudes may have exhibited a diminishing trend over several decades that would be consistent with CO2-mediated warming. The exact changes are probably less important than the concept that El Nino mediated warming and warming induced by atmospheric greenhouse gases will behave differently, and that their feedbacks will be different in magnitude and very possibly in sign.
Spencer, in analyzing short term data involving sea surface temperatue changes, is almost certainly looking at internal climate modes of the El Nino variety as a large component of his data. Even if short term analysis were a suitable instrument, drawing conclusions about climate sensitivity to CO2 from such an analysis would be misguided.
Relevant to this post, this new paper uses previous estimates of feedback strength to construct probability distributions for climate sensitivity. It shows how uncertainty about feedbacks can combine to create greater uncertainty about climate sensitivity.
Climate sensitivity distributions depend on the possibility that models share biases
Lemoine, D.M. 2010. Journal of Climate 23(16):4395-4415. doi:10.1175/2010JCLI3503.1
http://journals.ametsoc.org/doi/abs/10.1175/2010JCLI3503.1
http://www.dereklemoine.com/lemoine_2010_climate_probabilities.pdf
Some responses
#17 (Icarus): You’re right about sensitivity being a function of the base climate, but this is a decidely secondary effect over the range of climate change in consideration (i.e., so one can use the the LGM to pre-industrial change as a good proxy for the sensitivity from the preindustrial to 2xCO2). This probably isn’t very valid on the more extreme ends, like snowball Earth. Some people (Colman and McAvaney 2009, GRL) have looked at this issue, noting a general tendency for the water vapor and lapse rate changes to offset each other, and albedo feedbacks declining in importance in the warm limit. Clouds are always tough to diagnose, but they become less important in the very cold limit. Most importantly I think is the timescale and making sure you’re comparing two periods at equilibrium and over which feedbacks have similar time scale importances.
#19 (cervantes): I am not talking about carbon-cycle feedbacks, and this won’t be part of the second post either. It’s an interesting issue of course. This is really a good appeal though for defining your sensitivity to being at a doubling of CO2 rather than, say, “by 2100” or by emission scenarios, etc. The radiative transfer is concerned with the concentrations, but the actual timescale to get at some specific concentration is of course dependent on a wide variety of socio-economic or carbon feedback scenarios which are not greatly known.
#23 (Geoff): I don’t know much about the respiration issue, but I agree with gavin that the first-order effect is the ocean. You can understand a lot about the water vapor feedback from thermodynamic limitations, and the large-scale dynamics…but more in Part 2.
#24(Edward): Sorry! Creating a post (even two parts) that is appealing to everyone is difficult. I think RC has a certain standard though, and third-grade descriptions of feedbacks can be found elsewhere on the web. I’m hoping this one is readable to a wide audience though.
Harald Korneliussen @4 — The technical term \positive feedback\ means an amplifying or enhancing effect, not necessarily runaway. An example is found in what is called \power assisted sterring\ or simply \power steering\ whereby the driver does not supply all the force necessary to turn the wheels when the engine is running. No runaway effect there (beyond careless driving).
Excellent post! It addresses the middle of the road level of understanding many people seem to have when approaching this subject in earnest.
ATTN: ALL
When posting a comment, please use short paragraphs of no more than 8 lines. Otherwise the text is too difficult to read in particular for old folks like me with vision problems. About a decade ago I had a detached retina and after repair the plane of vision my right eye has downward slope of a few degrees from left to right.
If a comment is a long solid block of text, I often don’t read it.
ATTN: Mods
When reading a comment can you break up long paragraph into shorter ones?
Great post. Many thanks.
It still appears that increasing clouds decreases temperature.
http://www.atmos-chem-phys-discuss.net/10/1595/2010/acpd-10-1595-2010.html
Re #4 (Harald Korneliussen): From interactions with engineers versed in systems control theory over at Anthony Watts’ blog (WUWT), I think part of the problem is a difference in definition as to what constitutes “net positive feedback” between most climate scientists and the control theory folks.
Basically, it comes down to this: Most climate scientists think of the temperature response to the original forcing that is implied by the Steffan-Boltzmann Equation as being the zeroth-order effect and anything on top of that as being a feedback. In that usage of the term, it is possible to have a net positive feedback and still not have instability because you can just end up magnifying the temperature change implied by the S-B Equation as applied to the original radiative forcing.
On the other hand (and more in line with what is apparently done in control theory), you can also think of the increase in emitted terrestrial radiation implied by the S-B Equation as being a feedback on the zeroth order effect of the original radiative forcing. In that context, it is a negative feedback since the original radiative imbalance is reduced as the temperature rises. If you count the S-B response as a feedback then it is in fact true that a net positive feedback leads to instability. Translating what climate scientists are talking about into this language, what you really ought to be saying is that the net feedbacks EXLUDING that implied by the S-B effect are expected to be positive, but still not as large in magnitude as the negative S-B feedback, so that the net overall feedback to the initial radiative balance is still expected to be negative.
It is interesting to note that a few climate scientists have used the feedback terminology more in line with the control theory usage. For example, Dennis Hartmann in “Global Physical Climatology” talks about the “Steffan-Boltzmann Feedback” noting that it is “the most important negative feedback controlling the temperature of the earth.”
Of course, this all just boils down to a difference in terminology…The physics of what happens is unaffected. Nonetheless, I think it is important to understand the different usages of the term “net positive feedback” so that one can try to explain it to frantic engineering-types who will insistingly misunderstand what climate scientists are saying. (In my experience, explaining this to them doesn’t really help since most of them over at WUWT are wedded to disbelieving in AGW as a matter of principle and won’t believe what you tell them, but at least one can say that one tried!)
In your otherwise excellent article, you say “Although it is not feasible to trigger a runaway greenhouse like Venus even if we burned all the coal today”, I think you are being far too sanguine about it. Here’s why.
James Hansen, in his book “Storms of my grandchildren”, writes (Ch. 10): “Until recently, I did not worry much about that (a runaway greenhouse) …warmer climates may have a larger climate sensitivity; indeed, that today’s climate is not terribly far from the runaway situation.” Hansen began his career as a climatologist by studying Venus’s atmosphere.
Quoting the recent PETM study by Zeebe, Zachos & Dickens, “Our results imply a fundamental gap in our understanding of the amplitude of global warming associated with large and abrupt climate perturbations.” He concludes: ” If we burn all reserves of oil, gas and coal, there is a substantial chance we will initiate the runaway greenhouse. If we also burn the tar sands and tar shale, I believe the Venus Syndrome is a dead certainty.”
In effect, you are implying there will be no nasty surprises (apart from methane clathrates) awaiting us in the resulting heat age to trigger a runaway event (such as from high-level cirrus clouds).
It seems to me that even mainstream climatologists are now spooked by the thought of being labelled as “alarmist”, by even talking about the Venus Syndrome. However, since it would be the infinite catastrophy, it should not be swept aside so readily. Even a small risk – like 1% – is unacceptable. You don’t reassure me at all, Chris.
I agree with Harold Pierce Jr regarding breaking text into short paragraphs. Not only is this good for us old farts, it is also good practice for general communication. Steve
CC says:
\If you could remove all of the CO2 from our atmosphere, aside from making the planet more efficient at losing its heat to space (thus cooling) you would do a couple of things. First, you’d lose much of the water vapor and cloud greenhouse effects since temperatures would be too cold for them to exist in appreciable amounts. Secondly, you would also get temperatures cold enough to the point where expanding ice cover greatly enhances the surface albedo of the planet and triggers a snowball Earth.\
This is speculation and is just flat out wrong. The earth’s atmosphere is an insulating gas which retards the escape of heat from its surface. In \Weather\ by Lehr, Burnett and Zim on page 9, they state \The atmosphere as a thermostat controls the earth’s heat as automatically as in any heating system\ and \Its acts as an insulating blanket which keeps most of the heat from escaping at night.\ Presumably this refers to the 24 hour day.
The lowest temperature recorded on earh is about -90 C deg at the South pole in winter. During the lunar night (2 weeks) the moon’s surface temperature drops to -114 deg C. Since the winter at the South pole is much longer, we would expect the temperaure to drop lower than the surface of the moon, to the first approximation
The atmosphere also prevents the earth from overheating by removing heat via conduction and convection. During the lunar day, the moon’s surface is about 100 deg C whereas the highest temperature on earth is about 55-60 deg C.
The wind is the mechanism that transport water into the atmosphere and is far more important than simple evaporation of water in still air. This is why we use fans to keep cool on hot days and nights.
> some imagination
Science Fiction Atmospheres. R.T. Pierrehumbert.
http://geosci.uchicago.edu/~rtp1/papers/BAMS_SFatm.pdf
I’ll attempt some explanations:
(4) For a simple demonstration of how a simple feedback loop works, consider the following problem: We have an initial perturbation of 1unit, and a positive feeback coeffieient of F, which takes time delta T to be realized. So at time delta T we have changed the system by 1+F. Now the feedback is also driven by the change due to the feedback, so at time two delta T we have F operating on F, and our response is 1+f +f*f.
We can continue this indefintely and get an infinite series. We know from high school algebra that the sum of this series is 1/(1-f). If F is less than one, the series converges. If F=-1 it won’t converge, but our response would be to oscillate between plus and minus one. Obvious if F is greater than or equal to one, our system is unconditionally unstable.
(12) Jacques, regarding Spencer.
Do the following thought experiment, construct an atmosphere in a box, initially in equilibrium. Now heat the left side, but not the right. You get a convection cell (Hadley) where the air is rising in the left hand side, and sinking in the right hand side. The (warmer) lefthand side will be cloudier than the right hand side. Spencer measures the radiative differences between the two, and assumes that the sensitivity of radiation balance as a function of surface temperature is the same as if we hade raised the temperature equally on both sides. Is there any justification for this assumption?
You could also do a similar though experiment by heating and cooling the whole box on a short time scale. Again is the observed sensitivity of the radiation balance the same as for a slow heating?
Can you help us promote the official cop16 twitter and facebook account, will be deeply grateful!!. thanks
its facebook.com/cop16
and twitter.com/cop16
“To put this in perspective, it would take about five doublings of CO2 or a 7% increase in the total solar radiation hitting the Earth to produce the magnitude of climate change typical of glacial-to-interglacial transitions. Changes of this sort are well outside the bounds of what is characteristic of proxy records and observations, so this must mean that various feedbacks act to change the temperature much more than 0.3°C for a watt per square meter forcing”
Chris, I think that there is a logical loophole here : the fact that the observed variation is much larger than the linear response doesn’t mean automatically that this is due to a feedback. It can be attributed also to a change of the linear response itself – if this change is not due to the temperature , it is not a feedback. Another thing is that you speak of glaciations, but nothing insures that the feedbacks are the same for colder temperature and for hotter ones (an obvious reason is that the change of albedo due to ice -covered land doesn’t exist anymore during warm periods) . The simple idea of a linear system with a single amplification factor is obviously oversimplified.
re #29, feedback, etc…
From an electronics viewpoint: Positive feedback IN ITSELF basical DOES describe a runaway effect POTENTIAL, however in a given system, limits inherent to the system may prevent an actual Systen Runaway EVENT. An example is a comparator, where positive feedback causes a rapid shift to an extreme(stable)state. The device can go no further, but will quickly go to that limit.
The proposed “analogy” of power steering in an automobile is probably closer to an example of Negative Feedback, as used in a typical AMPLIFYING SYSTEM; the amplifier has extreme system gain (perhaps 1 million) but through using negative feedback of the proper amount we can set that gain(damping)to a ratio usably smaller (perhaps 2 hundred)and therefore a few lbs of pressure on the steering wheel is amplified SOMEWHAT to the force necessary to easily stear the vehicle.
In actuality such a hydraulic/mechanical system’s gain is probably set by simple physics laws and (feedback) doesn’t apply as much to the amplifying stage, however to maintain stability and smoothness of the actual operating process details, there are negative feedbacks (damping)in the design, even such simple ones as rubber bushings.
My 2 cents worth :)
Another article on cloud feedback. This tends to support the notion that decreasing cloud cover leads to increases in temperature. Hence, those models that predict decreasing cloud cover in a warming world have a positive feedback associated with clouds, while those that predict increases in cloudiness have a negative feedback parameter. Both, however, maintain that clouds are still are large uncertainty in the whole equation.
http://www.arm.gov/publications/proceedings/conf08/extended_abs/delgenio_ad.pdf
Dan, the paper you cite explicitly suggests a positive feedback. I appreciate that there are large uncertainties with cloud, but the paper cited notes in particular (last paragraph)
So I *still* don’t see any good basis for disputing the usual characterization of cloud feedback as positive, but with uncertainty in the magnitude, as described in this post by Chris Colose.
If models give increasing low cloud with warming, then those models have a problem matching the empirical data from ISCCP (International Satellite Cloud Climatology Program). The paper you cite identifies the GISS GCP as a model that does match the empirical observations low cloud change with rising temperature, and consequent positive feedback.
You said earlier (#10) that “Many other publication show a negative feedback associated with clouds.” I’m skeptical on that, frankly, though I’ll retract in a heartbeat if these publications can be shown. I’m not an expert on the topic. The conference paper you cite by Del Genio and Wolf does not show negative feedback, however.
I’m still speculating that perhaps you may have mixed negative feedback with the negative aerosol forcing which is indeed given in many publications and which is described in the IPCC 4AR; but this is a pure guess on my part. I’m not trying to be presumptuous, but it simply doesn’t match my reading to say that there are many publications showing negative cloud feedback.
Cheers — Chris
to Chris:
I liked this article but have have a point which will probably be redundant after Part 2.
The key sentence is very clear:
But accounts vary as to what constitutes the input and output. For example you say:
This looks like a very common choice of ‘system’ i.e. with input in watts/m^2 and output in units of global warming i.e. degs.C
In this choice, referring to global warming as negative feedback would be double counting because the Planck function (Stefan Boltzmann to Joel Shore) has been included … at least the first part of it has already and the rest will soon be…
You get:
global warming so far > more water vapour >…more global warming again in degs.C
Finished. You have now accounted for all of the Planck radiative feedback; you can’t add it again. You understand this, but will beginners? especially when you include this correct remark associated with an alternative definition of output:
“Cooling” ? Thats right because you are now losing more energy. Your output variable is now watts/m^2 instead of degs.C
But to the beginner the previous sentence could look a bit Orwellian, i.e.
“warming is the ultimate cooling”
which is right but a confusing consequence of the ambiguous nature of the terms cooling and warming.
I hope that you will mention that there is more than one way of doing this.
#Re 34
I have my doubts. Anyone who understands feedback in control theory and who wants to understand climate feedback should have no problem with definitions. In fact the climatologist Michael Schlesinger claimed somewhere to have imported some of these ideas from control theory.
Chris,
The paper I cited suggests a positive cloud feedback because they content that clouds will decrease in a warming world. The decreasing cloud cover would allow more solar energy to reach the surface, hence, warming.
If the models indicate that clouds increase, then (according to this paper) that would cause a negative feedback. Remember, the knock on Swenmark’s theory is not that increasing cloudiness leads to cooler temperatures, but that incoming galactic rays lead to more cloudiness.
Based on the research, it appears that increases in temperature will lead to increased water vapor which will lead to increased cloudiness, which will lead to a decrease in temperature. The contention appears to center on cloud formation.
My understanding is that nighttime clouds are warming, due to trapping heat, and daytime clouds have both warming effects through the same mechanism, and cooling effects b/c of reflecting sunlight back out of the atmosphere. Now I don’t have much else on that topic stored between my ears at the moment, and i’m dashing this off on my lunch break so I’m not in a position to look it up. But it occurs to me that it’s probably not nearly so simple as “increasing clouds” or “decreasing clouds.” Different cloud types presumably have different albedo and different heat trapping characteristics. Clouds at different times of day are going to have different impacts. Clouds at different latitudes–if only because of the amount of sunlight, but perhaps for other reasons–are going to have different impacts.
So I suspect that one would need to know a lot more detail about the cloud changes that are likely to happen as the world warms than “more or less of them” in order to say what the net feedback would be. And possibly even to be able to say what the sign of the feedback would be.
I do understand the hypothesis of negative cloud feedback, Dan… I am simply skeptical that it has much published support. I’m not trying to be a pain, but for clarity — what I questioned is the many publications. The sentence I queried is specifically this:
Is that really true? I don’t think so, but I’m happy to see examples, if there are any.
I’ve not seem many such publications, apart from the old and by now pretty solidly falsified IRIS proposal, from Richard Lindzen and colleagues. “Does the Earth have an adaptive infrared iris?”, by Lindzen, Chou, and Hou (2001). (Bulletin AMS 82: 417–432).
This was not model based, and it has been pretty solidly tested and falsified.
I can believe some models might have a negative cloud feedback, but I don’t know of any off hand and am not aware of a large number of publications on such models.
Cheers — Chris
Kevin,
Agreed. Much research has been which suggests that high, cirrus clouds have minimal daytime cooling, but larger nighttime warming. Low-level, stratus clouds have some nighttime warming, but a much higher daytime cooling. The biggest uncertainty concerns the storm clouds. Much work has been done describing the tropical cooling effects associated with the daytime buikdup of these clouds, but very little is known in the mad-latitude regions.
Clouds have the potential to greatly enhance global warming or completely moderate it. The literature appears to lean towards a net cooling affect with increasing cloudiness, with a very large uncertainty.