RealClimate logo

What triggers ice ages?

Filed under: — rasmus @ 16 February 2007 - (Português) (Türkçe) (Français)

by Rasmus Benestad, with contributions from Caspar & Eric

In a recent article in Climatic Change, D.G. Martinson and W.C. Pitman III discuss a new hypothesis explaining how the climate could change abruptly between ice ages and inter-glacial (warm) periods. They argue that the changes in Earth’s orbit around the Sun in isolation is not sufficient to explain the estimated high rate of change, and that there must be an amplifying feedback process kicking in. The necessity for a feedback is not new, as the Swedish Nobel Prize winner (Chemistry), Svante Arrhenius, suggested already in 1896 that CO2 could act as an amplification mechanism. In addition, there is the albedo feedback, where the amount of solar radiation that is reflected back into space, scales with the area of the ice- and snow-cover. And are clouds as well as other aspects playing a role.

Orbital forcing Martinson & Pitman III’s hypothesis states that the fresh water input works in concert with the Milankovitch cycle and the albedo feedback. They conclude that ‘major’ terminations can only follow from glacial build-up of sufficient magnitude to isolate the Arctic, inhibiting the inflow of fresh water to the point that salinity buildup in the surface layer from slow but continuous growth of sea-ice, causes overturn of the Arctic (through the effect on the atmospheric circulation and the ocean currents). The vertical overturning brings warmer water up from below, setting conditions that are more favourable for ice metling. Salinity plays a role too, but the hypothesis does not mention variations in the greenhouse gases (GHGs). A few questions: Did Martinson and Pitman III forget this last point? Or did the GHGs only represent a minor contribution? And, could not changes in GHGs explain much of the variability? On the other hand, it sounds plausible that changes in salinity and fresh water input may affect the sea-ice formation and the deep convection. However, so far, the hypothesis proposed by Martinson and Pitman III is merely a speculation, and we are waiting to see if the hypothesis can be tested through numerical model experiments (which would require higher resolution sea-ice and ocean models than used in todays global climate models). It would be interesting to carry out experiments to assess the significance of the fresh water only, GHGs, and the combined effect.

One reaction to the Martison and Pittman paper is: Where is the calculation of energy? Greenhouse gases only contribute a couple of W/m2, vs. the seasonal Milankovich forcing of >40. For this new idea to have merit, it had better have heat fluxes at least on par with the radiative forcing from CO2. Previous modeling studies find that GHG make up roughly 50% of the total LGM to present temperature response (see e.g. Broccoli & Manabe), the other part being albedo etc that respond to the seasonal cycle of irradiance. It is tricky to completely isolate the individual causes because changes in GHG may produce altered cloud and sea ice distribution. But roughly speaking, if you do an LGM run and only reduce sea level, put in the ice sheets, change the vegetation, add some dust (though that one is still rough), then you get about 50% the way you want to go. Change the GHG concentrations and you get close. This is more or less what Manabe and Stouffer showed 15 years ago. The question is do we need anything else, really, and does that ‘anything else’ pack sufficient punch.

139 Responses to “What triggers ice ages?”

  1. 101
    Hank Roberts says:

    Barton, I looked. Don’t go there, it’s newage stuff.

  2. 102
    mark s says:

    RE #99,

    Lynn, i agree that you are at one end of a continuum of opinion, that exists on this website, with our climate science friends being somewhat closer to the centre. I think it is very useful to have a range of legitimate opinions.

    I have had a look at Mark Lynas’s website, i would say that he is lining up along side Fred Pearce and James Lovelock. Perhaps the key point in the ‘super-alarmed’ arsenal, as i see it, is the extent to which the 1.5 to 2 DegC of warming we are commited to, begats further feedbacks.

    There must be a curve which describes the amount of feedback that is created by each 1 degC of extra warming. Is there any plotted historical data on this one?

    Is ‘super-alarmed’ a legitimate, scientifically valid veiwpoint? I feel that AAAS president (Prof Holdren) is setting the right tone, in saying that we are driving towards a cliff in a car with bad brakes, in the fog. So how well are the scientists among us sleeping? Holdren says we are now fighting to save our planet, not from dagerous CC (which is here), but fromcatastrophic CC.

    Our moderators have been doing a tough job (ie fighting the deniers propoganda machine), for a long time. I think this site is a demonstration of their commitment to getting the ‘message’ out. Maximum respect for that. :-)

    So now the battle against the deniers is nearly won, will our climate scientists be ‘coming out’, as having more in common with the ‘super-alarmed’ than previously stated? Should we all cut the flannel, because 2007 is gonna be even hotter, and therefore accept that Greenland and the WAIS are as good as gone?

    I think that the battle-lines, on this blog, and in the world in general, may be redrawn, as the pressure on the scientists, in the future, comes from the ‘super-alarmed’, not the deniers.

    I am keen to hear that people tell us we are hysterical, Lynn, and to metaphorically slap us around the face. This may be necessary on an increasingly frequent basis!

    You might be interested to know, especially if you live in the US, that I am applying to study for a degree, starting in September, in Flood Risk Management, as part of an initiative by the Environment Agency!Fortunately, my socialist (and probably KGB inspired) government is prepared to pay my course fees, for my accomodation whilst on placement, and it is also offering £12k a year, which would cover living costs. Is that very generous, or merely telling? Here in Europe, the wallets are being opened! Any way, i might not get in, so fingers crossed…

    My apologies for filling your blog with unscientific, off topic ‘noise’, but RC is the only quality forum I have. It has undoubtably inspired and informed me, in thinking about my future.

    Regards, Mark Schneeweiss

  3. 103
    Hank Roberts says:

    >97, recommends

    This is a good one I’ve read about before — there’s solid science done on past use of charcoal as a soil amendment, and also good work done on extracting hydrogen from carbohydrates, leaving charcoal behind that can be added to soil.

    Following that link led me to a cite Erich posted there, that I think merits posting here:

    [12] Day, D., Evans, B., Lee, J., Reicosky, D.
    “Economical CO2, SOx ,and NOx capture from fossil-fuel utilization with combined renewable hydrogen production and large-scale carbon sequestration”
    in press for Energy: The International Journal; (pre-press copy)

  4. 104
    Matt says:

    Did someone mention the effect of the ice mass on the Arctic water passage. I understand that the ice can push down the land masses substantially. The ice builds up, the arctic passage opens as land mass is squished, the ice warms up as warm water inflows.

  5. 105
    David Price says:

    I can’t access the Science article, but the theory concerned the cause of ice house periods generally, not the 100,000 year periods within the cycle, where the dust theory does seem a bit dubious.
    Some mechanism that dims the sun at certain intervals enough for orbital cycles to plunge the Earth in to ice ages seems necessary. Any good theories about?

  6. 106

    Reid, you’re absolutely right. I misspoke, what I should have said was “padding the frequency domain with zeroes is essentially the same as interpolatig the time series” (in MRI, we acquire data in frequency domain, so I always think “backwards”.) And you are also right that you will essentially get some error, I assume from truncation or ringing of some sort.

    I knew you werent critiquing tamino- i just found the discussion a fun tangent :)

    For the purposes of these graphs, a crude analysis would be simply interpolate the time series onto a power of 2 and evenly-sampled axis, and then slam an FFT on it. I doubt a more rigorous approach would give substantially different results. The idea really is just to see where the frequency spikes overlap, if at all.

  7. 107
    David B. Benson says:

    The simple harmonic oscillator provides an ultra-simple model of ice ages. I tuned the oscillator to have a natural period of about 125 ky and then added orbital forcing to vary the tuning. The resulting system is a more complex form of the Mathieu equation in that the orbital forcing occurs at several different periods, not just one. Nontheless, the results are similar. I ran the program for a simulated 2.688 My.

    The response has 48% of the power in the 85.3–128 ky band, 40% of the power in the 64–85.3 ky band and only 8% of the power in the forcing bands. Interestingly, about about 4.6% of the power is in bands longer than 128 ky. (Numbers do not add to 100% due to rounding.)

    The response is ordinarily rather slow. The fastest rates of change are about 5% per ky, so the fast recovery from LGM is not well modeled.

    Nonetheless, this little exercise shows that even frequency modulation of a harmonic oscillator suffices to demonstrate some aspects of ice age climate.

  8. 108
    tamino says:

    Re: playing with models of glaciation

    The simplest model of ice volume which gives anything like a realistic picture is the “Imbrie model” (Imbrie & Imbrie 1980, Science, 207, 943). Ice volume V is actual expressed as its deviation from a rather high value, so the modelled parameter V is usually negative. The model is

    dV/dt = – (1 +/- b) (I(t) + V) / T

    The function I(t) is the driving function, often taken as the midsummer solar insolation at latitude 65deg. (June 21st for the northern hemisphere, December 21st for the southern). T is a time constant, giving the timescale of the response to forcing (10 ky is not an unreasonable value). The constant b gets added during deglaciation, and subtracted during glaciation, to simulate the observed fact that deglaciation tends to happen faster than glaciation; b=0.3 is not an atypical value. If you enjoy playing with models, try a variety of values for the parameters, and try different parameters for the different hemispheres.

    One of the problems is that the forcing function (midsummer insolation) responds to the precession cycle more strongly than to the obliquity cycle, while observed global ice volume shows greater response to obliquity than precession. But Raymo et al. (2006, Science, 313, 492) have made a strong case that the reason is that obliquity affects both hemispheres the same while the effect of precession is 180 degrees out of phase between the hemispheres. When precession causes ice to grow in one hemisphere, it shrinks in the other, so for the impact on global ice volume the opposite hemispheres mostly cancel out the precession effect. For obliquity, growth and decay are in phase between hemispheres.

    Of course there are still unanswered questions. Glacial changes are much stronger during the last 800 ky or so, than in the preceding 2 million or so; this may be due to slow but steady decrease in the “zero point” of CO2 concentration. Also in the last 800 ky or so, there is a strong 100 ky cycle, originally attributed to eccentricity (rather than obliquity or precession), but that idea is now out of favor (but not dead). The best idea I’ve heard is that recent glaciations (global ice valume) are driven by obliquity, but for some reason deglaciation is only triggered every 2 or 3 obliquity cycles (Huybers 2007, Quaternary Science Reviews, 26, 37). The reason is unclear (but I’ve just submitted a theory on that to GRL).

  9. 109
    Richard Allan says:

    Regarding the driving force between changes in the Earth’s orbit around the sun and glacial cycles, I have seen no comment on the recent paper by Gerard Roe
    (available draft version | final GRL article)
    in which he clearly shows that 65N solar radiation is linked to the time rate of change in ice volume (dV/dt) rather than the ice volume, V, that is commonly used.

    I am no expert but his arguments seemed convincing to me. Add in ice albedo feedback, CO2 feedback (e.g. colder water can dissolve more CO2) and water vapour feedback and dust then the large swings in global mean temperature would appear reasonably consistent with theory.

  10. 110
    Margo says:

    Re 87:

    [Response: I get yo….. We know to a pretty high accuracy what the man-made greenhouse gases have done, there is no comparable estimate for GCR effects. That has nothing to do with heuristic arguments about the glacial-interglacial cycle. -gavin]

    Thanks for the more detailed answer on the GCR effects; this is much more satisfying than the “space alien” bit.

    As to this bit: “We know to a pretty high accuracy what the man-made greenhouse gases have done..”

    Steve Bloom will be stunned to learn that I agree with that to a large extent particularly since you don’t need a GCM to predict that there should be some effect, and many of the uncertainties have nothing to do with what’s involved with coding GCM’s. ( Example: clouds. )

    (BTW. I think given Steve’s comments, I think it’s worth nothing that the just because I think some of the narratives I read about GCMs seem to oversell, that doesn’t mean I think they are useless. My reading of the papers suggests many things are done with degrees of approximation just somewhat beyond the level done back when Launder and Spaulding were working on models. In many ways I’m even ok with that but certain types of overstatement set my teeth on edge. )

    Re 88: [Response: ….. eff explained this in words, to be more accessible, but it’s quite easy to cook up a pair of differential equations with an external periodic forcing (think Milankovic) which illustrates the point in quantitative terms. By the way, many scientists have proposed specific testable mechanisms for the CO2 glacial-interglacial cycle. They have all been falsified, but that’s just science. …. …. –raypierre]

    Raypierre: I recognize that what Severinghaus considered to be his main point. For what it’s worth, I realize the fact that CO2 lags doesn’t necessarily mean T “causes” CO2 to rise. I am familiar with positive feedback (like the kind that caused this bridge to fail.)

    I also know it should be easy to cook up a pair of differential equations that reflect the narrative in that article. Or at least it should be easy if we understand the effect of temperature on CO2 uptake or release and if we also understand the effect of CO2 on factors that affect the energy balance.

    One thing I was trying to find out was has anyone done it? Have the solutions of these DE’s differential equations been compared to the data and found to work? (Or not?) Do we have mathematical models for the effect of temperature on the CO2 uptake and release?

    You seem to be saying maybe some mathematical models involving the CO2 Temperature linkage have been proposed but found failure? I agree failure is part science. :) And, as it happens I’d be interested in reading papers describing the failed models. (And not, because I’m trying to find holes– I’m curious about what someone tried to come up with in equation form rather than just words. Do you have references? Names? Titles? I can find them, but it’s a bit hard to google with random words.)

    Re 88: Steve: I did read what you said. I asked for something specific. You decided to suggest a reference that did not discuss that specific thing. Thanks anyway.

  11. 111
    Steve Bloom says:

    Re #110: Margo, it seemed to me that a list of various modeling efforts would be helpful in that you could go check them out. Of course each of those modeling efforts would need to involve some sort of algorithm along the lines of what you say you want to see, but I guess what you really wanted was a direct pointer to one or more of them. Sorry, I’m interested in this subject but not interested enough to devote time to such a search.

    By coincidence, though, I did just happen across this new paper (not through peer review yet). It uses an AR4 GCM (recalling that as of two years ago GCMs weren’t being used for this sort of thing, but these authors have access to loads of time on an advanced supercomputer) and seems to get pretty good results. I haven’t read through it carefully yet, but it does include an equation or two.

  12. 112
    Margo says:

    Re 111: Steve Bloom. I don’t know why it seems to you that student paper will help me find what I want. I’ve downloaded many of the papers discussed in section 2.4 in the paper you linked, and let me assure you, they don’t describe the sort of thing I am looking for. The paper you found today is even further from the mark.

    Raypierre can very close to giving a qualitative description of what I want when he said:

    [[…..eff explained this in words, to be more accessible, but it’s quite easy to cook up a pair of differential equations with an external periodic forcing (think Milankovic) which illustrates the point in quantitative terms. By the way, many scientists have proposed specific testable mechanisms for the CO2 glacial-interglacial cycle. They have all been falsified, but that’s just science.]]

    I agree that — if the physical processes are sufficiently well understood — then it would be quite easy to cook up these equations. If I’m understanding Raypierre correctly, he is suggesting he is aware of some people who have attempted the “quite easy” task of cooking up equations of this sort, but their models didn’t pan out — that is to say, they were falsified.

    If Raypierre knows the names of authors who documented their efforts to develop these models and compare them to data , I’d like to know those so I can search for the papers. (Falsified models are not generally widely cited. So, if Raypierre doesn’t mention some names to give me a place to start, I really don’t think I’m going to find them.)

  13. 113
    Hank Roberts says:

    … a close temporal coupling between events in the tropical and high latitude North Atlantic during the last deglaciation [1]. We recently extended this record to approximately 120,000 years BP in order to track vegetation change over a full glacial cycle at millennial to orbital timescales. High frequency oscillations in the δ13C composition of long chain fatty acids during MIS 3 appear to coincide with Dansgaard/Oeschger variability in high latitude ice cores, with positive (negative) excursions occurring during stadial (interstadial) periods. The largest enrichments (up to 8 per mil) are associated with Heinrich Events in the North Atlantic. After a relatively stable MIS 2 period, Termination 1 is marked by a rapid 13C depletion over the Glacial-Bolling transition followed by a return to somewhat heavier values during the Younger Dryas, similar to earlier observations [1]. These high frequency fluctuations are superimposed upon a long-term trend that tracks the variation in overhead insolation…..

  14. 114
    Hank Roberts says:

    There’s always something interesting — setting aside the fact that these astronomical variations are minuscule compared to the rate of the current human contribution to climate change, there’s plenty of interesting science.

    Someone can now add this variable to all the others:

    “One pole of the sun is cooler than the other. … data from the ESA-NASA Ulysses spacecraft.”

    (One _magnetic_ pole; oooler temp. difference flipped along with the field direction, the one time so far that’s been observed.)

    I don’t know if that temperature difference detectable from the unique orbital perspective over the Sun’s poles is also detectable in the plane of Earth’s orbit now that we know to look for it (that is, I don’t know if the whole Solar System is warmer on one side of the Sun’s magnetic field — it’s a funny asymmetry, eh?).

    And on the cosmic ray stuff:

    And New Scientist last week had a good article on the upcoming chance to slingshot another deep space probe — there’s a configuration coming soon that would allow a spacecraft to pass Jupiter and be slung very fast out along the line the Sun will be traveling, which could get among other things advance info on any change in dust levels; one’s certain to happen when we leave the local ‘bubble’; there may be streams or clouds of dust. And yes, Svensmark is mentioned — changes in dust will have local consequences, changing cosmic ray rates, though no one’s quite sure yet how. No online text that I am aware of; library/sub required.

  15. 115
    Reid says:

    Aziz – not to beat a dead horse, but just be aware that interpolation is like low pass filtering and it will smudge and attenuate the higher frequencies. Unless it’s a high order interpolation, which might amplify portions of the spectrum.

  16. 116
    Hank Roberts says:

    There’s a (distorted) echo?


    and this:

    Both appear to be the exact same topic, attributed to RC.

    But the first link takes you into an independent and different thread of responses:

    [This copy of this response is posted at RC]

  17. 117
    Lynn Vincentnathan says:

    Re #102, thanks, Mark. And I, too, would like to know about:

    “There must be a curve which describes the amount of feedback that is created by each 1 degC of extra warming. Is there any plotted historical data on this one?”

    I know Mark Lynas’s book SIX DEGREES is coming out March 19th, and I’m hoping he’s addressed this very question.

    GW is a big, complex issue, & we need some people to bring it all together in lay terms.

    I just read that the Texas Water Dept Board projected that 85% of Texans will not have adequate water by 2060 during drought conditions….and they didn’t even take GW into consideration, which would imply a still worse scenario.

    So glad you are going into a field that could use a GW perspective.

  18. 118
    David B. Benson says:

    Re #102 & #117: Mark Schneeweiss & Lynn Vincentnathan — The paper “Positive feedback between global warming and atmospheric CO2 concentration”, by Marten Scheffer, Victor Brovkin and Peter Cox, describes an important part of what you wish to know.

    You can either web trawl for the preprint version of the paper or else find the comment here on RealClimate where the link was posted on the 19th or 20th of this month.

  19. 119
    Steve Bloom says:

    Re #112: I imagine it’s the case that not every piece of a model appears in a paper, Margo; if anything I would expect just the opposite until such time as one of the teams thinks they have it all nailed own, at which time I would expect all of the details to be carefully described in papers. Your next obvious step is to email the modelers, since the one thing we can be confident about is that each of those models must contain a version of what you’re after. Also, not to put words in Ray’s mouth, but when he said “falsified” I don’t think that was the same thing as saying that there aren’t valid algorithms that describe the relationships between the things we have paleo records of (e.g. CO2 levels, ice sheet mass and location, and vegetative response per the paper Hank just linked) or can calculate with exactness (Milankovitch forcing). The Japanese team does appear to be quite close (with e.g. far better results than the McGill model was getting a couple of years ago).

    What’s still missing, and the reason why their model too continues to be falsified, is that there are literally dozens of important feedbacks and physical parameters that must be gotten right in order for the model not to be falsified. They named a few that they know they need to continue to work on, and I have to say it seemed like a pretty short list. Comparing their overall results to what the McGill team was getting a couple of years ago, it seems likely that they have the guts of it right. (That said, I’d love to know what other modelers think about this.) Anyway, we’ll see what happens over the next couple of years as they plug in those changes.

  20. 120
    Steve Bloom says:

    Just to add to the above a little: I’m obviously no expert, but it seems to me that it would be a comparatively trivial exercise to develop a set of equations that describe, e.g., the relationship between CO2 levels, insolation and ice sheet mass. Could you use that set of equations in any direct way to describe the timing and location of the ice sheets? I suspect not. As I think about it, it’s not entirely clear to me that the modelers would need to go through a formal step of developing such a set of equations unless those steps were amenable to plugging into the model in a fairly direct way. I did notice that the Japanese team had a fair number of equations they used to describe various relationships within their model, but those may not be taken from the model in a strict sense. IOW, did they write those equations after they wrote the code simply as a convenient way to describe what they did?

    Putting this whole thing another way, would it be possible as to state an entire GCM as a set of formal equations (as opposed to code)? Whether it is or not, it seems clear that it would be pointless.

  21. 121
    David B. Benson says:

    Re #120: Steve Bloom —

    B. Saltzman
    Dynamical Paleoclimatolgy
    Academic Press, 2002

    is an attempt to develop equations to explain ice ages in a rigorous manner. I am finding the book quite useful, so at least I do not find Saltzman’s effort to be pointless…

  22. 122
    William Astley says:

    In Reply to Robert’s Comment 67 to my comment 66. “How about this one?
    “… (All researchers agree that the very, very rapid field changes could not possibly be due to changes in the earth’s core. If the problematic data is correct, the earth’s magnetic field is not generated in the core.)”

    Attached is a link to a summary that provides papers and a discussion of the salient scientific issues.

    Physical issues with explaining rapid field changes:

    1. Did the very, very, rapid geomagnetic field changes occur? Yes three papers, two different authors confirm, the event happened.

    2. How rapid is the observed geomagnetic field change of 6 degree/day. Current field change is 0.5 degree per year.

    3. Physically, why is it difficult to explain this rapid change? If the geomagnetic field change was caused by changes in core, the outer core fluid velocity would have needed to increase a 1000 times (what is the physical reason for a 1000 times increase?) or there would need to be a more complex field configuration in the core than is currently used in computer models. (i.e. A theoretical configuration of core currents which is believed to be physically not be possible. No mechanism to move the core fluid to create the observed pattern.)

    4. Regardless and most important, the mantle is slightly conductive, rapid changes in the core magnetic field would create currents in the mantle that would resist the changes. Based on the theoretical geomagnetic field computer models (Which are all incorrect as the geomagnetic field is not generated in the core. Other theoretical problems with core as source, such as what is driving the core fluid motion? Heat flux problem.) it is believed (and it is stated in paper’s) that reversals take a couple of 1000 years to complete.

    5. As the sea floor sediment proxy data, filter’s the geomagmetic field changes, everyone assumed field changes took 1000’s of years to complete.

  23. 123
  24. 124
  25. 125
    Hank Roberts says:

    Gah. I thought I made this up as fiction a few weeks ago. But no.
    Can’t get the page to open, just Google Scholar hit:

    External forcing of the geomagnetic field? Implications for the cosmic ray ux�climate variability
    – J Wendler – Journal of Atmospheric and Solar-Terrestrial Physics, 2004 –

    “… Thus, the reversal rate of our planet is correlated … This could mean that the geomagnetic
    field is not … The polarity change between the Kiaman-Reversed and the … “

  26. 126
    Steve Bloom says:

    Re #121: Interesting. Was he then using those equations in a model, and if so could you expand on that a bit?

  27. 127
    David B. Benson says:

    Re #126: Steve Bloom — With pleasure. The late Professor Satzman, with colleagues, develop a system of three first order nonlinear differential equations to represent ice mass, carbon dioxide and ocean temperature, all globally. Using reasonable values for the parameters, the model is tuned to naturally resonate at a 100+ ky period. Then orbital forcings are added. The result gives a qualitatively good match to the most recent four ice age cycles, according to the 1990s data that they used.

    They then go on to somewhat more sophisticated models which assume a secular downwards trend in carbon dioxide over the last 5000 ky. Running this model gives quite a decent, but qualitative fit, for the entire Late Cenezioc ice age. The fits for the latest 700 ky are quite respectable.

    The conclusion is that the ice age cycles occur because the climate system has a natural period of about 100 ky. The orbital forcings just set the phase. I interpret his view as implying that eccentricity has essentially nothing to do with the cycle length.

    I am such an amateur at this that I am not in a position, yet, to suggest revising his constitutive equation for the rate of change of ice mass. But either I am missing something or his equation is overly simple. Either way, he has done a most impressive piece of work in reducing the climate system into a form which can be expressed with only 4 free parameters. (9 for the longer term model.)

    It is interesting to note that in this work, even the carbon dioxide feedback does not suffice alone to quickly end the major glaciations. There is a calving effect included once there is sufficent bedrock depression. With this, the modeled ice masses quickly and appropriately decay.

  28. 128
    Sashka says:

    Satzman did interesting work but ultimately it was a blind alley because the equations that he and his colleagues were solving could not have been rigorously derived. Their counter-argument would be the same point that I keep trying to make here: the GCMs are even worse, actually much worse because in addition to arbitrariness of parameterizations and complexity they use a lot more “free” parameters.

    [Response: Not so. The difference between GCMs and more heuristic models is precisely because they include more real physics that can be indpendently measured. The number of ‘free’ parameters is actually not that great, and since only a handful are used to tune the models ‘holisitically’ the arbitrariness is limited. As time goes on, assumptions that were initially made are tightened up based on real observations, and as that has happened simulations have got better – and that is not because they are better ‘tuned’. It is in some sense optimistic, but the principle that drives GCM development – that improvements in basic physics lead to improvements in climatology – seems in fact to be empirically true. – gavin]

  29. 129
    David B. Benson says:

    Re #128: Sashka — Most important, I committed a typo. His name was Barry Saltzman. The book is cited in an earlier post of this thread.

    I, at least, find this work useful in obtaining a fairly rigorous derivation of equations which approximately describe ice age climate. There are simplifying assumptions, of course. But of greatest interest are two assumptions. (1) Carbon dioxide feedback is important. (By now I suppose most are convinced of the reality of this.) (2) Earth’s climate has had a natural period of about 100 ky for about 900 ky and a shorter period before that. Saltzman and coworkers offer an explanation of this shift in natural period.

    Being a relative new-comer to amateur paleoclimatology, I am impressed with how well they did without the consumption of vast quantities of computer resources. The paper by A. Abe-Ouchi et al. linked in a comment earlier on this thread seems to largely confirm that Saltzman and his colleagues are correct.

    Of course, there may still be debate regarding the existence of a long natural period for ice age climate, since it seems to be much longer than the periods for any of the feedbacks and forcings (other than the trivial variations due to eccentricity). This is an interesting question still, at least for me.

  30. 130
    tamino says:

    Re: #129

    A good deal of research focuses on the roughly 100 ky “cycle” not actually being a period, but being a “characteristic timescale.” Essentially, the system exhibits stochastic behavior, but because of the nature of the beast, this leads to major deglaciations on a roughly 100 ky timescale. Carl Wunsch and colleagues are probably the strongest proponents of such a viewpoint.

    The “shorter period” which precedes the mid-Pleistocene transition is 41 ky, and is most assuredly due to the cycle of changes in earth’s obliquity (axial tilt); there’s no doubt.

    As for eccentricity forcing, its importance is an idea which is indeed out of favor, but by no means dead.

  31. 131
    Sashka says:

    Gavin, it’s not only the number of free parameters. The functional form of the closure hypotheses is not justifiable, for example, neither for momentum nor property turbulent fluxes on subgrid scale. To be clear, I don’t blame anybody it’s just the problem is too darn hard. But I disagree with your optimistic assessment. You are giving the impression that GCMs develop towards some ultimate truth. In reality the GCMs do improve incrementally but they are doomed to be off-mark. BTW, while I’m stating my own position, this is more or less what I heard Saltzman’s people in person. While they were more active you must have had a chance to rebuff their claims face-to-face.

    David, you’re mistaken. Gavin is absolutely correct in referring to this kind of models as heuristic, as opposed to rigorous.

  32. 132
    David B. Benson says:

    Re #130: tamino — Thank you for hinting that I ought to look into Carl Wunsch’s work. I don’t mind stochastic models, but I do prefer deterministic ones when I can get them.

    Re #131: Sashka — It depends on what one means by rigorous. In other fields what are developed, rigorously from simplifying assumptions, are useable constitutive laws since the microscale and mesoscale physics and chemistry are too messy for actual applications. I view what Saltzman and colleagues have done in the cited book as a form of constitutive law for ice age climate.

    Anyway, it sounds better than heuristic. :-)

  33. 133
    David B. Benson says:

    I went to Carl Wunsch’s publications page and found

    Eli Tziperman et al.
    Consequences of pacing the Pleistocene 100 ky ice ages by nonlinear phase locking to Milankovitch forcing
    PALEOCEANOGRAPHY vol. 21, PA4206, 2006.

    Therein the four authors careful define what they mean by nonlinear phase locking and then point out that model identification is impossible by comparing model runs to the proxy data. All even partially realistic models will phase lock, they state. So it is not possible to even to distinguish those nonlinear models which have a natural period from those which would be static without Milankovitch forcing.

  34. 134
    Pat says:

    Re 92 and 105: First I just want to reitterate more generally what 105 said – Milankovitch cycles have had climate signals, in ice ages or otherwise, – well probably ever since the Moon formed, although the signal from times past will not always reach us, but I’ve read of evidence of Milankovitch precession cycle forcing of monsoons in lakes in Pangea (PS over geologic time the periods of some of the Milankovitch cycles have changed as the Moon recedes from the Earth due to tides). There have not and will not always be ice ages because the Earth can be just too warm or ill-conditioned to respond in that manner (The Earth might also become too cold to have interglacials within an ice house period – though I’m not sure if that has ever happenned outside the Proterozoic snowball/slushball episodes).

    Why sometimes ice house and sometimes not – As far as I’m aware:

    Faster sea floor spreading, presumably associated with more volcanic activity at subduction zones, and/or other increases in volcanic activity or geologic outgassing, or faster oxidation of exposed fossil organic C (as in shales) – greater geologic CO2 emissions (I think another way of looking at the inorganic part is that any given region of sea floor has less time to accumulate carbonate minerals from chemical weathering, so that C reservoir could shrink while others, including the atmosphere, can grow).

    The silicate + CO2 -> different silicate + carbonate chemical weathering rate tends to increase with temperature globally, and so is a negative feedback (but is too slow to damp out short term changes) – but chemical weathering is also affected by vegetation, land area, and terrain (and minerology, though I’m not sure how much that varies among entire mountain ranges or climate zones) – ie mountanous regions which are in the vicinity of a warm rainy climate are ideal for enhancing chemical weathering (see Appalachians in the Paleozoic, more recently the Himalayas). There is also an idea that mountain glaciers, and repeated continental glaciation and deglaciation, may actually enhance chemical weathering by their mechanical weathering, even though the cold climate generally inhibits it.

    Organic C burial – affected by topography and climate (wind driven upwelling and dust fertilization by wind may favor burial in the ocean, while a wet climate on flat land may favor burial on land (as in peat bogs)).

    In the above and in other ways, the arrangements of the continents and oceans.


    Re 130: Isn’t there an idea of a threshold, that the eccentricity cycle’s modulation of the precession cycle is such that some but not all obliquity cycle peaks occur in tandem with a precession cycle forcing strong enough to end an ice age.

    Also, one idea I’ve read of is that through ice age after ice age, the bedrock would be scraped bare of looser material, which might have lubricated the ice sheets. As that lubrication might be lost after several ice ages, the ice would not spread as fast, it would build up into thicker sheets, and the higher elevation of the ice surface would cause it to be colder than otherwise and so less likely to start melting.

    I also read once of an idea that ecological succession could play a role, that in some forested areas with enough moisture, bogs would start to take over. There may be influences on the C cycle, but the part I remember well is that this would make winter snow more reflective as it would be on a flatter surface, as opposed to snows upon trees.

    I’m curious about the mechanisms responsible for the CO2 feedback in glaciation and deglaciation. What I’m aware of is that:

    CO2 is less soluble in saltier water, but more soluble in colder water, and the second effect would win out for globally averaged changes. But for the reduction of atmospheric CO2, the surface ocean CO2 must also have fallen even given the greater solubility.

    Likely less biomass in the ice ages, too. It may be reasonable to expect less C in soils, so the C increase must have been in the deeper ocean.

    I’m vaguely aware that faster deep water formation could be a factor. I also know that wind driven upwelling and fertilization of phytoplankton by dust could also be factors.

    What I’m wondering about is –
    What happens to the C stored in soil underneath a young ice sheet? I’m thinking it wouldn’t decay right away. It would be buried, and then eventually carried to the margins of the ice sheet. There, it may still be quite cold, so I’m imagining that much of it may have stayed in moraines (providing a convenient source of CO2 during deglaciation) or been washed into the ocean (perhaps decaying later during warming?) This C was out of the atmosphere before glaciation, so it doesn’t directly help accelerate the glaciation, but could it have accelerated deglaciation? Of course, once in an interglacial, the boreal soil C must be stored up again…

    What might the significance of methane hydrates be?

    Any thoughts? (PS I was recently paging through “Earth’s Climate Past and Future” by Ruddiman)

    Also, I was wondering – how certain is the prediction that the next ice age, absent anthropogenic effects, will start in 50,000 years?

    Re 122 – if the magnetic field is weak than it’s direction could change faster with the same vector change in the field.

  35. 135

    Pat — You have some good insight into this. I recommend highly that you look at the 1981 paper by Walker et al. that first came up with the idea of the carbonate-silicate feedback. Here’s a link to it:

  36. 136
    David B. Benson says:

    Re #134: Pat — From orbital forcing theory, some state that future climate ought to have a stab at an ice age (stadial) in 20 ky and again in another 30 ky. Some opine that the first is too weak a forcing, hence 50 ky.

    All of the above ignoring anthropogenic effects, of course.

  37. 137
    Pat says:

    Re 135,136: Thanks.

  38. 138
    Pat says:

    Re 122 – also, if a magnetic pole where nearby, small movements could cause large directional changes in the horizontal component of the magnetic field, although the horizontal component should be quite small in that case.

  39. 139
    Bill DiPuccio says:

    I am an elementary school science teacher. Having read this article and other articles on ice ages and climate changes in the past, it is apparent to me that climatologists are still wrestling with the causes of such changes. If we do not fully understand the mechanisms of past climate changes, how can we be sure that the current rise in global temperature is primarily anthropogenic rather than primarily natural? Is is possible that both factors are working together?