RealClimate logo

Back to the future

Filed under: — gavin @ 30 April 2008 - (Español)

A few weeks ago I was at a meeting in Cambridge that discussed how (or whether) paleo-climate information can reduce the known uncertainties in future climate simulations.

The uncertainties in the impacts of rising greenhouse gases on multiple systems are significant: the potential impact on ENSO or the overturning circulation in the North Atlantic, probable feedbacks on atmospheric composition (CO2, CH4, N2O, aerosols), the predictability of decadal climate change, global climate sensitivity itself, and perhaps most importantly, what will happen to ice sheets and regional rainfall in a warming climate.

The reason why paleo-climate information may be key in these cases is because all of these climate components have changed in the past. If we can understand why and how those changes occurred then, that might inform our projections of changes in the future. Unfortunately, the simplest use of the record – just going back to a point that had similar conditions to what we expect for the future – doesn’t work very well because there are no good analogs for the perturbations we are making. The world has never before seen such a rapid rise in greenhouse gases with the present-day configuration of the continents and with large amounts of polar ice. So more sophisticated approaches must be developed and this meeting was devoted to examining them.

The first point that can be made is a simple one. If something happened in the past, that means it’s possible! Thus evidence for past climate changes in ENSO, ice sheets and the carbon cycle (for instance) demonstrate quite clearly that these systems are indeed sensitive to external changes. Therefore, assuming that they can’t change in the future would be foolish. This is basic, but not really useful in a practical sense.

All future projections rely on models of some sort. Dominant in the climate issue are the large scale ocean-atmosphere GCMs that were discussed extensively in the latest IPCC report, but other kinds of simpler or more specialised or more conceptual models can also be used. The reason those other models are still useful is that the GCMs are not complete. That is, they do not contain all the possible interactions that we know from the paleo record and modern observations can occur. This is a second point – interactions seen in the record, say between carbon dioxide levels or dust amounts and Milankovitch forcing imply that there are mechanisms that connect them. Those mechanisms may be only imperfectly known, but the paleo-record does highlight the need to quantify these mechanisms for models to be more complete.

The third point, and possibly the most important, is that the paleo-record is useful for model evaluation. All episodes in climate history (in principle) should allow us to quantify how good the models are and how appropriate are our hypotheses for climate change in the past. It’s vital to note the connection though – models embody much data and assumptions about how climate works, but for their climate to change you need a hypothesis – like a change in the Earth’s orbit, or volcanic activity, or solar changes etc. Comparing model simulations to observational data is then a test of the two factors together. Even if the hypothesis is that a change is due to intrinsic variability, a simulation of a model to look for the magnitude of intrinsic changes (possibly due to multiple steady states or similar) is still a test both of the model and the hypothesis. If the test fails, it shows that one or other elements (or both) must be lacking or that the data may be incomplete or mis-interpreted. If it passes, then we a have a self-consistent explanation of the observed change that may, however, not be unique (but it’s a good start!).

But what is the relevance of these tests? What can a successful model of the impacts of a change in the North Atlantic overturning circulation or a shift in the Earth’s orbit really do for future projections? This is where most of the attention is being directed. The key unknown is whether the skill of a model on a paleo-climate question is correlated to the magnitude of change in a scenario. If there is no correlation – i.e. the projections of the models that do well on the paleo-climate test span the same range as the models that did badly, then nothing much has been gained. If however, one could show that the models that did best, for instance at mid-Holocene rainfall changes, systematically gave a different projection, for instance, of greater changes in the Indian Monsoon under increasing GHGs, then we would have reason to weight the different model projections to come up with a revised assessment. Similarly, if an ice sheet model can’t match the rapid melt seen during the deglaciation, then its credibility in projecting future melt rates would/should be lessened.

Unfortunately apart from a few coordinated experiments for the last glacial period and the mid-Holocene (i.e. PMIP) with models that don’t necessarily overlap with those in the AR4 archive, this database of model results and tests just doesn’t exist. Of course, individual models have looked at many various paleo-climate events ranging from the Little Ice Age to the Cretaceous, but this serves mainly as an advance scouting party to determine the lay of the land rather than a full road map. Thus we are faced with two problems – we do not yet know which paleo-climate events are likely to be most useful (though everyone has their ideas), and we do not have the databases that allow you to match the paleo simulations with the future projections.

In looking at the paleo record for useful model tests, there are two classes of problems: what happened at a specific time, or what the response is to a specific forcing or event. The first requires a full description of the different forcings at one time, the second a collection of data over many time periods associated with one forcing. An example of the first approach would be the last glacial maximum where the changes in orbit, greenhouse gases, dust, ice sheets and vegetation (at least) all need to be included. The second class is typified by looking for the response to volcanoes by lumping together all the years after big eruptions. Similar approaches could be developed in the first class for the mid-Pliocene, the 8.2 kyr event, the Eemian (last inter-glacial), early Holocene, the deglaciation, the early Eocene, the PETM, the Little Ice Age etc. and for the second class, orbital forcing, solar forcing, Dansgaard-Oeschger events, Heinrich events etc.

But there is still one element lacking. For most of these cases, our knowledge of changes at these times is fragmentary, spread over dozens to hundreds of papers and subject to multiple interpretations. In short, it’s a mess. The missing element is the work required to pull all of that together and produce a synthesis that can be easily compared to the models. That this synthesis is only rarely done underlines the difficulties involved. To be sure there are good examples – CLIMAP (and its recent update, MARGO) for the LGM ocean temperatures, the vegetation and precipitation databases for the mid-Holocene at PMIP, the spatially resolved temperature patterns over the last few hundred years from multiple proxies, etc. Each of these have been used very successfully in model-data comparisons and have been hugely influential inside and outside the paleo-community.

It may seem odd that this kind of study is not undertaken more often, but there are reasons. Most fundamentally it is because the tools and techniques required for doing good synthesis work are not the same as those for making measurements or for developing models. It could in fact be described as a new kind of science (though in essence it is not new at all) requiring, perhaps, a new kind of scientist. One who is at ease in dealing with the disparate sources of paleo-data and aware of the problems, and yet conscious of what is needed (and why) by modellers. Or additionally modellers who understand what the proxy data depends on and who can build that into the models themselves making for more direct model-data comparisons.

Should the paleo-community therefore increase the emphasis on synthesis and allocate more funds and positions accordingly? This is often a contentious issue since whenever people discuss the need for work to be done to integrate existing information, some will question whether the primacy of new data gathering is being threatened. This meeting was no exception. However, I am convinced that this debate isn’t the zero sum game implied by the argument. On the contrary, synthesising the information from a highly technical field and making it useful for others outside is a fundamental part of increasing respect for the field as a whole and actually increases the size of the pot available in the long term. Yet the lack of appropriately skilled people who can gain the respect of the data gatherers and deliver the ‘value added’ products to the modellers remains a serious obstacle.

Despite the problems and the undoubted challenges in bringing paleo-data/model comparisons up to a new level, it was heartening to see these issues tackled head on. The desire to turn throwaway lines in grant applications into real science was actually quite inspiring – so much so that I should probably stop writing blog posts and get on with it.

The above condensed version of the meeting is heavily influenced by conversations and talks there, particularly with Peter Huybers, Paul Valdes, Eric Wolff and Sandy Harrison among others.

523 Responses to “Back to the future”

  1. 1
    Charles Raguse says:

    Is there an ultimate limit that arises from human intellectual capability, irrespective of model simulations and databases? After all, even with the aid of meetings and model simulations and conversations there is a limit. Unless one subscribes to the notion of “artificial” intelligence, this would come down to an analog of a “world” chessmaster, whose intellect is (?) superior to all others, and (?) all of the parameters of the contest (in this case “prediction” of a winner) are known. A derivative of this kind of pondering leads to the “flap of a butterflies wing” analogy, i.e., how do climatologists reconcile to the chaos theory?

  2. 2
    Eric (skeptic) says:

    Are there any proxies for clouds (e.g. average cloudiness or thickness, rainfall patterns, mountaintop rime ice, etc) that can be picked up on an annual or finer basis? Seems like that would provide validation for, or inputs to a model.

  3. 3
    Jon says:


    Has anyone gone about compiling a “wishlist” of sorts in the modeling community of studies (other than syntheses) they’d like to see come out of the paleo community along the lines of what you mention in your ninth paragraph?


    [Response: Wishlists compiled without appreciation of the reality of paleo studies tend to be a little pointless. For instance, if asked, modellers often request proxies for clouds, or the thickness of the ozone layer – both would be great – but this is just wishful thinking. Instead, modellers really need help with more forward modelling (including the proxy system in the models directly) and on downscaling larger patterns to specific proxy records (i.e. for a specific proxy record, how does the local environment influence what gets recorded). These issues require input from the data-gathering community (since they know their proxies and sites better than anyone) and would go some way to making the comparisons more direct. – gavin]

  4. 4
    Jim Galasyn says:

    Assuming we keep beating Moore’s law, computation should soon be “too cheap to meter”. Given essentially limitless computational power, how will GCMs be affected? Will our abilities to model and predict weather and climate become limited only by the power to collect data? Will we be able to push our predictions to the very edge of the complexity horizon?

  5. 5
    Ray Ladbury says:

    Jim Galasyn,
    Technically, Moore’s law derived from scaling of one particular type of semiconductor technology–Complementary Metal Oxide Semiconductor or CMOS. The next generation could be designed from the former using fairly similar models–departing from the recipes only when technical roadblocks were encountered. That worked well down to about 0.5 microns minimum feature size. Every generation since then has been a struggle, but we are still more or less on track of Moore’s Law. The law has taken on a life of its own–driven more by competition and economic necessity. The result is that I have sitting on my desk DDR3 SDRAMs for testing and we will soon be testing 45 nm feature size parts.
    If you want to peer into the crystal ball a bit, the International Technology Roadmap for Semiconductors (ITRS) makes interesting reading. What you find is that there are lots of roadblocks, but also lots of interesting alternative technologies that could take their place.
    Bottom line, I don’t think we’re done as far as increasing computing power for GCMs and other applications for at least a couple of decades yet. However, if economic hard times hit, progress will be slower.

  6. 6
    Mark J. Fiore says:

    Excellent post.
    However, the current changes are so quick and fast that prior changes in the Earth’s climate, all created by nature, do not significantly model the true extent of the negative feedback mechanisms currently ongoing, created my humans. Take the pine beetle crisis in Canadian forests, for example. The trees are dying because the yearly low frost temperature does not fall below the level to keep the beetles in check.As the trees die they release carbon. There is no precedent for the quickness and severity of the current rise in CO2.280 ppm to 380 ppm in 150 years?And current estimates show 450 ppm in 50 years or so.Such a quick climate change has never occurred, even with huge volcanic events of the past. Even with the meteor strike 65 million years ago. Also, from a moral standpoint, these events were caused by nature, and were unavoidable. Mankind has through negligence and disrepect caused the current changes.Also, if you look carefully at the current data, the shutdown of the North Atlantic current has already begun. Upwelling, and salinity levels, are already negatively effected by the meling of the ice at the poles.I am no scientist, but I know what I read, and I read a lot.
    Mark J. Fiore

  7. 7
    Kevin says:

    re: Charles Raguse
    It’s interesting to think about. I think that by definition climate is less dependent on initial conditions than is weather. The reason is that the term “climate” is not meaningful for any one point in time–it is by definition a more generalized state of the system over a period of time. So I think in a way your question about chaos is more for meteorologists than climatologists, i.e., will we ever have the understanding of enough variables, and enough computing power, to accurately predict weather many years into the future (or even months?)? Never say never, but that level of understanding of the system seems mighty distant right now.

  8. 8
    Lou Grinzo says:

    Ray and Jim: Another key point is that we’re still in the early stages of learning how to really leverage mesh or cloud computing, e.g. using a very large number of networked computers of various speeds to work together on a single problem. There have been some notable projects with people “donating” CPU cycles over the ‘net for protein folding and SETI research, so I would expect that a similar thing could be done with climate modeling, assuming the code could be written in a reasonable compartmentalized form.

    [Response: It’s already been done. – gavin]

  9. 9
    Jim Steele says:

    Excellent topic. Using paleoclimate data is exactly what made me a skeptic.

    The change in forcing by CO2 as calculated by this equation = 6.3 ln (C/Co) ( perhaps you may use a different coefficient but the principle remains) and using a base line of say 200 ppm for Co , we can see that at 280 ppm we get an increase of 2.119 W/m*2 from the periods of glaciation.

    Now if we look at the paleoclimatic record graphed in the Nature article we see that sometimes at 280 ppm the temperature rises and other times it falls regardless that the atmospheric concentration is 280 ppm.

    This suggests there is a force that is much more powerful driver than CO2 to overcome the forcing at 280 ppm. Most likely the sun. Assuming the sun’s effects are typically cyclical we are faced with the problem that as temperatures rises during an interglacial the effect of the sun is less powerful because it is being assisted by the increased forcing of CO2 concentrations. However as temperatures fall as we approach the next period of glaciation the sun’s decrease in output must decrease much more rapidly than it increased to offset the upward temperature pressure that would be created by the forcing of CO2 at 280 ppm.

    But such a conclusion seems at odds with known solar behavior. So it suggests that 1) either the sun is much more variable than we are aware or 2) if the sun’s output is cyclical, the effect of CO2 is negligible as temperatures rise and fall. It has been argued that the sun starts the temperature rise and then the increased forcing by increased CO2 is a feedback mechanism to increase temperatures. However that same story can not hold true when the tmepratures are falling, otherwise if CO2 is the driver, the feedback would raise the temperatures, but instead we fall back to a period of glaciation.

    [Response: Ahem. If you this is all that convinced you to be a skeptic, you can breathe a sigh of relief and let the scales drop from your eyes. CO2 is an important part of what is going on during the glacial/interglacial cycles, but it is not the only part. Can you say “Milankovic?” Read up on that in wikipedia, then come back. Needless to say, Milankovic forcing is too slowly varying to be playing any significant role in the warming we are seeing now. –raypierre]

  10. 10
    David B. Benson says:

    Mark J. Fiore (6) — Using the GISP2 Central Greenland temperature data averaged over 30-year intervals, the temperature there went up 0.72 K from 8170 to 8140 kpb. This was during the recovery from the 8.2 kya event:

    From data there and also the HadCRUTv3 global temperatures since 1850 CE, this is close to the warming experienced during the last thirty years. The temperature recovery at the end of Younger Dryas was even more dramatic, at least in Central Greenland.

    That said, this current rate of warming is certainly hard on already over-stressed organisms.

  11. 11
    Phil Scadden says:

    #4 – limitless computer power? I doubt it. My models (for basin thermal evolution) have benefited enormously from increased computing power but what you can practically do is still limited heavily by available power and your idea of a “reasonable” run time. An 8-day run time limits the no. of model iteration you can do a year somewhat! With access to 1000 cpu’s, we can now do montecarlo runs to look at sensitivities but can realistically look at only a few variables and a limited no. of samples. Increasing computer power would allow more runs, more variables, and ultimately finer meshes but its hard to imagine having enough computer power in my lifetime, even assuming it doubles every year, to feel the model is not compromised by computing limitations. I suspect this is true in spades for GCMs.

  12. 12
    Hank Roberts says:

    David (10), Mark Fiore’s (6) refers to rate of change of CO2 in the atmosphere, not rate of temperature change.

    Tamino blogged on the rate here:

    “… the fastest rise or fall in the entire Vostok ice core record amounts to a rate of increase of 0.06 ppmv/year.”

  13. 13
    Vern Johnson says:

    It’s a “mess” alright! What is unknown far far exceeds what IS known or just probable given our extreme lack of perspective inherent in our chaotic history as a species.

    As the sun burns it’s hydrogen, what is happening to it’s total mass and therefore it’s gravitational attraction? If the earth is gradually increasing it’s mass thru dust and meteorite deposition, what does this do to our orbital radius from the sun? What do radar-distance studies tell us vis a vis Venus, Mars, Mercury inside the “snow-line”?

    Gavin, please tell us what can be measured with some accuracy and then tell us what you think those measurements mean.

  14. 14
    pete best says:

    Appalling journalism quite frankly. Typical right wing media. A decade lull, its is not a consensus view and only preliminary view.

    Arctic sea ice melt is looking slightly worrying.

  15. 15
    Ike Solem says:

    Ray and Jim, I think you have to set Moore’s Law side by side with the computational demand increase for every stepwise increase in model resolution. The European Centre weather model uses a grid scale about 25 km per side, horizontally, and it divides the atmosphere into 91 layers:

    What climate modelers do is try and make their code as efficient as possible while preserving the basic physical processes – I’m not sure what law that follows. I believe that the current view is that it is better to spend any extra computing power on running more ensembles of models, and possibly on improving simulation of specific physical processes, rather than on increasing the grid resolution of current models.

    Lorenz (1993): “If the models could ever include so many variables that individual thunderstorms and other smaller-scale structures would be properly represented, and parameterization would no longer be needed, it would be totally unreasonable to expect that errors in the details of these structures would require two days to double. Individual thunderstorms only last a few hours… since a thunderstorm can in reality double its severity in less than one hour, we should expect that the difference between two rather similar thunderstorms would double just as rapidly…

    The apparent drop in returns with continued increases in resolution has led some forecasters to propose that the anticipated additional computer power in the middle nineties can be more advantageously used to carry out some Monte Carlo procedure.

    With all these obstacles around, it may surprise us to learn that within our chaotic atmosphere there are certain weather elements at a few locations that can be rather accurately predicted not just two weeks in advance but two months or even two years ahead. The most spectacularly predictable of these are the high winds in equatorial regions…

    I was fortunate enough to be present at the meeting in 1960 when Reed announced his findings, and I could see members of the audience shaking their heads as he maintained that at these heights the equatorial winds would blow continually from the east for about a year, and then from the west for a year, and then from the east again for another year… the subsequent years have fully confirmed his claims.”

    RIght now it is hard to say what will happen to the North Atlantic thermohaline circulation, which now appears to vary by large amounts on a monthly to yearly basis (not the Gulf Stream, but the other components). However, the multi-model ensemble approach is being applied here as well:

    Collins et. al 2006, Interannual to Decadal Climate Predictability in the North Atlantic: A Multimodel-Ensemble Study, JC

    Ensemble experiments are performed with five coupled atmosphere–ocean models to investigate the potential for initial-value climate forecasts on interannual to decadal time scales. Experiments are started from similar model-generated initial states, and common diagnostics of predictability are used…

    If one wanted to do that with paleoclimate, one of the well-studied time regions is the Younger Dryas, the sudden cold freeze about 12,900 – 11,500 years ago, which is thought to have happened rapidly, within a decade, and an associated cooling of the North Atlantic – though I don’t know if it really has been tied down to a halt in thermohaline circulation or not – and as usual, this is a topic that many people have tackled:

    Tarasov & Peltier (2005), Arctic freshwater forcing of the Younger Dryas cold reversal, Nature

    For those who are interested in doing a little research, you can look at all the papers that have cited that paper, courtesy of Google Scholar.

  16. 16
    Greg van Paassen says:

    @ #11. Back in the 70s computer scientists realised that the increasing speed of hardware made the design of algorithms, code tuning, etc, *more* important, not less so. A famous example of paying attention to these matters reduced the run time of a simulation from a year to a day.

    It would be great to get some of the best CS people helping out with climate modelling! (Perhaps some are already, but they’re keeping their heads down.)

  17. 17
    Steve L says:

    It’s sometimes hard to go back and get all of the previously recorded information sorted, cleaned, verified, put into the format you want, and then used in a synthesis. My limited experience with meta-analyses leads me to recommend asking the researchers who do the primary data collection to try and use the same measures, formats, etc in their future reporting so that useful analyses can then be easily done. I’ll give one example with which I am familiar: in the Province of British Columbia, Canada, to research fish, part of the requirements to obtain sampling permits was to put results into a Provincial data base on the web. The standardization of methods (gear used, linear and area estimates of habitat surveyed, etc) and details to be recorded (GPS coordinates, species classification) was a benefit in mapping distributions of species. Because negative results were also reported, it removed some biases that would show up if only published work was surveyed for the mapping.
    Perhaps you can request similar standardizations in paleoclimate?

  18. 18
    John Mashey says:

    re: Moore’s Law and such
    OR: Amdahl’s Law is becoming much more relevant
    OR: don’t expect magic for every kind of simulation

    0) Moore said nothing (in 1965) about specific technology (Bipolar, NMOS, PMOS, CMOS, BiCMOS), and in fact, it wasn’t even until ~1982/1983 that Intel shipped its first CMOS DRAM’s and micros (80286).

    1) Moore’s law *really* was the 2X # transistors/chip every 2 years, or equivalent, gotten by shrinking
    a) Shrinking transistors
    b) Shrinking wires
    c) Adding more layers [which also costs more mask steps, i.e., more cost]

    2) In “the good old days”, this tended to give you more performance as well [not necessarily 2X, as there were memory delays as well], but transistor-switching speed was the limitation, and transistors switch ~ in proportion to size, so we not only got 2X transistors, but got a speed boost “for free”.

    In the really old days, one could do an optical shrink from one generation to the next (or maybe at least to a half-generation) and the same design would work, just use smaller chips and run faster, and the same software would run, just faster.

    3) But for various reasons, wires don’t shrink as fast as the transistors, and a lot of the time is “in the wires”, and people have already used most of the clever well-known architectural tricks for making a single CPU go faster … which is why you may have noticed:

    a) the end of massive touting of big, easy GHz improvements … because those are over (at least for CMOS).
    b) there are lots of 2-core or 4-core chips around, because it is way easier just to replicate simpler CPU cores.

    Unfortunately, while 2 CPU cores might give you ~2X (or somewhat less) throughput of independent tasks, they don’t automagically make any given program run 2X faster.

    4) Some algorithms parallelize well, some don’t, but Amdahl’s Law says that the unparallelizable part matters. For example, if one CPU runs a program in X time, and:
    90% is parallelizable onto multiple CPUs
    10% is serial code

    Then, no matter how many CPUs you have, the best you’d ever do is approach a 10X speedup. Actually, in practice, most codes get faster with each additional CPU, but each CPU adds less performance, and after a while, adding a CPU actually makes the program run slower. Fortunately, some important classes of simulations can use *many* CPUs, unlike, say Microsft Word.

    5) Bandwidth, latency, and programmability matter as well: some algorithms work OK on distributed PCs, others really want lots of CPUs connected with low-latency, high-bandwidth shared-memory. Even with multi-core CPUs, some algorithms bottleneck on access to external memory, in any of several different ways.

    Good parallel programming is not easy, and the free ride we got for many years on single-CPU performance is over until some post-CMOS technology happens. We will still get more transistors/die for a while (someone mentioned ITRS). However, at least multi-core chips (including nVidia chips being used for parallel programming) are so widespread that maybe more students will learn.

    Lately, a huge amount of effort has been going into power reduction, not just for mobile devices, but for server chips, albeit for different reasons.

    re: #16 there are some pretty good CS people out here already, i.e., I suspect the long-hanging fruit has been pretty well picked over.

  19. 19
    David B. Benson says:

    Hank Roberts (12) — Thank you for the clarification.

    To a man with a hammer, every problem looks like …

    Greg van Paassen (16) — Some computer scientists are interested in designing machines capable of considerable parallelism. Others are then interested in the question of how one ought to program in order to best utilize such machines. This gives rise to better programming languages for parallel computation, and some of these features can even be stuffed into the Fortran straight-jacket. :-)

    Anyway, the computer science assistance with climate modeling is largely under the surface, so to speak.

  20. 20
    gusbobb says:

    #9 Jim Steele, CO2 is a result of solar changes that increase ocean temperatures not the driver of changes. Raypierre mentions the Milankovitch cycles but their periodicity is too long too explain the current warming or things like Dansgaard-Oeschger events that Svensmark connects to cosmic rays. RC writers will allow for some solar warming but then attribute the rest to increased CO2 feedback. But as you suggest that makes no sense because how does temperatures decrease back to the low baseline with that CO2 now in the atmosphere unless something else is the driver, reducing temperatures and allowing CO2 to dissolve into the oceans.

    Raypierre failed to mention short term solar variability. The sunspots activity is way lower than the consensus predicted, very similar to a Maunder Minimum, and several researchers like Khabibulo Absudamatov have been saying the sun will undergo decreased activity and create a cooling trend. I too have predicted lower solar output and a cooling ocean. Now there is a paper that will be published in Nature telling us that due to ocean currents that we can expect a ten year cooling trend. (Hardly what the GCM’s have predicted) and once again evidence that natural variations overwhelm CO2. Here is a snip and a link.

    “Those natural climate variations could be stronger than the global-warming trend over the next 10-year period,” Wood said in an interview.

    According to earlier discussions with Gavin on the Galactic Glitch he says that the ocean will respond to the atmospheric forcing to achieve equilibrium with radiative input, so if this cooling trend further supported, by his own logic he would have to admit that the there is less input.

    [Response: Not really. The Nature study is talking about changes associated with ocean circulation even while CO2, and the global imbalance, and global temperature, is increasing. It is exactly what we’ve been trying to explain. – gavin]

  21. 21
    Timothy says:

    The reason those other models are still useful is that the GCMs are not complete.

    I don’t think that’s really fair, it’s not the only reason, and one of the main reasons for using simplified models, is because they are cheaper, and so can be run for longer.

    If I’m correct, what counts as a really long control run for a typical GCM is a few thousand years, while the standard centennial timescale experiments for the IPCC reports each run from 1860-2100, ie 240 years.

    I think it’s correct to say that the GCMs are still too expensive, because they’ve got so much in them, to use them for glacial cycle experiments. You just can’t run a GCM for a hundred thousand years of model time.

    This brings me to the discussion about computer power. My impression was that, with even desktop PCs now having to go the route of parallel processing to improve performance, the modellers now face the problem of re-writing their code to get it to work across not just dozens of processors, but hundreds, (or thousands?), if they want to see the increases in performance that they’ve achieved in the past. This will be hard because of the problem of exchanging data between the processors, and reducing that data flow to a minimum will be a key challenge.

    [Response: Tim, you make a good point and I should have expanded that line. In relation to using massively parallel computers, the scaling is ok for relatively high resolution models and for intelligent domain decomposition. It does take work to get the models configured appropriately though. But it always pays to remember that mutliple ensembles (initial conditions and perturbed physics) are likely to play an ever larger role in the modelling, and these scale perfectly on MPP machines! – gavin]

  22. 22
    gusbobb says:

    Gavin said “The Nature study is talking about changes associated with ocean circulation even while CO2, and the global imbalance, and global temperature, is increasing. It is exactly what we’ve been trying to explain.”

    I guess I need to fix my glasses because no where did I read that “global temperature, is increasing”. Could you provide a quote?

    This is what I read:
    “Average temperatures in areas such as California and France may drop over the next 10 years, influenced by colder flows in the North Atlantic, said a report today by the institution based in Kiel, Germany. Temperatures worldwide may stabilize in the period.”

    And when you say that is “exactly what we’ve been trying to explain”

    can you link me to the quote where you or other RC writers predicted the next 10 years temperatures will stabilize worldwide and California and France will get colder?

    And the article contradicts predictions made by the NASA folks in GISS Surface Temperature Analysis
    Global Temperature Trends: 2007 Summation

    “Based on these considerations, it is unlikely that 2008 will be a year with truly exceptional global mean temperature. These considerations also suggest that, barring the unlikely event of a large volcanic eruption, a record global temperature clearly exceeding that of 2005 can be expected within the next 2-3 years.”

    [Response: The global temperature in the models used in this study are increasing (see figure 4) – the changes are initially less in their SST-restored run, but they are still increasing and after a decade or two are at the same level that the standard run has. These are however 5 year trends, and so don’t include variations related to ENSO and the like. My guess is that the next record breaking year will coincide with the next moderate-to-high El Nino, a position that is not contradicted by these new results – which focus on the predictability of the N. Atlantic, not the Pacific. Since I have never presented any initialised climate forecasts, you will find that I have never predicted any short term temperature trends at the regional scale. This paper (and a previous study by Smith et al earlier this year) are some of the first people to do this and so it’s not yet clear how good it is as a method. I would recommend suspending judgement for a while. – gavin]

  23. 23
    gusbobb says:

    I have suspended judgment and only speculated that “if the trend is supported”. I figure in 5-10 years we will be able to decide who had better predictions.

    Gavin said “My guess is that the next record breaking year will coincide with the next moderate-to-high El Nino, a position that is not contradicted by these new results ”

    But their expectation of stabilized and decreasing temerpatures doesn’t exactly support the prediction of new global highs in the near future El Nino or not. Are the equations and data for calculating global temperatures available to the public. I understand they are not.

    In earlier conversation you adamantly said heat does not leave the ocean. So where does the heat come from so that El Nino will cause a record breaking year?

  24. 24


    Given all the media attention the new Nature article has been getting (see tomorrow’s NYT, for example, or National Geographic, the CSM, etc.), it may merit a full RC post. Its also interesting to compare this prediction with that of the Hadley Center last year, which also suggested a stagnation of temperatures (till 2009-2010 in their case, if I recall).

    [Response: We’ll try and have something up at the weekend. The NYT story is ok, but the Telegraph is appalling. – gavin]

  25. 25
    tharanga says:

    The number of enthusiastic posts about Moore’s law gives me some hints as to the professions of the other lay readers of the site.

    Back to paleoclimate: is it possible to pick out large volcanic events, of the Krakatoa or Laki scale, in the ice core CO2 data? I’m having a look at some data from Taylor Dome, and it looks like the temporal resolution is only about 100 years or so, which would make it tough.

    I realise that currently, anthropogenic sources far outweigh volcanic; just wondering how much CO2 is released by the absolute biggest events, and if that can then be seen in the record.

    [Response: Volcanos affect climate almost exclusively through the aerosols they pump into the atmosphere, which reflect sunlight. The amount of CO2 pumped out by even a big eruption is pretty trivial, and would hardly make a blip in the observed CO2 record. To put things in perspective, the anthropogenic source of CO2 at present is something like 20 times the estimate of the long term mean flux of CO2 out of
    the interior of the earth by all volcanos and all undersea vents. –raypierre]

  26. 26
    Thomas says:

    I put in about two decades worth of tuning mostly engineering code for modern computing applications. We have to recognize that modern computing cores will run at much higher throughput when low level parallelism, what computer scientists would call Instruction Level Parallelism, is exposed to the hardware/software combination. Even with all data local the rate of operation can vary by as much as an order of magnitude for well designed single thread code, than otherwise poorly designed code. During the past few years CPU clock rates have stagnated, mainly because of power consumption/heat dissipation issues, but peak computation per core rates of well designed code has continued to increase. At the same time the number of cores per chip has been rising, with six to eight coming within the next couple of years. Unfortunately off chip data bandwidth is not keeping up with on chip computation rates. This means that unless your program makes very efficient usage of the cache memory hierarchy that even without Amdahl’s law the parallel efficiency of the chip can be a lot less than anticipated. This trend towards an increasing ratio of chip capability to off chip bandwidth is expected to continue.

    Personally, I’m not as sanguine about the prospects for as rapid an increase beyond the next couple of chip shrinks, we are already at 45nm, and two more shrinks brings us to 22nm at which point both quantum effects, and the fact that the number of atoms per part starts getting small enough that part to part variation starts getting pretty severe. I wouldn’t be at all surprised that after another five years or so that progress slows (but doesn’t stop).

  27. 27

    ITRS is practically a conspiracy to make sure Moore’s law keeps happening. The semiconductor companies have parceled out research tasks so that each company takes a particular technology to research They have signed cross-licensing agreements. If any technology makes a breakthrough, the whole world makes the breakthrough. Some semiconductors other than silicon are up to 1000 times faster than silicon. There were about 20 semiconductor materials under consideration last time I checked. Raymond Kurzweil says some time in the near future [say 2020 or 2025] computer IQ will exceed human IQ. See “The Age of Spiritual Machines.” Once that happens, computer IQ will fly way past human IQ and soon exceed the thinking power of all of the people. We use computers to design computers. A better computer can better design its own successor. A computer smarter than a human can design a computer 10 times smarter than a human or itself. Keep on iterating.
    So don’t count computers out as far as better climate predictions go. Future computers will be programmed by computers. If Kurzweil is right, computers will take on all of the tasks presently done by human climatologists. That includes the synthesis of fragmentary data task. The new kind of scientist who is at ease in dealing with the disparate sources of paleo-data and aware of the problems, and yet conscious of what is needed (and why) by modellers could be a machine. It would be a 2025 machine, not a 2008 machine. The question is, will the 2025 computer be smart enough to figure out how to undo the climate screw-up that will have happened by then? Will the humans be savable or will only computers survive? The computers will survive the extinction of humans if they become conscious/self aware because computers don’t need food or breathable air. Raymond Kurzweil is one of those super talented software people, not a science fiction writer, so I don’t count his ideas out.

  28. 28
    Ike Solem says:


    The ocean contains far more stored heat than the atmosphere does, and is also continually exchanging heat with the atmosphere – unless covered by sea ice.

    It’s very hard to predict what the net ocean-atmosphere heat exchange will be over any given year. Much depends on the strength of the wind-driven upwelling, which brings cold deep water to the surface.

    At this point, model forecasts of ocean circulation at the decadal scale should probably be viewed cautiously, since we cannot even predict an El Nino more than 9 months in advance or so.

    Predictions of the Pacific Decadal Oscillation are even more unlikely – claims that it has entered its “cool phase” are basically completely unsupported. No mechanism for the PDO is known, and whether it has regular phases that will continue into the future is also unknown. El Nino is far better understood.

    I would wait for the paper to come out before leaping to promote it.

  29. 29

    Holland’s old (1978) figure for CO2 from volcanism and metamorphism was 330 million tons per year. According to T.M. Gehrels at USGS, volcanoes contribute about 200 MT of that. Last year, human technology added about 30 billion tons of CO2 to the air. So human output dwarfs natural by a factor of 91 and volcanoes by a factor of 150.

  30. 30
    Harold Pierce Jr says:

    RE: #6 Mountain Pine Beetle Outbreak

    Global warming has little to do with the massive mountain pine beetle outbreak in BC. A cold snap of -35 to -40 deg C for at least 2-3 days is required to kill larva and eggs under the thin bark of a lodgepole pine in winter and -25 deg C in early fall. Cold snaps of this magnitude don’t happen all that often.

    The other natural control of pine beetle infestions is wildfire started by lightning strikes. Lodgepole pine is the only common pine that has serrotinous or heat sensitive cones, which can hang on the branches for long periods of time. The scales on these cones only open and drop their seeds when exposed to the intense heat of fire which melts the resin. This occurs just before the cone is consummed by fire. After the forest burn downs, the seeds germinate in soil enriched from the ash of the burned trees, and the cycle start over again with a period of about 100-150 years. This is also why there is a mono culture of LP that goes on forever and ever.

    Pine forest in north western NA just don’t last that long as compared to the coastal rainforest where there are giant Skita spruce, Doug fir, and western red cedar. Old-growth WCR is more valuable than platinium, and the BC forest service is in a constant on-going, cat-and-mouse struggle with tree rustlers.

    Vigorous suppression of forest fires beginning with the Smokey the Bear Era has led to large mature LP forest and lots of fuel on the forest floor. This is the reason the Yellowstone pine forest burned down in 1988. I recall on the cover of Newsweek mag a phrase,”Is global warming the cause?”, or something like that.

    There is one other curious weather pattern that has contributed the current outbreak. Starting in the late ’90 and early ’00’s, late spring and early summer were unsually cool and rainy. This led to a low incident of forest fires. Many fires which start in remote areas with beetle-infested trees are often left burn out because there is no easy access for fire fighting crews and equipment.

    Mountain pine beetles are very picky critters and only mass attack mature LP pines that are about 40 cm or larger in diameter at breast height. Prof John Borden (Bio. Sci., SFU) and I worked for about 25 years tying to find a hypothetical attractant host chemical emitted only by mature trees. We never found the “magic bullet”.

    For all the interesting info on NA’s No. 1 tree-killing beetle, GO:

    And checkout this awesome video that shows the growth of the infestation in the BC LP forest, GO:

    Note how the cold snap in 1985 wiped out the beetles in the Chilcotin.

  31. 31
    Harold Pierce Jr says:


    When posting comments, please limit paragraphs to less than ca 10 lines. Otherwise the text is too hard to read especially for us old guys with vision problems.

  32. 32
    Ray Ladbury says:

    John Mashey–I should have been more specific. True Moore’s original formulation was not for CMOS. However, most of the period over which the law has been valid relied on scaling of CMOS to achieve the increased density and speed. I agree that the age of CMOS scaling is coming to an end–and it has been for the past 3 or 4 generations. I think we’ll get to 22 nm with great difficulty, but beyond that I think we’ll need to rely on new technologies–to wit:

    The thing that is interesting here is not just that the technology could provide a large boost in memory density, but that it could provide an entirely new type of circuit that could “learn”. Coupled with genetic algorithms and other programming advances, this could very well revolutionize programming for complex models such as GCM.

    Now, if we can only fly it in a satellite, I’ve got job security until I retire at age 80.

  33. 33
    Scott Reid says:

    Re: #30

    “Lodgepole pine is the only common pine that has serrotinous or heat sensitive cones, which can hang on the branches for long periods of time.”

    Just to set things straight, jack pine (Pinus banksiana) is a very common pine that has serotinous cones. Jack pine is also under threat from the spread of the mountain pine beetle.

  34. 34
    mg says:

    27 It is not all obvious that the world can afford much further development of ITC. It has been clear for nearly a decade (eg see Paul Strassmann’s article here ) that the corporate world would hit the critical point of being unable to afford further expense of scrap & build cycles. The critical point is now in terms of cycle confluence and it is no coincidence that world markets are edging towards fundamental restructuring.

    The IT world, in addition to confronting an increasingly cash-strapped client base, also has to comes to terms with its contributions to climate change. Further, it is worth doing some supply chain analyses for the semiconductor industry and watch out for that wild card: sea level rise. The conditions are right for a fundamental revolution to sweep through the ITC industries, driven by financial evaporation.

  35. 35
    gusbobb says:

    Ike Solem Says:
    At this point, model forecasts of ocean circulation at the decadal scale should probably be viewed cautiously, since we cannot even predict an El Nino more than 9 months in advance or so.

    Ike I agree with you. I think the ocean is a wild card regards heat distribution and climate predictions. I have advocated that since my first post.

    I was not predicting an El Nino or its effects. It was Gavin who predicted the next record high will be in a few years with an El Nino. My question is where does the heat come from to generate the spike in temperatures during the El Nino.

    I also find it amusing that cool spells due to ocean circulation and oscillations are not seen as “climate change” as the reports on the new cooling predictions suggest but road bumps to inevitable warming. The distinctions seem arbitrary and biased. If models of past climates were done correctly then these oscillations were accounted for. Attribution for increased temeperatures from warm phases of various oscillations were supposedly already accounted for. The IPCC predicted a consensus .3 degree rise in temperature in the next decade. Now the predicted decade of cooling is reported as an anomaly that was not part of the models?

    And what drives these oscillations? Before it was suggested that AGW was increasing the probabilities of warm phases. Then by prior logic a cold phase must be affected by less radiative input. I eagerly await new papers.

  36. 36
    BBP says:

    Vern (13)
    The atomic mass of Hydrogen is 1.00794 and Helium is 4.002602 – so fusing 4H -> 1He you only convert 0.029 atomic mass units into engery, which is only 0.7%. During its 10 billion year lifetime the sun will fuse of order 10% of its hydrogen, so over any timescale of interest to climate science the mass change is negligible. The sun is slowly incresing in luminosity, but this is easy to take into account.


  37. 37
    Martin Vermeer says:


    My question is where
    does the heat come from to generate the
    spike in temperatures during the El Nino.

    The global mean temperatures are computed by areal averaging of the weather station data on land, and sea surface temperature data observed by satellite. The temperature variation of El Nino doesn’t require heat as much as it requires a very large area in the Pacific to be a little bit warmer at the surface.

    The software for computing global temperature (one version of it) is freely downloadable from the GIStemp website.

    About GCMs including El Nino etc., no they don’t. But they do produce their own El-Nino -like natural variations (IOW, physical realism!) which however are not synchronized with the real world. IIUC, the computations described in the article aim at precisely achieving this sync by setting realistic initial conditions.

  38. 38
    tamino says:

    Re: #31 (Harold Pierce)

    I too would urge all commenters to help us old guys whose vision isn’t what it used to be. Very long paragraphs can make your comment harder to read.

  39. 39
    Ray Ladbury says:

    Gusbob, the classification of oceanic variability as weather rather than climate is not at all arbitrary. Oceanic variability typically persists on timescales of less than a decade. Since the solar cycle is 11 years, anything less than this is weather. And you seem to be missing the point entirely. OK, so let’s say there is a flux of cold water to the surface to suck up heat for a decade. All that means is that you won’t see the type of warming we’ve had for the last 20 years for a decade. At the end of that time, the CO2 is still there and warming kicks off again with a vengeance. What is so hard to understand about this? Climate consists of long term trends, neglecting short-term variability (

  40. 40

    RE #20 & Gavin’s “The Nature study is talking about changes associated with ocean circulation even while CO2, and the global imbalance, and global temperature, is increasing.”

    I just read a news article about this, and the impression I got was that the team was using a new approach (something like between the climate & weather timescales???), and that the ocean circulation (a slow down ??) and some natural variability would work in sync to make it a bit cooler in the N. Atlantic region, while the natural (down turn) variability working against the GW would keep it about constant in the tropical zones.

    Then after this next decade GW might kick in with even more ferocity than ever.

    I’m just paraphrazing what might be a lot of mistakes via a news article. However, I remember in high school physics being in charge of the ripple tank, and how upswing waves when they converge can create bigger waves, while downswing and upswing waves crossing each other cancel out each other.

    And what worries me is that the denialists will spring forth from the woodwork to denounce GW, and this may impact public perception, and our GHG emissions will continue to soar, and we’ll really be in hot water after that decade of slight cooling (or stable average temps) and beyond into ??? hysteresis.

    From what I understand our GHG emissions are already higher than the worst-case IPCC scenario.

    And there are other considerations, aside from the warming — like the oceans becoming acidic from CO2, and the ocean conveyor slow down reducing the upwelling of nurishment for sea life…..etc.

  41. 41
    Chris says:

    Re #35 gusbobb

    You ask with respect to El Nino’s “where does the heat come from to generate the spike in temperatures during the El Nino.”

    It comes from the sun. Within a world completely in energy balance with a stable equilibrium temperature, there are El Nino’s and La Nina’s. These correspond to periods in which heat energy is redistributed. In an El Nino event warm surface waters (warmed by the sun!) spreads across the central tropical Pacific, and cold water upwelling off the West coast of S. America is suppressed. So the Earth’s surface temperature warms for a short spell until the circulation patterns return to their “normal” state.

    Presumably all of these ocean circulations act to take solar energy into the oceans and if the Earth is in positive heat imbalance (as it seems to be under the influence of an enhanced greenhouse effect), will contribute to ocean warming, even if occasionally (due to fluctuations such as La Nina events), the ocean surface temperature is overall rather cooler for a short while.

    So these truly are fluctuations or oscillations that act to modulate either the equilibrium surface temperature (if the Earth is in “heat balance”) or the transition to a new equilibrium temperature if the Earth is in imbalance with respect to temperature forcings (as in the case of an enhanced greenhouse effect).

    That’s all very straightforward isn’t it? It seems rather likely that in a world warming under the influence of enhanced greenhouse forcing, that many of the “record years” will coincide with positive modulations from the various internal “oscillations” (e.g. El Nino’s in this case) that transiently redistribute heat to or from the Earth’s surface.

    So I don’t see why cool spells due to ocean circulation and oscillations should be seen as “climate change” (after all they’re small and short lived), and if these are a consequence of generally “heat-neutral” and short lived redistributions of ocean surface warmth, it’s not surprising if these are considered to be “bumps” on the year on year variation of the surface temperature anomaly as it responds to enhanced greenhouse forcing.

    Why should we expect otherwise? That’s how things have been during the warming of the last 30-odd years. Why should they be any different now? And many of the “bumps” can be accounted for in the past record (El Nino’s, La Nina’s, volcanic contributions, solar cycle; i.e. all of the factors that result in transient modulations of the long term trend). What’s “arbitrary” or “biased” about that??

  42. 42
    catman306 says:

    I’m probably over my head here, but I’d like to point out that all of these advances in computers and writing faster running code won’t solve a basic problem with modeling: the assumption that non-included variables will stay linear and can be safely ignored. Anything that can be measured might be one of these suddenly exponential variables. Among the thousands, millions, or more ignored variables there’s probably a few that can unexpectedly go exponential and ruin the model when compared to the real world. The effect on our climate by being struck by an asteroid is an obvious example. But so too might be some undiscovered chemical reaction that only takes place at the extremely pressure found at the bottom of the deepest oceans. This variable could quantify anything imaginable (or unimaginable!) where scientists are not currently looking.

  43. 43
    Nick Gotts says:

    RE #27 Edward Greisch “Raymond Kurzweil is one of those super talented software people, not a science fiction writer, so I don’t count his ideas out.”

    [edit] Compare actual technical progress over the past half century with what techno-optimists were predicting in the 1950s, in areas such as space travel, medicine, nuclear power, transportation, and above all, artificial intelligence.

    [Response: This is getting way off topic. Bring it back to climate models or drop it. – gavin]

  44. 44
    Hank Roberts says:

    > 31, 38, “old guys with vision problems”

    I’m another, same plea for line breaks, same reason.

    aTdHvAaNnKcSe (THANKS in advance)

    Contributors writing inline replies too, please?
    Return — twice for a blank line — between thoughts.

  45. 45
    Pekka Kostamo says:

    Re 35: “I also find it amusing that cool spells due to ocean circulation and oscillations are not seen as “climate change” as the reports on the new cooling predictions suggest but road bumps to inevitable warming. The distinctions seem arbitrary and biased.”

    This is just a standard strawman practice claim. It is the deniers who consistently choose the “bump” year of 1998 as their reference in all kinds of calculations. I have not seen one of the climate scientists claim that year as proof of global warming.

    Any record of 20th century global temperature measurements shows the warming trend as well as the short term noise, both up and down.

  46. 46
    Hank Roberts says:

    From the first post:

    > the tools and techniques required for doing good
    > synthesis work are not the same as those for making
    > measurements or for developing models. It could in
    > fact be described as a new kind of science (though
    > in essence it is not new at all) requiring, perhaps,
    > a new kind of scientist. One who is at ease in
    > dealing with the disparate sources of paleo-data
    > and aware of the problems, and yet conscious of
    > what is needed (and why) by modellers. Or
    > additionally modellers who understand what the
    > proxy data depends on and who can build that into
    > the models themselves

    This seems like a chance to create a new academic course, to start with. Would you all consider starting a Wiki or something in which possible study guides, coursework, references and such could be collected?

    Possibly there’s a department chair or academic dean watching or you could attract the attention of one.

    This sort of thing is being done for flu pandemics:

  47. 47
    Aaron Lewis says:


    This was one of your very best posts. The only thing missing from your post was a paragraph or two that we could copy and send to our congressman to get you (plural) some money for a paleo-climate information data warehouse.

    (Aside to US readers; Gavin has other things to do. Do not wait. Re-read the post and go to and send your congressional representative a missive. Then, have all of your family members, friends, and coworkers send similar missives.

    Aside to other readers; Tell your representatives that the people that emit (or plan to emit) greenhouse gases should take full responsibility for their actions; including the research necessary to determine the extent of the problem.)

  48. 48
    Aaron Lewis says:

    In the old days we called this “Interdisciplinary Environmental Studies”.

    While climate modeling with Paleo data is important, climate modeling with an understanding of infrastructure engineering and economics is just as important. I would say that the fact that the climate models fail to account for land ice sheet dynamics speaks volumes that our climate modeling effort needs to be vastly broader and deeper.

    When the IPCC was formed, there were 2 questions that needed to be answered. The IPCC has danced around these questions 4 times, and never offered an answer to either question. The questions are: “How fast can heat be transferred to the major ice sheets?” and, “How will the ice sheets respond?” Paleo data can help answer these questions. Then, we will need infrastructure engineering and economics to decide what, if anything, we should or can do to adapt or mitigate.

    The ice sheets represent a global threat at a risk level above VP Chevy’s 2% rule. Therefore, asset allocation to climate threat analysis (climate modeling) should be similar to asset allocation to assessing and fighting global terrorism – call it US $100 billion per year. Gavin should be able to run a paleo data center on some slice of that.

  49. 49
    Lawrence Brown says:

    If, as some of the above references to the article in the May 1 issue of “Nature” point out there is a temporary cooling in the near term due to natural variability, this will reinforce the need to take immediate action to reduce GHGs. It’s an opportunity to take advantage of the fact that less glacial, sea surface and (at least) northern hemisphere continental ice will melt. This will mitigate the positive feedback effects of the loss of albedo(I almost said libido-that ship has almost sailed- I’m another one of those old guys, who prefers shorter paragraphs).
    A cool phase coupled with accelerated steps,starting very soon, to reduce CO2,CH4,NO2 and other heat trapping gases can benefit the planet in this regard.

  50. 50
    Hank Roberts says:

    Aaron, I have big hopes for the International Polar Year science — once we start seeing it.

    I’ve seen only one journal article from the ANDRILL cores. The first of these were only shipped to Florida late last year, and sliced up and bits of them passed out to scientists all around the world. They must be writing papers by the dozen.

    I think those two questions really have been waiting on a whole lot of research that’s wel under way now.