With ENSO, chaos rules, models drool

A new paper in Nature from the Department of Oceanography, School of Ocean and Earth Science and Technology, University of Hawaii, makes a somewhat surprising claim about predicting ENSO events. This is probably one of the shortest abstracts ever, but then, there’s not much to be said beyond this simple statement.

Climate science: A high bar for decadal forecasts of El Niño

Pedro DiNezio

Nature 507,437–439 (27 March 2014) doi:10.1038/507437a
Published online
26 March 2014

Climate simulations suggest that multi-decadal periods of high and low variability in the phenomenon known as the El Niño-Southern Oscillation in the tropical Pacific Ocean may be entirely unpredictable.


I suppose this explains why this model has been doing so poorly for the last year in predicting a new El Niño, it has been showing an El Niño just months away for almost a year.

NINO 3.4 SST Anomalies Forecast

Will we see an El Niño this year? Only chaos knows for sure.

More at the WUWT ENSO page

About these ads

59 thoughts on “With ENSO, chaos rules, models drool

  1. Now we can roll up the rest of the AGW clap trap and get on with actually helping people?
    Naw, that would be too simplistic.

  2. El Niño only occurs around Christmas, that’s why it’s called El Niño.
    It’s looking increasingly likely.
    The no warming in 15,16,17 years meme might really break next time and then we can all agree that global warming is real and occurring.

  3. Someone should do a multi-year study and analysis of the forecasts. Are they any better than darts or dice? I look at them every week. They make the Hansen and the IPCC temperature model projections appear skillful. I think the shortest abstract ever may also be the most accurate.

  4. Simon says:
    March 26, 2014 at 5:07 pm
    =====
    CO2 has punched out it’s dance card…..so it would be perfectly normal warming if it did

  5. if people assume chaos theory then its all just dice games and no one can predict anything which means ipcc is just self indulgent seat warmers club?

    co2 dogma assumes predictability so it can’t assume chaos? they assume a hierarchy of processes with co2 at the top as the most active catalyst.More catalyst more reaction.

    taxes can’t determine chaos.

    however i have seen people predict the uk winter storms from 3 months out using just normal meteorological reasoning [pressures]. The fact the co2 dogma can’t do that and even deny the possibility despite the evidence suggest they have their hierarchy and correlations wrong. The long range forecasters will destroy the co2 dogma because they have a better understanding of processes.

  6. ‘Simon says:
    March 26, 2014 at 5:07 pm
    …. The no warming in 15,16,17 years meme might really break next time and then we can all agree that global warming is real and occurring.’

    I think we all already would agree that global warming has been occurring since the end of the LIA. Nothing controversial there.

    Except it’s neither catastrophic, nor unprecedented, nor caused by C02. You should read ‘The Pursuit Of The Millennium’ by Cohn. The end times have been around for a long time.

  7. Per this 2011 paper;

    “Real-time model predictions of ENSO conditions during the 2002–11 period are evaluated and compared to skill levels documented in studies of the 1990s. ENSO conditions are represented by the Niño- 3.4 SST index in the east-central tropical Pacific. The skills of 20 prediction models (12 dynamical, 8 statistical) are examined. Results indicate skills somewhat lower than those found for the less advanced models of the 1980s and 1990s. Using hindcasts spanning 1981–2011, this finding is explained by the relatively greater predictive challenge posed by the 2002–11 period and suggests that decadal variations in the character of ENSO variability are a greater skill-determining factor than the steady but gradual trend toward improved ENSO prediction science and models. After adjusting for the varying difficulty level, the skills of 2002–11 are slightly higher than those of earlier decades. Unlike earlier results, the average skill of dynamical models slightly, but statistically significantly, exceeds that of statistical models for start times just before the middle of the year when prediction has proven most difficult. The greater skill of dynamical models is largely attributable to the subset of dynamical models with the most advanced, highresolution, fully coupled ocean–atmosphere prediction systems using sophisticated data assimilation systems and large ensembles. This finding suggests that additional advances in skill remain likely, with the expected implementation of better physics, numeric and assimilation schemes, finer resolution, and larger ensemble sizes.”

    http://journals.ametsoc.org/doi/abs/10.1175/BAMS-D-11-00111.1

    ENSO forecasts are apparently getting worse. I built the WUWT ENSO Forecast Page a month ago;

    http://wattsupwiththat.com/reference-pages/climatic-phenomena-pages/enso/enso-forecast-page/

    but am hesitant to introduce given how bad we are at forecasting ENSO. The problem is that “The period from February through May is commonly referred to as the spring barrier. During this time, models generally have the least skill to predict the coming season.”

    http://iri.columbia.edu/news/la-nina-still-hanging-on/

    “Using predictions for the sea surface temperature (SST) generated by a Flexible Global Ocean-Atmosphere-Land System model of IAP/LASG (FGOALS-g), the season-dependent predictability of SST anomalies for El Nino/La Nina events is investigated by analyzing the forecast error growth in an imperfect model scenario. The results indicate that, for the predictions through the spring season in the growth phase of El Nino events, the prediction errors induced by both initial errors and model errors tend to have a prominent season-dependent evolution and yield a prominent spring predictability barrier (SPB). For the decay-phase predictions of El Nino events, a less prominent season-dependent evolution of prediction errors and then a less prominent SPB are observed. For the growth- and decay-phase predictions of La Nina events, the prediction errors do not exhibit a significant season-dependent evolution and yield a less prominent SPB phenomenon. These results indicate that the SPB phenomenon depends remarkably on the ENSO events themselves, particularly the phases of the El Nino/La Nina events. We also report that the initial SST errors that correspond to a significant SPB for El Nino events tend to have the dominant modes in a large-scale dipolar pattern with negative anomalies in the equatorial central-western Pacific and positive anomalies in the eastern Pacific, or vice versa. We further demonstrate that the error growth related to a significant SPB for El Nino prediction generated by the FGOALS-g model can result from two dynamical mechanisms: in one case, the prediction errors grow in a manner similar to El Niño; in the other, the prediction errors develop with a tendency opposite to El Niño.”

    http://onlinelibrary.wiley.com/doi/10.1002/joc.3513/abstract

    “Using the sea surface temperature (SST) predicted for the equatorial Pacific Ocean by the Flexible Global Ocean-Atmosphere-Land System Model-gamil (FGOALS-g), an analysis of the prediction errors was performed for the seasonally dependent predictability of SST anomalies both for neutral years and for the growth/decay phase of El Niño/La Niña events. The study results indicated that for the SST predictions relating to the growth phase and the decay phase of El Niño events, the prediction errors have a seasonally dependent evolution. The largest increase in errors occurred in the spring season, which indicates that a prominent spring predictability barrier (SPB) occurs during an El Niño-Southern Oscillation (ENSO) warming episode. Furthermore, the SPB associated with the growth-phase prediction is higher than that associated with the decay-phase prediction. However, for the neutral years and for the growth and decay phases of La Niña events, the SPB phenomenon was less prominent. These results indicate that the SPB phenomenon depends extensively on the ENSO events themselves. In particular, the SPB depends on the phases of the ENSO events. These results may provide useful knowledge for improving ENSO forecasting. ”

    http://www.iapjournals.ac.cn/aosl/EN/abstract/abstract116.shtml

  8. The word ‘chaos’ only appears in the retort at the end. I think it reflects a misunderstanding of what it means to be ‘chaotic’. Chaotic things are predictable, just not very far ahead. Random things are unpredictable and weather/climate is not random at all. In fact weather is quite predictable over short time scales.

    Do the authors think it is ‘unpredictable’ because modellers have failed to model? Trying to predict El Ninos using CO2 is laughable. We can presume tbe modellers are not that silly, right?

    I think Bob Tisdale has a good grip on how ENSO works but I don’t recall him making ‘predictions’. Did he?

  9. Right, so there is a complex, multivariate, non-linear system with definite but unknown feedback and some of the variables as yet also unknown. That abstract, clear and concise as it is (in stark contrast to almost every scientific paper I have ever read, and already leading one to suspect that the authors are actually intelligent) must be one of the most unsurprising in the history of science.

    So, for all the geeks out there, what does the “B” stand for in Benoit B Mandelbrot?

    Answer: Benoit B Mandelbrot.

    I thank you.

    Your humble servant,

    The Master of Bad, Nerdy Jokes

  10. Oh guys give Simon a break. This is the period when alarmists get to gloat, when someone PREDICTS the temperature record MAY be broken (like when they predicted the arctic ice would disappear by 2013). Because, obviously, when the record ISN’T broken and Simon goes back to being a butthurt warmist all he’ll have is the ability to call everyone laughing at him a “denier.”

    He still won’t learn that believing the people who are always wrong may be an indication that those people don’t know what the hell they’re talking about though.

  11. What’s the sun doing right now. It’s up to her. El Nino is really dependent on cloud cover.

  12. When something is this unpredictable, it means that there are significant inputs to the process that have not been identified. We are probably seeing the inputs and are disregarding them.

  13. Crispin in Waterloo but really in Johannesburg says:
    I think Bob Tisdale has a good grip on how ENSO works but I don’t recall him making ‘predictions’. Did he?

    “Just about all indicators are pointing to a moderately strong El Niño for the 2014/15 ENSO season…The subsurface temperature anomalies along the equatorial Pacific associated with the downwelling (warm) Kelvin wave are quite warm…Eventually, some (but not all) of that warm water will rise (be drawn) to the surface. ”

    http://wattsupwiththat.com/2014/03/24/enso-update-outlook-suggests-a-moderately-strong-el-nino-for-the-201415-enso-season/

    That’s what Bob wrote. Me, I’ve voting for La Nada.

  14. Simon sez….just what exactly?
    – El Nino is associated with Christmas, but not linked.
    Even if some warming starts back up, the problem for the AGW believer is that the climate is still not behaving as predicted. More importantly, it is clear the climate is not undergoing some catastrophic rate of change. Happily, the hiatus, which by the way is by some measures actually more like 20 years long, supports the skeptic position that climate sensitivity to CO2 is much lower than the catastrophist hype would have us believe.

  15. ‘Shaken champagne bottle’… It is coming fast, it’s going to pop and over flow, everyone will say WOW and it will fade just as fast. Fade by November

  16. “Here come de heap big warmy. Bigtime warmy warmy. Is big big hot. Plenty big warm burny hot. Hot! Hot hot! But now not hot. Not hot now. De hot come go, come go. Now Is Coldy Coldy. Is ice. Hot den cold. Frreeeezy ice til hot again. Den de rain. It faaaalllll. Make pasty

    [...]

    So who are we to believe? For a final word, I turned to the greatest climate change scientist of all, Dr David Viner, one-time senior research scientist at the climatic research unit of the University of East Anglia, who predicted in 2000 that, within a few years, winter snowfall would become “a very rare and exciting event”.

    However, he was trapped under a glacier in Stockport, so was unable to comment at the time the Telegraph went to press.

    http://blogs.telegraph.co.uk/news/seanthomas/100222487/when-it-comes-to-climate-change-we-have-to-trust-our-scientists-because-they-know-lots-of-big-scary-words/

  17. If they can make Hockey sticks out of historic climate, what stops them from making Hockey sticks out of present weather?

  18. Craig says:March 26, 2014 at 7:39 pm
    “Entirely unpredictable?”
    The correlation that you show, is it an input or output of the process?

  19. “Will we see an El Niño this year? Only chaos knows for sure.”
    =================================
    Well, if anyone knows it would be Bob Tisdale.
    He never clues us in though.
    Makes you think he is still trying to just bury us with data, so we can individually make fools of ourselves.

  20. Crispin in Waterloo but really in Johannesburg says:

    “The word ‘chaos’ only appears in the retort at the end. I think it reflects a misunderstanding of what it means to be ‘chaotic’. Chaotic things are predictable, just not very far ahead. Random things are unpredictable and weather/climate is not random at all. In fact weather is quite predictable over short time scales.”

    Really predictable, not really it can only be a few hours out with any degree of accuracy and that may be only for a few variables such overall wind speed temperature, as to precipitation it generally on a bush basket guess never can you nail in tenths of an inch. most of the time it plus a minus a half inch sometimes off by several inches. The three day out forecast can often vary widely some time it arrives a day early or a day late and then sometime not at all. If the atmosphere is in stable state such as a blocking high often the computer prediction works out well but when the atmosphere is unstable all bets are off. Of course years ago just working from weather maps a experienced meteorologist often did just as well as todays weather models.

  21. The statistical models do a slightly better job, given that they are based on matching analogue years demonstrating similar predisposing parameters. The dynamical models aren’t worth crap on a sidewalk.

  22. I see that the abstract says:

    Climate simulations suggest that multi-decadal periods of high and low variability in the phenomenon known as the El Niño-Southern Oscillation in the tropical Pacific Ocean may be entirely unpredictable.

    I love that. They say that climate simulations suggest that we may not be able to simulate the climate … you sure this wasn’t published in The Onion?

  23. “””””…..Climate simulations suggest that multi-decadal periods of high and low variability in the phenomenon known as the El Niño-Southern Oscillation in the tropical Pacific Ocean may be entirely unpredictable……”””””

    The hell you say !

    So just what is it then, that you ARE simulating ??

  24. This is pure nonsense. ENSO is not caused by Gaussian noise and it is not caused by some butterfly that is flapping its wings in the Amazon as sometimes are suggested with chaos theory.
    The main driving forces for ENSO are tidal forcing, magnetic forcing and the inertia in the affected sea currents. An El Niño condition is not going to exist at the end of this year. It’s going to be in La Niña or neutral condition.

    The reason that they haven’t figured this out is that they haven’t looked.

  25. Reality imitates comedic art.

    From the Wikipedia entry on “Get Smart” the TV comedy from the 60’s.
    “The nemesis of CONTROL is KAOS, described as “an international organization of evil.”

    Al Gore and his CAGW followers are “Kontrol” and global climate chaos (KAOS) really does confound their efforts at world control. Controlled chaos… an oxymoron.

  26. Per Strandberg, I agree WHOLE HEARTEDLY. It sounds trivial but these jokers, who should know better, just want to keep being heard. Like clairvoyants or star gazers (astrologers). e.g. There is a 50-50 chance I might die in the next 20 years. You know folks, instead of all these carbon credits and taxes etc., why aren’t they putting in money to prevent extreme weather events killing people. Flood levys,(or is it levies) sea wall breaks, cyclone proof homes, shelters, advanced warning systems (Well if Vesuvius blows again, they won’t get much warning if 79 AD is an example). Stop growing palm oil and bio-fuels instead of crops. Improve degrading soils naturally. Stop growing mono crops that increase diseases in plants, and deplete the soils eventually. Build more dams to store fresh water. But we can adapt to these and maybe save lives, especially that landslide in America.
    We had something similiar in Australia in one of our ski resorts, only one man survived. There was warning prior to this event the land was unstable. And bush fires? You know most of our bush fire fighters are volunteers, although the governments do pay compensation of course. Bit late if your home and possession or live stock is lost.

  27. quoting “from behind the paywall”:

    “Climate simulations suggest that multi-decadal periods of high and low variability in the phenomenon known as the El Niño-Southern Oscillation in the tropical Pacific Ocean may be entirely unpredictable.

    The episodic warming and cooling of the surface temperature of the tropical Pacific Ocean, known as the El Niño–Southern Oscillation (ENSO), causes year-to-year climate fluctuations, affecting weather, ecosystems and economies around the world. The occurrence of these episodes is not regular. For example, whereas the period covering the years 1970–2000 witnessed the strongest El Niño (warming) events on record, the years since 2000 have experienced fewer and weaker such events. Writing in the Journal of Climate, Wittenberg et al.1 make the case that these multi-decadal epochs of enhanced and subdued ENSO activity occur randomly and therefore may be unpredictable.

    Changes in ENSO behaviour from decade to decade are commonly seen in historical observations and palaeoclimate proxy records2, 3, 4, 5. These variations were first put into context by Wittenberg in an earlier study6, which examined a 2,000-year simulation based on a fairly realistic climate model, known as GFDL-CM2.1. This concluded that decadal- to centennial-scale changes in ENSO behaviour can be internally generated by the model in the absence of any external forcing, such as increases in greenhouse-gas concentration or variations in solar output.

    Predicting whether the coming decades will bring an onslaught of strong ENSO events — or none at all — is crucial because of the impact of such events on weather patterns around the world. Individual episodes may be predicted up to two years in advance7, but on larger timescales our ability to forecast ENSO behaviour accurately may hinge on how ENSO responds to changes in the background climate system. This idea is supported by studies suggesting that the level of activity could be related to natural or man-made changes in the climate of the tropical Pacific8, 9.

    In their latest study, Wittenberg et al.1 used the same GFDL-CM2.1 model, this time to forecast epochs of high and low ENSO activity. For each epoch of activity in the model’s control run, the authors performed 40 forecasts, each differing by a tiny perturbation of the size of the computer’s rounding error to one of the model’s numerical grid points. These ‘perfect model’ forecasts have the best chance of reproducing the extreme ENSO epochs seen in the control run, and permit assessment of the model’s intrinsic ability to predict them.

    The authors found that, beyond the first two to four years after initialization of the forecasts, the multi-decadal epochs of high and low ENSO activity are completely unpredictable. For each epoch, the model forecasts either active or quiet events with the same probability. That is, the perturbations can alter the forecasts in such a way that the model is capable of forecasting an inactive ENSO decade where it originally simulated a highly active one (Fig. 1a), or an active decade where it simulated a quiet one (Fig. 1b).

    Figure 1: Decadal forecasts of El Niño–Southern Oscillation

    Using their GFDL-CM2.1 climate model, Wittenberg et al.1 performed simulations of the variability of the El Niño–Southern Oscillation (ENSO), here quantified using the Niño-3.4 sea surface temperature (SST) index. Two distinct epochs are shown, characterized by high (a) or low (b) ENSO variability in the model’s control run (black curves). For each epoch, 3 sets of 40 simulations are initialized from model years 1721, 1731 and 1741, and from 1151, 1161 and 1171, respectively (faint grey curves). Model years do not coincide with historical years. The simulations differ in only one tiny perturbation applied to the model’s initial conditions. For both epochs, at least one simulation out of 40 exhibits high (red curves) or low (blue curves) ENSO variability, indicating that the level of activity is unpredictable. (Graphic courtesy of Andrew Wittenberg.)

    Full size image (104 KB)

    This is the ‘butterfly effect’ of chaos theory applied to ENSO events. Seemingly small perturbations to a system, such as the flapping of a butterfly’s wings, may lead to large changes in that system. It is a sobering finding, because it suggests that the changes observed in ENSO behaviour during the twentieth century could very well be random fluctuations unrelated to natural or man-made changes in the climate of the tropical Pacific.

    Further research is needed to determine whether the study’s conclusions can be extrapolated from the model world to the real world. During the past decade, climate models have progressed substantially in their ability to simulate ENSO events. Many models can now emulate the long-term modulation first seen in the GFDL-CM2.1 simulations, possibly owing to the inclusion of improved wind patterns11. But it is not known whether even the best climate models simulate the correct mix of the myriad processes that influence ENSO. One cause of uncertainty might be that the decadal fluctuations in the background climate, which are thought to be the source of ENSO predictability12, are too weak in the models’ simulations13. Conversely, models simulate activity that is much stronger than observed6, 14, so this too-strong ENSO might be oblivious to the too-weak changes in background climate, resulting in decreased predictability. The realism of the simulations must be improved if model-based conclusions are to be applied to the real world.

    Existing observational records are not yet long enough for us to investigate whether, and how, ENSO responds to long-term climate fluctuations that could be sources of predictability. Progress on this front depends on maintaining and expanding our observational capability in the ocean, which relies on arrays of autonomous profiling floats and tropical moorings. In the meantime, results such as those of Wittenberg et al. are reminders of the challenges associated with forecasting ENSO changes. Future attempts to attribute the causes of individual events and their decadal variations now face a much higher bar.”

  28. John Doyle surely holds the record for a short pithy abstract when he dropped a bombshell on the feedback control field in his paper “Guaranteed Stability Margins for LQG Regulators”, IEEE Trans. Auto. Control, 1978, Vol. AC-23. The abstract read: “There aren’t any”.
    The author(s) could have recycled Doyle’s abstract if they’d titled this new paper with something along the lines of “Useful methods for predicting ENSO”.

  29. @RichieP March 26, 2014 at 5:44 pm:
    “I think we all already would agree that global warming has been occurring since the end of the LIA. Nothing controversial there.”

    The question might be if the “has been occurring since the end of the LIA” is true or not. Let’s consider…

    Going back, there were steep rises in ~1970 to 1998, ~1910 to 1940, ~1850 to 1880, and the LIA ended about 1800 (some put it as late as 1900, some as early as 1762 or so). In between those rises there were “hiatuses” more or less like the pause right now. Each hiatus wasn’t exactly a decline, though the 1940-1970 one dropped a little. Each hiatus began at then end of one of the steep rises, so each hiatus was at a higher level than the one preceding it.

    As I understand her paper, Judy Curry calls it the “Stadium Wave.” But she wasn’t the first to point it out. The first I saw of it was back about 2006 in a YouTube video where a lecturer noticed it on camera and wondered about it. It was basically, starting at the end of the LIA, a sine curve with an incline to the whole thing, so that rises were steep and downlsopes were more or less horizontal. He commented that maybe he should do a paper on it. If he did I dno’t know about it.

    So, allowing for those small drops during the hiatuses, yes, global warming has been going on since the end of the LIA. How anyone wants to explain the rises and falls, that is not the question. The rise has existed since about 1800 – 150 years before the advent of the increase in industrial CO2 emissions. Any global warming-CO2 hypothesis thus is required to explain why those rises existed for those 150 years.

    In addition, they need to also explain why tree rings began diverging from the CO2 and temps at about that same time – 1940 or 1950.

  30. @Willis

    Well spotted!

    “They say that climate simulations suggest that we may not be able to simulate the climate … you sure this wasn’t published in The Onion?”

    Their ability to model reality suggests that reality is not susceptible to being modeled. That is right up there with, “When reality doesn’t match the models, reality is wrong.”

  31. bushbunny says:

    March 26, 2014 at 10:11 pm
    ” There is a 50-50 chance I might die in the next 20 years.”
    =========================
    I’ve built a model that indicates a 95% chance I’ll die within the next 20 years.
    The model parameters are alcohol, cigarettes, gambling and free time.
    (not necessarily in that order, at all times).
    Other than tornadoes, I think I’m safe from the ravages of Global Warming (or whatever).
    I’m lucky to have lived thru these first 51 years in a healthy state.
    I’ve got no complaints.

  32. It does seem likely that the relative strength of El Nino events and La Nina events can vary in line with changes in the level of solar activity.

    A quiet sun gives a cloudier globe due to more wavy jet stream tracks which reduces solar energy into the oceans and so weakens El Ninos compared to La Ninas.

    Hence the decline in strong El Ninos of late.

  33. This Nature item is about the following new (modelling!) paper in the Journal of Climate.

    http://journals.ametsoc.org/doi/abs/10.1175/JCLI-D-13-00577.1

    Wittenberg et al., 2014: ENSO Modulation: Is It Decadally Predictable?. J. Climate, 27, 2667–2681. doi: http://dx.doi.org/10.1175/JCLI-D-13-00577.1

    From the abstract:

    ” …no decadal-scale predictability of ENSO behavior is found. This indicates that multidecadal epochs of extreme ENSO behavior can arise not only intrinsically but also delicately and entirely at random. … … those slow variations appear not to lend significant decadal predictability to this model’s ENSO behavior … … the potential implications of these results are sobering for decadal predictability …”

  34. Retired Engineer John says:
    March 26, 2014 at 7:09 pm
    When something is this unpredictable, it means that there are significant inputs to the process that have not been identified. We are probably seeing the inputs and are disregarding them.

    Unfortunately, that is not necessarily the case. Chaotic systems vary strongly based on very small changes in intial conditions. As summarized by Lorenz:

    “Chaos: When the present determines the future, but the approximate present does not approximately determine the future. ”

    We could probably list all the inputs for which way a pencil, balanced on its point, will topple. But because very small changes in those inputs could result in the pencil toppleing in a different direction, it is for all practical purposes unpredictable.

  35. From the original Wittenburg paper abstract:

    These 40-member reforecast ensembles display potential predictability of the ENSO trajectory, extending up to several years ahead. However, no decadal-scale predictability of ENSO behavior is found. This indicates that multidecadal epochs of extreme ENSO behavior can arise not only intrinsically but also delicately and entirely at random.

    They could have saved themselves a lot of effort by just reading “Deterministic Nonperiodic Flow by Ed Lorenz (1962):

    http://www.astro.puc.cl/~rparra/tools/PAPERS/lorenz1962.pdf

    This paper is the foundation of any meaningful climate science.

  36. When something is this unpredictable, it means that there are significant inputs to the process that have not been identified. We are probably seeing the inputs and are disregarding them.

    Well, no, often it means that the process involves solution to stiff systems of differential equations or chaotic dynamical systems, either one of which are highly sensitive to small tweaks of their initial conditions because a bundle of solutions started from nearby points tend to rapidly diverge to fill a phase space of possible outcomes, possible future histories.

    Weather is not only such a system, it is the dynamical system where this (in retrospect) unsurprising fact was discovered, even though at this point it has long been reduced to simple classroom demonstrations. It’s like the feedback loop I often observe when boiling down my wort making beer — under the right conditions of heating and cooling, the entire system undergoes an irregular oscillation where it suddenly boils up, releasing a bunch of heat and cooling the bottom with new, infalling cooler wort from the top. The boil then slows or ceases for some time while the new wort makes it across the boiling threshold. It then surges up, displaced by the cooler wort on top and the sides, and a burst of bubbles and foam arrive on top. But this again quenches the boil, and the cycle repeats itself.

    The time in between and the specific structure of the surges as the liquid “chuffs” in this way are “unpredictable”, except for the certain knowledge that they are going to occur and some very rough bounds on the times involved — the liquid is being heated and cooled at different places and this phenomena is going to occur until its temperature is uniformly close enough to boiling that a single boiling convective roll establishes itself that picks up heat at the bottom, boils up to the top, dumps its steam, and redescends (basically a Prigogene-style self-organized pattern that optimizes heat transport from the bottom to the top, a heat engine).

    Note that this is in a stationary, unstirred liquid. If I stir it, I can completely alter or suppress the cycle. I can even alter it in specific ways (this is what physicists do when making beer, we can’t just down an ale or two and watch, we have to figure things out:-). If I stir the liquid into a spin in the round pan, the center fluid is nearly stationary and there is a tipping angle on the buoyant force that makes the center axis of the pan bottom the location of the densest fluid. This trapped fluid is the stuff that accumulates heat to go over the boiling point so that the surge comes up in the middle of the pot, pushing a burst of foam out across the rotating surface in a coriolis-driven spiral. This lasts only a second or two, however, and as the steam is released there is a general collapse back towards the pan axis as new cool fluid is pushed in on the bottom and trapped in the middle. The boil stops. A few seconds later it surges up again, stops, again, stops, each time making a spiral galaxy of foam first appear, then collapse back to the center. Eventually the spin slows, and as it slows I often get a boil that establishes on one side and the usual stable convective boil appears.

    It is my guess that ENSO and the other decadal oscillations are in many respects quite similar conceptually to this simply stovetop “system”, with obviously greater complexity. The system is a differentially heated spherical system with an interface that is bottom heated for one fluid and top heated for another (much denser) fluid. The sphere is rotating, so that there is a much more complex pattern of differential coriolis forces on the surface in addition to a reasonably but not perfectly consistent gravitational field. The entire system is being differentially heated and cooled with two primary harmonic components — one with a (TOA) strength of 100% of peak and a period of 24 hours, and a secondary annual variation of the peak with an overall amplitude of around 3.5% (7% total range). However, even this insufficient to describe the sphere surface, as entire substantial regions of the tipped planet have a 100% annual cycle of TOA strength, and the entire planet has a latitude dependent roughly sinusoidal modulation with this period, meaning that the fourier transform of the effective TOA flux to the surface isn’t sharp at 1 year and 1 day, but is indeed rather broad with lots of contributing frequencies and artifacts.

    Then, the fluids being heated and convected are a lot more complex as well. The lower fluid is almost all at or very near 4 C, its temperature at peak density, in a state of very, very slow non-thermally driven motion that arises because of surface phenomena involving evaporation and density variation due to a substantial load of solute salts. The top few hundred meters, however, deviates from this temperature due to a complex process of differential flow fluid dynamic turnover (the fluid is not in laminar flow and hence develops large scale eddies and rolls even when surface heated) that — like my pot — absorbs heat near the equator, reducing density, displacing the entire surface generically poleward, so that it cools as it moves, dumps its heat where it is much cooler than where it is absorbed, becomes denser than the surface waters, descends, and displaces an absolutely enormous volume of water that moves in a very, very slow convective roll around the entire planet. This roll is so long that it actually makes multiple trips to the top along the way to warm, move, cool, and sink, often in less than obvious and possibly long term unstable locations.

    A second problem is that the pot isn’t nice and cylindrical and shiny — the fluid flow is constrained by continental borders that look like they were drawn by van Gogh during his electric I’m-losing-my-mind phase. There are whole patches of nearly or completely circumscribed water — the Gulf of Mexico and the Mediterranean and its the linked saltwater seas — as well as whole pole to pole stretches with extremely variable width — the two primary oceans. The bottom depth is similarly fractally irregular, dotted with islands and intruding whole continents (like Australia). As fluid tries to flow poleward past these continental boundaries and over this bouncy bottom, eddies, traps, whorls are set up that cause e.g. a certain amount of “chuffing” of gulf stream water as it is heated in the summer time.

    A third problem is that this isn’t the only fluid. On top of this fluid is a fluid that itself is heated at the BOTTOM, but not just by surface contact (that would be too easy!). It is heated at the bottom by surface contact, latent heat transport from the heavier liquid at the semi-sharp interface in between them, and (sigh) is heated in depth and in opposite directions by radiation both coming in from the TOA and going back up from the surface and the surrounding fluid itself in entirely different spectral band based on the specific chemistry of the fluid itself, which is not particularly constant as the concentration of key absorptive components is highly variable with temperature, density, location, and dynamically, with the very specific history the vary specific parcel of fluid in question. Its surface interaction is similarly constrained by an irregularly shaped continential profile where the surface is covered with things like mountain ranges that reach up to the top of the troposphere and arbitrary mixtures of smooth, vegetation free sand or rock, grasslands, forested lands, farm lands, lakes and rivers, plains and foothills and cities and… all of which interact differentially with the fluid as it passes over it in contact.

    I wish we were done, but of course we are not. This upper fluid is cooled almost exclusively at the top (as is the lower fluid!) and by a single mode — radiation. The catch is that the fluid is differentially penetrable by radiation consistent with the general thermal range of the fluid and surface, in a way that is also highly chemistry dependent. Worse, we have completely ignored the extreme variability and stationary variation of the surface’s absorptivity to incident radiation in the various wavelengths of delivery — parts of the surface are comparatively highly reflective, parts are highly absorptive. Then there are phase change linked changes in reflectivity. Finally, the pesky lower fluid has an annoying habit of transporting latent surface heat upwards to where it can be radiated away and then enabling a phase change of the lower fluid that first makes it into fluffy, high albedo clouds that differentially reflect a very high fraction of any incoming radiation (in the day) and also differentially reflect a very high fraction of any outgoing surface radiation back (day or night) that otherwise might escape the system to keep it in thermal balance. The fluid, cooled, then falls back down to eventually rejoin the lower fluid, as a schizophrenic addition — it is lower in the dissolved salts and hence dilutes the baseline fluid density, but is also cooler and hence increases it as the fluid re-mixes. Instabilities galore are created, enabling and driving long-length scale transport process that get pushed this way and that and up and down as the lower fluid moves.

    The upper fluid also moves. It picks up heat, rises, spreads out and moves polewards, being deflected to spinward as it moves, even as the cooled fluid on the return loop returns to the hottest (equatorial) patch and is deflected to antispinwards (relative to the surface) as it does so. These general motions are modulated by the annual differential polar heating and instabilities are easily created by interactions with the specific state of the underlying fluid surface, by the eddies created by surface and boundary structure, by small variations in the many not-really-constant drivers, and by one particular butterfly in Brazil that controls both hurricane season on the East Coast and ENSO by flapping its wings — in a chaotic system a perturbation that small will create an entirely different — substantially different — future history in a completely deterministic evaluation of the physics at very high spatiotemporal resolution (something we are utterly incapable of at this point).

    At this particular moment, we cannot integrate equations of motion at a spatiotemporal resolution sufficient to predict a chaotic thermally drive chuffing caused by a complex process of heat absorption, transport, and release in this system — ENSO. That is not surprising! We cannot integrate the equations of motion of the vastly simpler fluid in my heated, smooth, cylindrical boiling pot that describe the timing of the chuffing of the boil up from the center under nearly ideal conditions, let alone conditions that might hold if I leave the spoon in to create rotational eddies or periodically turn the heat on or off, up or down, or just give the liquid a bit of a random stir.

    We cannot even determine the initial conditions to use to solve the coupled equations of motion that describe ENSO more than crudely, from scattered, comparatively sparse measurements compared to the volume and surface of the heavier fluid involved, and have absolutely no hope of decribing the detailed pattern of fluid motion of the lighter fluid on the surface, which couples to the entire past thermal/climate history of the planet (which is even more unknown).

    So it is not just a matter of not knowing some key aspect of the dynamics. Even if we had the exact equation of motion for the system, even if we had exact knowledge of the actual initial conditions on some comparatively fine spatiotemporal grid, specified to six significant digits, we would see a substantial spread of final solutions to the coupled equations due to variations in the seventh significant digit, and the actual evolution might or might not resemble any of the solutions generated in an ensemble of a few hundred independent runs with these initial conditions because of an unpredicted weather/climate event — a tropical storm, say — that rapidly kicks the system outside of the entire phase space range of the trajectories predicted.

    That isn’t to say nothing can be done with models. The closer you get to what you are predicting, the less time the initial bundle of solutions has to diverge based on butterfly wing flaps. Hurricane models, for example, often such badly a week out, but one or two days out they are pretty good as the space of accessible futures collapses. For even larger scale, slower phenomena like ENSO, predicting a year out is probably impossible because a single large tropical storm in the pacific basin could alter temperatures enough to change everything. But a few months out? Sure, why not. It is interesting that there is a barrier to the range of prediction — it suggests a high degree of sensitivity to chaos elsewhere that is still indeterminate at that time — a question of just which of the vast number of potentially nucleating fluctuations grow and which ones damp out. After one has nucleated and grown past some critical point, however, the system probably slaves to it and becomes comparatively predictable.

    rgb

  37. The model parameters are alcohol, cigarettes, gambling and free time.
    (not necessarily in that order, at all times).
    Other than tornadoes, I think I’m safe from the ravages of Global Warming (or whatever).

    Don’t forget sharks. And meteors. But especially sharks.

    rgb

  38. Thanks Joel O’Bryan @ 10:22 last night for the peekhole through the paywall. I particularly liked: ‘This concluded that decadal- to centennial-scale changes in ENSO behaviour can be internally generated by the model in the absence of any external forcing, such as increases in greenhouse-gas concentration or variations in solar output.’
    ==============================

  39. rgbatduke says:
    March 27, 2014 at 7:32 am

    When something is this unpredictable, it means that there are significant inputs to the process that have not been identified. We are probably seeing the inputs and are disregarding them.

    Well, no, often it means that the process involves solution to stiff systems of differential equations or chaotic dynamical systems, either one of which are highly sensitive to small tweaks of their initial conditions because a bundle of solutions started from nearby points tend to rapidly diverge to fill a phase space of possible outcomes, possible future histories.

    Weather is not only such a system, it is the dynamical system where this (in retrospect) unsurprising fact was discovered, even though at this point it has long been reduced to simple classroom demonstrations. It’s like the feedback loop I often observe when boiling down my wort making beer — under the right conditions of heating and cooling, the entire system undergoes an irregular oscillation where it suddenly boils up, releasing a bunch of heat and cooling the bottom with new, infalling cooler wort from the top. The boil then slows or ceases for some time while the new wort makes it across the boiling threshold. It then surges up, displaced by the cooler wort on top and the sides, and a burst of bubbles and foam arrive on top. But this again quenches the boil, and the cycle repeats itself.

    So your beer boiler is a version of the Belousov-Zhabotionsky reactor? How do these dynamics affect the beer’s taste?

  40. BS (bad science)!

    ENSO has a clear ~300 months +/- ~3 months cycle. Bob Tisdale can confirm I contacted him about this and that I am still working on the data analyses. This cycle links past, current and future ENSO events consistently. Using this cycle one is likely able to accurately predict ENSO events with great accuracy up to 25 years ahead of time. I still need to finalize the data analyzes, because right now I am more or less able to identify the exact ONI-seasons (3 months period) when these events are going to happen, but not yet what type; especially further ahead in time (2016 and beyond). The data will likely tell me that eventually too.

    For now, the cycles show
    2014 MAM-season: ONI turns +?
    2014 MJJ-season: el nino starts?
    2014 OND-season: el nino peak?
    2015 JFM-season: el nino ends?
    2015 NDJ-season: ONI turns -?
    2016 JAS-season: la nina starts?
    2016 NDJ-season: la nina peak?
    etc up to 2037
    Again, this is still preliminary as I need to link all the different cycles to better understand which event links to the next and the next etc, so don’t hang your hat on it just yet, but there’s nothing unpredictable in the ONI data at all.

  41. Climate science: A high bar for decadal forecasts of El Niño

    – Pedro DiNezio1

    One more data point demonstrating that Climate Science has all the predictive powers of Botnay.

  42. Geoff Withnell says: March 27, 2014 at 7:21 am
    rgbatduke says: March 27, 2014 at 7:32 am
    phlogiston says:March 27, 2014 at 8:13 am
    When I read you comments, I realized that there were two areas of difference.
    1. I did not reference Chaos theory; although, it could be appropriate to this discussion. Chaos theory should produce boundaries to where the pencil falls, even if does not tell you which direction. These predictions are short term enough that significant error should not accumulate as to make the predications false almost from the start.
    2. My engineering background rejects random events that do not have causes. Generally random events are not random, they are simply events where the reasons are not known. There are many things that are not known about the Ocean and the processes that occur deep in the Ocean. Here is an example: http://gyre.umeoce.maine.edu/physicalocean/Tomczak/regoc/pdffiles/colour/double/08P-Pacific-right.pdf
    Currents below 2000 meters, page 119
    “At the same latitudes, the cores of the South Subsurface Countercurrent and the North Subsurface Countercurrent are usually located near 600 meters. An explanation for the existence of these currents is still lacking. Recent observations indicate that the banded structure of currents at the equator continues to great depth (Figure 8.15). …(the figure shows currents at depths greater than 2,000 meters.)… the dynamics of the equatorial region cannot be explained by our 1 1/2 layer model. The EIC, NSCC, and SSCC are integral part of a dynamic system that reaches much deeper than the thermocline.”
    When you find that there are unknowns about the Ocean currents that would directly impact movement of heat, etc. in the area most sensitive to the ENSO, you cannot expect that predictions will be reliable.

  43. Thanks- Joel O’Bryan says:
    March 26, 2014 at 10:22 pm
    for the bit more detail. The paper is only looking at one models ability to predict increasing or decreasing strength ENSO events. The model obviously obviously can’t predict in its own simulated results, let alone reality. At least they were honest in saying that.
    rgbatduke says:
    March 27, 2014 at 7:32 am
    “Well, no, often it means that the process involves solution to stiff systems of differential equations or chaotic dynamical systems, either one of which are highly sensitive to small tweaks of their initial conditions”
    I can’t help but think that you and Retired Engineer John are talking about the same thing.
    In the real world I see the Kelud eruption (Feb 2014 VEI5?) as likely more than just a small tweak of “initial conditions”. The satellite temperature data (and the long/shortwave data) for the next couple of years will be interesting.

  44. Retired Engineer John says:
    March 27, 2014 at 5:14 pm
    Geoff Withnell says: March 27, 2014 at 7:21 am
    rgbatduke says: March 27, 2014 at 7:32 am
    phlogiston says:March 27, 2014 at 8:13 am
    When I read you comments, I realized that there were two areas of difference.
    1. I did not reference Chaos theory; although, it could be appropriate to this discussion. Chaos theory should produce boundaries to where the pencil falls, even if does not tell you which direction. These predictions are short term enough that significant error should not accumulate as to make the predications false almost from the start.
    2. My engineering background rejects random events that do not have causes. Generally random events are not random, they are simply events where the reasons are not known. There are many things that are not known about the Ocean and the processes that occur deep in the Ocean. Here is an example: …

    The position you outline is known as determinism. If we know everything we can predict everything. It was well articulated by Simon Laplace in 1814:

    We may regard the present state of the universe as the effect of its past and the cause of its future. An intellect which at a certain moment would know all forces that set nature in motion, and all positions of all items of which nature is composed, if this intellect were also vast enough to submit these data to analysis, it would embrace in a single formula the movements of the greatest bodies of the universe and those of the tiniest atom; for such an intellect nothing would be uncertain and the future just like the past would be present before its eyes.

    —Pierre Simon Laplace, A Philosophical Essay on Probabilities

    However this statement of determinism is false, and has been entirely falsified by two principal scientific discoveries of the 20th century. One of these is quantum physics – for example at the quantum scale, particles continually appear out of nothing then disappear – giving rise to radiation at the event horizons of black holes for instance when one of a pair of such spontaneous particles is sucked into the black hole and the other not.

    The other equally fundamental and important discovery is chaos and nonlinearity. The Russians discovered this first with the likes of Kolmogorov in the early-mid 20th century, in the west Lorenz first articulated the unpredictability of a chaotic system in DNP62, the period doubling phenomenon underlying it was illucidated by Feigenbaum. (In the west Lorenz was largely ignored for several decades and even now many still labour in futile analysis of complex systems unaware of his work – such as all climate modellers.)

    In summary – determinism is false. It doesn’t matter how much you know, quantum physics together with chaos together absolutely prevent any chance of determinism in a complex system. For certain systems, no amount of knowledge will allow you to predict them.

  45. phlogiston says: March 28, 2014 at 9:16 am
    “The position you outline is known as determinism. If we know everything we can predict everything. It was well articulated by Simon Laplace in 1814:

    We may regard the present state of the universe as the effect of its past and the cause of its future. An intellect which at a certain moment would know all forces that set nature in motion, and all positions of all items of which nature is composed, if this intellect were also vast enough to submit these data to analysis, it would embrace in a single formula the movements of the greatest bodies of the universe and those of the tiniest atom; for such an intellect nothing would be uncertain and the future just like the past would be present before its eyes.
    —Pierre Simon Laplace, A Philosophical Essay on Probabilities

    In summary – determinism is false. It doesn’t matter how much you know, quantum physics together with chaos together absolutely prevent any chance of determinism in a complex system. For certain systems, no amount of knowledge will allow you to predict them.”

    I agree determinism is false. I feel sure if Simon Laplace were alive today, he would also say it was false. My position is not determinism. We cannot determine when a radioactive atom will breakdown and emit some form of radiation or when certain things in quantum physics will occur. We can make estimates based on bulk properties and measured performance. The problem with designating ENSO as a chaotic system without thoroughly exhausting the search for causes means we can be overlooking processes that can be used to predict at least short time behavior.

    Random events do not necessarily fall in the category of chaos. A random event is not necessarily based on a series of events that multiply into a random event. The event can be caused by a stress that exceeded the design limits of the device. The event is considered random when the stress cannot be identified.

    I am interested in quantum mechanics and I would be interested in a paper or book that gives the details of your statement: ” One of these is quantum physics – for example at the quantum scale, particles continually appear out of nothing then disappear”. I have seen where Hawkings is working with these particles; however, there are no technical details.

  46. “mwhite says: March 29, 2014 at 5:08 am If Joe is right, not all El Ninos are equal.”
    I don’t get an article on El Ninos, I get a temperature map. Please check your URL.

  47. During the balmy Pliocene, before closure of the Isthmus of Panama, the tropical East Pacific was more or less in permanent El Niño mode.

    During the Holocene Interglacial, the millennial, centennial & decadal scale warm & cold fluctuations (I’ll call the longer term variations Bond Cycles & the shorter PDO/AMO phases, while praying for deliverance from the fury of the Northman Leif) appear to me associated with relative frequencies of El Niño & La Niña events. There seems to be more energy in Earth’s climate system during the warm phases. It would be surprising if solar cycles were not largely responsible for these observations.

    The same has probably held for prior interglacials & maybe even more so during the longer glacial intervals of the Pleistocene.

  48. Retired Engineer John says:
    March 28, 2014 at 5:36 pm
    phlogiston says: March 28, 2014 at 9:16 am
    “The position you outline is known as determinism. If we know everything we can predict everything. It was well articulated by Simon Laplace in 1814: …

    I agree determinism is false. I feel sure if Simon Laplace were alive today, he would also say it was false. My position is not determinism. We cannot determine when a radioactive atom will breakdown and emit some form of radiation or when certain things in quantum physics will occur. We can make estimates based on bulk properties and measured performance. The problem with designating ENSO as a chaotic system without thoroughly exhausting the search for causes means we can be overlooking processes that can be used to predict at least short time behavior.

    Random events do not necessarily fall in the category of chaos. A random event is not necessarily based on a series of events that multiply into a random event. The event can be caused by a stress that exceeded the design limits of the device. The event is considered random when the stress cannot be identified.

    I am interested in quantum mechanics and I would be interested in a paper or book that gives the details of your statement: ” One of these is quantum physics – for example at the quantum scale, particles continually appear out of nothing then disappear”. I have seen where Hawkings is working with these particles; however, there are no technical details.

    The wiki article on Hawking radiation cites as the original reference just a “conversation” with Charlie Rose. However there are other technical citations addressing particle-antiparticle spontaneous appearance and then one of the pair zipping into the black hole before it has the chance to disappear, the basis of Hawking black hole radiation, with the connected problem of trans-Plankian wavelength:

    http://en.wikipedia.org/wiki/Hawking_radiation

    Chaos by itself is not strictly speaking (philosophically) an argument against determinism, but, together with quantum mechanics, it becomes one. Quantum mechanics forbids the degree of precision of knowledge that would be needed to predict a chaotic system.

Comments are closed.