Gavin Schmidt, supermodeller: the emergent patterns of climate change

In this new TED talk, Gavin Schmidt, NASA climate modeller and juggler extraordinaire, talks about the climate system, how we use models, how they’re put together, and how the great swirls of earth’s atmosphere emerge from a million lines of Fortran code. It’s a great exposition, and the graphics he calls up in support are magnificent.

51 thoughts on “Gavin Schmidt, supermodeller: the emergent patterns of climate change”

      1. That was actually quite funny. Legacy Fortran is often really poorly coded. Regardless of how well it is written, debugged, and documented, there is bound to be coding errors in a million lines of Fortran. The question is whether they give unintended results that no-one picks up on. It would be a nightmare to maintain, Fortran does (did?) not have concepts like encapsulation and abstraction to reduce unintended side-effects.

        1. Fortran had the most evil of programming constructs, the computed GOTO.

          However, it was and is very good at array processing, hence its use in GCM simulations and seismic data processing.

          Back in the day, we used Fortran with dedicated array processing hardware module.

          The reason NASA and others still use it is probably because there isn’t a good business case to rewrite a million lines of code in another language. Today’s new thing is tomorrow’s legacy app.

          1. Yes “GOTO” as you say – evil!
            Actually the reason it is still used is because being pretty close to “assembler” – (now that was evil) it is still the fastest floating point language around.
            “Predicting weather and predicting climate are actually quite separate beasts. Sure, they both use stonking big computers and telephone-book-thick reams of Fortran code (yes, Fortran, for those sniggering up the back – it’s still the fastest floating point arithmetic language). But the divergence is in how the forecasts are made.”
            http://theconversation.com/a-chaotic-beast-probably-wacky-weather-and-climate-forecasting-5182

            1. On Assembler: When I started our company in 84 writing for the first generation Apple Macs I wrote a whole text processor drop in for our Desktop Publishing app in Assembler. Anything written in a high level language was simply agonizingly slow….
              I still remember dreaming in Hex code back then. Oh the good old times… 😉

            2. You evil person you! :)
              Wasn’t MacWrite good enough? lol I think I still have a 3.5in floppy with that on someplace…

            3. Yep! We squeezed our “RagTime”, a Desktop Publishing application with build in spreadsheets, graphics, a calculating text processor (spreadsheet expressions with their results floating in the text flow – something MSFT has not done to-date), and even an SQL DB interface for catalog publishing into a single 3.5″ floppy back in the 80ties and early 90ties… :-)

    1. Good question!

      My take on it is this: There are people who have early in the “game” declared their position (one of Denial) publicly on bogs and elsewhere, at a time when they had not actually considered the science carefully yet. But people find it incredibly hard to admit to being wrong. Instead they double down again and again, and dig themselves deeper and deeper into the dodo as this is the only way they see or know how to proceed. We have prime exemplars spamming this blog all the time…

      Then there are people who realize quite clearly that admitting to the severity and the cause of AGW will immediately lay blame at the door of an out-of-control exponential growth system called “libertarian Capitalism”. It is the cancer that is eating our future. However these people are totally married to the paradigms of this system and a global threat that outs the cancer for what it is, must be denied because the cancer diagnosis is inconceivable to them. They equate alternative suggestions of how society should govern itself to navigate round the abyss as some form of communism, which must be defeated at all cost.

      In the end, both these dispositions are frequently coming together in the same individuals. They will deny the the science and weasel around just like the young Earth bible sort will weasel around any discussion of Evolution or Paleo-Sciences with the most outrageous never ending nonsense. And it is also quite often the case of cause that people trained in science denial due to their religious fantasies are well placed to deny climate science and often do so to.

      Many Humans are so silly!

  1. Reasons for climate denial? I agree with Thomas some people dig themselves into positions.

    You also have the fossil fuel lobby who are quite powerful behind the scenes.

    You also have people who are very capitalist and see government regulation as a threat. I feel this tendency can go too far.

    Another reason may be more interesting. Apparently conservatives are much more likely to be climate sceptics. Conservatism and liberalism are believed to be innate traits people are born with, and both clearly have value. However if you are conservative, you may be uncomfortable about change, so climate change and changes to lifestyles or fossil fuel use would be seen as a major threat best denied.

  2. “Apparently conservatives are much more likely to be climate sceptics”

    1) The IPCC started when Reagan / George H.W. Bush were Presidents, and about that time, climate science in US wasn’t particularly politicized. That split was manufactured over the last 25 years … and I know folks affiliated with the Hoover Institution (not exactly a left-leaning place) who are quite concerned about climate change.

    2) “skeptics”: SeePseudoskeptics Are Not Skeptics. Jo Nova claims to own the word “skeptic,” but I don’t think so.

    3) But Gavin was off on one thing (well, really an in-joke that I suspect audience missed):
    Fortran, not “newfangled languages like C.” C is 40+ years old. :-)

    1. Yes it is! But then Fortran is even older! I started my programming on Fortran IV in the 1960’s on an IBM 360 with punch cards :) The task was to see which milking shed was the more efficient, herring bone or circular.

  3. Yes, Fortran dates to the 1950s, so nearly 20 years older than C.
    If you’re ever in San Francisco area, at the Computer History Museum in Mountain View, we still have working IBM 026 keypunches, in a room with working IBM 1401s and all their gear.

    Computers have come a long way since then, as Gavin’s models would not get far on those, or even bigger S/360s. We had a fairly large one, 1MB of fast memory, 8MB of Large Core Storage.

    Although I learned C in 1973, Fortran was still a good thing when I was designing supercomputers and helping sell them to NASA, NCAR, etc … where they spent most of their cycles running Fortran.

    1. Ours was big by NZ standards – The NZ Dairy Board as it was then (Now Fontara) held every cow and bull in NZ on tape. Each cows milk production – butter fat, carotin, etc etc. Selective breeding from the top producers and bulls lead to increasing milk production. We in the research branch only had the odd hour or two to run our small programmes two or three times a week.

  4. Quick question if someone doesn’t mind. I was reading something by a certain Judith Curry and she was talking about sea level rise slowing in recent years (from 3.5mm to 2.5mm/yr by memory). It has been well publicised that increased heat is being trapped in the deep ocean. I would think the resulting thermal expansion would lead to sea level rise increasing at a greater rate rather than slow down?

    I can’t find an explaination online and it doesn’t make sense to me. Any thoughts most welcome.

          1. Oh, I don’t know, I was sort-of wondering where Mr. ‘I’m Not Denying the Science, Me!’ was going to go with this. I understand that ‘the seas aren’t rising’ is a currently active meme in the ‘Quantum Climate Denial’ community…

  5. A space to watch:

    Climate change is clear and present danger, says landmark US report
    National Climate Assessment, to be launched at White House on Tuesday, says effects of climate change are now being felt.

    Climate change has moved from distant threat to present-day danger and no American will be left unscathed, according to a landmark report due to be unveiled on Tuesday.

    http://www.theguardian.com/environment/2014/may/04/climate-change-present-us-national-assessment

  6. SIgh. As often happens when climate and computing intersect, people say all sorts of things.

    IMPORTANT
    Of course models are large, but physics models need to follow conservation laws that tend to keep bugs from propagating too much.
    See Gavin’s excellent FAQs on models.

    If people reject the possibility that large Fortran codes can be “good enough”, they should never ride in cars designed from the 1990s onward, fly in new airplanes, cross newer bridges, go up in newer skyscrapers, use drugs designed with computers, or in fact use anything designed with software from these folks. Also, they can reject the idea that the design of Black Magic had anything to do with computers.

    I took a quick look at Gavin’s GISS code a few years back, it’s Fortran 90, not Fortran II or IV or even Fortran 77, and it looked decent to me.

    For people who would like a reasonably-general introduction, at the Scientific American level, I’d strongly recommend a 1993 (still relevant, and beautiful) book, Supercomputing and the Transformation of Science, which can be gotten fro not much more than cost of shipping. It’s well done, and one of the authors, Larry Smarr, is a distinguished supercomputing expert … and even better, who bought a lot of gear from us at SGI. :-)

    MINOR, BUT SILLY, AND ONLY FOR COMPUTER FOLKS:
    There was nothing particularly evil about Fortran’s Computed GOTO, a useful construct whose near-equivalents are found in most algorithmic languages. (example: “switch” in C and its derivatives), and has been deprecated/deleted in more recent Fortran standards that include SELECT CASE,
    The evil one was ASSIGNED GOTO, an absolute nightmare for optimizing compilers, which cannot be sure of the branch target at the GOTO itself. That was obsolescent in f90, deleted from 95.

    1. I was thinking that the computed GOTO was something along the lines of the old Basic construct GOTO i*100 where i is a variable. I checked on this last night and you are correct, John that the Fortran computed GOTO is more akin to the C switch statement. Nevertheless, the potential for evilness is still there,
      http://en.wikipedia.org/wiki/Considered_harmful
      as per Djjkstra’s famous letter, on the general GOTO statement, and in this cartoon
      http://xkcd.com/292/

      I was not implying that Gavin’s code was bad, or wrong, as a result, by the way. Developers like to knock their own products more than most, as exemplified by sites such as “The Daily WTF” and “Coding Horror”

  7. Its a repeating theme of deniers that the models are not right; and they are only models, and they were predicting we should all be dead by now; or underwater, or anything. Like most deniers themes they are not interested in the truth but if you keep repeating the same story it becomes an accepted urban myth.

        1. Remember how andy ‘missed the biblical references’ in a post entitled ‘Postcards from la la land: David Archibald and the four horsemen of the cooling apocalypse’ which contained as its first quote from the said Mr. Archibald:

          The Four Horsemen of the Apocalypse come from the Book of Revelation, the last chapter of the Bible. The Book of Revelation also warns of another beast with these words:

          …and rather went on from there?

          I don’t think andy pays all that much attention to the OP, frankly.

          We think the models have skill, andy.

          1. In this case, I have done my homework

            Gavin says at 6:35 in the presentation:

            “Models are not right or wrong. They are always wrong, they are approximations..”

            and then goes on to discuss model “skill”

            In the Realclimate FAQ that John Mashey linked to upthread. Gavin asks the following questions

            What is being done to address the considerable uncertainty associated with cloud and aerosol forcings?

            and

            Are clouds included in models? How are they parameterised?
            Models do indeed include clouds, and do allow changes in clouds as a response to forcings. There are certainly questions about how realistic those clouds are and whether they have the right sensitivity

            From my limited understanding of the shortcomings of climate models, aerosols and clouds are some of the biggest uncertainties.

            This is not to say that models are wrong or a waste of time

            However, the uncertainties expressed in the FAQ weren’t really expressed in the TED talk, perhaps because TED is aimed at a non-technical audience

            1. This is not to say that models are wrong or a waste of time

              …but I’m going to carry on as if that were the case.

              Only a fool would assume that models that have performed well in hindcasting can be dismissed on the basis of their political implications – and the gotcha! revelation that the future hasn’t happened yet (well, duh!) – when forecasting.

              Which is where you come in…

            2. I notice that you, based on your comments at Slater’s blog, seem to have no qualms about the modelling involved in the satellite-based temperature time series.

            3. Rob, I am unaware of the modelling issues in RSS, as I pointed out in my reply to you at WO

              My only real comment about models was the “it’s models all the way” which was a reference to the Potsdam paper.

              The models may or may not be correct, but they are still, nonetheless, just models, and should be treated with the same caution as any other model that we use to approximate a real world entity

            4. I would have thought that historical data is used to calibrate the models, so hindcasting seems to be just validating that they fit the historical data.

              However, it seems that hindcasting doesn’t work that well across the board.

              e.g

              The Atlantic Multidecadal Oscillation (AMO) index is predicted in most of the models with significant skill, while the Pacific Decadal Oscillation (PDO) index shows relatively low predictive skill. The multi-model ensemble has in general better-forecast quality than the single-model systems for global mean surface temperature, AMO and PDO.

              http://judithcurry.com/2012/05/17/cmip5-decadal-hindcasts/

  8. My fault, I realise it is not relevant to this thread. I was not having much luck myself so just thought I would ask here. Will keep it on topic in future, no problem.

    [Snipped. You were warned. GR]

    1. If you look at the NASA sea level graph there was a big dip in 2010/11 when it rained like hell in Pakistan and Australia and it took at least a year for the water to get to the sea.Like the temperature it is not an even progression.

    1. A quick insert of ‘meaning that each jump in the scale gets larger (or longer) by a factor of ten’ wouldn’t go astray, I agree. Though then people often don’t grasp what a vast transition just a handful of such jumps can represent, but that is, at least, explained anecdotally in the presentation.

  9. On this sea level rise issue Judith Curry claims sea level rise has decreased from 3.5mm to 2.5mm per year. Probably true if you look at just the last 5 years, as there was a big drop in sea level rise some years ago due to heavy rains in South America, but since then rates have jumed back up, but an average for the period could be about 2.5mm.

    So what does all this mean? Typical sceptic picking a short period that is coloured by natural variability and too short to conclude anything. They do it with temperatures and arctic ice, so why not try the same things with sea level rise.

  10. The big issue with sea level rise is that the models are not good at predicting sudden collapse. They can work out the melt rate of a glacier, just as anyone can with a big block of ice, but Greenland is riddled with holes like a rotten cheese and when the holes join up it collapses. The scientists who work there believe that the situation is dire and much worse than the IPCC can prove but predicting it is rather difficult. A lot hangs on the answer as millions of people live within one metre of sea level. We can offer a new home to the people of Tokyo, Shanghai, Bangladesh, Florida and of course Kiribati.

    1. A sobering consideration:

      a mere 1% loss of ice from these three sources (Greenland, East Antarctic, West Antarctic ice sheets) would produce a likely increase in sea levels of around 83cm – from these ice formations alone….

      The best fitting trend finds that Greenland ice loss is accelerating at a rate of 30 Gigatonnes/yr2. Greenland’s mass loss doubled over the 9 8 year period. (2002-2009)

      http://www.skepticalscience.com/stable-greenland-ice-sheet-basic.htm

      and

      http://www.skepticalscience.com/stable-greenland-ice-sheet-intermediate.htm

Leave a Reply