by Judith Curry
Isaac Held’s new article in Science raises some interesting questions.
Isaac Held in Science
Isaac Held has published a Perspective in Science entitled Simplicity amid complexity. Excerpts:
In addition to these challenges, the turbulent and chaotic atmospheric and oceanic flows seemingly limit predictability on various time scales. Is the climate system just too complex for useful prediction?
More fundamentally, an emphasis on complexity in the climate system must be balanced by recognition of emergent simplicity. The seasonal cycle provides a useful counterpoint. An individual year’s temperature record is a consequence of chaotic weather superposed on a relatively simple and smooth underlying cycle. Watching temperatures change during a few weeks in the spring does not affect confidence that typical summer temperatures will eventually emerge. Climate models paint an analogous picture for the evolution of climate over decades to centuries: a superposition of internal variability and an externally forced component containing a natural part (solar variations and volcanoes) and a part due to human activities that, to first approximation, is a linear superposition of responses to different forcing agents such as CO2, methane, and aerosols.
Thus, the attribution problem—the separation of the forced change from internal variability in observations and the partitioning of this forced component into parts due to different agents such as the well-mixed greenhouse gases or aerosols—is a tough challenge. But attribution leads directly to prediction of the forced response if the case can be made that the response is simple enough for past trends to strongly constrain future evolution.
Any skill in predicting multidecadal internal variability on top of this forced response is icing on the cake.
A creative tension between simulation and understanding, between accepting complexity and searching for simplicity, is present in many challenging scientific problems. Climate science provides an excellent example of this tension. The most advanced comprehensive climate models effectively represent the current ability to simulate the climate system, and it is natural and appropriate to take the output of those models as the basis for predicting the future climate. However, it is the understanding of these responses—an understanding that depends on the presence of an emergent simplicity in the forced response—which provides a level of confidence that justifies advising policy-makers and the public to pay heed to these predictions.
Held’s APS comments
For further insights into Held’s argument, see the transcript from the recent APS Workshop, excerpts starting on p 324:
I don’t like this argument from complexity saying oh, it’s a chaotic system. There is all sorts – you can get a nonlinear system to do anything you want. That just doesn’t tell me anything.
But everything that I looked at on the climate, I look at the forced response of the climate system. It looks linear to me. And what is the best example we have of forced responses? The seasonal cycle. Seasonal cycles are remarkably linear-looking.
There is an awful lot of nonlinear fluid dynamics and cloud formation stuff going on underneath this. My analogy here is the thermodynamics limit, statistical mechanics.
The smaller response, you seem to worry about the fact that the external forcing is so small, but that just makes it more likely to be linear.
The whole language, the whole forcing-feedback language we look at is assuming that this linear picture is useful. Otherwise, what is forcing and what is feedback? I don’t even know where to start.
The models look pretty linear. That observed seasonal cycle that looks linear. Even if in the Ice Age times, things look pretty linear. We don’t know that much about it.
So, why should I assume that things are, gee, the anthropogenic CO2 pulse is going to interact in some exotic way with internal modes of variability? Well, it’s conceivable. I am not convinced. I don’t think that is particularly relevant.
JC reflections
Held’s article raises a very important issue – whether climate change is predominantly linear and dominated by external forcing, or whether natural internal variability is the intrinsic mode of variability on decadal to century timescales. In other words, is natural internal variability the icing on the cake, or the cake itself?
While I like Held’s article in the sense that I find it to be provocative, I disagree with much of it.
First, I don’t think the seasonal cycle works very well as an argument for external forcing. To be used in an analogy for CO2 forcing, you should consider the globally averaged seasonal cycle, which is not very large (NH winter season is globally slightly warmer than SH winter season over differences in distance from the sun), and a key issue in this is the different amount of land vs ocean in the different hemispheres. If you use a local example (Held used Minnesota in his APS talk), it is very interesting to compare the annual cycle response to the diurnal cycle response. I don’t think the temperature change per change in W m-2 scales linearly between these two timescales in the mid-latitudes (but I haven’t done the calculation).
Second, it is not at all clear to me that natural internal variability and forced variability are easily separable in a linear way.
Third, a truly complex system cannot be understood as a linear superposition of individual elements (discussed in my Uncertainty Monster paper).
Fourth, the climate response to the relatively small greenhouse forcing may well be linear, but this linear response may be swamped by the natural internal variability.
And finally, I find this statement to be particularly telling:
The whole language, the whole forcing-feedback language we look at is assuming that this linear picture is useful. Otherwise, what is forcing and what is feedback? I don’t even know where to start.
The linear model of forcing and feedback I think was useful at zeroth order, in context of conceptually thinking about the problem in early days. I think this concept is also useful in comparing the climate of different planets. But I am increasingly of the opinion that this concept is not all that useful in the interpretation of climate variations on decade to century timescales, and maybe even millennial – the timescales where ocean circulations are a dominant player. The debate around the sensitivity to CO2 forcing is a symptom of this problem.
So which vision is correct – the linear model whereby climate variations are forced externally (with noise from internal variability), or the complexity model (e.g. climate shifts) whereby natural internal variability is the intrinsic signal, with external forcing projecting onto the internal modes?
To me, this is the heart of the scientific debate on climate change, and why the hiatus (and how long it will last) is so important.