Early in my career I trained as a systems analyst. My PhD was about the ability to identify and make use of multiple perspectives on a system when understanding people’s needs, and designing new information systems to meet them. I became a “systems thinker”, although I didn’t encounter the term until later.

I also didn’t really appreciate until recently how much systems thinking changes everything about how you perceive the world. Perhaps the best analogy is the scene in The Matrix, when Morpheus offers Neo the choice of the red pill or the blue pill. One of these choices will allow him to step outside of the system and see it in a new way. Once he has done that he can never go back to seeing the world the way he used to (although there’s an interesting subplot in the movie where one of the characters tries to do exactly that).

When I think about climate change, I approach it as a systems thinker. I look for parts of the problem that I can characterize as a system: where are the inputs and outputs, boundaries and control mechanisms, positive and negative feedbacks, interactions with other systems? I want to build systems dynamics models that capture a system as a set of stocks and flows, and explore how cycles and delays affect the overall behaviour of the system. And of course, I’m always looking out for emergent properties: things that arise as a result of interactions across a system but that cannot be studied through reductionism.

It’s not surprising then, that I’m fascinated by Earth System Models (ESMs). These capture some of the most complex systems interactions ever described in a computational model – on a planetary scale! ESMs can be used to explore how processes at small scales give rise to emergent properties on a global scale. They provide a test-bed for what-if questions, to explore whether our understanding of the physical systems makes sense. And fundamentally, they’re used to probe questions of stability of the system: the relationship between the size of a “forcing” (which tends to push the system out of equilibrium) and the size of its “effect” (e.g. how sensitive is the global average temperature to a doubling of CO2?). To connect the two, you have to explore the positive and negative feedbacks that amplify (or dampen) the effects. And of course, we’d like to understand the nature of tipping points, thresholds beyond which positive feedbacks can push the system towards entirely different equilibrium points.

People who don’t understand climate change tend to lack a grasp of how complex systems work, and that’s unfortunate because for any system of sufficient complexity, most of its behaviour is counter-intuitive. People ask how a gas that forms such a tiny fraction of the atmosphere can have such a large effect, because they don’t understand that the earth constantly receives and emits huge amounts of energy into space, and that it only takes a tiny imbalance between the input and output to disrupt the planet’s equilibrium. People assume the climate system will always tend to revert to the stable pattern it has exhibited in the past, because they don’t understand positive feedbacks and exponential change. People assume we can wait to fix the climate system once we’ve seen how bad it might get, because they don’t understand the ideas of inertia and overshoot when a system has a delayed response to a stimulus. And people wonder how we can predict anything at all about climate dynamics, because they confuse chaos with randomness.

Climate science (and especially climate modeling) is inherently a systems discipline. However, climate scientists tend to hail from the physical sciences, and hence sometimes seem to miss an important aspect of systems analysis. In the physical sciences, you learn how to observe and experiment with physical systems in order to understand and explain them. But you’re not trained to re-design them to work better – that’s generally left to the engineers. Unfortunately, most engineering disciplines don’t cover systems thinking either. They concern themselves with the properties of families of devices (e.g. electrical circuits), and how such devices can be applied to solve problems. Engineers are not usually trained to re-conceptualize systems in entirely new ways, to understand how they can be changed. (Systems Engineering would be the exception here, but it’s a very young discipline).

So systems thinkers are quite rare, both across the physical sciences and the engineering disciplines. You actually encounter more of them in the social sciences, because social systems tend to defy attempts at understanding them through reductionism, and because social scientists are often more comfortable with constructivism: the idea that the systems we describe as existing in the world are really only mental constructs, arrived at through social processes. My favourite definition of a system, from Gerald Weinberg is “a way of looking at the world”. In a sense, systems aren’t “out there” in the world, waiting to be studied. Systems are a convenient mental tool for making sense of how things in the world interact with one another. This means there’s no such thing as the “climate system”, just lots of interacting thermodynamic and chemical processes. That we choose to call it a ‘system’, name its parts, and treat it as a whole, is a convenience. But it’s a very useful one, because it offers rich insights for understanding, for example, how human activities alter the climate. Modelling the climate as a system means that we have to decide which clusters of things in the world to include in the models, and where we might usefully draw system boundaries. And if we’re doing this right, we ought to acknowledge that there are other ways of viewing these systems – no decision about where to draw system boundaries can ever be ‘correct’, but some decisions lead to more insights than others (compare with Box’s famous saying about models: “All models are wrong, but some are useful”).

While traditional branches of science offer tools and methods for understanding each of the pieces of the climate system, the study of the climate system as a whole requires a different approach. It is a trans-disciplinary field, because the interactions that matter include physical, chemical, biological, geographic, social, and economic processes. It goes beyond traditional methodologies of the physical sciences because it is anti-reductionist: it must grapple with understanding holistic properties of systems, even when the detailed behaviour of those systems is not sufficiently understood. In other words it’s a systems science, and climate modellers have to be systems thinkers.

All this leads me to argue that we should incorporate more of the key ideas from systems thinking into discussions about climate change and sustainability. I think that a better understanding of systems dynamics would help a lot in giving people the right intuitions about climate change. And I think a better understanding of critical systems approaches would give people a better understanding of how to improve collective decision-making around climate policy.

Note: This is the first of a series of posts exploring the systems dynamics of climate change. Here’s the rest of the series, so far:

The Climate as a System, part 1: the central equilibrium loop
The Internet: Saving Civilization or Trashing the Planet?

4 Comments

  1. I agree with you, but then I would. I’ve been a very large systems designer and integrator for forty years. What would your suggestions be for introducing systems dynamics and systems approaches for people who’ve never gotten as far as calculus? Those are the people I struggle most to reach about how complex systems work. The paper you note on systems dynamics is not something I can give to non-technical people, and the second is too vague. IT’s an important topic – how do we make it more accessible?

  2. Hi, Steve! Thanks for the bait.

    Hmm, about engineering, I’d put it this way.

    Engineers (EEs & chem Es especially, which is most of us) do commonly think in terms of whole systems, but most of their skill comes with small-signal perturbations because most of what they do relates to designed systems. Even if the system is not linear, it is designed to be predictable. And an engineered system which is nonlinear in complex ways is an engineered system which is broken, even if the nonlinearities are well-understood.

    There’s a lot to be learned from engineers, but there is no general algorithm to study large signal behavior of complex systems mathematically other than by computational experiment.

    But I believe ESMs are underconstrained and we need to understand the relationship between climate models and the real climate much better before we put too much confidence in ESM dynamics.

    Weather models simulate; earth systems models (and economic IAMs) model more or less by extrapolation. Climate models occupy an interesting intermediate ground but they can be well worth all those cycles. I don’t know that the computationally intensive ESMs can say the same; I am not convinced they are worth their salt.

  3. Pingback: The Climate as a System, part 3: greenhouse gases | Serendipity

  4. Pingback: The Climate as a System, part 1: the central equilibrium loop | Serendipity

  5. Pingback: The Climate as a System, part 2: energy consumption | Serendipity

  6. Pingback: The Climate as a System, part 4: earth system feedbacks | Serendipity

  7. Had a conversation with my (aerospace engineer) brother that supports what MT says: standards like http://en.wikipedia.org/wiki/ARP4754 are used for achieving a robust systems-wide link between every single component and the whole system; he described it as a v-shape process down to each component and back up again, linking across system levels too. But like MT says, that’s all about ironing out unpredictability completely.

  8. Dan: I used to teach the V-model in my software engineering classes. It’s great for reminding engineers that just because each of the components has been tested, that doesn’t mean they’ll still do the right thing when you integrate them into assemblies of components. So integration and system level testing check that end-to-end system behaviours work as expected. Unfortunately, the V-model is tied to the idea of specifying the system properties in advance, and then testing they do hold in the integrated system. It’s not good at detecting emergent (unexpected) dynamics, except by accident.

  9. Pingback: The Climate as a System, part 5: clouds | Serendipity

Join the discussion: