Thursday, June 26, 2008

Climate science after Kyoto

With the Official Science monkey gone from our backs, what can we do with climate and climate science?

Over the last year, I've outlined on this blog a set of open questions that are frequently ignored and often not even seen correctly as questions, but that need to be answered if we're going to talk sense about "climate" and "climate change." Not able to predict weather beyond about two weeks ahead, we need a simplified and abstracted definition of "climate" whose "state" can be defined, analyzed, and predicted with some confidence. The fallacies of temperature averaging and "climate parameterizations" were a failed attempt to do this. Whatever notions of "climate," "climate state," and "change of climate state" we end up with will have to withstand - as temperature averages and the "hockey stick" cannot - probing criticism and emerge free of the hefty list of fallacies associated with the "global warming" hysteria.

In retrospect, the generational dead-end of climate and allied sciences with "global warming" and general circulation models (GCMs) can be viewed as an attempt to force a premature integration of theory and observation. It would be best for all involved if theory and observation were to remain aware of one another, but go their separate ways until such time as they have sufficient means to meet each other honestly. Until such a synthesis is possible, it's best to maintain a pluralistic and agnostic attitude about the Big Picture, resisting the forces bent on our "salvation" from the "evils" of industrial civilization or plying us with supposed alternatives to "knowing" nature.

What follows is a personal and partial view, informed by this failure to get a grip on these issues and by my own scientific experience along the edges of the problem and in related fields. It should not be taken as definitive or complete.

Theory: It's a really hard problem. The Earth's climate is the most complex scientific problem ever posed and almost certainly unsolvable in its full generality. Any progress we make with this problem will therefore necessarily involve approximations. The essential point is that, if we're going to attempt actual predictions, we need better and controlled approximations at every step. These are currently lacking.

Theory: Science is hypothesis and deduction. That is, it's not just a piling up of facts. Certainty of conclusions requires control of assumptions and reasoning.

There is thus an important role for mathematical deductive modeling, with simplified and controlled approximations applied at every step. The ideal should be to make these mathematically simplified and controlled models closer and closer, with each step, to the real climate. The failure of the "parameterized" GCM approach underscores the need to keep the modeling within controlled approximations at each step and not jump into the deep end of the pool right away.

Theory: Don't BS - simplify and smooth. A revealing way to look at the "climate state" problem is to grasp the motive behind "climate parameterizations": it was to "force closure" on the dynamical-structural equations of climate. In general, there are never enough equations to match the number of unknown variables. "Forcing closure" on the system means guessing or making up extra equations to close the gap.

But the gap could equally well be closed the other way: reduce the number of variables. A simplified "climate state," less complex than "the exact, instantaneous state of the whole atmosphere," is just such a proposal. It's also likely that such a state will not only involve flows and topology in space, but time and space integrals of the basic variables (equivalent to what statisticians call cumulants). Such integrals are usually better-behaved than the original variables.

Theory: Boil, mist, and trouble. Climate is chaotic, in the technical sense: exponential sensitivity to errors in initial conditions. Alternatively, climate is essentially nonperiodic, and not all climate disturbances die away. The atmosphere is a fluid, in the physicist's sense; its chaos is turbulence. Turbulence is the largest unsolved problem in physics. A partial or complete solution would have immediate impact on many areas of science and engineering, pure and applied, theoretical and practical - everything from understanding convection in planetary atmospheres and stars to improving your airplane or boat ride to reducing turbulence losses in your car engine.*

Climate needs new and better techniques for coping with chaos. Many such techniques have been developed in the last 25 years in various areas of science, but they haven't penetrated far into the climate world, partly because of the paralysis induced by Official Science. They include exceptionally relevant techniques like the following.
  • Renormalization. This technique is a powerful generalization of the dimensional analysis we learned in school. (It's sometimes goes under the guise of "homology" or "rescaling.") It relates one mathematical problem posed at one set of space and time scales to a different problem at a different set of scales. Sometimes, impossible problems posed at one scales can be recast into other problems at different scales, and those different problems are solvable, either exactly or by controlled approximation.

    Renormalization for climate means imagining a scale at which decades, centuries, or even millennia seem modest and slow cycles like El Niño, say, wink by in rapid succession. On those scales, we can see more clearly the invariant and almost-invariant structure that must define, at a deeper level than everyday weather, what "climate" is.

  • Dynamical reconstruction of phase space. This requires some contact with observed climate (see below), but the essential technique amounts to isolating the relevant degrees of freedom in the very complex climate system, the ones that operate on scales of tens to thousands of miles. Only a small subset of the possible changes in the climate system are actually important. Isolating them is a big step toward defining "climate" in a simplified sense. It will undoubtedly involve flows of heat, water, etc. (not local temperatures or humidities) and how they're connected in space (their topology).

  • Non-Gaussian statistics, for extreme weather analysis. This is an application of the great progress that has been made in understanding how energy and other conserved physical quantities move through "open" systems, like the climate. Again, the issue straddles both theory and observation. People just have to stop assuming Gaussian (classical central limit or bell-curve) conditions in analyzing weather "events." There's never been any reason to do so.

  • Pattern formation. This is an intersection of renormalization, non-Gaussian statistics, and "complexity," as an earlier posting discussed. The locus classicus for these techniques is understanding the perpetually landsliding sand pile. (There's even a cute book on the subject by Per Bak.) In complex, open systems with "flow-through" of air, water, and heat (or sand grains for that matter), long-range patterns with "almost" (but never quite!) repetitive behavior form and dissipate over and over - just like the weather: cyclones, storms, fronts, and so on.

    Pattern formation is especially germane to understanding clouds - their nature and lifecycle - better. Clouds are the most important feature of climate not easily captured by simplified models; convective turbulence is actually secondary in importance, at least for heat flow, although it's still crucial for the complete picture. And the big, difficult pieces of climate - clouds, turbulence, water transformations - are all linked together. Convection doesn't just transport heat; it lifts water vapor to higher altitudes than it would otherwise go, making clouds form more often and last longer than they would otherwise.

    The presence of clouds in turn transforms the climate by changing how radiation flows into and out of the atmosphere and providing a greatly enhanced form of upward heat convection. The main source of IR-active gas in the clear air is not CO2 or CH4, but the feedback effect of enhanced, clear-air water vapor. But even limited condensation of the enhanced water vapor into clouds changes the radiation flow drastically.
Observation: Go forth and squint hard. In the end, a chaotic system is its own best computer: no model we devise or limited set of observations we make can ever capture every aspect of its behavior. But observation nonetheless remains essential for understanding climate. Modern scientific observation of the atmosphere, increasingly detailed since the 18th century, has a lot to tell us about the repertory of possibilities.

In understanding actual climate, we must always keep in mind the proviso that chaotic systems feature an unending stream of unique events. We also have to face repetitive trends that repeat on time scales longer than the modern scientific record captures. Climate is, in this sense, a unique problem, in that we're inside the system being studied, and we're myopic observers with only hints and partial clues about the long term. Although laboratory experiments are essential for isolating general physical laws, the actual conditions of climate do not constitute a laboratory experiment. It's not controlled, and we're not outside the system in a position to aspire to know and control everything about it.

Observation: The Sun will have its say. It always does. It's the ultimate factor in charge of Earth's climate. Like other stars, the Sun is variable, at a small but measurable level. What limited observations have been made of our Sun already strongly hint at important solar modulations of Earth climate. The more basic solar physics in control here, and how the Earth responds to solar changes, are still poorly understood, and the whole problem remains at the frontier of research. But the base of raw data needed is now available in a way not true 20 or 30 years ago. Studying other planets' response to the Sun's variability will help.

Observation: All things green and blue. Plants and oceans need to be understood better as well. Over scales of decades and longer, they play a critical role in absorbing and recycling carbon dioxide. Current climate models capture the ocean part only imperfectly and plants barely at all. Yet there's a 0.2/0.00038 = 530 ratio of diatomic oxygen (O2) to CO2 in the air, which large ratio is made entirely possible by plant metabolism.** The annual plant-driven variations in atmospheric CO2 concentrations are about eight percent of the total. Since 100/(8/year) ~ 12 years, every CO2 molecule in the atmosphere gets captured by a plant in a little more than a decade.

What's not understood is how plants are responding in their annual cycle of growth and decay to increasing CO2 in the atmosphere. With more "food," there will be more and bigger plants. How much is unknown, although the Ice Age results give a very rough idea. What little research here has been done so far has been strongly tainted by people out to "prove" that plants aren't important - even though they clearly are. It's an obvious place for Gaia-philes to speak up. One of the few geoengineering ideas with any merit involves humans enhancing an already old and thoroughly proven means for removing CO2 from the atmosphere: more plants, bigger plants, maybe even über-plants. More on that next.
* Indeed, the wild claims made by the IPCC for GCMs and "climate parametrizations" can be put into sharp relief when we consider that, were these results actually definitive answers for climate, we would also have a solution to fluid turbulence.

In fact, we don't. Over at the Clay Institute web site, you'll see there's a Clay Millennium prize for solving turbulence (Navier-Stokes equations) - and it remains unclaimed. Given the true state of affairs (turbulence remains an unsolved problem in physics and engineering), we can then rightly reason backward and conclude that the climate problem remains unsolved as well, since the turbulence problem is embedded within it.

** Without constant plant replenishment, the O2 would rapidly disappear from the atmosphere by oxidation weathering and water absorption. Animal metabolism would be impossible without plants, although plants can and, long ago, did do fine without us and our animal relatives.

Labels: , , , , , , , , ,


Post a Comment

Links to this post:

Create a Link

<< Home