Is the climate computable?

Our WUWT thread on Antarctic Sea Ice Losses has spurred quite an interesting discussion. Dr. Robert G. Brown of the Physics Department at Duke University responds to a comment on  ice albedo with a summary of water vapor action, the greenhouse effect, and the chaotic nature of the atmosphere. He ends with his view of why he’s not a betting man.

Well worth a read.

rgbatduke says:

May 21, 2014 at 5:49 am

phlogiston: I do realise that over the Antarctic land mass albedo from surface snow is anomalously higher than that from cloud, since the snow presents such a pure white surface. However this is probably not the case for sea ice whose surface is more irregular and cracked with patches of dark sea in between.

The trouble is that water vapor is literally a two-edged sword. As vapor, it is the strongest greenhouse gas in the atmosphere by (IIRC) around an order of magnitude, so increasing water vapor can and does measurably increase the GHE — a lot, when considering dry air versus saturated air. In arid deserts, temperatures skyrocket during the day and plummet at night because of the absence of a water vapor driven GHE — CO_2 alone isn’t nearly enough to keep upward facing surfaces from rapidly losing their heat due to radiation. In very humid tropical climates, the nights are consistently warm because of the GHE.

However, water vapor is also the mediating agent for two major cooling mechanisms. One is the bulk transport of latent heat — sunlight and LWIR hit the sea surface and cause rapid evaporation of surface molecules of water. Wind blows over the ocean surface, stripping off water molecules as it goes. This evaporated water has a huge heat content relative to liquid water — the latent heat of vaporization. As the warm water vapor is carried aloft by convection, it carries the heat along with it. It also cools as it rides the adiabatic lapse rate upward, and further cools by radiating its heat content away (some of which returns to the Earth as GHE back radiation). Eventually the partial pressure of water vapor in the moist air becomes saturated relative to the temperature and the dew point is reached, making it comparatively probable that the water vapor will recondense into water. In order to do so, though, several things have to be “just right”. The water vapor has to be able to lose the latent heat of vaporization that it picked up at the water surface when it evaporated. The future water droplets have to be able to nucleate — which is a lot more likely to occur when there are ionic aerosols in the atmosphere as water (a polar molecule) is attracted to bare charge of either sign.

Once a water droplet is nucleated and grows past a critical size (that depends weakly on humidity and temperature) its surface becomes large enough that growth due to increased surface deposition outweighs loss due to surface evaporation, and the droplet stabilizes as a single droplet of condensation in a cloud or continues to grow to fall as rain. Either way the water, now high in the troposphere and hence through most of the optically opaque greenhouse layer, releases heat that is “short circuited” through the greenhouse mechanism and lost to space via radiation.

The cloud, as you note, has a very high albedo. High albedo means that it strongly reflects short-wave (e.g. visible) radiation without ever absorbing it and being heated by it. During the day, clouds outside of the polar regions act as a cooling agent, reflecting sunlight before it has a chance to reach the ground and lower troposphere to warm either one. During the day and the night, however, the cloud also acts as a powerful greenhouse blanket, directly reflecting LWIR as well as visible back down towards the Earth’s surface. In the tropics, daytime reflection wins by a landslide — reducing the incident sunlight by a huge fraction for a large fraction of the day beats the comparatively small modulation of surface radiative losses both day and night. In the temperate zone (again, IIRC) albedo still wins, but by a smaller and smaller margin as one creeps north (and in ways that are increasingly dependent on seasonal weather patterns — in the winter clouds can easily be net warming where in the summer they can be net cooling).

However — and this is key and the reason I’m replying to you — in the polar regions clouds are generally net warming, at least most of the year. You’ve already indicated some of the reasons — the polar regions are already often or permanently ice covered, and the gain in daytime albedo from clouds vs ice is not so great. The real problem, however, is that nighttime warming from the enhanced GHE from clouds scales with the fraction of the day that it is nighttime, and of course inside the arctic circles that can be as long as 100% of it. High albedo doesn’t cool when there is no incident sunlight to reflect, and even in the arctic summer, the sun comes in at a substantial angle so that direct solar warming is weak (so that clouds can reflect only a proportionally smaller amount of heat). A lot of polar temperature is determined by heat transport, not direct heating, explaining the substantial difference in mean temperatures of the North and South poles. In the north, there is substantial heat transport and heat exchange via the ocean; in central Antarctica there is only the atmosphere to carry heat in from the warmer latitudes and it just can’t do the job half as well.

That’s why I hesitated to assign a sign to the net feedback from any sort of local modulation of e.g. ocean-air humidity or sea ice coverage. The processes are COMPLEX and can have either sign, and they are NON-LOCAL as adding humidity in one place can increase albedo someplace else thousands of miles away is it finally concentrates enough to form clouds. A large part of the rain that falls over North Carolina comes up from the Gulf of Mexico maybe 1000 miles away. Some of it comes all the way over from the Pacific, where some of that might have originated in e.g. the growing El Nino. Heat from the tropical Pacific can be transported all the way to NC before it finally releases its heat and falls as rain, before it finally creates clouds that cause NC to cool after helping to greenhouse warm much of the surface area it crossed in between.

This is the kind of thing that the models are supposedly trying to model, but they perforce replace all of the small-length scale detail of this description with presumptive averages over cells 100-300 km square (where weather phenomena such as thunderstorms are order of 1 to 10 km square, where the details of front structure and development are much finer than this). They are excruciatingly tuned to aerosol levels and albedo — they have to be to stabilize anywhere near the correct/observed temperatures and preserve the central tenet that CO_2 causes X amount of baseline warming that is on average augmented by additional water vapor.

This last assumption is finally dying a quiet and well deserved death. AFAIK, it is due to Hansen, who in his original papers predicting disaster assumed universally positive water vapor feedback (and for no particularly scientifically motivated reason that I can see, hypothesized truly absurd levels of water vapor feedback that doubled or tripled the CO_2-only warming of his then very simple models). Naturally, some of the GCMs out there have built into them parametric assumptions that preserve this much “climate sensitivity” — total ACO_2 warming plus feedback, usually at the expense of an overdriven response to e.g. volcanic aerosols necessary to explain periods of global cooling and to keep the model from having a runaway exponential instability (because one has to have a mechanism that keeps positive feedback water vapor from causing increase of water vapor without bound just from FLUCTUATIONS in water vapor content or global temperature — the climate cannot be a biased random walk where every time the temperature goes up a bit, average water vapor increases and hence resets the Earth’s average temperature a bit higher unless a competing process can completely erase the gain when the temperature fluctuates down a bit).

At the moment, estimates of climate sensitivity are struggling to retain any net positive feedback from water vapor in the face of data that already solidly excludes the kind of absurd feedback levels Hansen originally hypothesized. Even the question of net negative feedback from water vapor, long considered to be anathema in climate science (except for a few mavericks who managed to publish papers suggesting that clouds could easily lead to net negative feedback through the dual mechanism of latent heat transport and modulation of albedo) is no longer completely off of the table. I don’t know that people will start to take it too seriously unless/until the Earth actually cools (several tenths of a degree, sustained, not just vary up or down or weakly downward trend) but obviously if this happened it would truly be the only likely catastrophe associated with global warming to all of those that have invested their professional careers, hundreds of billions of dollars of global wealth, and their political and/or scientific reputation on shaky claims in poor agreement (so far) with observational data.

IF there is a super-ENSO, perhaps it will help their arguments survive a bit longer, or perhaps it will truly kick up the temperature to where the models become believable again. Perhaps not. ENSO is not the only factor in climate evolution, and while it has been dominant for the last half century or so in mediating positive jumps as documented by Bob Tisdale, its ability to do so could easily be predicated by the phases and states of the other decadal oscillations, the state of the Sun, the state of baseline vulcanism, the immediate past climate history, and the price of tea in China. A chaotic nonlinear system can be quasiperiodic and apparently causal for a while and then for no computable reason change to an entirely different mode of behavior where a significant quasiparticle/process becomes insignificant and some other process becomes the critical driver. We could still watch as the developing ENSO discharges all that heat in such a way that it never manages to raise global average temperatures by much because of some confounding wave that causes the heat to be efficiently transported up and quickly lost rather than persisting to spread out over the globe at high altitude, or by a mere modulation of the winds that causes albedo over the warm(ing) patch to be higher than expected so that the delivery of solar energy to the ocean is effectively interrupted. It’s not like we can properly predict ENSO (although we can do pretty well with forward projective hindsight once an ENSO process has started).

No matter what, I expect the next year to be highly informative. If we have a super El Nino that heats the planet by 0.3C very rapidly, that certainly makes GCMs more, not less, plausible on average as it kicks global average temperatures at least in the right direction for them not to be as egregiously wrong as they currently appear to be. If it only kicks the temperature up by 0 to 0.1 C, and that only transiently so that temperature in a year are again pretty much flat relative to 1998-2000, it is very bad news for the models. If it fizzles altogether — short-circuited, perhaps, by the downhill side of solar cycle 24 that maybe be beginning and which will proceed with poorly predictable speed and which may or may not have a competitive local effect on the climate and produces no gain at all and cycles immediately into a cooling La Nina that augments any solar cycle cooling to actually drop global average temperatures, that too will be very informative.

Personally, I won’t even place a bet. I don’t think the climate is computable, which means that I think one is basically betting on the output of a (possibly biased) random number generator. I’d rather play Mumbledy-peg for money.