Robert Brown: What we don’t know about energy flow

I saw this comment on WUWT and was so impressed by it that I’m making a separate post of it here. Dr Brown (who is a physicist at Duke University) quotes another commenter and then gives us all an erudite lesson. If Nikolov and Zeller feel they need to take any of the complaints on WUWT about the way  they handle heat distribution from day to night side Earth seriously, they probably need to study this post carefully. this is also highly relevant to the reasons why Hans Jelbring used a simplified model for his paper, please see the new PREFACE added to his post for further elucidation.

———————————————————————————-

I can’t speak for your program, but I will stand by mine for correctly computing the ‘mean effective radiative temperature’ of a massless gray body as a perfect radiator. Remember, there is no real temperature in such of an example for there is no mass. It takes mass to even define temperature. (but most climate scientist have no problem with it and therefore they are all wrong, sorry)

I’d like to chime in and support this statement, without necessarily endorsing the results of the computation (since I’d have to look at code and results directly to do that:-). Let’s just think about scaling for a moment. There are several equations involved here:

P = (4\pi R^2)\epsilon\sigma T^4

is the total power radiated from a sphere of radius R at uniform temperature T. \sigma is the Stefan-Boltzmann constant and can be ignored for the moment in a scaling discussion. \epsilon describes the emissivity of the body and is a constant of order unity (unity for a black body, less for a “grey” body, more generally still a function of wavelength and not a constant at all). Again, for scaling we will ignore \epsilon.

Now let’s assume that the temperature is not uniform. To make life simple, we will model a non-uniform temperature as a sphere with a uniform “hot side” at temperature T + dT and a “cold side” at uniform temperature T – dT. Half of the sphere will be hot, half cold. The spatial mean temperature, note well, is still T. Then:

P’ = (4\pi R^2)\epsilon\sigma ( 0.5*(T + dT)^4 + 0.5(T – dT)^4)

is the power radiated away now. We only care how this scales, so we: a) Do a binomial expansion of P’ to second order (the first order terms in dT cancel); and b) form the ratio P’/P to get:

P’/P = 1 + 6 (dT/T)^2

This lets us make one observation and perform an estimate. The observation is that P’ is strictly larger than P — a non-uniform distribution of temperature on the sphere radiates energy away strictly faster than it is radiated away by a uniform sphere of the same radius with the same mean temperature. This is perfectly understandable — the fourth power of the hot side goes up much faster than the fourth power of the cold side goes down, never even mind that the cold side temperature is bounded from below at T_c = 0.

The estimate: dT/T \approx 0.03 for the Earth. This isn’t too important — it is an order of magnitude estimate, with T \approx 300K and dT \approx 10K. (0.03^2 = 0.0009 \approx 0.001 so that 6(0.03)^2 \approx 0.006. Of course, if you use latitude instead of day/night side stratification for dT, it is much larger. Really, one should use both and integrate the real temperature distribution (snapshot) — or work even harder — but we’re just trying to get a feel for how things vary here, not produce a credible quantitative computation.

For the Earth to be in equilibrium, S/4 must equal P’ — as much heat as is incident must be radiated away. I’m not concerned with the model, only with the magnitude of the scaling ratio — 1375 * 0.006 = 8.25 W/m^2, divided by four suggests that the fact that the temperature of the earth is not uniform increases the rate at which heat is lost (overall) by roughly 2 W/m^2. This is not a negligible amount in this game. It is even less negligible when one considers the difference not between mean daytime and mean nighttime temperatures but between equatorial and polar latitudes! There dT is more like 0.2, and the effect is far more pronounced!

The point is that as temperatures increase, the rate at which the Earth loses heat goes strictly up, all things being equal. Hot bodies lose heat (to radiation) much faster than cold bodies due to Stefan-Boltzmann’s T^4 straight up; then anything that increases the inhomogeneity of the temperature distribution around the (increased) mean tends to increase it further still. Note well that the former scales like:

P’/P = 1 + 4 dT/T + …

straight up! (This assumes T’ = T + dT, with dT << T the warming.) At the high end of the IPCC doom scale, a temperature increase of 5.6C is 5.6/280 \approx 0.02. That increases the rate of Stefan-Boltzmann radiative power loss by a factor of 0.08 or nearly 10%. I would argue that this is absurd — there is basically no way in hell doubling CO_2 (to a concentration that is still < 0.1%) is going to alter the radiative energy balance of the Earth by 10%.

The beauty of considering P’/P in all of these discussions is that it loses all of the annoying (and often unknown!) factors such as \epsilon. All that they require is that \epsilon itself not vary in first order, faster than the relevant term in the scaling relation. They also give one a number of “sanity checks”. The sanity checks suggest that one simply cannot assume that the Earth is a ball at some uniform temperature without making important errors, They also suggest that changes of more than 1-2C around some geological-time mean temperature are nearly absurdly unlikely, given the fundamental T^4 in the Stefan-Boltzmann equation. Basically, given T = 288, every 1K increase in T corresponds to a 1.4% increase in total radiated power. If one wants a “smoking gun” to explain global temperature variation, it needs to be smoking at a level where net power is modulated at the same scale as the temperature in degrees Kelvin.

Are there candidates for this sort of a gun? Sure. Albedo, for one. 1% changes in (absolute) albedo can modulate temperature by roughly 1K. An even better one is modulation of temperature distribution. If we learn anything from the decadal oscillations, it is that altering the way temperature is distributed on the surface of the planet has a profound and sometimes immediate effect on the net heating or cooling. This is especially true at the top of the troposphere. Alteration of greenhouse gas concentrations — especially water — have the right order of magnitude. Oceanic trapping and release and redistribution of heat is important — Europe isn’t cold not just because of CO_2 but because the Gulf Stream transports equatorial heat to warm it up! Interrupt the “global conveyor belt” and watch Europe freeze (and then North Asia freeze, and then North America freeze, and then…).

But best of all is a complex, nonlinear mix of all of the above! Albedo, global circulation (convection), Oceanic transport of heat, atmospheric water content, all change the way temperature is distributed (and hence lost to radiation) and all contribute, I’m quite certain, in nontrivial ways to the average global temperature. When heat is concentrated in the tropics, T_h is higher (and T_c is lower) compared to T and the world cools faster. When heat is distributed (convected) to the poles, T_h is closer to T_c and the world cools overall more slowly, closer to a baseline blackbody. When daytime temperatures are much higher than nighttime tempratures, the world cools relatively quickly; when they are more the same it is closer to baseline black/grey body. When dayside albedo is high less power is absorbed in the first place, and net cooling occurs; when nightside albedo is high there is less night cooling, less temperature differential, and so on.

The point is that this is a complex problem, not a simple one. When anyone claims that it is simple, they are probably trying to sell you something. It isn’t a simple physics problem, and it is nearly certain that we don’t yet know how all of the physics is laid out. The really annoying thing about the entire climate debate is the presumption by everyone that the science is settled. It is not. It is not even close to being settled. We will still be learning important things about the climate a decade from now. Until all of the physics is known, and there are no more watt/m^2 scale surprises, we won’t be able to build an accurate model, and until we can build an accurate model on a geological time scale, we won’t be able to answer the one simple question that must be answered before we can even estimate AGW:

What is the temperature that it would be outside right now, if CO_2 were still at its pre-industrial level?

I don’t think we can begin to answer this question based on what we know right now. We can’t explain why the MWP happened (without CO_2 modulation). We can’t explain why the LIA happened (without CO_2 modulation). We can’t explain all of the other significant climate changes all the way back to the Holocene Optimum (much warmer than today) or the Younger Dryas (much colder than today) even in just the Holocene. We can’t explain why there are ice ages 90,000 years out of every 100,000, why it was much warmer 15 million years ago, why geological time hot and cold periods come along and last for millions to hundreds of millions of years. We don’t know when the Holocene will end, or why it will end when it ends, or how long it will take to go from warm to cold conditions. We are pretty sure the Sun has a lot to do with all of this but we don’t know how, or whether or not it involves more than just the Sun. We cannot predict solar state decades in advance, let alone centuries, and don’t do that well predicting it on a timescale of merely years in advance. We cannot predict when or how strong the decadal oscillations will occur. We don’t know when continental drift will alter e.g. oceanic or atmospheric circulation patterns “enough” for new modes to emerge (modes which could lead to abrupt and violent changes in climate all over the world).

Finally, we don’t know how to build a faithful global climate model, in part because we need answers to many of these questions before we can do so! Until we can, we’re just building nonlinear function fitters that do OK at interpolation, and are lousy at extrapolation.

rgb