Yes, that’s what I meant. The energy gates open and close randomly, but what about the amount of energy sat inside the gates?

Remember that compared to weather, which can vary 5-10C from day to day, the change represented by observed global warming (~0.8C/century) is comparatively small. The cumulative element need not be very obvious to give rise to it.

Oh, and in talking about energy balances, don’t forget about convection.

“I don’t really see this being a logical fallacy, rather just arithmetic.”

I can see what you mean. But wouldn’t you agree that the same argument could be applied to any of the positive terms, and the sum of the rest – including the CO2 term – likewise cancel out?

I stand corrected and embarrassed. You’ve pointed out a life-long misconception I’ve had about random walks, which was that they behaved like white noise. I never knew they gave 1/f spectra. After doing some reading, I see that while the expectation value is zero, the variance is potentially infinite. I’m going to be processing that for weeks to come and owe you thanks .

But my real discomfort was with the idea that heat or temperature randomly walk. They’re constrained by the energy budget. Rather, quantities like cloud cover, humidity, etc – those things which directly affect the radiative balance can randomly vary (and can be part of complex feedback cycles, and can be chaotic). The energy gates can open and close randomly.

“There’s not some huge well of heat thats been hidden away in some deep place for hundreds of years thats now emerging and making things warmer.”

I thought it followed from the previous assertion, that its difficult to bury energy for more than decades.

When the global temperature spikes during ENSO events, like 1998, where does the heat come from?
I confess a nearly total ignorance of meteorology, ocean currents and such, especially when it comes to El Niño, but I think that the hot water which is the predominant component of the phenomena is largely near the surface, no? I would think it was heated by sunlight, and relatively recently beforehand, based on our previous discussion. A large geographic increase as well as a temperature increase of equatorial water that unpredictably appears in one year could be explained by one of your random/chaotic drifts of cloud cover, prevailing winds, ocean currents, humidity or something else that would temporarily (over months) favor a net increase in solar heating of the ocean surface in that locale. That seems more likely than hot water somehow accumulating in the deep for some long period and suddenly rising from depths.

And even if, as you say, the temperature rise is due to an input/output change, why is it CO2; rather than clouds, evaporation, equator-to-pole circulation, land use changes, or absorption by black soot?

This question succinctly frames the issue from my point of view. Without more measurements, I can’t rule out any of the non-CO2 terms, known or unknown. But if heating response times are less than decades, and if we observe a net temperature increase lasting decades or longer, that tracks (very roughly) to what must occur from CO2 trapping alone, then all those other terms must cancel out with each other (again very roughly), even though they could be much larger in magnitude than the CO2 term. I don’t really see this being a logical fallacy, rather just arithmetic.

That’s a common sceptic position, actually – that CO2 will cause warming, but that the feedbacks that supposedly turn it into a disaster scenario are unproven and unlikely.

I can’t really argue with this. (That doesn’t mean that I don’t believe a disaster scenario is possible.) The ultimate test is to fully monitor the radiation budget, reflection and emission, which I believe is feasible. I thought it was underway already. Spectroscopic imaging of the globe from UV to say mid IR by satellites should not be a problem given today’s detectors. A couple of satellites in suitable polar orbits could cover day and night sides continuously. (As my boss used to often say to me “Why wasn’t it done already?”)

By the way, may I compliment you on the quality of the debate?

Likewise.

No, it’s not related to having Normal distributions, it’s related to having a decaying autocorrelation function. If you take a simple cumulative sum of zero-mean unit-variance Normal distributions (a standard random walk), then the distribution at time t is Normal with mean zero and variance t. Normal plus Normal gives Normal. It’s Normally distributed at every step, and has a 1/f spectrum.

That said, it is very likely that on a sufficiently long time scale the values do average out. But if it is thousands or millions of years, then from the point of view of our hundred year span what we see would be indistinguishable from unconstrained accumulation. In particular, any rises seen could easily be stochastic.

“There’s radiation in from the Sun, and thermal radiation out, and any net positive difference must accumulate in the Earth as heat,”

Don’t forget that as soon as the temperature rises, you radiate more. The heat only accumulates until the temperature reaches the equilibrium point again, and then heat flow in is once more equal to heat flow out. Since the surface can cool 10-20C overnight, we evidently don’t have a problem getting heat in and out of the system. It’s not immediately obvious why we cannot reach a new equilibrium very fast. The question is, where is that new equilibrium going to be?

“I don’t see how one can bury huge amounts of energy in the oceans for more than decades.”

Agreed. Although that does seem to be what some people are claiming. We’re half way to doubling CO2 over pre-industrial levels, so naively we ought to have achieved half of the warming. This doesn’t fit with the most dire of the predictions. So they have had to propose reasons for the result of the change to be delayed, of which ocean heat content is a prominent leader. However, I don’t see how the heat can get into the oceans fast enough to constitute a significant power sink. Yes, there’s ultimately lots of capacity there, but it doesn’t seem very accessible.

“There’s not some huge well of heat thats been hidden away in some deep place for hundreds of years thats now emerging and making things warmer.”

How do you know?

When the global temperature spikes during ENSO events, like 1998, where does the heat come from?

And even if, as you say, the temperature rise is due to an input/output change, why is it CO2; rather than clouds, evaporation, equator-to-pole circulation, land use changes, or absorption by black soot?

“there’s still an excess of CO2 in the trapping side of the budget that roughly accounts (within a factor of two) for the observed increase”

This is confirming the consequent. The CO2 may be being cancelled by some feedback mechanism, while whatever is causing the warming isn’t. But I won’t argue, because I think the idea is reasonable. 40% increase in CO2 alone would cause 0.5C surface temperature rise, which is roughly what we’ve seen, and if we saw another 0.5C over the next century, nobody would be worried.

That’s a common sceptic position, actually – that CO2 will cause warming, but that the feedbacks that supposedly turn it into a disaster scenario are unproven and unlikely.

By the way, may I compliment you on the quality of the debate? It’s one of the best I’ve had on here for quite a while.

Be careful here – random processes governed by normal distributions do smooth out, which is why things like averages work. Processes that don’t smooth out have certain spectral properties (like 1/f dependence).

The point I’m trying to make is that beyond a certain distance out in space from the Earth, the only energy transfer mechanisms are radiative. There’s radiation in from the Sun, and thermal radiation out, and any net positive difference must accumulate in the Earth as heat, regardless of the chaotic complexities of the air, oceans and land. Furthermore, that heat doesn’t stick around for very long, therefore any long term changes in global temperatures must be due to long term changes in the radiative energy balance.

I estimated the radiative cooling time for oceans by looking at the wattage out and the total heat capacity of the ocean, which is naive to be sure – completely ignoring thermal diffusion – thinking that convection would be more important anyway. You point out that surface water cools off even faster, and that just strengthens my point. Regarding diffusion, I’m looking at fig 1 in this paper showing diffusion times/depths which shows the heat being largely limited to shallower depths. I don’t see how one can bury huge amounts of energy in the oceans for more than decades.

I don’t see how one can get around this, that the global increase in the last 30 years boils down to a net increase in the radiation balance. There’s not some huge well of heat thats been hidden away in some deep place for hundreds of years thats now emerging and making things warmer. Its getting warmer because we’re getting more energy from the Sun than we’re radiating in the IR. Granted that there’s a complex system response time of months/years and that its chaotic. The budget change can also be do to variations in cloud cover or other albedo effects, variations in water vapor etc.. Even if these variations were somehow long term responses to stimuli that occured thousands of years ago, there’s still an excess of CO2 in the trapping side of the budget that roughly accounts (within a factor of two) for the observed increase, which means that all the other terms, uncertain though they may be, roughly cancel out.

Unfortunately, we don’t. There are quite a few data sources that have only come on line in the past five years, and many parts of the system that even today we can’t measure with much accuracy. The Earth is a big planet.

And we’re certainly a long way from understanding all the natural effects. Take a look at the IPCC’s “level of scientific understanding” numbers for things like clouds and aerosols. And it certainly isn’t clear from what we can measure that we can “account” for the energy budget. (See here and here for one example.)

“the radiative cooling times for the oceans is on the order of years per degree”

No, it isn’t.

This is the trouble with simplified scientific explanations. In trying to explain a really quite complicated and speculative hypothesis to the general public, they have inadvertently given the impression that it’s simple.

Here’s some sea surface temperature data.

Note the scale. The sea temperature can change 7-8C in six months, as anyone who has tried swimming in it in both January and August would know. That’s a rate of 15C/year.

The effect is more involved. Heat is conducted into a body in a complicated way. If it’s simple conduction (which of course for water it is not) the temperature change penetrates a depth proportional to the square root of time. So if the heat comes and goes, it can only penetrate a thin layer before it has to turn around again. The faster it varies the thinner the layer. Hence the heat capacity of the oceans is time dependent. For longer periods of sustained change, there is more heat capacity available, but remember that penetration depth is proportional to the square root of time, so this is an ever-decreasing amount being added as time passes, and the rate of heating that can be sustained is slow. It therefore can’t take much power, although it can take it for a long time.

Water works differently, because it convects, circulates, and is stirred up by weather, tides, and sea life. So it might not be a square root, and you can access more capacity more quickly. But you evidently can’t access it too fast, or the surface would not warm and cool so. And measurements by scientists have failed to find the expected warming.

“I’m actually familiar with the effect”

Excellent!

“Perhaps this is wrong, but it strikes me that by claiming that a process is random, we’re saying we don’t and can’t know the details, and so we assume its random.”

What we’re saying is that climate is an accumulation of weather, and that weather is chaotic, which often has many of the same properties as random. It’s a good question as to whether it has all the ones we need, but it’s the best guess we’ve got.

Accumulations don’t smooth out. Whether the weather does and on what time scale is, so far as I know, an open question. But some statisticians recently have done some analysis that seems to show that a cumulative process is indicated by the limited data we’ve got. (The Beenstock and Reingewertz paper on polynomial cointegration.) It’s too early to say whether their result will stand up to scrutiny, but I wouldn’t reject the possibility just yet.

To say that all the noise cancels out after some fairly short interval is commonly an article of faith. Take a look at the HadCET Central England temperature series, and tell me if you can figure out what that time scale is. How do you calculate it?

My problem with historical record arguments are that we don’t have all the energy budget data. There are too many unknowns. For example, it seems reasonable to attribute the large scale swings in temperature to contemporaneous changes in albedo, IR trapping vapor (humidity) or possibly even solar flux. Unfortunately we don’t have that historical data as well, so there’s no way to know.

We do however have much of this data this data for the last thirty years or so, and it seems reasonable to expect that all the terms can be accounted for during that period, including all the natural effects. Temperatures have been rising during that period, the energy (heat) increase in the oceans had to come from someplace, the solar flux has remained essentially constant during that time, so the cause seems to boil down to either a net decrease in the reflectance of sunlight during that time (clouds) or an increase in IR trapping. Have satellites measured a net albedo decrease during that time? Have water vapor or methane been increasing? If all these causes can be ruled out, CO2 would win by elimination.

Regarding the random runs: I’m actually familiar with the effect and agree with you that much of the public probably is not and often sees trends in things like random walks, etc. However, I’m very uncomfortable with applying it to long term, bulk climatic changes. Perhaps this is wrong, but it strikes me that by claiming that a process is random, we’re saying we don’t and can’t know the details, and so we assume its random. This has worked very well for statistical and quantum physics, for example. I can also see it being appropriate for short term weather (i.e the butterfly effect) and perhaps other global processes like oceanic circulation. But applying that kind of thinking to long term climate bothers me because it seems like the climatic timescales involved are much longer than bulk motion or energy transfer timescales (the radiative cooling times for the oceans is on the order of years per degree), which would mean that the random fluctuations would smooth out. Any long term variations would therefore have to be connected to corresponding variations in the energy transfer described above. Long term (i.e. thousands of years or more) increases or decreases in temperature would have to follow corresponding changes in albedo or trapping.

Yes, CO2 absorption of IR is well-established (although it is actually emission of IR that matters in the greenhouse effect), and energy is conserved. But the climate is incredibly complex, and there are hundreds of other influences and mechanisms that affect climate. How do you know which one of them is responsible?

A few examples – during the ice ages there are temperature fluctuations called Dansgaard-Oeschger events, roughly every 1500 years, give or take 500, in which the temperature rises 5C for a hundred years or so before dropping again. The equivalent during the Holocene are known as Bond events. It appears to be some sort of oscillation effect in long-term ocean circulation, but nobody really knows what caused them, or even whether the apparent regularity is coincidental or not.

The Atlantic Multidecadal Oscillation (AMO) and the Pacific Multidecadal Oscillation (PDO) are large scale switches in air pressure systems over the oceans, associated with changes of surface temperature. They have a period of about 50-60 years, so they switch from warm phase to cold phase and back again every 30 years. During the period from 1910-1940 the PDO was warm, and the temperature rose. During the 50s-70s the PDO was cool, and the temperature dropped slightly. During the 80s and 90s it flipped warm again, and the global temperature rose. Now of course that might be a complete coincidence – correlation doesn’t imply causation – but as we don’t know exactly how the ocean oscillations work, or what their ultimate consequences might be, we can’t say. Ocean currents circulate with periods of thousands of years.

Clouds have a dramatic effect on the surface temperature, which varies depending on their height, thickness, consistency, time of day, and many other factors. Since climate models simulate the atmosphere at a coarse resolution – grid points are hundreds of kilometres apart – smaller scale effects have to be incorporated with approximations to their effect averaged over a grid box. Cloud formation is affected by the temperature profile of the atmosphere, moisture, wind, dust, and possibly even cosmic rays. (Ionisation causes droplets to coalesce.) In the tropics, it is generally sunny in the morning, the heat evaporates water to form clouds during the day, blocking further input, and eventually you get tropical storms late in the day. The hotter it is, the earlier the clouds form, the more incoming radiation is blocked.

Irrigation for agriculture increases humidity, and land clearance changes the surface colour and hence how much heat is absorbed. Even if the average darkness stays the same but the contrasts vary, this can change the amount of convection you get due to different patterns of surface heating.

It is even possible that there is no cause, as such, and the rise is simply random noise. Real random processes in nature are often more complicated than the simple forms of randomness you see in school. To illustrate, let’s say we represent the temperature at time t by T(t), and suppose the temperature at time t+1 depends partly on the previous temperature, with a drift back towards some central value, and a random heat input, which may be positive or negative with equal probability. Lets say we write this T(t+1)=0.99T(t)+random. If you plot such a sequence out in Excel, you will see that over short periods there appear to be steady rises and falls, like systematic trends. But if you keep going long enough, they all eventually average out to zero. It’s called a “stochastic trend” and has been known about by statisticians since about 1910.

It’s definitely worth actually doing this. People don’t understand until they actually see it, and have a chance to experiment with it.

The Eemian interglacial was about 5C warmer than today, with forests growing up to the Arctic coastline. The Holocene optimum was warmer than today, as probably were the Minoan warm period, the Roman warm period, and maybe too the Medieval warm period (although some people will argue about that one, and say it was about the same as today). If the current temperature is within the bounds of natural variation, and none of those others caused by man-made CO2, then how can you be so sure that the current variation is not natural? And that it has this one specific cause, out of so many? (And so many others still unknown.)

The burden of evidence is the other way round. It is for climatologists to say why it is and can only be CO2, not for them to suggest CO2 and for us to have to prove them wrong. What real-world observations (not models) are the supporting evidence for it?

Given that CO2 IR absorption is well established, and that terms add and subtract in the energy budget,
if the claim is that CO2 is not the cause, doesn’t that mean that there must be some negative term(s) that cancel out the CO2 trapping contribution, but also some other positive term(s) re-compensate by enough to account for the observed temperature increase? If those effects are triggered by increased CO2 in the first place, then isn’t CO2 still the causative factor? And if its not a cause, then the other effects are unrelated and therefore coincidental? It seems like the latter should be ruled out by Occam’s razor until supporting evidence is presented.

(maybe I’m misunderstanding something)