Guest Post by Willis Eschenbach
In a recent post, I described how the El Nino/La Nina alteration operates as a giant pump. Whenever the Pacific Ocean gets too warm across its surface, the Nino/Nina pump kicks in and removes the warm water from the Pacific, pumping it first west and thence poleward. I also wrote about the dolphins in a piece called “Here There Be Dragons“.
Fulfilling an obligation I incurred in the latter paper by saying I would write about emergence and climate, let me take a larger overview of the situation by noting that both the El Nino pump and the dolphins are examples of a special class of things that are called “emergent” phenomena.
Figure 1. Hands emerging from the paper …
Emergence is a very important concept. Systems with emergent phenomena operate under radically different rules than those without. Today I want to talk about emergent systems, and why they need to be analyzed in different ways than systems which do not contain emergent phenomena.
Examples of natural emergent phenomena with which we are familiar include sand dunes, the behavior of flocks of birds, vortexes of all kinds, termite mounds, consciousness, and indeed, life itself. Familiar emergent climate phenomena include thunderstorms, tornadoes, clouds, cyclones, El Ninos, and dust devils.
Generally speaking, we recognize emergent phenomena because they surprise us. By that, I mean emergent phenomena are those which are not readily predictable from the underlying configuration and physics of the situation. Looking at a termite, if you didn’t know about their mounds there’s no way you’d say “I bet these bugs build highly complex structures a thousand times taller than they are, with special air passages designed to keep them cool”. You wouldn’t predict mounds from looking at termites, no way. They are an emergent phenomenon.
The El Nino phenomenon is another excellent example of emergent phenomena. Looking at a basin of water like the Pacific, there’s no way you would say “Hey, I’ll bet that ocean has this complex natural system that kicks in whenever the ocean overheats, and pumps warm water up to the poles.” You wouldn’t predict the existence of the El Nino from the existence of the Pacific Ocean. It also is an emergent phenomenon.
In addition to their surprising emergence from the background, what other characteristics do emergent phenomena possess to allow us to tell them from other non-emergent phenomena?
One common property of emergent phenomena is that they are flow systems which are far from equilibrium. As a result, they need to evolve and change in order to survive. They are mobile and mutable, not fixed and unchanging. And locally (but of course not globally) they can reverse entropy (organize the local environment). Indeed, another name for emergent phenomena is “self-organized phenomena”.
Another key to recognizing emergent phenomena is that they arise spontaneously when conditions are right. They don’t have to be artificially generated. They create themselves in response to external stimuli.
Next, they often have a lifespan. By a “lifespan”, I mean that they come into existent at a certain time and place, often when some natural threshold is exceeded. Thereafter they are in continuous existence for a certain length of time, and at the end of that time, they dissipate or disappear.
Another characteristic of emergent phenomena is that they are not cyclical, or are at best pseudo-cyclical. They do not repeat or move in any regular or ordered or repetitive fashion. Often they can move about independently, and when they can do so, their movements can be very hard to predict.
Another feature of emergent phenomena is that they are often threshold-based. By that I mean that they rarely emerge below that threshold, but above it their numbers increase rapidly.
Another attribute of emergent systems is that they are often associated with phase changes in the relevant fluids, e.g. clouds occur because of a phase change of water.
One final attribute of threshold-based emergent systems is crucial to this discussion—they exhibit “overshoot” or hysteresis. In the Rayleigh-Bénard circulation shown below, it takes a certain threshold temperature difference from top to bottom to cause the emergence of the circulation pattern. But once that circulation is established, it will persist even though you turn the heat down far below the initiation threshold temperature.
So those are some of the characteristic features of emergent phenomena. They are flow systems far from equilibrium which arise spontaneously, often upon crossing a critical threshold. They are not obviously predictable from the underlying conditions. They move and act unpredictably, they are often associated with phase changes, and they exhibit “overshoot” (hysteresis).
However, not all emergent systems are created equal. Some of them are what might be termed “field-wide”. An example of this is the spontaneous emergence of “Rayleigh-Bénard” natural circulation in a fluid heated from the bottom and cooled from the top.
Figure 2. Rayleigh-Bénard circulation. Curiously, the time reads first down the right column then down the left. At the end an organized series of rising and falling areas is emerging. It is characterized by narrower rapidly upwelling sections, separated by larger, slower-moving downwelling sections. Original Caption: Onset and development of thermal convection cells in Rayleigh-Benard convection. Note the regularity of initial “bubbles” and their coalescence to form larger loops. SOURCE
Another, more complex type of emergent systems are what might be called “independent”. Examples of these in the climate world are thunderstorms and dust devils. Unlike the field-wide emergent phenomena, these are free to roam about the landscape. Like all flow systems far from equilibrium, they are constantly adjusting and evolving to met the physical constraints. For example, thunderstorms move preferentially across the surface to warmer areas.
As I said above, I want to highlight the difference between the analysis of systems that do and do not contain emergent phenomena. My thesis is that systems with emergent phenomena cannot be analyzed in the same manner as systems without emergent phenomena. The corollary is that climate models are appropriate only for systems without emergent phenomena. Let me give an example of each kind of system, so you can see the difference.
For the first system, let me consider a flat slab of iron that is warmed by the sun or some other heat source in a vacuum. As the heat source varies, the temperature of the slab of iron varies as well. This variation in temperature with energy input is quite regular and predictable. If we graph the changes, we’d see that there are no sharp bends in the graph. In addition, the more energy that the iron is receiving, the hotter it gets, with an unchanging mathematical relationship between downwelling radiation and the temperature of the iron slab. So we could approximate it by a straight line.
Now, lets replace the flat slab of iron with a flat slab of cool water, and we’ll add the possibility of clouds and thunderstorms as the emergent phenomena. Starting with cool water, at first we’d see basically the same thing as with the iron slab—the more energy we add, the warmer the water gets. Everything is all nicely proportional, the water is acting just like the iron. (Yes, there’s a million details, but work with me here. It’s a thought experiment.)
But at a certain point, a curious and surprising thing happens. A threshold is passed, and clouds form.And when they do, it reflects some of the incoming energy back to space. So we get a “knuckle” in the graph of incoming energy versus temperature. We’re no longer warming as fast as we were.
If the incoming energy continues to rise, however, a more surprising thing happens. Another threshold is passed, and thunderstorms begin to form. These cool the surface in a host of ways, most importantly by piping the warm surface air through the middle of the thunderstorm up to high altitudes. This avoids almost all of the greenhouse gases (H2O and CO2) in the lower troposphere, and allows for free radiation of huge amounts of thermal energy to space. Not only that, but thunderstorms are radically different from a feedback, because they cool the surface down to well below the thunderstorm initiation threshold temperature. This means that they can not only slow down a temperature increase, they can stop it in its tracks.
And at that point, when thunderstorms start forming, the water basically stops warming. Further increases in incoming energy are simply equalled by further increases in thunderstorms and changes in their orientation, so that the surface temperature hardly warms after that.
Now, one of the claims of the AGW supporters is that there is a linear relationship between downwelling energy and temperature. They say that any increase in incoming energy must be matched by an increase in surface temperature. Despite the known non-linearity of the system, the claim is made that over a narrow interval, a linear approximation of the relationship between energy and temperature is a very reasonable approximation to the reality.
But in the thunderstorm part of the tropical thermal regime, it is important to note that not only is the relationship between incoming energy and temperature non-linear, but in fact there is no relationship between incoming energy and temperature. So you cannot even approximate it with a linear relationship. In that regime, increases in incoming energy are generally balanced out by increases in thunderstorm numbers and associated increased evaporation and convection, leaving only small residual temperature changes.
So one reason you can’t simply map a linear approximation to a non-linear relationship is because in the thunderstorm regime, there is almost no relationship, non-linear or otherwise, between incoming energy and temperature. Given the number of phase changes of water that are involved in the thunderstorm system, this should be no surprise at all—the same exact situation occurs when water is boiling. The temperature of the boiling water can no longer be even approximated by looking at how much energy is going into the water. The boiling water system simply moves energy through it at a faster rate, it doesn’t run any hotter. The exact same thing is going on in the thunderstorm regime. If you increase the solar radiation, all you get is more thunderstorms moving faster. The surface doesn’t get hotter, the energy and the water just circulate faster.
There is a second reason that you can’t just take an average, note that the average doesn’t move much, and assume linearity. The problem is that in the tropics, the climate sensitivity is very different depending on the time of day. Here’s why. First, without reference to anything else, climate sensitivity is an inversely proportional function of temperature for several reasons.
• Radiation is a function of T^4.
• Parasitic losses increase with temperature.
• Emergent cooling mechanisms (thunderstorms, dust devils, rain) are temperature based with high numbers above a threshold of emergence.
So clearly, climate sensitivity is inversely proportional to temperature, falling as temperature rises. It is not a constant in any sense of the word.
Next, climate sensitivity varies over both space and time. In the early morning in the all-critical tropics where the energy enters the planet-sized heat engine we call “climate”, the temperature rises rapidly because of the lack of clouds—a high change in temperature per change in watts (high sensitivity). In the late morning the watts are still rising but the clouds greatly reduce the temperature rise—smaller change in temperature per change in watts (low sensitivity). And indeed, certain areas at certain times can show negative sensitivity, and some areas of the planet are not sensitive to the forcing at all.
Now, the global average climate sensitivity, the one that people take as a constant, is no more than the average of these highly varying sensitivities. But the average is greatly misleading, because it is taken as constant or semi-constant. In the real world, however, climate sensitivity not constant in any sense. It is both inversely proportional to temperature and highly non-linear.
For example, in Figure 3 above, the “climate sensitivity” is taken as the average slope of the linear trend line relating temperature and incoming radiation. As you can see, if the earth were like an iron slab with no emergent phenomena, a straight light approximates the curve extremely well at every temperature. But in the real world with water and clouds, the trend line is meaningless—it doesn’t represent the actual climate sensitivity at any temperature.
As a result you can’t just say well, because the global average surface temperature doesn’t vary much, we can treat it as a constant. The average is not real, it is a mathematical chimera. In the real world, we don’t see an average temperature. If the “average temperature” goes up by one degree, and it happens to be evenly spread out, let’s say the morning temperature goes from say 7°C to 8°C, while the afternoon goes from 22°C to 23°C.
But both the climate sensitivity, and the change in climate sensitivity with temperature, are very, very different in the two temperature regimes of morning and afternoon. It takes much, much more energy to go from 22°C to 23°C than it does to go from 7°C to 8°C. So while the average temperature doesn’t change much, that is highly deceptive. In reality, the dependence of sensitivity on temperature makes a huge difference in how the system actually reacts to changes in forcing.
To explain this in detail, I’m going to shamelessly steal, re-heat, and re-forge a section from my earlier post called “It’s Not About Feedback”, because it is highly relevant to the questions I’m discussing. To understand why emergent phenomena are critical to understanding the climate, here is the evolution of the day and night in the tropical ocean. The tropical ocean is where the majority of the sun’s energy enters the huge heat engine we call the climate. So naturally, it is also where the major homeostatic mechanism are located.
At dawn, the atmosphere is stratified, with the coolest air nearest the surface. The nocturnal overturning of the ocean is coming to an end. The sun is free to heat the ocean. The air near the surface eddies randomly.
Figure 4. Average conditions over the tropical ocean shortly after dawn.
As you can see, there are no emergent phenomena in this regime. Looking at this peaceful scene, you wouldn’t guess that you could be struck by lightning in a few hours … emergence roolz. As the sun continues to heat the ocean, around ten or eleven o’clock in the morning there is a sudden regime shift. A new circulation pattern replaces the random eddying. As soon as a critical temperature/humidity threshold is passed, local circulation cells spring up everywhere. These cells transport water vapor upwards to the local lifting condensation level. At that level, the water vapor condenses into clouds as shown in Figure 3.
Figure 5. Average conditions over the tropical ocean when cumulus threshold is passed.
Note that this area-wide shift to an organized circulation pattern is not a change in feedback. It has nothing to do with feedback. It is a self-organized emergent phenomenon. It is threshold-based, meaning that it emerges spontaneously when a certain threshold is passed. In the “wet” deep tropics there’s plenty of water vapor, so the major variable in the threshold is the temperature. In addition, note that there are actually two distinct emergent phenomena in the drawing, the Rayleigh-Bénard circulation which emerges prior to the cumulus formation, and which is enhanced and strengthened by the emergence of the clouds.
Note also that we now have a change of state involved as well, with evaporation from the surface and condensation and re-evaporation at altitude.
Under this new late-morning cumulus circulation regime, much less surface warming goes on. Part of the sunlight is reflected back to space, so less energy makes it into the system to begin with. Then the increasing surface wind due to the cumulus-based circulation pattern increases the evaporation, reducing the surface warming even more by moving latent energy up to the lifting condensation level.
Note that the system is self-controlling. If the ocean is a bit warmer, the new circulation regime starts earlier in the morning, and cuts down the total daily warming. On the other hand, if the ocean is cooler than usual, clear morning skies last later into the day, allowing increased warming. The system is regulated by the time of onset of the regime change.
Let’s stop at this point in our examination of the tropical day and consider the idea of “climate sensitivity”. The solar forcing is constantly increasing as the sun rises higher in the sky. In the morning before the onset of cumulus circulation, the sun comes through the clear atmosphere and rapidly warms the surface. So the thermal response is large, and the climate sensitivity is high.
After the onset of the cumulus regime, on the other hand, much of the sunlight is reflected back to space. Less sunlight remains to warm the ocean. In addition to reduced sunlight there is enhanced evaporative cooling. Compared to the morning, the climate sensitivity is much lower. The heating of the surface slows down.
So here we have two situations with very different climate sensitivities. In the early morning, climate sensitivity is high, and the temperature rises quickly with the increasing solar insolation. In the late morning, a regime change occurs to a situation with much lower climate sensitivity. Adding extra solar energy doesn’t raise the temperature anywhere near as fast as it did earlier.
Moving along through the day, at some point in the afternoon there is a good chance that the cumulus circulation pattern is not enough to stop the continued surface temperature increase. When the temperature exceeds a certain higher threshold, another complete regime shift takes place. Some of the innocent cumulus clouds suddenly mutate and grow rapidly into towering monsters. The regime shift involves the spontaneous generation of those magical, independently mobile heat engines called thunderstorms.
Thunderstorms are dual-fuel heat engines. They run on low-density air. That air rises and condenses out the moisture. The condensation releases heat that re-warms the air, which rises deep into the troposphere.
Figure 6. Afternoon thunderstorm circulation over the tropical ocean.
There are a couple of ways to get low density air. One is to heat the air. This is how a thunderstorm gets started, as a strong cumulus cloud. The sun plus GHG radiation combine to heat the surface, warming the air. The low density air rises. When that gets strong enough, a thunderstorm starts to form.
Once the thunderstorm is started, the second fuel is added to the fire — water vapor. Counter-intuitively, the more water vapor there is in the air, the lighter it becomes. The thunderstorm generates strong winds around its base. Evaporation is proportional to wind speed, so this greatly increases the local evaporation.
This, of course, makes the air lighter, and makes the air rise faster, which makes the thunderstorm stronger, which in turn increases the wind speed around the thunderstorm base, which increases the evaporation even more … a thunderstorm is a regenerative system, much like a fire where part of the energy is used to run a bellows to make the fire burn even hotter. Once it is started, it is much harder to stop.
This gives thunderstorms a unique ability that, as far as I know, is not represented in any of the climate models. It is capable of driving the surface temperature well below the temperature that was needed to get it going. It can run on into the evening, and at times well into the night, on its combination of thermal and evaporation energy sources.
Thunderstorms can be thought of as local leakages that transport heat rapidly from the surface to the upper atmosphere. They cool the surface in a host of ways, utilizing a combination of cold water, shade, wind, spray, evaporation, albedo changes and cold air.
And just like the onset of the cumulus circulation, the onset of thunderstorms occurs earlier on days when it is warmer, and it occurs later (and sometimes not at all) on days that are cooler than usual.
So again, we see that there is no way to assign an average climate sensitivity. The warmer it gets, the less each additional watt per metre actually warms the surface.
Finally, once all of the fireworks of the daytime changes are over, first the cumulus and then the thunderstorms decay and dissipate. A final and again different regime ensues. The main feature of this regime is that during this time, the ocean radiates about the amount of the energy that it absorbed during all of the previously described regimes. How does it do this? Another emergent phenomenon …
Figure 8. Conditions prevailing after the night-time dissipation of the daytime clouds.
During the nighttime, the surface is still receiving energy from the GHGs. This has the effect of delaying the onset of oceanic overturning, and of reducing the rate of cooling. Note that the oceanic overturning is once again the emergent Rayleigh-Bénard circulation. Because there are no clouds, the ocean can radiate to space more freely. In addition, the overturning of the ocean constantly brings new water to the surface, to radiate and to cool. This increases the heat transfer across the interface.
As with the previous thresholds, the timing of this final transition is temperature dependent. Once a critical threshold is passed, oceanic overturning kicks in. Stratification is replaced by circulation, bringing new water to radiate, cool, and sink. In this way, heat is removed, not just from the surface as during the day, but from the entire body of the upper layer of the ocean.
There are a few things worth pointing out about this whole system.
First, this is what occurs in the tropics, which is where the energy enters the hot end of the great heat engine we call the climate.
Next, sometimes increases in incoming energy are turned mostly into temperature. Other times, incoming energy increases are turned mostly into physical work (the circulation of the ocean and atmosphere that transports energy to the poles. And other times, increasing energy is mostly just moved from the tropics to the poles.
Next, note that this whole series of changes is totally and completely dependent on temperature threshold based emergent phenomena. It is a mistake to think of these as being feedback. It’s more a drunk walking on a narrow walkway, the guardrails are not feedback—they are a place where the rules change. The various thresholds in the climate system are like that—when you cross them, everything changes. The ocean before and after the onset of overturning are very different places.
And this, in turn, all points to one of the most important control features of the climate—time of onset. How much energy the ocean loses overnight depends critically on what time the overturning starts. The temperature of the tropical afternoon depends on what time the cumulus kick in, and what time the thunderstorms start
Finally, look at the difficulty in analyzing or modeling this kind of situation. You have a grid box that is far larger than any cloud or thunderstorm. And all you have to go on, the only things in your model, are the average statistics of that gridbox. And the main control system is the timing of the initiation of threshold based phenomena that are below your gridcell size …
Think about say the average humidity of the tropical Pacific where there are thunderstorms. As soon as the thunderstorms kick in, they start discharging dry air up high. This dry air cools and descends in the area between the thunderstorms. So if you were to average the relative humidity of the bulk of the atmosphere across say one gridcell of a climate model, a hundred miles square or so, you’d see humidity falling as thunderstorms develop.
But this bulk drying of the downwelling air masks what is really happening. Under the thunderstorms, the storm-driven winds kick the evaporation into overdrive. The dry surrounding air is drawn in, loaded to the brim with moisture via the increased evaporation, and shot skyward at rates up to 10 m/sec. In a few minutes it’s moved up to the LCL, the “lifting condensation level”, where it condenses as clouds and rain.
As a result, despite the fact that the bulk atmosphere is drying, in fact huge amounts of moisture are being moved vertically through the system. So simple averages are useless. The system is moving more water but the average relative humidity of the bulk atmosphere has dropped.
So increasing input may only increase the throughput, rather than increasing the temperature. Not all of the energy that hits the tropical ocean is immediately radiated back to space. A large amount of it is moved, via the ocean and the atmosphere, towards the polar regions before finally returning to space. This means that one of the crucial determinants of the temperature of the tropical regions, as well as of the polar regions, is the rate of energy throughput—how much energy is moved from the tropics to the poles. Once the system is into the thunderstorm regime, almost all of the incoming energy goes to simply turning the wheel faster, moving more energy from the surface to the upper troposphere, moving more air and water from the tropics to the polar regions. So instead of warming up the surface, the energy is moved skywards and polewards.
Again, however, these changes in throughput make the situation difficult to analyze. The dang system won’t ever stand still, it responds to everything that happens. How can one accurately measure how much energy is being moved and transformed by a thunderstorm?
An allied difficulty is with the size of the phenomena. The thunderstorm is the most common natural heat engine on the surface of the planet. But they are way, way below the typical grid size of a climate model. As a result, they simply cannot be simulated in modern global climate models. This means that they must be “parametrized”, which as near as I can tell comes from the Latin and means “made up to fit the programmer’s preconceptions”. But while parametrizing a simple system is not difficult, parametrizing a system containing emergent phenomena is a very hard thing to do well.
In part, this problem arises from the very thing causing the need to parameterize—the small size of the thunderstorms. The problem is that those small thunderstorms cool down small hot spots before they ever get large. I have seen, for example, a single solitary thunderstorm in the morning, sitting over some warm spot in the ocean, with not another cloud in the sky. It was feeding off of some very local hot spot which had persisted through the night, and as long as it was hot, the thunderstorm stayed and cooled it down.
How on earth can one parametrize such an instantaneous response to excess warmth? Thunderstorms spring up over hot spots and cool them down to below the initiation temperature of the thunderstorm. And that kind of quick proactive response containing overshoot is not easily put into parameters.
And given that all you have are grid box averages, how will you model the critical changes in the time of onset of the various emergent phenomena? If the cumulus doesn’t appear until an hour later, or shows up an hour earlier, it makes a huge difference. And of course, the clouds and thunderstorms never shows up off-time, it emerges only as and when required, because its appearance is set by the immutable laws of wind and water and evaporation and condensation. It can’t occur late or early, it’s always right on time. But in the models, there are no thunderstorms …
As I mentioned above, there are a range of emergent climate phenomena. In general, they work together to maintain the temperature of the planet within fairly narrow bounds. The most important one of these is the tropical thunderstorm system described above. And there is something very critical about this system, something you may not have noticed so let me repeat it. A main control on the temperature is exerted by the timing and strength of the emergent phenomena, particularly clouds and thunderstorms. Now, here’s the important part. The time of day when a cloud forms is a function of the physics ruling the winds and the waves and the water and evaporation and condensation and the air and how they react to temperature.
Did you get it that time? Here’s why that statement is important. It is important because of what is missing—there is no mention of CO2, because CO2 doesn’t exert any direct effect on when clouds form. Clouds form in response to temperature and pressure and the like, not CO2.
So if there is a bit of additional forcing and the surface is a bit warm, the clouds simply form earlier, and the thunderstorms form earlier, and the nightly overturning of the ocean starts earlier … and that balances out the additional forcing, just like it has done for millions of years.
Nor is this just theory. I’ve shown that at the TAO buoys, days that start out colder than average end up warmer than average, and days that start out colder end up warmer … just as this theory predicts. See here and here for further discussion of the effect of emergent systems as seen in the TAO buoy records.
Now, note that I didn’t say that this kind of system containing emergent temperature control systems was impossible to model … just that it is hard. I’ve done a lot of computer modeling myself, both iterative and non-iterative models, and so I’ve both written and used physics based models, economic models, models using neural nets, machine learning algorithms, computerized evolution models, tidal models, I’ve played the game a lot in a lot of fields and a lot of ways. It could be done. But it can’t be done the way that they are doing it, because their way doesn’t account for the emergent phenomena.
The emergence of clouds and thunderstorms radically cooling the surface, plus the increase in convection and evaporation with temperature, plus the thermal radiation going up as the fourth power of the temperature, all combine to put a serious barrier in the way of any increases in temperature. As near as I can tell, the climate models have no such barrier. In the model world, going up six degrees or even ten degrees seems to be no big deal, model runs achieve that without breaking a sweat.
But in the real world, of course, Murphy conspires with nature to make sure that every single additional degree is harder and harder to achieve … and emergent phenomena not only stop warming, they actively cool the surface down. Until both the theory and the models robustly embrace the emergent phenomena, the models will continue to be a funhouse-mirror version of reality … you can recognize it as some kind of climate but with all the distortions, you can’t use that as a guide for anything.
One last question—how would I recognize a good climate model? Well, in a good model all of the emergent phenomena we know about would actually emerge, not be parametrized … because the free actions of those emergent phenomena, the variations and changes in their times and locations of appearance are what control the temperature, not the CO2 “control knob”. So when the forcing from CO2 increases a watt or two, in an accurate model the clouds will emerge a few minutes earlier on average across the tropics, and the balance will be restored. This system of control by emergent phenomena has worked very well for billions of years, and it handles large swings in radiation every single day—it won’t be altered by a few watts of extra forcing from CO2.