Drying The Sky

By Willis Eschenbach – Re-Blogged From WUWT

Eleven years ago I published a post here on Watts Up With That entitled “The Thermostat Hypothesis“. About a year after the post, the journal Energy and Environment published my rewrite of the post entitled “THE THUNDERSTORM THERMOSTAT HYPOTHESIS: HOW CLOUDS AND THUNDERSTORMS CONTROL THE EARTH’S TEMPERATURE“.

When I started studying the climate, what I found surprising was not the warming. For me, the oddity was how stable the temperature of the earth has been. The system is ruled by nothing more substantial than wind, wave, and cloud. All of these are changing on both long and short time cycles all of the time. In addition, the surface temperature is running some thirty degrees C or more warmer than would be expected given the strength of the sun.

Despite that, the earth’s temperature has stayed in a surprisingly narrow range. The HadCRUT global surface temperature shows that the range of the temperature trend was 13.9°C ± 0.3°C over the entire 20th Century. This represents a temperature variation of ±0.1% during a hundred years. That stability was the curiosity of curiosities for me, because to me that temperature stability was clear evidence of some kind of a strong thermoregulatory system. But where and what was the regulating mechanism?

The short version of my hypothesis is that a variety of emergent phenomena operate in an overlapping fashion to keep the earth’s temperature stable beyond expectations. These phenomena include tropical cumulus clouds, thunderstorms, dust devils, squall lines, tornadoes, the La Nina pump moving warm water to the Poles, tropical cyclones, and the Julian-Madden, Pacific Decadal and North Atlantic Oscillations. In addition, I’ve adduced a large body of evidence supporting my hypothesis.

So I was interested to see Judith Curry, in her marvelous weekly post entitled “Week in review – science edition“, had linked to a paper I’d never seen. It’s a paper from 2010 by Marat F. Khairoutdinov and Kerry A. Emanuel (hereinafter K&E) entitled AGGREGATED CONVECTION AND THE REGULATION OF TROPICAL CLIMATE, available here. Inter alia they say:

Moist convection in the Earth’s atmosphere is mostly composed of relatively small convective clouds that are typically a few kilometers in horizontal dimension (Byers and Braham, 1948, Malkus, 1954). These often merge into bigger clusters of ~10 km in horizontal dimension, such as air-mass showers. More rarely, under special circumstances, moist convection is organized on even larger scales; this includes squall lines (e.g. Houze, 1977), mesoscale convective complexes, (e.g. Maddox, 1980), and tropical cyclones.

One of the robust characteristics of self-aggregation is the rather dramatic change in the mean state that accompanies it. In particular, in all non-rotating experiments (Bretherton et al., 2005) and an experiment on an f-plane (Nolan et al., 2007), self-aggregation leads to dramatic drying of the domain-averaged environment above the boundary layer. This appears to be the result of more efficient precipitation within the convective clump as more of the condensed water falls out as rain and less is detrained to the environment, per unit updraft mass flux. Such dramatic drying would reduce the greenhouse effect associated with the water vapor, and thus, would lead to cooling of the SST, which in turn may disaggregate convection. This would re-moisten the atmosphere, increasing the water-vapor greenhouse effect, and, consequently, warming the system. So, as in self-organized criticality (SOC), the tropical state would be attracted to the transition critical state between the aggregated and disaggregated states.

Let me point out a few things about their most interesting study. First, they are clear that a strong effect of the aggregated thunderstorms is to regulate the tropical temperature … just as I’ve been saying for years.

Unfortunately, their study is model-based. This is always frustrating to me because there is no way to check either the quality of their models or how many runs ended up on the cutting room floor …

However, given that shortcoming, their study points to something I noted in my original post—not just aggregated thunderstorms but also individual thunderstorms dry out the air in between them. This has two big cooling effects on the surface.

First, the dry descending air allows for increased evaporation from the surface, because the dry air can pick up more moisture from the surface. This increased evaporation cools the surface.

In addition to the increased evaporation, the effect they discussed is that the dryer air descending around the thunderstorms reduces the amount of the world’s main greenhouse gas, water vapor, that is between the surface and outer space. This allows the surface to radiate more freely to space, which also tends to cool the surface.

In their summary they say:

Idealized simulations of radiative-convective equilibrium suggest that the tropical atmosphere may have at least two stable equilibrium states or phases, one is convection that is random in time and space, and the second is the spontaneously aggregated convection. In this study, we have demonstrated using a simplified and full-physics cloud-system-resolving models that there is an abrupt phase transition between these two equilibrium states depending on the surface temperature, with higher SST being conducive to the aggregation. A significant drying of the free troposphere and consequent reduction of the greenhouse effect accompany self-aggregation; thus, the sea-surface temperature in the aggregated state tends to fall until convection is forced to disaggregate.

So, big credit to them for noticing the thermostatic effect in the tropics. However, their look is tightly focused. They have looked only at one cooling mechanism. In addition, they have only looked at two of what are at least four of what they call stable equilibrium states or phases. However, again to their credit they’ve said “at least” two stable states, acknowledging the existence of others.

Since Khairoutdinov and Emanuel had demonstrated using models that dry air increased with increasing aggregated thunderstorms, I thought I’d take a look at, you know … observations. Data. Crazy, I know, since so much attention is paid to models, but I’ve been a computer programmer far too long to put much faith in models.


Let me start by saying that they are looking at the third and fourth stable equilibrium states in the entire spectrum of the daily tropical thermally-driven threshold-based atmospheric response to increasing surface temperature. Each of these steps involves self-organized criticality.

In the tropics, by dawn, particularly over the ocean, the night-time atmosphere is generally stable and thermally stratified, with clear skies at dawn.

The first step is when the solar warming of the surface warms the air above it enough to initiate the stable equilibrium state called Rayleigh-Benard convection. As is common with such self-organized transitions, once the critical transition temperature is exceeded, the change between states is rapid.

Once Rayleigh-Benard circulation is established, areas of ascending air are interspersed with areas of descending air. The areas of rising air, often called “thermals”, transfer surface heat and surface water vapor upwards. This cools the surface directly through conduction, because the air traveling across the surface picks up heat from the surface. The R-B circulation also increases thermal radiation to space from the upward movement of the warm air above the lowest atmosphere, which contains the greatest amount of greenhouse gases.

Finally, the R-B circulation increases evaporation by moving the surface moisture upwards and mixing some of it into the lowest part of the troposphere.  This transition to R-B circulation is generally invisible, although the onset of daily overturning can sometimes be felt in the wind.

The second transition is again temperature-based. It occurs when the surface temperature is large enough to drive the Rayleigh-Benard circulation higher into the troposphere. In the tropics, this transition typically happens in the late morning. When the water vapor in the ascending columns of the Rayleigh-Bernard circulation is moved upwards to the “LCL”, the “lifting condensation level” where water vapor condenses, at that altitude cumulus clouds form. The water vapor in the air condenses into the familiar puffy cotton-ball cumulus clouds. Each individual cumulus cloud group sits like a flag marking an ascending part of the Rayleigh-Benard circulation shown above.

Again, the transition is rapid. In the space of about a half-hour, the entire tropical atmospheric horizon to horizon can go from clear air to a fully developed cumulus field. And again, the transition is temperature-based. Below a certain temperature, there are hardly any cumulus clouds at all. Above that temperature, suddenly there are lots of cumulus clouds.

The third transition occurs when a somewhat higher temperature threshold is exceeded. The third stage of development is when individual cumulus clouds self-aggregate into scattered cumulonimbus. They build tall cloud towers, and the rain starts.

After this transition to the thunderstorm state, large areas of descending dry air form around each thunderstorm. This is the return path of the air that was first stripped of water in the base of the thunderstorm. When the water vapor condenses it gives up heat. The heated air then moves up the thunderstorm tower, emerges at the top, and descends as dry air in the areas around the thunderstorm.

This stage, of active thunderstorms, is well illustrated in the most entrancing simulation shown below. The colored layer added at one minute twenty seconds shows the temperature of that layer, with dark blue being coldest and red/orange being warmest.

The fourth and final transition occurs only in certain conditions at the highest transition temperature, when individual thunderstorms self-aggregate into squall lines and supercells, medium-scale convective complexes, and tropical cyclones. This is the only one of the four stable equilibrium states studied by Khairoutdinov and Emanuel. As with the other transitions, they point out that it is associated with a transition temperature. Like the thunderstorm regime, areas of descending dry air form around the aggregated phenomena. Here’s a photo of a single squall line from space.

It is worth noting that each of these succeeding stages exhibits an increase in the rate at which the surface loses heat. With each transition, the rate of surface heat loss increases from a variety of causes. The cause that is discussed by K&E, increased radiation to space through drier air, is only one among many.

The first transition, from quiescent stratified night-time atmosphere to Rayleigh-Benard circulation, increases surface heat loss to the atmosphere through conduction and convection of both latent and sensible heat. It encourages atmospheric loss to space by moving the surface heat up above the lowest atmospheric levels with their denser concentration of the greenhouse gases, mostly water vapor and CO2. It mixes surface heat and surface water vapor upwards. Because water vapor is lighter than air, the ascending areas are moister and the descending areas are dryer in the R-B circulation.

The second transition, to the cumulus field, adds two new methods of cooling the surface. First, energy is moved from the surface aloft in the form of latent heat. This heat is released when the rising columns of air condense into clouds. The sun then re-evaporates the water from the upper surface of the clouds, and the water vapor mixes upwards. This moves the surface heat well up into the lower troposphere.

The cumulus field also cools the surface by reflecting sunlight back to space. This is a very large change in the energy balance, on the order of a couple of hundred watts per square metre or so. The timing and density of the emergence of the cumulus field is one of the major thermal regulation mechanisms. How strong is this regulatory action? Here’s a typical day’s available solar energy, measured at ten-minute intervals at a TAO buoy in the Equatorial Pacific Ocean.

The deep notch in the available solar energy from clouds covering the sun at around 11:30 AM in the graphic above is quite typical of the drop when clouds cover the sun. On this day it lasted about half an hour. It reduced the available solar energy flux by about six watts per square metre averaged over that 24 hour period.

By comparison, a theoretical doubling of CO2 from the present, which is highly unlikely to happen, would add a flux of about 3.7 watts per square metre during that 24 hour period.

So in that area, that one cloud would be more than enough to cancel out even a doubling of CO2 for that day … and that is just one of the many ways the surface is being cooled by emergent phenomena.

The third transition, from developed cumulus field to scattered thunderstorms, adds the whole range of new surface cooling methods that I list in the endnotes. And unlike the first two transitions, thunderstorms can actually cool the surface to a temperature below the temperature needed to initiate the thunderstorms. This allows thunderstorms to maintain surface temperatures. When any location gets hot a thunderstorm forms and cools the surface back down, not just to where it started, but down below the onset temperature. This “overshoot” is the signature of a governor as opposed to a simple linear or similar feedback. Simple feedback can only reduce a warming tendency. A governor, on the other hand, can turn warming into cooling.

In the fourth transition, the transition to the larger self-aggregated phenomena like squall lines, supercells, and the like, no new surface cooling methods are added. What happens instead is that the previous methods move to a new level of efficiency. For example, thunderstorms self-organize into squall lines as shown in the photo above.

Instead of individual areas of descending air around each individual thunderstorm, in a thermally-driven squall line you get long rolls of dry descending air along the flanks of the squall line. Because the carpet-roll-type circulation is streamlined, with the air smoothly rolling in a long tube, the squall line moves more energy from the surface to the upper troposphere than would be moved by the same number of individual thunderstorms.

To summarize the discussion so far:

There are four distinct successive emergent transitions from a quiescent stratified atmosphere to fully developed squall lines. Each is the result of self-organized criticality. Each one is a separate emergent phenomenon, coming into existence, persisting for some longer or shorter time, and then disappearing. In order, the transitions and the new emergent phenomena are:

  • Still air to Rayleigh-Benard circulation
  • Rayleigh-Benard circulation to cumulus field.
  • Cumulus field to scattered thunderstorms
  • Scattered thunderstorms to aggregated thunderstorms.

Each transition removes more energy from the surface to the atmosphere and thus eventually from the system.

Khairoutdinov and Emanuel discuss drying of descending air in only one of the states, the fourth one where thunderstorms aggregate. They are correct. However, this does not begin at the fourth stage. All the stages dry the descending air. And after each succeeding transition, the air becomes dryer and dryer.

I’ve demonstrated the close dependence of thunderstorms and “aggregated” thunderstorms on the surface temperature. I made up a movie showing this a while back using the CERES data, hang on … OK, here it is. I am using the extent of deep convection as measured by the cloud top heights as a measure of the strength of the thunderstorms and aggregates.

In the movie, you can see the thunderstorms and aggregated thunderstorms (color) following the warm water (gray lines) around the Pacific throughout the year.

And if we take a scatterplot of average cloud top altitude versus sea surface temperature, we find the following relationship:

Just as we saw in the movie above, when the sea surface temperature goes over about 26°C thunderstorms explode vertically, getting taller and taller. This is clear support for the idea that the transition between states is temperature-threshold based.

With all of that as prologue, let me move to the question of the descending dry air between the thunderstorms. I realized that we actually have some very good information about the amount of water in the air. This is data from the string of what are called the TAO/TRITON buoys and other moored buoys that stretch on both sides of the Equator around the world. Here are their locations.

Let me begin with another look at rainfall and temperature. Here’s a scatterplot of the sea surface temperature versus the rainfall in the equatorial Pacific area shown by the yellow box above (130°E – 90°W, 10°N/S). The blue dots below show results from the TAO buoys in the yellow box. The red dots show gridcell results from the Tropical Rainfall Measuring Mission (TRMM) satellite rainfall data and Reynolds OI sea surface temperatures.

Man, I do love it when several totally independent datasets agree so well. In the graph above the blue dots are co-located measurements of average rainfall and sea surface temperature at individual TAO/Triton buoys. The red dots are 1° latitude by 1° longitude averages of Reynolds OI Sea Surface Temperatures, and Tropical Rainfall Measuring Mission (TRMM) satellite-based rainfall data. And in both datasets, we see once again that thunderstorms start forming in numbers only when sea surface temperatures get above about 26°C.

It’s also interesting that once the sea surface temperature gets into the upper-temperature range, there are no dry areas. Every place gets at least a certain minimum amount of rain. Not only that, but the minimum amount of annual rainfall increases smoothly and exponentially as the average sea surface temperature goes up.

Why is it so important that this threshold is temperature based? It’s important for what it is NOT. It is not forcing based. In other words, the great global thunderstorm-based air-conditioning and refrigeration system kicks in at about 26°C, no matter what the forcing is doing. No matter what the CO2 is doing. No matter what the volcanoes are doing. The regime shift from puffy white cumulus clouds to scattered thunderstorm towers kicks in when the temperature passes a temperature threshold, and not before, regardless of what CO2 does.

And this, in turn, means that these successive regime shifts, first to Rayleigh-Benard circulation, then to the cumulus field, then to scattered thunderstorms, and finally to aggregated thunderstorms, are functioning in a host of different ways to regulate and cap the surface temperature.

And finally, by a fairly circuitous but interesting route, we’ve arrived back at the question of the drying of the air in between the thunderstorms and thunderstorm aggregations.

There are eight TAO buoys that are directly on the Equator across the Pacific. It’s an interesting group because they all get identical sunshine. Despite getting identical solar energy, there is a temperature gradient from Central America across to Asia, with the Asian end at about 29°C and the South American end at about 24°C. So looking at these eight buoys gives us a look at how some phenomena vary by temperature.

Using the temperature and the relative humidity measurements from these buoys, I calculated the absolute humidity for each of them. This is the amount of water that is present per cubic metre of air. That number is important because the absorption of long-wave radiation by water vapor varies proportionally to the absolute humidity, not the relative humidity. Less absolute humidity means more surface heat loss by long-wave radiation to space.

These observations from the buoys are done every ten minutes. This allows me to calculate what an average day’s variations look like. To understand the daily variations, I aligned them at the morning minimums. Here are the records of those eight TAO buoys that are directly on the Equator.

In this graph, note that the warmer that the sea surface temperatures are, the smaller the 10 AM peak, and the more the afternoon absolute humidity drops from the 10 AM peak.

This is because as the thunderstorms form and increase the local area moisture is concentrated in the small area in and under the thunderstorms, with descending dry air between the thunderstorms making up the bulk of the lower troposphere. And in areas with warmer sea surface temperatures, shown in red above, clouds and thunderstorms form earlier, are denser, and at times form even larger aggregations of thunderstorms.

Now, what I’ve shown above are long term full-dataset averages. So it’s tempting to think “well, thunderstorms only happen where the average temperature is over 26°C”. But thunderstorms are not touched by averages. These temperature-regulating phenomena can appear, persist, and disappear at any time of day. All that matters are the instantaneous conditions. Whenever the tropical ocean gets warm enough, regardless of the longer-term averages for that location, you are likely to see thunderstorms form. All the averages mean is that the surface gets sufficiently hot to create thunderstorms on more or fewer days of the year.

My conclusions?

K&E were right about the drying power of aggregated thunderstorms.

It is also true that individual thunderstorms, as well as cumulus clouds and Rayleigh-Benard circulation, dry out the descending air.

This lower level of water vapor cools the surface by increasing radiation loss to space and by increasing evaporation.

This is only one of the host of ways that cumulus clouds and thunderstorms keep the tropics from overheating

Rayleigh-Benard circulation, cumulus fields, scattered thunderstorms, and aggregated thunderstorms are all emergent phenomena. They emerge wherever there is sufficient surface heat, meaning when the temperature exceeds some local threshold. Each succeeding state, in turn, starts removing more heat from the surface. This is an extremely efficient temperature regulating system because they emerge only as and where there are local concentrations of surface heat.

Finally, I want to emphasize one of K&E’s interesting claims:

Such dramatic drying would reduce the greenhouse effect associated with the water vapor, and thus, would lead to cooling of the SST, which in turn may disaggregate convection. This would re-moisten the atmosphere, increasing the water-vapor greenhouse effect, and, consequently, warming the system. So, as in self-organized criticality (SOC), the tropical state would be attracted to the transition critical state between the aggregated and disaggregated states.

In other words, all of these phenomena act to stabilize the temperature.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s