Dehumidifying the Tropics

Guest Post by Willis Eschenbach

I once had the good fortune to fly over an amazing spectacle, where I saw all of the various stages of emergent phenomena involving thunderstorms. It happened on a flight over the Coral Sea from the Solomon Islands, which are near the Equator, south to Brisbane. Brisbane is at 27° South, and as a result, it is much colder than the Solomons.

It was overcast when we took off that early afternoon from Honiara, the capital of the Solomons. When we got up to altitude we crossed over the mountainous spine of the island and started south over the ocean. My attention was immediately drawn to the scene below me. I could see that for mile after mile over the ocean, the thunderstorms were all arranged in the long serried rows called “squall lines”.

Figure 1. Honiara to Brisbane. Honiara, the capital of the Solomon Islands, is on the north side of the island of Guadalcanal. Distance is about 2,000 km (1,200 miles).

Squall lines arranged in row after row are the final emergent circulation pattern in the chain of events involving the formation and subsequent intensification of thunderstorms. At the start of the process, the first emergent phenomena to appear are cumulus clouds and a change in circulation patterns. Rather than random movements in the lowest atmosphere, the cumulus have Rayleigh-Benard circulation associated with them.

Figure 2. The first stage in the evolution of thunderstorms. The cumulus clouds could be thought of as flags marking the upwelling section of the Rayleigh-Benard circulation. SOURCE

Of course, the increased albedo due to the clouds initially cools the surface. However, over the next hours as the surface continues to heat up despite the reduced incoming energy due to the formation of the cumulus, above a certain temperature threshold a scattered few of these cumulus clouds develop into thunderstorms. If surface temperatures continue to rise, the number of thunderstorms continue to rise, and quite rapidly. The circulation pattern changes, with surface air being lifted to condensation level. There, the moisture falls as rain, and the heat of condensation drives the deep circulation to the upper troposphere. From there, dry air descends again to the surface.

Figure 3. Further circulation changes as thunderstorms develop.

The next emergent pattern in this process is that the thunderstorms begin to line up in a row, shoulder to shoulder. This allows for a more dense packing of thunderstorms into any given area, basically doubling the possible areal density of thunderstorms and greatly increasing their power. Here’s a really big example of that:

Figure 4. A radar observation of a long squall line, with few breaks, stretching a thousand kilometres (600 miles) from Oklahoma to Indiana. SOURCE

Until the day I took that flight, however, I had never seen or even imagined that there was a further possible emergent circulation pattern.  I’d seen plenty of squall lines during my time sailing and motoring over the tropical ocean. But I never guessed that the squall lines could stack up, one after the other, in relatively straight lines, for as far as the eye could see. And curiously, even now I can’t find a single clear photo showing the phenomenon, I can only report what I saw. I was shocked, it was a totally new cloud formation that I’d never witnessed or at least never noticed, endless rows of walled thunderstorms reaching well up towards flight altitude, with clear air canyons  to the surface between the towering sides. It was awe-inspiring.

And flying that afternoon from Honiara to Brisbane, going over the Coral Sea and slowly moving southwards from warm ocean to cool ocean, I saw the whole process leading from cumulus clouds up to those endless stacked squall lines, only I saw it in reverse because I was flying from warm to cold. It started with flying for a few hundred miles over row after row after row of thunderstorms. These rows stretched from horizon to horizon, which from jet elevation is a long ways.

One of the clear advantages of this physical arrangement is that it is more efficient, with less turbulence than individual storms. There are huge long areas, canyons between the squall lines, where all of the air is smoothly descending. From the plane I could see the long cirrus anvils spilling downwind from the tops of each of the squall lines. The air flowed out there at the top and began the long descent into the canyon below. This long rolling cylinder of air allows the efficient movement of huge amounts of air containing both latent and sensible energy into the base of the thunderstorms, up through cumulonimbus towers making up the the squall lines to the upper troposphere, and then rolling out at the anvils and downwards again, turning the waterwheel over and over.

As we flew south, the squall lines weakened. First there started to be gaps in the squall lines, narrow breaks in what had been continuous sections of thunderstorm after thunderstorm to the horizon. After that, the breaks got wider. Then areas of squall lines began to be interspersed with areas of dense, closely packed thunderstorms. Beyond a certain clear surface temperature threshold, there were no more squall lines, just dense thunderstorms. From altitude I could see the two separate regimes. To the north, squall line after squall line, with fewer and fewer breaks the further north I looked.

And to the south, not one squall line in sight, just masses and masses of densely packed individual thunderstorms.

As we continued flying south over cooler and cooler water, the number of thunderstorms in sight decreased. Eventually, there was another clear line, another temperature threshold beyond which there were no further thunderstorms, just cumulus clouds. It was fascinating to see the towering thunderstorm clouds suddenly come to an end at a definite line. Out my window I could see dozens and dozens of thunderstorms in view north of the line, and not one thunderstorm in the slightly cooler area south of the line. And finally, by the time we got to Brisbane, there were only scattered cumulus.

Now, I bring all of this up because in the comments to a previous thread I’d said I would write a post about humidity. And when you write about humidity, you have to write about thunderstorms. I wrote before about how thunderstorms function as natural refrigerators, springing into existence as needed to cool off the hot surface spots around the planet. In addition to acting as refrigerators, however, they also function as immense natural solar-powered dehumidifiers. And as I saw, they are capable of covering huge areas and removing an almost unimaginable amount of moisture from the air.

Thunderstorms do this in the same manner as one class of human-made dehumidifiers. They cool the air until the water condenses out, and then re-warm the dehumidified air. Here’s a graphic from my post on thunderstorms as refrigerators, which also shows how they dehumidify the air.

Figure 5. Thunderstorm acting as a de-humidifier. Moist air rises under the base of the thunderstorm until it starts to condense. At a certain point it begins to fall as rain. More moisture is removed, as both ice and water, within the tower. Finally, the now-dry air descends between the thunderstorms to repeat the cycle. SOURCE

Note that after the moisture has been wrung out of the air and has fallen as rain, the resulting dry air then returns to the surface, warming as it descends. So paradoxically, although the thunderstorm greatly increases local evaporation under and near the base, it actually decreases the overall bulk average local humidity. When the thunderstorm regime is fully established, most of the bulk atmosphere is composed of dry downwelling (descending) air. In other words, the thunderstorm is a natural dehumidifier that is constantly stripping the moisture out of the lowest troposphere.

========================

Now, why is this important to the climate? Well, it has to do with the amount and nature of the water vapor feedback, which is said to amplify the warming from any source.

The Earth intercepts about three hundred and forty watts of solar energy for every square metre of its surface averaged on a 24/7 basis.

The upwards radiation from the surface, on the other hand, averages more than that, about three hundred and ninety watts per square metre, again on a 24/7 basis.

This implies a net gain of about 15% (390 / 340 – 1) from the intercepted radiation to the surface radiation, including all systems—clouds, evaporation, winds, surface albedo, parasitic losses from sensible heat transport, radiation “window” to space, everything. In large measure this elevation of the surface temperature is due to the absorption and radiation of infrared (longwave) radiation by the poorly-named “greenhouse gases” in the atmosphere. First among these greenhouse gases, of course, is water vapor.

Now me, I don’t think the climate works linearly. I don’t think that a change in forcing will necessarily result in a proportional change in surface temperature. However, that’s the current paradigm, so let me run with their assumption and see where it goes. The doubling of CO2 in the atmosphere is stated to bring a change of 3.7 W/m2 at the top-of-atmosphere (TOA). Other things being equal (which they never are, but we’re assuming here in order to get rough numbers), this should result in a surface change on the order of 15% larger than the TOA change of 3.7 W/m2, or about four and a quarter watts per square metre.

And this, by Stefan-Boltzmann’s relationship and assuming blackbody for convenience, should give us about eight-tenths of a degree (C) of warming for a doubling of CO2.

How do we get from that basic calculation to the dread range of three degrees or so from a doubling of CO2 that the global warming enthusiasts talk about? Ah, that’s where the handwaving comes in. It is supposed to be from two main sources, which they call “cloud feedback” and “water vapor feedback”. The modelers say that both of these are positive, i.e., they both act to amplify the amount of warming from a given change in TOA radiation.

There is a terminology problem to start with. Only one of the two, the effect of water vapor, is a simple linear or semi-linear feedback. Clouds are not just simple feedback, they act as governors, applying both positive and negative feedback in response to different situations. However, let’s set that question aside as well to follow the main point, that of water vapor.

Using the TAO buoy dataset, I have demonstrated how at times in the tropics cloud feedback actually reverses the effects of increasing amounts of incoming energy, so that the surface air temperature cools despite an increase in total downwelling energy. (See One , Two , and Three .) And that’s just the net energy (albedo minus longwave) change in total downwelling energy, it doesn’t include the effect of emergent homeostatic climate phenomena like thunderstorms …

And other research of mine has shown that in general clouds cool the earth in the summer and warm it in the winter. Again, the clouds are not a feedback but act as a governor, tending towards a homeostatic state. Despite that, the climate models show a positive cloud feedback … go figure.

In any case, while I was writing the above story about thunderstorms and humidity as I had said I would do, I got to musing the other day about the second leg of the modelers’ claim, water vapor. The reasonable idea put forward by the climate modelers is that as temperatures warm, the amount of water vapor in the air goes up in some quasi-linear fashion. Since water vapor is the main greenhouse gas, this of course increases the downwelling infrared (longwave) radiation. This positive feedback is claimed to greatly enhance the surface warming, as a result of increased radiation-trapping water vapor.

The problem is, given my understanding of the tropical ocean and the weather therein, I thought that when the thunderstorms kicked in, the humidity would drop.

So after I had written the description above, about how thunderstorms constantly dehumidify the tropical atmosphere, I realized that I had the data I needed to actually see if my hypothesis held water, so to speak. I could see if my understanding of thunderstorms was correct.

You see, my understanding of thunderstorms and their actions comes from what I might term “first principles”—but not the first principles of physics usually referred to by that term. Nor are the first principles what I learned in a class or from a book.

I say I know thunderstorms from first principles because I lived for seventeen years in the tropical Pacific, and I was either in view of or being rained on by thunderstorms most of those days. So I know they dry the air around them because you can feel it. When the thunderstorms kick in, usually sometime during the afternoon, the air in the neighborhood gets much fresher and less sticky. I don’t mean the downwelling cold wind that comes with the rain, that’s different. I mean when the circulation changed and the dry air starts descending, the atmosphere in the whole area feels drier … but I’d never actually looked at the data to verify my experience. So, trapped by my own writing, I set out to see if I was right.

For this I turned again to the trusty TAO buoy dataset, which has thousands of hourly observations. For no other reasons than that it is in an interesting region, the “Pacific Warm Pool”, and because I used to live on this map, on a tiny island above and to the right of the “a’ in “Sea”, and so I’m very familiar with the weather there, I looked at a TAO buoy I’d considered previously. It’s on the Equator up north of the Solomon Islands. Among the information that is measured and recorded there, hour by hour, are the surface air temperature and the relative humidity.

Figure 6. Location of the TAO buoy, moored to the abyssal plain in water of about 2,200 metres (1.3 miles) depth. Nearest land is ~ 600 km away, no urban warming there.

I started by looking at the correlation between the actual measured variables, the surface air temperature and the relative humidity. It’s not the comparison I’m looking for, we’ll get to that, but I can use relative humidity for verification, because I expect it to show a negative correlation (an inverse relationship) between relative humidity (RH) and temperature. This is because if a parcel of air warms, it can hold more moisture. If there is the same amount of moisture in the parcel, when the parcel warms the relative humidity drops. So I’d expect the RH to drop with increasing temperature. Figure 7 shows those results from the TAO buoy at 0°N, 156°E.

Figure 7. Relative humidity versus air temperature, from the TAO buoy located as shown in Figure 6. N =119,359.

As expected, the RH drops with increasing temperature. So I would say the dataset is valid and acting as expected.

However, I wasn’t interested in relative humidity. Relative humidity is the relative amount of water vapor in the air, expressed as a percentage of what the air could potentially hold if it were fully saturated.

But the amount of radiation absorbed by water vapor doesn’t vary with the relative amount of water in the air (relative humidity). Instead, absorption varies with the actual amount of water vapor in the air, which is called the “absolute humidity” or AH. The amount of outgoing radiation intercepted by water vapor depends on the actual amount of water vapor in the air (AH), not the relative amount of water (RH). However, you can calculate the absolute humidity from the temperature and the relative humidity, so I did that.

Figure 8 shows a scatterplot of the temperature (bottom scale) versus the absolute humidity (AH, vertical scale). The AH for each hourly observation is calculated from the hourly temperature and relative humidity (RH) using the method shown here. (The pressure adjustment is too small to be of interest at the earth’s surface, it can be assumed to be constant.)

Figure 8. Temperature (°C) versus absolute humidity (kg/m^3). The individual gold lines show the trend for each degree celsius` interval. As an aside, the maximum air temperature of ~ 30°C reflects the general global oceanic temperature maximum, with few values over that in open oceans anywhere. In this case, only 0.5% of the data is above 30°C, and only a tenth of that is above 30.5°C (0.05%, five hundredths of a percent).However, although only 0.04% of the observations are below 24°C, there is no hard lower limit visible in the data.

Given the results shown in Figure 8, I’m gonna have to say that my experience is vindicated. As you can see, at cooler temperatures around 24 or 25°C the amount of water vapor in the air goes up fairly rapidly with increasing temperature, just as people claim.

However, as the temperature continues to rise, the rate of water vapor increase slows down, then levels off from 26 to 27 C. Then, between 27 and 28 C, the amount of water vapor in the air actually decreases with increasing temperature.

Above that temperature, water vapor resumes increasing with temperature, but quite slowly.

My interpretation is that the overall leveling off of the trend curve shown in gold, and the decrease in absolute humidity from 27° to 28°, are clear signs of the dehumidifying action of the thunderstorms. The thunderstorms form as a response to surface hot spots, and when they form, they wring the water out of the air.

What can we draw as more general conclusions from this? Well, not a whole lot, it’s only one buoy in one location, we need to look much further … but it does support the dehumidifying nature of thunderstorms. Assuming that the finding is general, the first conclusion would be that in the wet tropics, where most of the energy enters the climate system, the amount of water vapor in the air doesn’t vary a whole lot with temperature. At this location, as temperatures rise, thunderstorms start dehumidifying the air and the resulting curve is pretty flat, with little change in the region from 26°C to 29°C

And that, of course, means that the amount of amplification of any warming due to water vapor feedback will be smaller than would be calculated under the assumption of a relatively linear or constant rise in absolute humidity with temperature.

Of course, more study is required, and there is lots more to learn, both using the TAO dataset and elsewhere. All conclusions are provisional and subject to future falsification, void where prohibited or taxed, you know the drill. I was just happy to have my understanding confirmed by this preliminary look at the data.

Regards to all,

w.

============

DATA AND CODE: Well, the code is a mess. By a mess I mean it’s in my usual form, in that I write in bits and chunks, and in general I only run selected lines. So while the code contains all the information and calculations needed to do the analysis above, they are not necessarily in order. Plus there’s a lot of other code that isn’t used, or didn’t work, or worked but wasn’t needed for the post. And far from being user-friendly, code I write might be described as “user-agressive”, I wrestle with it constantly.

In any case, there are two R files and a data file, zipped, downloadable here. The data file contains (as an R “save” file) the two NetCDF (ncdf) files, one for humidity called “rh” and one for temperature called “nc”, for the given buoy 0N156E. The ncdf files are the 4-bit files available here. The two R programs are “buoy temperatures” and “buoy humidity”. The “buoy humidity” file contains the calculations for the graphics shown above. The temperature data is calculated in the “buoy temperature” file, and then utilized by the “buoy humidity” file. So good luck, and as the poet said, “Lasciate ogni speranza voi ch’entrate” …