Guest Post by Willis Eschenbach
For some time now I’ve been wondering what kind of new evidence I could come up with to add support to my Thunderstorm Thermostat hypothesis (q.v.). This is the idea that cumulus clouds and thunderstorms combine to cap the rise of tropical temperatures. In particular, thunderstorms are able to drive local temperatures to a level below that necessary for thunderstorm initiation. This allows them to act as a true governing mechanism.
Recently I came across a fascinating study by Zeng et al called “A Multiyear Hourly Sea Surface Temperature Data Set“(PDF). Figure 5 of their paper contains a host of information in three panels, from which I have extracted the average hourly changes in surface temperature in three areas of the Pacific shown below. The information is from moored buoys.
Figure 1. Multi-year hour by hour long-term average surface air temperatures for three areas in the tropical Pacific (2°N 95°W, 0°N 155°W, and 0°N 156°E, shown by colored circles). Two full days are shown, in order to see the complete temperature change during the night.
Note the strange shape of the hour-by-hour temperature record, and yet how similar it is between the various sites. Of course the question arises … where in Figure 1 is the evidence for a homeostatic system regulating the temperature? To investigate that, I looked at the data in a slightly different way.
Figure 2 shows exactly the same data as in figure 1, but shown as anomalies about the average, and only covering one day rather than two. There is information in the similarities as well as in the differences between the three locations.
Figure 2. Anomalies in hourly Pacific temperatures. Locations and colors are the same as in Figure 1. Intensity of the theoretical clear-sky solar energy at the surface (surface insolation) is shown by the line running through the sun (right scale). Blue oval shows general cumulus initiation time, green oval shows thunderstorm initiation time, and the brown oval shows the cloud dissipation time.
Let me give the timeline of the tropical day that is shown in Figure 2, pointing out the important time periods.
At dawn the day is clear and the air is calm. As a result, the air temperature starts rising quite quickly after sunrise.
As the day warms, at some point before noon a threshold is passed. Cumulus clouds (the small puffy white summer clouds) start to form, and the lower tropospheric circulation switches to a new pattern. The cumulus clouds can be thought of as flags that mark an area of rising air. Around each of them, there is a ring of descending air. This change has a couple of immediate effects.
The first is an immediate reduction in the incoming solar energy because of reflections from the cloud tops. This is accompanied by a rise in the surface wind. Since the evaporation varies linearly with the wind, if the wind doubles from 1 metre/second (2 knots) to 2 metres/second, evaporation doubles. This also cools the surface.
As a result of the changes in the lower troposphere, from calm with no clouds to increasing numbers of circulation cells with cumulus clouds, in the central and western Pacific the temperature actually starts dropping. This is happening despite the continuing increase in the amount of incoming solar energy (negative climate sensitivity, go figure). In the eastern Pacific, the air (and the ocean) is much cooler. As a result the cumulus coverage is less complete and develops more slowly, and the temperature rise is not reversed, but the temperature rise is significantly slowed.
However, even in the western and central Pacific the steady input of solar energy overwhelms the cumulus, and the temperature starts to rise again. At some point, usually early afternoon (and earlier in the day if the temperature is warmer), another threshold is passed and thunderstorms start to form. The temperatures continue to rise, but at a much slower rate than earlier in the day.
At some point in the late afternoon, the amount of incoming energy is less than the amount of outgoing energy, and the air starts to cool. It is aided in this cooling by the thunderstorms, which in warmer seas often continue until several hours after dark.
After the sun goes down, of course, clouds have a warming effect because they greatly increase the amount of downwelling “greenhouse” radiation. This can be seen in the kink in the curve between about 19:00 and 21:00 hours.
Once the clouds dissipate, however, the air is much freer to radiate out to space, and rapid cooling sets in which lasts until dawn. By that time, there are no clouds and the atmosphere is generally calm … and the cycle starts over again.
DISCUSSION
There are two main effects to having this daily cycle. The first is the rapid warming in the morning. The second is the abrupt slowing, and in warmer waters the reversing, of the rapidly rising temperatures by clouds and thunderstorms. These effects work together to keep the daily temperatures within a fairly narrow band.
The differences between the three locations are revealing. The onset of the cumulus is nowhere near as effective at preventing the temperature rise in the warmer waters of the western Pacific “hot pool”. In this warmer area the temperature starts rising again around noon, where in the central Pacific this doesn’t happen until about 2PM (14:00). In the much cooler waters of the eastern Pacific, on the other hand, it appears that not many clouds form, and the result is to push the peak daily excursion higher than that of the warmer waters. And because there are fewer thunderstorms in the cooler eastern Pacific, the temperature stays warmer after dark and does not show as pronounced a kink after dark.
Also, in the hottest region the peak temperature occurs earlier in the day than in the two cooler areas. Presumably this is because of the more complete development of the thunderstorm regime in the warmer waters, leading to more rapid cooling.
One unexpected curiosity in this data involves the fact that it takes energy to warm the air. If we start at the coldest point in the day and sum the temperature anomalies around that point, it will give us a relative measure of how much energy it took to warm the air during the 24 hours.
Here’s the odd part. Despite the differences in the local average air temperature, all three areas warm the local air the same amount, within about ± 5%. I have no idea why this might be so. Whatever the combination of clouds and thunderstorms is doing, it leads to about the same amount of warming of the air in all three areas. I have no idea why.
CONCLUSION
The hourly air temperatures over the tropical Pacific show a clear pattern that is due to the specific timing, formation rate, and number of cumulus and cumulonimbus (thunderstorm) clouds. This pattern is completely congruent with the idea that cumulus and cumulonimbus clouds act as a homeostatic mechanism to prevent the oceanic air temperatures from getting too warm or too cold.
I have not yet taken a look at the underlying dataset, I’m sure that there’s much more to learn. So to make this an official science paper, let me say that additional studies are needed …
All ideas welcome, be nice, attack the ideas all you want but please don’t attack individuals or cast aspersions on motives. That angrifies my blood mightily and I tend to do things I later regret. I’d prefer to avoid that.
Best regards to all,
w.
jae says:
June 7, 2011 at 2:42 pm
I suspect the difference is that the Guam station is on an island 110 metres above the ocean, and the information from Zeng et al. is buoy data from a couple metres above the ocean. I’ve checked the Zeng figures against the TAO/TRITON buoy data and they agree.
Good call.
w.
Richard Scott says:
June 7, 2011 at 5:38 pm
Surely one of the ways of testing a hypothesis is to see if it is supported by experimental data … not sure what your point is here. I had a hypothesis. I looked at the observational data that applied (hourly ocean air temps), data that might or might not support my hypothesis. In this instance the data supported my hypothesis.
How on earth is that not science?
w.
Willis,
Try downloading a few months of these satellite photo animations
http://agora.ex.nii.ac.jp/digital-typhoon/archive/monthly/
and watch the difference between the Pacific basin trade wind and prevailing westerlies wind patterns and how much different they are from patterns over the USA and the North Atlantic, that is due to mountain topography and lunar tides in the atmosphere.
The difference in the temperature climb rates are due to the dew point difference from cooler Eastern Pacific, lowering the specific heat index of the warmed parcel of air increases the temperature greater for the same joules of input, compared to the warmer and thus more humid air in the western takes more joules to get same amount of temperature rise.
The reason the three areas all have the same basic spread of temp gain is due to the additional moisture from evaporation needed to move the air upwards and pull in more surface air for the thermometer, is the same relative shift in dew point to effect the convection drive for that ambient temperature, at the uniform height of the gauges.
that magical point in mid day when the clouds start to form, is due to the convection meeting air at the height where the radiative pressure from the SST combined with the starting over night minimum dew point (that fell over night to match the SST), reaches the height where the aerobatic cooling lowers the condensational threshold past the dew point of the rising air parcel.
This explains the flat bottom base of clouds, the stronger the convective lifting the easier it is to break through CAPE that form at the top of the radiative limit of the energy from the surface temperature, on land soil moisture and the dew point minimum from the night before regulate this balance point. Over the ocean the SST is uniform and at the same latitude the thickness of the troposphere will be almost the same, hence the common reading of the total temperature shift. I would venture to say that you will see shifts in the spread with increases in latitude due to relative height of atmosphere lowers toward the poles.
This in no way disarms your hypothesis, just qualifies the changes predictable due to latitude changes of data base. Over the ocean if the air is warmer than the sea it will drop till the dew point of the surface air matches the SST (with maybe some adjustment for salinity, and its hygroscopic properties) that difference might be what you are seeing in the range of shift in min to max air temp, now that i think about it. Any time the dew point of the air tries to cool and drop below the dew condensational point, it condenses on the temp gauge and releases the heat of condensation giving that smooth bottom to the beginning of the 6:00am climb.
Willis Eschenbach says:
June 8, 2011 at 12:19 am
Richard Scott says:
June 7, 2011 at 5:38 pm
“For some time now I’ve been wondering what kind of new evidence I could come up with to add support to my Thunderstorm Thermostat hypothesis (q.v.). ”
Wrong, wrong, wrong! Searching for evidence to support your hypothesis is not the way science is done. That’s the kind of subjective nonsense we see from the alarmists.
Science is done by TESTING a hypothesis, not looking for evidence to support it.
Surely one of the ways of testing a hypothesis is to see if it is supported by experimental data … not sure what your point is here. I had a hypothesis. I looked at the observational data that applied (hourly ocean air temps), data that might or might not support my hypothesis. In this instance the data supported my hypothesis.
How on earth is that not science?
w.
IF you look to see if there is a signal in the height of the sea tides due to the orbital parameters of the Earth Moon system, that is science, BUT If I take the patterns of the daily weather and compare that set of data against the same orbital parameters, that is judged to be Astrology and banned from further reading or consideration?WTF?
I get the same reasoning, that I am just picking data that fits the hypothesis, and they say correlation is not causation, even when it is predictive for the regulation of the atmospheric tides , just as it is for the sea tides.
Willis
Tom, what “sinusoidal shape” are you referring to? The shape of the temperature curves are not sinusoidal in the slightest, which is why I say it supports my hypothesis.
I am referring to the Fourrier transform of a periodic function with a period of 24 h (that’s what your T is) .
If you do that with a suitable origin of time (around sunrise) you will not be surprised that the first term of the Fourier development which is sin[(2.Pi/24).t] (yes, a sinus !) has the by far biggest coefficient.
That is what I meant by saying that the SST has a “sinusoidal shape”.
Everything else, as I already explained in my first comment are just corrections due to harmonics at higher orders.
This curve (SST) is so strongly sinusoidal that the corrections are relatively small.
If you have a Fourier transform software, just feed in it the coordinates of the curves you have and you will see how far you were from the truth when you thought that the shape “was not sinusoidal in the slightest” .
I stress one more time that I have no criticism against your qualitative description of what’s happening with SST in the tropics AT A 24H SCALE.
I am sure that it’s indeed what happens .
And I just wrote some simple equations which are a quantitative equivalent of your qualitative description .
If you wanted you could get a very good fit for the day part with this model. The night fit would be good too but not exceptional because the backradiation of the clouds during the first night hours is neglected . But one can easily add it if one wants to .
Btw it’s true that there are “only” 2 degrees of freedom . However once you realize that one of the degrees of freedom is a function (the cloud cover – a(t)) , it is in fact equivalent to have an infinity of degrees of freedom 🙂
That’s why I don’t need to do the computer work to know that the fit of the model to the curves (temperatures not anomalies !) would be almost perfect for the day half .
And I wouldn’t do it anyway because as I said , I can’t see what’s new or interesting in there.
Willis Eschenbach says:
June 7, 2011 at 2:04 pm
“Actually that part is pretty well understood. See CAPE. However, that just shows the conditions of instability and will give an approximate time for cumulus or thunderstorm formation. It doesn’t specify the spatiotemporal distribution of the storms … a fancy way of saying we can say approximately what time of day the storms are likely to start, but we can only give probabilities about the wetness of your golf outfit …”
Thanks for the CAPE link. While I was being somewhat tongue in cheek about the golf, thunderstorms are a serious threat to golfers because of lightning. Most golfers I know are pretty keen about weather conditions during our hot months and understand the consequences of ignorance. As I once wrote on a golf web site I had,
“There is no penalty for heading to the clubhouse when lightning is around. The beer tastes much better when you are alive.”
Now, if you can just predict over which course those thunderstorms are going to pop up ……….
Many thanks for your courteous reply. I have just tried again for 10 mins to get my heat gun to make any change to temperature of the water in the bucket.No joy. I am applying serious heat here. If I can’t make any impression with a heat gun what chance does a molecule of co2 stand.Science appears to me to have completely underestimated the strength of surface tension or maybe nobody thought of it. If I’m right the whole agw thing is toast. Give it a try. kind rgds r.m.b
Richard Holle says:
June 8, 2011 at 3:41 am
“Wrong, wrong, wrong! Searching for evidence to support your hypothesis is not the way science is done. That’s the kind of subjective nonsense we see from the alarmists. Science is done by TESTING a hypothesis, not looking for evidence to support it.”
You do not know what your hypothesis states (implies) until you know all the evidence that it implies. So, looking for empirical evidence is just a matter of fleshing out the hypothesis. And, as you find true predictions from the hypothesis, you build a record of confirmation. Of course, if you discover predictions that are false then you have most likely found that the hypothesis has been falsified.
I think you are working with a rather simplified version of Popper’s thesis that hypotheses should be bold and that scientists should attempt to falsify them. Good suggestions, but to ignore the evidence is to stunt your understanding of the hypothesis. Popper was a very good thinker in the area of hypothesis testing and such but his work is not up to the level of that of Carl G Hempel, Israel Scheffler, Isaac Levi, and some others. Scheffler’s “Anatomy of Inquiry” discusses Popper in detail.
steven mosher says:
June 7, 2011 at 10:53 am
“2. The thermostat [hasn’t] keep the planet from warming. that is, unless it changes the way heat eventually returns to space (radiation), then it’s not really doing anything interesting on a global scale.”
I think that your conclusion is too strong. The observed warming undercuts Willis’ analogy to a “thermostat”, but leaves open the possibility that the mechanism he describes might be a net negative feedback to CO2-induced global temperature increase. He has not elaborated all the mechanisms (I expect he’s working on it), but the mechanism by which increased CO2 and increased temp cause increased heat transport to above cloud level should reduce what’s called the “climate sensitivity” even if the basic mechanism of CO2-induced warming is correct.
In general the AGW theory depends too much on globally and temporally averaged temperatures, and not enough on the careful examination of particular heat transfer mechanisms at particular times, temperatures, and places. The mechanism that Willis explores here corrects that defect, to a small degree at least. Careful measurement and mathematical analysis are required, but the mechanism that Willis has been addressing opens up the possibility that, starting with temperatures as they are now, future CO2 increases and associated low level temp increases will produce no future net heating, and even that the last 150 years of temp increases have been independent of CO2 all along.
I wrote “to a small degree at least”, but the heat flows in the mechanism are considerable, and could represent sufficient heat transfer to completely counteract the hypothesized H2O feedback that boosts the estimated CO2 sensitivity from about 0.6C (by some estimates) to 3.5C (the warmers’ favorite). More work is needed, I repeat, but the mechanism that Willis is addressing is really interesting.
Sooooo,
to Willis I say,
thanks, and keep up the good work.
Willis, I would still like to know your thoughts on the Lindzen & Rondanelli paper I referenced above, in case you missed my comment. Thanks!
Willis:
“I suspect the difference is that the Guam station is on an island 110 metres above the ocean, and the information from Zeng et al. is buoy data from a couple metres above the ocean. I’ve checked the Zeng figures against the TAO/TRITON buoy data and they agree.”
Fascinating! I had no idea that the durnal variation was so small over water (but I should have). Thanks.
timetochooseagain says:
June 8, 2011 at 3:32 pm
I found it interesting but not ultimately convincing. Perhaps that’s because I’m still not sure exactly what they found out. They say that with increasing sst, the convective precipitation increases and the stratiform precipitation decreases. However, the signal is quite weak, or perhaps just buried in a lot of noise. I’m not convinced that this is the easiest or clearest way to study the issue.
I do think that the “Iris effect” is real, and is one of the various homeostatic mechanisms that regulate the temperature of the planet.
w.
here willis
http://www.newton.ac.uk/programmes/CLP/seminars/120617001.html
Thanks, Steven. I have to say, that was the most confusing pile of information I’ve come across in a while. Pages of equations with no explanation of the symbols … what good is that? It seems like slides to accompany a lecture, but without the lecture it’s real, real hard to follow. The author of the paper seems to say that the answer is … drum roll … supercomputers.
I say that what supercomputers do is increase the rate at which you can iterate incorrect assumptions.
If the answer were bigger, faster computers, surely we would have seen some progress in fundamental questions like climate sensitivity. We have seen huge advances in computer size and speed. We have seen no such advances in the estimates of climate sensitivity.
I hold that this lack of advance is because the concept of a constant average “climate sensitivity” is an incorrect description of a dynamic situation. How much the system responds to a given change in forcing ( the sensitivity) depends on the temperature. The warmer the system gets, the less it responds to further forcing (reduction in climate sensitivity).
This means that any global average climate sensitivity is temperature dependent. One result of this is that sinusoidal forcing will result in non-sinusoidal response, with the tops clipped on the high side (daytime), and deep sinusoidal curves on the low temperature side (nighttime).
Or in other words … Figure 2 above.
w.
I had to ponder this for a week. I like the thermostat hypothesis. If true then this would show how the recent tornado outbreak is related to climate change. Your hypothesis says that the tropics will try to maintain the same temperature regardless of other forcing such as solar changes. If the net energy in the globe increases then this will show up as increased polar warming as the tropics will not warm due to increased thunderstorms. The temperature gradient from the equator to the poles will be reduced so there will be less turbulent motion in the temperate zones and milder weather. Thus the warmists, shouting more extreme warmcold and wetdry have it exactly wrong. Global warming leads to less violent and extreme weather.
Conversely, a drop in global energy will result in a greater temperature gradient as mostly the poles cool and more violent weather will occur as the turbulence seen in the frontal storm systems increase. Thus, the quiet suggestion that the recent tornado outbreak is due to cooling becomes evidence for the thermostat hypothesis. I do not for a moment believe Hansen’s single urbanized thermometer is telling us what is happening at the North pole. It has to be getting colder.
This is scary. More turbulent frontal storms will result in shorter growing seasons as cold weather storm systems will end the growing season with early frosts even if the local average temperature does not drop that much. A week a warmer weather after a hard killing frost does not help the growing season even though the average monthly temperature may not show much of anything. I have always thought a measurement of the growing season would be a lot more useful than meaningless average of the intrinsic temperature variable.
RE: Gary Palmgren: (June 16, 2011 at 6:48 pm)
I had to ponder this for a week. I like the thermostat hypothesis. If true then this would show how the recent tornado outbreak is related to climate change. Your hypothesis says that the tropics will try to maintain the same temperature regardless of other forcing such as solar changes.
This disregards the fact that the severe tornado activity this year is due to more cold air with a lower thermostat setting sliding south over warmer air. The reason this activity peaks out in the spring is because the tropics and northern subtropics are rapidly warming from direct overhead sunshine while the arctic is still very cold from the winter.
It may help to consider how this proposed temperature control mechanism may work. Note that the CO2 concentration is largely a side issue. At the tropopause the CO2 band is not as wide as at ground level but it is still saturated and one can assume that most CO2 emissions will be at those same wavelengths where it is also most absorptive.
http://homeclimateanalysis.blogspot.com/2010/09/earths-tropopause.html
I think that the thickness of the troposphere is probably related to the *absolute* humidity at ground level. In the troposphere, a normal atmosphere has a lapse rate of 6.5 degrees C colder per kilometer increase in altitude. There may be discontinuities in this slope where the air aloft is from a different source from the ground level. When a warm parcel of dry air rises, it cools adiabatically (without energy exchange) at a rate of 9.8 degrees C per kilometer increase in altitude so that it will quickly reach an altitude where it is as cold and as dense as the surrounding air and thus come to a stop.
If that air is damp, however, it may rise to an altitude where condensation begins. Once this happens, the latent heat of vaporization reduces the cooling rate to 5 degrees C per kilometer increase of altitude. When this happens, the rising air parcel is actually getting warmer than the surrounding air at a nominal rate of 1.5 degrees C per kilometer increase in altitude. This is the positive-feedback engine that can cause this parcel of air to rise to the top of the troposphere. In the tropics, where the tropopause altitude is some 17 kilometers, we might expect this air parcel to be over 24 degrees C warmer than the surrounding air if the 6.5 degree C per kilometer lapse rate and 5 degree C per kilometer wet cooling rate continued almost up to that level. Perhaps the tropopause may be thought of as the ‘dry-out’ altitude.
It would seem that this process would be most effective at regulating oceanic temperatures as condensation would tend to begin at relatively low altitudes.
Willis
Please see Curry’s
Critique of the HADSST3 uncertainty analysis, especially the shift in wartime sampling from every 6 hours to 8 AM, Noon & 8 PM. See:
Reassessing biases and other uncertainties in sea-surface temperature observations measured in situ since 1850, part 2:
biases and homogenisation Journal of Geophysical Research
From eyeball guestimates of your diurnal temperature variations this wartime change in sampling time appears to give a substantial wartime sampling time Type B (bias) error of 0.17 deg C or 0.09 C.
Look forward to your quantitative evaluation of this time sampling bias error.