Guest Post by Willis Eschenbach
This is the third in a series ( Part 1, Part 2 ) of occasional posts regarding my somewhat peripatetic analysis of the data from the TAO moored buoys in the Western Pacific. I’m doing construction work these days, and so in between pounding nails into the frame of a building I continue to pound on the TAO dataset. I noticed that a few of the buoys collect data on both shortwave (solar) radiation and longwave (infrared or greenhouse) radiation at two-minute intervals. For a data junkie like myself, two-minute intervals is heaven. I decided to look at the data from one of those buoys, one located on the Equator. at 165° East.
Figure 1. Location of the buoy (red square) which recorded the data used in this study. Solid blue squares show which of all the buoys have the two-minute data. DATA SOURCE
It was a fascinating wander through the data, and I found that it strongly supports my contention, which is that the net effect of clouds in the tropics is one of strong cooling (negative feedback).
To start with, I looked (as always) at a number of the individual records. I began with the shortwave records. Here is a typical day’s record of the sun hitting the buoy, taken at two-minute intervals:
In Figure 2 we can see that when clouds come over the sun, there is an immediate and large reduction in the incoming solar energy. On the other hand, Figure 3 shows that clouds have the opposite effect on the downwelling longwave radiation (DLR, also called downwelling infrared or “greenhouse” radiation). Clouds increase the DLR. Clouds are black-body absorbers for longwave radiation. After they absorb the radiation coming up from the ground, they radiate about half of it back towards the ground, while the other half is radiated upwards The effect is very perceptible on a cold winter night. Clear nights are the coldest, the radiation from the ground is freer to escape to space. With clouds the nights are warmer, because clouds increase the DLR. Figure 3 shows a typical 24 hour record, showing periods of increased DLR when clouds pass over the buoy sensors.
Once again we see the sudden changes in the radiation when the clouds pass overhead. In the longwave case, however, the changes are in the other direction. Clouds cause an increase in the DLR.
So, here was my plan of attack. Consider the solar (shortwave) data, a typical day of which is shown in Figure 2. I averaged the data for every 2-minute interval over the 24 hours, to give me the average changes in solar radiation on a typical day, clouds and all. This is shown in gray in Figure 4.
Then, in addition to averaging the data for each time of day, I also took the highest value for that time of day. This maximum value gives me the strength of the solar radiation when the sky is as clear as it gets. Figure 4 shows those two curves, one for the maximum solar clear-sky conditions, and the second one the all-sky values.
As expected, the clouds cut down the amount of solar radiation by a large amount. On a 24-hour basis, the reduction in solar radiation is about 210 watts per square metre.
However, that’s just the shortwave radiation. Figure 5 shows the comparable figures for the longwave radiation at the same scale, with the difference discussed above that the clear-sky numbers are the minimum rather than the maximum values.
As you can see, the longwave doesn’t vary much from clouds. Looking at Figure 3, there’s only about a 40 W/m2 difference between cloud and no cloud conditions, and we find the same in the averages, a difference of 36 W/m2 on a 24-hour basis between the clear-sky and all-sky conditions.
At this location, clouds strongly cool the surface via reflection of solar radiation (- 210 W/m2) and only weakly warm the surface through increased downwelling longwave radiation (+ 36 W/m2). The net effect of clouds on radiation at this location, therefore, is a strong cooling of – 174 W/m2.
This likely slightly overstates the radiation contribution of the clouds. This is because, although unraveling the effect on shortwave is simple, the effect on longwave is more complex. In addition to the clouds, the water vapor itself affects the downwelling longwave radiation. However, we can get an idea of the size of this effect by looking at the daily variation of longwave with and without clouds in more detail. Figure 6 shows the same data as in Figure 5, except the scale is different.
Note that the minimum (clear-sky) DLR varies by about 10 W/m2 during the 24 hours of the day. Presumably, this variation is from changes in water vapor. (The data is there in the TAO dataset to confirm or falsify that presumption, another challenge for the endless list. So many musicians … so little time …). Curiously, the effect of the clouds is to reduce the underlying variations in the DLR.
This warming due to water vapor, of course, reduces the warming effect of the clouds by about half the swings, or 5 W/m2, to something on the order of 30 W/m2.
Finally, to the perplexing question of the so-called “cloud feedback”. Here’s the problem, a long-time issue of mine, the question of averages. Averages conceal as much as they reveal. For example, suppose we know that the average cloud cover for one 24 hour period was forty percent, and for the next 24 hours it was fifty percent. Since there were more clouds, would we expect less net radiation?
The difficulty is, the value and even the sign of the change in radiation is determined by the time of day when the clouds are present. At night, increasing clouds warm the planet, while during the day, increasing clouds have the opposite effect. Unfortunately, when we take a daily average of cloud cover, that information is lost. This means that averages, even daily averages, must be treated with great caution. For example, the average cloud cover could stay exactly the same, say 40%, but if the timing of the clouds shifts, the net radiation can vary greatly. How greatly? Figure 7 show the change in net radiation caused by clouds.
In this location, the clouds are most common at the time they reduce the net radiation the most (mid-day to evening). At night, when they have a warming effect, the clouds die away. This temporal dependence is lost if we use a daily average.
So I’m not sure that some kind of 24-hour average feedback value is going to tell us a lot. I need to think about this question some more. I’ll likely look next at splitting the dataset in two, warm dawns versus cool dawns, as I did before. This should reveal something about the cloud feedback question … although I’m not sure what.
In any case, the net cloud radiative forcing in this area is strongly negative, and we know that increasing cloud coverage and earlier time of cloud onset are functions of temperature. So my expectation is that I’ll find that the average cloud feedback (whatever that means) to be strongly negative as well … but in the meantime, my day job is calling.
A final note. This is a calculation of the variation in incoming radiation. As such, we are looking at the throttle of the huge heat engine which is the climate. This throttle controls the incoming energy that enters the system. As shown in Figure 7, in the tropics it routinely varies the incoming energy by up to half a kilowatt … but it’s just the throttle. It cools the surface by cutting down incoming fuel.
The other parts of the system are the tropical thunderstorms, which further cool the surface in a host of other ways detailed elsewhere. So the analysis above, which is strictly about radiation, actually underestimates the cooling effect of tropical clouds on surface temperature.
All the best, please don’t bother questioning my motives, I sometimes bite back when bitten, or I’ll simply ignore your post. I’m just a fool like you, trying to figure this all out. I don’t have time to respond to every question and statement. Your odds of getting a reply go way up if you are supportive, on topic, provide citations, and stick to the science. And yes, I know I don’t always practice that, I’m learning too …
PS — Here’s a final bonus chart and digression. Figure 8 shows the average of the actual, observed, measured variation in total downwelling radiation of both types, solar (also called shortwave) radiation and longwave (also called infrared or “greenhouse”) radiation.
Here’s the digression. I find it useful to divide forcings into three kinds, “first order”, “second order”, and “third order”. Variations in first order forcings have an effect greater than 10% of the average forcing of the system. For the system above, this would be something with an effect greater than about seventy W/m2. Figure 7 shows that the cooling from clouds is a first order forcing during the daytime.
Variations in second order forcings have an effect between 1% and 10% of the average. For Figure 8 that would be between say seven and seventy W/m2. They are smaller, but too big to be ignored in a serious analysis. With an average value of 36 W/m2, the warming from night-time clouds is an example of a second order forcing.
Finally, variations from third order forcings are less than 1%, or less than about seven W/m2 for this system. These can often be ignored. As an example of why a third order forcing can be ignored in an overall analysis, I have overlaid the Total Radiation (red line in Figure 8) with what total radiation would look like with an additional 7 W/m2 of radiation from some hypothetical CO2 increase (black line in Figure 8). This seven watts is about 1% of the 670 W/m2 average energy flowing through the system. The lines are one pixel wide, and you can scarcely see the difference.
Which is why I say that the natural governing mechanisms that have controlled the tropical temperatures for millions of years will have no problem adjusting for a change in CO2 forcing. Compared to the temperature-controlled cloud forcing, which averages more than one hundred and fifty W/m2, the CO2 change is trivial.