Guest Post by Willis Eschenbach (@weschenbach on eX-Twitter)
I love science because of the surprises. Today I had several. My first surprise today was evidence of strong negative feedback in surface temperature. Let me note that I’m not claiming I’m the first person to make these observations. I’m simply saying, it was surprising to me.
My method of scientific investigation is graphics-based. I take chunks of numbers, sometimes tens of thousands of them, and display them graphically. And at times the result is what I expected, or even hoped for.
Other times, though, my latest graph comes up on the silver screen and I say “Whaaa?” … those are the surprises that make it all worthwhile. And those are where interesting meandering paths start. Come amble with me along one of those paths.
As the result of a series of misunderstandings and coincidences, I ended up taking a look at the month-by-month changes in the net effect of clouds on radiation. “Net effect” refers to the fact that clouds both warm and cool the surface.
Cooling occurs as a result of clouds blocking the sunlight from hitting the ground by reflecting sunlight back out to space or by absorbing the sunlight. Either way cools the surface.
Warming occurs from that part of the thermal radiation emitted by the clouds that hits and is absorbed by the ground.
“Net effect” is the difference between the two opposing effects—including both effects, are the clouds warming or cooling the surface, and by how much?
Unsurprisingly, this is known as the “surface net cloud radiative effect”, or the “surface net CRE”, hereinafter “CRE”. When the CRE is negative, it means the net radiative effect of the clouds iscooling the surface. Correspondingly, a positive CRE means clouds are warming the surface via radiation changes. Figure 1 shows the 24-year average of the CERES satellite record of the net surface CRE.
Figure 1. The effect of clouds on the net total of radiation (longwave and shortwave) absorbed by the earth’s surface. The horizontal dashed lines near the equator mark the edges of the tropics (23.5° N/S). The horizontal dashed lines near the poles are two polar circles (66.5° N/S). Units are watts per square meter (W/m2).
There are some interesting things about Figure 1.
• Overall, clouds cool the surface by about -19 watts per square meter (W/m2)
• The ocean is cooled almost three times as much as the land.
• The areas polewards of the two polar circles are warmed by clouds.
• The only areas warmed on average by the clouds are those polar regions and the deserts.
• The greatest cooling is at the inter-tropical convergence zones just above the Equator and the Pacific Warm Pool north of Australia.
What I’d never looked at, though, is the month-by-month record of the surface net CRE. Of course, to look at that we need to look at the hemispheres separately, to avoid the effects of the opposing seasons in the two hemispheres. Figure 2 below, showing northern hemisphere variation by month, was my first surprise.
Figure 2. Monthly net surface cloud radiative effect, northern hemisphere.
I did NOT expect the effect to vary from slight warming in the winter to -40 W/m2 cooling in the summer. That is a giant swing in the effect of the clouds.
It was also interesting to see the net cooling effect of -0.2 W/m2 per decade. The decadal increase in CO2 forcing was +0.27 W/m2 (95% CI: 0.22 W/m2 – 0.32 W/m2). So over the period of record, the small change in surface CRE is of the same order of magnitude and is acting in opposition (cooling) to any warming effects of the CO2 forcing.
Of course, that got me to wondering how much difference not having the radiative effect of clouds would make in the summer and winter temperatures … which led me to create Figure 3.
Figure 3. Current northern hemisphere summer temperatures (black), and theoretical temperatures without the clouds’ radiative effect (other things being equal, which of course they never are). Values in all instances were done in units of W/m2, and then converted to temperature using the Stefan-Bolzman equation with an assumed emissivity of 0.95.
So instead of northern hemisphere peak average summer temperatures being around 72°F (22°C), without the varying radiative effects of the clouds they would be about 84°F (29°C). Yikes! And winters would be slightly colder as well.
(And yes, I’m aware that without clouds a bunch of other things would change, so my graph is pure theory. I’m just trying to give folks a sense of just how huge a swing of cloud cooling from +5 W/m2 in winter to -40 W/m2 in summer actually is.)
Intrigued, I decided to take another look at the whole globe as in Figure 1, but this time for the northern hemisphere midwinter (December) and midsummer (June) separately. Here are those two graphics.
Figure 4. As in FIgure 1, but showing midsummer and midwinter surface net cloud radiative effect. December and June averages. The horizontal dashed lines mark the edges of the tropics (23.5° N/S), and the two polar circles (66.5° N/S).
Again, more things of interest. In the NH midwinter (December), clouds warm almost all of the area north of about 35°N or so. In the southern hemisphere midwinter (June), the same is true. The clouds warm areas south of about 35°S.
Another oddity. In many cases, the white/ black contour lines outline desert areas where according to CERES, clouds are warming regardless of the season. Why?
Next, I looked at scatterplots of the surface temperature versus the surface cloud radiative effect, utilizing 1° latitude by 1° longitude gridcell data. For each hemisphere there are 32,400 data points. I graphed the data by season and by hemisphere. And in doing so, I noticed a most curious oddity. This was my second surprise.
The graph of the relationship between the midwinter temperature and midwinter cloud radiative effect is very similar in the two hemispheres.
And the same is true of the relationship between midsummer cloud radiative effect and midsummer temperatures. The two hemispheres have similar summer relationships. Here are those comparisons.
Figure 5. Gridcell scatterplots. Upper panel shows midwinters—northern hemisphere midwinter (December) and southern hemisphere midwinter (June). Lower panel shows midsummers—northern hemisphere midsummer (June) and southern hemisphere midsummer (December).
There are some interesting points here. First, the correspondence between the two winters (upper frame) and between the two summers (lower frame) is surprisingly close.
The main difference is in the summers in the low-temperature gridcells. The southern hemisphere has open ocean almost all the way to the ice-covered high Antarctic Plateau. In both winter and summer, the clouds warm Antarctica. So in the summer, the change in the cloud radiative effect at Antarctica’s shoreline area is a sudden and almost vertical change to warming (left end of orange/black line, lower frame). In the Arctic, because the pole is covered with water rather than the high-elevation land of the South Pole, the change to the polar warming is slower and more gradual (left end of blue/black line, lower frame)
Other than that, however, the two hemispheres are quite similar. Most importantly, in both summer and winter, as temperatures go above about 26°C or so, the cloud cooling rapidly strengthens and increases faster with each additional degree of surface warming.
The seasonal similarity of the oceans of the two hemispheres is important to me for a curious reason. I’ve used a gridcell-based scatterplot analysis of the type in Figure 5 above for things like the following look at how temperature and CRE are related over the entire globe. See my post Observational and theoretical evidence that cloud feedback decreases global warming for a discussion of the implications of Figure 6 below.
Figure 6. Scatterplot, net surface cloud radiative effect versus surface temperature, all 1° latitude by 1° longitude surface gridcells.
The main objection that people have raised to my using a gridcell-based scatterplot analysis of the type in Figures 5 and 6 above is their claim that it’s investigating location-based relationships, and thus it does not demonstrate any direct relationships between the two variables.
Another way to state the objection is to say that of course certain locations have some given relationship between temperature and CRE—the relationship is ruled by the location-based characteristics of the gridcells in question. Maybe there are ocean currents or nearby mountains that are ruling both the temperature and the CRE.
Now, that didn’t seem logical to me, because in Figure 6, the CRE values are grouped by the average gridcell surface temperature. And there are lots of gridcells around the planet with very similar average temperatures. But I hadn’t figured out how to counter that objection, to show that it’s not location-based.
However, the similarity of the hemispheric ocean midwinters, and of the hemispheric ocean midsummers, shows that the relationship between temperature and cloud radiative effect is not due to some location-specific characteristics.
It can’t be location-specific, since there are no locations that are common to both hemispheres. These are entirely different gridcells in entirely different oceans in different hemispheres, with different currents, different depths, different adjacent landmasses … and yet the relationship between temperature and surface cloud radiation is surprisingly similar.
So by chance, to my third and greatest surprise of the day, after starting down a totally different path I stumbled across a way to counter the main objection I’ve gotten to my gridcell-based scatterplot analyses.
Funny how life works when I follow random byways with no guide star except my endless curiosity about the wonders of this world.
===
Moon rising up above the redwood trees. Must be time for me to go moon viewing. I just need someone with a uniform and a Glock to come by every few hours and say “Step back from the computer, sir, keep your hands away from the keyboard and nobody gets hurt!”
Onwards, my friends, and my best to you all—may your lives be full of marvels, adventures, and surprises of all kinds.
w.
As Usual: I politely request that when you comment you quote the exact words you’re discussing. Avoids endless misunderstandings.
Figures 5 (summer) and 6 show why the Earth’s average temperature is 15C and has no tipping point.
The change in CRE opposite to the change in CO2 blows
the IPCC theory of 3x water amplification out of the water.
The tipping points are not reached until earth’s average temperature reaches 25C or -20C.
Willis –
“Warming occurs from that part of the thermal radiation emitted by the clouds that hits and is absorbed by the ground.”
Not unless the cloud is warmer than the ground. Highly unlikely, as clouds actually decrease surface temperatures when they prevent sunlight (from a 5500 K source) reaching the ground.
Maybe you are confused about the IR reflective properties of clouds, which result in reduced rates of cooling at night. Cloudless deserts cool really quickly at night.
“Another oddity. In many cases, the white/ black contour lines outline desert areas where according to CERES, clouds are warming regardless of the season. Why?”
While day time Temperatures in the Sahara desert >40°C are common. Night time temperatures can drop to 0°C, the same would be impossible in Germany. The reason why the drop is so big, is that the air above the Sahara desert is so dry. Anything that is increasing water vapor will increase night time temperatures.
My take is that the water cycle is an evaporative cooling process. Where there is no water to evaporate, clouds have a net effect of insulation.
That’s an interesting one. Checking the temperatures for a number of desert locations in both north Africa and Australia, the average daily temperature range in any given month is about 20 degrees C. That’s quite a bit larger than the typical range in moister areas.
Having said that, some individual days in Spring and Autumn in dry areas can have larger ranges. For example, the largest forecast range for central Australia this week (as deserty a desert as one is likely to find) is around 23 degrees C – see http://www.bom.gov.au/nt/forecasts/map7day.shtml
I think people need to get over the idea that clouds only block radiation one way. They both block the radiation from the sun and they block it from the Earth. The first is obvious from shade under a cloud. The second is obvious from cloudy nights being warmer than clear-sky nights. (I can recall driving to work one day in temperatures of 28 deg C because the previous day had been sunny and hot but then clouds came over in the evening and blocked much of the IR escaping.)
There’s another slight qualification that needs to be made. In high northern latitudes, as winter nears, a cloudy day is often warmer than a sunny day simply because the cloud is blocking the release of heat.
How’s this blocking going to affect recorded temperature. The first thing to remember is that we’re only interested in the daily minimum and maximum temperatures (which are averaged across a given month and the mean of the two gives the mean monthly temperature). The second thing is that when the two are recorded at 9:00am each day the maximum temperature is logged for the previous day and the minimum temperature for the current day.
It’s often argued that the mean of clear sky day and night will be the same as the mean temperature for cloudy sky day and night; the latter will show a smaller diurnal range than the clear sky conditions but the mean be the same. I don’t believe this works. When the previous day is cloud free and warm and then cloud forms overnight and traps the heat, not as much radiation will be required the next day in order to get back up to the previous day’s maximum temperature. This is going to distort the maximum temperature on the second day and therefore the average maximum for the month and ultimately the mean temperature for the month.
Agreed. Willis touches on this timing issue when he describes his tropical cloud formation in the afternoon impacting the amount of energy reaching the tropics but I think its important at the poles too.
More cloud at the poles when the pole is in winter darkness ought to result in higher recorded temperatures and less ice formation than if it was clear sky.
But in summary, the timing of events is going to be every bit as important as CO2 and cant be modelled because our models simply aren’t capable of creating those changes. So we cant know the future impacts and hence future climate.
It is why the average should always have a standard deviation quoted along with it.
An average is is the midpoint of a normal distribution. A distribution will also have a standard deviation.
This is just totally ignored by climate science. Measurement uncertainty is just manipulated to get what looks good and allows milli-kelvin to be shown as accurate.
“ I don’t believe this works. When the previous day is cloud free and warm and then cloud forms overnight and traps the heat, not as much radiation will be required the next day in order to get back up to the previous day’s maximum temperature.”
I’m not sure this is correct. The sun’s insolation does include an IR component. The clouds will block some of this insolation meaning the surface will not receive as much warming radiation from the sun.
go here to see my temps for the last 30 days:
go here to see the rain for the last 30 eays:
The days around when we got rain were cloudy. The 3rd to the fifth the temps were high when it was clear before the rain on the 9th. The temps from the 25th to the 29th were high when it was clear before the rain on the 31st. It was cloudy from the 9th through the 20th when we were getting moisture (i.e. cloudy) over the whole period and that was when both max temps and low temps were the lowest.
This is why I have long advocated that temperatures are a *very* poor metric for determining climate. Although I haven’t done the calculations my guess is that even though the temps were low over the rainy period the enthalpy was probably higher due to pressure and humidity. Trying to measure climate using temps and radiation balance just doesn’t tell the whole story, especially because of all the unknowns associated with temps and radiation balance.