Guest Post by Willis Eschenbach
For a while now, I’ve been using a curious kind of scatterplot. Here’s an example. It shows the relationship between the surface temperature and the effects of clouds on surface radiation. Clouds can either warm or cool the surface, depending on location, time, and type. The phenomenon is called the “Cloud Radiative Effect” (CRE).
The amount of radiative cloud warming or cooling (the CRE) is measured in watts per square meter (W/m2). Positive means clouds warming the surface, and negative means clouds cooling the surface. Globally as an area-weighted average, clouds radiatively cool the surface by about – 21 W/m2.
Figure 1. Scatterplot, surface temperature (horizontal “x” axis) versus net surface cloud radiative effect (vertical “y” axis). Gives new meaning to the word “nonlinear”.
What are we looking at here? Well, each blue dot in Figure 1 represents a 1° latitude by 1° longitude gridcell somewhere on the earth’s surface. Each dot is placed horizontally with respect to its 21-year average temperature and vertically by its 21-year cloud radiative effect. The yellow/black line is a LOWESS smooth that shows the overall trend of the data.
And of particular interest, the slope of the yellow/black line shows how much the cloud radiative effect changes per 1°C change in temperature.
We can see from Figure 1 that clouds generally warm the coldest areas of the planet. Gridcells in the ~ 10% of the planet where the average annual temperature is below -5°C are warmed by clouds.
In warmer areas, on the other hand, clouds cool the surface. And when the temperature gets above about 25-26°C, cloud cooling increases strongly with increasing temperature. In those areas, for each additional degree of temperature, cloud cooling increases by up to -15 W/m2. This is because of the rapid increase above 26°C in the number, size, and strength of thermally driven thunderstorms in the warm wet tropics.
Here’s a video showing how the thunderstorms follow the warm water throughout the year.
Figure 2. Thunderstorm intensity is shown by colors (cloud top altitude is a measure of thunderstorm strength). Gray contour lines show temperatures of 27, 28, and 29°C.
From this, we can see that thunderstorms emerge preferentially over the hot spots, and they effectively put a cap on how far the temperature can rise in those areas. This is the reason that only 1% of the earth’s surface area, and virtually none of the open ocean, has an annual average temperature over 30°C.
With that as a prologue, since few people in climate use a gridcell-based scatterplot, let me discuss this kind of scatterplot. It has a very valuable property.
The value is that the method is looking at longer-term averages. In Figure 1, for example, these are the average temperatures that each of the gridcells has settled to after millennia. As a result, the gridcell-by-gridcell temperatures include all the possible various feedbacks and the majority of the slow responses to changing conditions.
And this allows us to answer questions like “what will be the response of the clouds if the temperatures warm slowly”? Alarmists would have you believe that the warming will be increased by the feedback of the clouds.
But Figure 1 tells a much more complex and nuanced story. The slope of the yellow/black line shows the change in CRE in response to a 1° change in temperature. If it slopes down to the right, it shows that the magnitude of the cloud-caused cooling is increasing with increasing temperature—the CRE is getting more negative, and clouds are doing more cooling..
There are only two places where the clouds act to increase an underlying warming. These are the areas in Figure 1 where the yellow/black line slopes upwards to the right. They are the 3% of the surface colder than -20°C, and the ~30% of the earth between 15°C and 25°C. These total about a third of the planet.
Gridcells at all other temperatures will have increasing cloud cooling as they warm, particularly the third of the globe that averages above 25°C.
Conclusion? Only a third of the globe has a warming cloud feedback and it is not that strong. Two-thirds of the globe has cooling cloud feedback, and in addition, the cooling feedback is far stronger than the warming feedback.
Thus, we can say that on average the cloud feedback is negative, not positive. An area-weighted average of the above data shows that globally, cloud cooling averages -3.2 W/m2 of cooling for each one degree C of warming. (In reality, the overall cloud response will be smaller than that, because the warmest areas of the earth where the cloud feedback is greatest are generally not going to warm much.)
Now, I’ve stated above that this method gives us the long-term answer after almost all of the various feedbacks, slow warmings, and adjustments have occurred. I’ve stated that this is not the short-term response of the clouds to surface temperature. It’s the long-term, basically steady-state response.
As a result, it can actually answer the question about the long-term response of clouds to 1°C of warming. And it can answer the question in detail, showing how cloud feedback varies from the poles to the tropics.
The only argument that I can see against this is that some slow thermal adjustment from the most recent warming hasn’t arrived yet. Possible, but here’s why that will likely make little difference—a rising tide generally lifts all boats.
In other words, if we have several nearby gridcells and one gets a slow residual thermal adjustment from recent warming, in all likelihood the other nearby cells will get a similar slow residual thermal adjustment as well.
And this will leave the slope of interest, the slope of the yellow line in Figure 1, pretty much unchanged.
Or at least, that’s what my logic said. However, I’ve always preferred data to logic. After some thought, I realized I could test this by taking shorter averages of the CERES data instead of the full 21-year average. I used 5-year averages of the same CERES data. For comparison, I’ve plotted them to the same scale as in Figure 1.
Figure 3. LOWESS smooths of the scatterplots of four selected subsets of the CERES data. Underlying scatterplot data not shown.
As you can see, the LOWESS smooth trend lines of all four gridcell scatterplots are so close that they cover each other up. This definitely shows that a gridcell scatterplot is indeed showing the long-term, all-inclusive relationship between the two variables of interest. It’s barely affected at all by the changes in CRE and temperature between the 5-year periods.
I’ll leave this here, and I will return to what I’ve learned from other gridcell scatterplots in the next post.
My very best to all,
PLEASE: When you comment, quote the exact words you are responding to. I can defend my own words. I can’t defend your rephrasing of my words. Thanks.