Arctic Layer Cake

Guest Post by Willis Eschenbach

There’s a recent paper paywalled here, called Arctic winter warming amplified by the thermal inversion and consequent low infrared cooling to space. Fortunately, the Supplementary Online Information is available here, and it contains much valuable information. The paper claims that during the arctic winter, the atmospheric radiation doesn’t go out to space … instead it is directed downwards, increasing the surface warming.

Now I haven’t figured out yet how that works, radiation being “directed downwards”. But that’s what they say. From their Abstract: 

We find that the surface inversion in fact intensifies Arctic amplification, because the ability of the Arctic wintertime clear-sky atmosphere to cool to space decreases with inversion strength. Specifically, we find that the cold layers close to the surface in Arctic winter, where most of the warming takes place, hardly contribute to the infrared radiation that goes out to space. Instead, the additional radiation that is generated by the warming of these layers is directed downwards, and thus amplifies the warming. We conclude that the predominant Arctic wintertime temperature inversion damps infrared cooling of the system, and thus constitutes a positive warming feedback.

Hmmm … so their basic claim is that the (poorly named) “greenhouse effect” is strengthened by the temperature inversion in the winter, that this slows the surface cooling, and that as a result the surface ends up warmer than it would otherwise be. A second claim is that the cause of additional Arctic winter downwelling radiation at the surface is a temperature inversion. The third claim is that this Arctic inversion is not unusual, but that there is a “predominate” winter temperature inversion in the Arctic.

Now, all of these claims can be investigated using the CERES satellite radiation dataset. To look at their first claim, I thought I’d follow the lead of the estimable Ramanathan and consider how much of the upwelling radiation from the surface is absorbed during the Arctic summer versus the Arctic winter. Ramanathan proposed the use of this atmospheric absorption of surface radiation as a measure of the strength of the greenhouse effect. Obviously, the more upwelling longwave that is absorbed by the atmosphere, the warmer the surface ends up. Figure 1 shows the strength of the greenhouse effect using Ramanathan’s measurement (absorbed radiation as a percentage of surface radiation) in June and in December.

Figure 1. Strength of the poorly-named “greenhouse effect”, as measured by the percentage of the surface upwelling longwave radiation (thermal infrared radiation) that is absorbed by the atmosphere. The situation is shown for the month of June (upper panel) and December (lower panel). Following Ramanathan, the absorbed radiation is calculated as the upwelling surface radiation minus the upwelling TOA radiation.

As you might imagine, and can see in Figure 1, the greenhouse effect is strongest where there is water. As a result, the effect is strongest in the tropics, and is stronger over the ocean than over the land. For the same reason, the greenhouse effect is weaker over the deserts and at the poles.

Now, their claim is that there is additional greenhouse warming in the Arctic in the wintertime compared to the summertime, slowing the radiative cooling of the surface. However, the CERES data disagrees, and indeed it shows the opposite. The CERES data says that at both poles, the greenhouse effect is stronger in the summertime, not weaker. This makes sense, because there is more water vapor in the air in the summer.

Note also that while there are areas of temperature inversions (shown in blue), and they do occur in a few areas in the Arctic winter(lower panel), they are not a general feature of the Arctic. On the other hand, large areas of the Antarctic do have a temperature inversion in winter (upper panel, blue). 

So the CERES data doesn’t agree with the study regarding the slowed cooling in winter. The CERES data says the opposite, that cooling is easier in winter because less upwelling surface longwave is absorbed by the atmosphere. Nor does the Arctic temperature inversion seem to be as widespread or pervasive as the authors state.

Next, they claim increased downwelling longwave at the surface in the Arctic winter. To investigate this claim, Figure 2 shows the June and December downwelling longwave surface radiation, once again as a percentage of the upwelling longwave surface radiation.

Figure 2. Downwelling surface longwave radiation as a percentage of the upwelling longwave surface radiation, June (upper panel) and December (lower panel).

The main oddity in Figure 2 is that most places, most of the time, the downwelling radiation is about 86-88%, with not much difference summer to winter or place to place, particularly in the ocean. I wouldn’t have guessed that. Note that Figure 2 also reveals the widespread winter temperature inversion in the Antarctic winter (upper panel, red) indicated by downwelling longwave radiation exceeding upwelling surface radiation, and the lack of such a widespread inversion in the Arctic winter (lower panel, red).

More to the current point, we have a curiosity related to the authors’ claims about the Arctic. Note that in Antarctica in the wintertime (upper panel) there is a marked increase in the downwelling radiation as a percentage of the surface radiation compared to their summer (lower panel). The difference is large, 98% versus 64%. Presumably, this is the increased downwelling that they describe in their paper (although as expected the upwelling also increases).

But in the Arctic, where the paper claims this phenomenon of increased downwelling radiation is occurring, there is no difference between the downwelling surface longwave in the summer and the winter (88% in both cases).

So we do in fact find the phenomenon they point to of increasing downwelling radiation  … but we don’t find it in the Arctic as they claim, we find it at the opposite pole.

Summary

1. Their claim, that there is “reduced cooling” in the arctic in wintertime that affects the surface temperature, is not supported by the CERES data. To the contrary, the CERES data shows the Arctic radiative cooling is much more rapid in the winter than the summer, because the atmosphere is absorbing much less radiation. Note that this is what we’d expect, due to the reduced amount of water vapor in winter.

2. Their claim, that the Arctic temperature inversion is widespread, is not supported by the CERES data. It shows general wintertime temperature inversion in the Antarctic, but not in the Arctic.

3. Their claim, that the Arctic downwelling longwave radiation increases in the winter, is not supported by the CERES data. Curiously, it is true in the Antarctic. In the Arctic, however, there is almost no difference between summer and winter.

Now, how did they get this so wrong? From their methods section (emphasis mine):

An often used method to increase the signal-to-noise (i.e. climate change- to-variability) ratio is to study multi-model output, such as those obtained in the CMIP3 initiative for ‘realistic’ forcing scenarios. The general idea then is to apply statistics on the multitude of independent members (individual models) to reduce the noise, and also to use intermodel differences to relate climate processes to feedbacks2.

Another method, the one employed here, is to use one climate model and apply a sufficiently large forcing (e.g. 2xCO2) to obtain a climate change signal that is much larger than the noise. The advantage of this approach is that dedicated experiments can be carried out, including changing certain model processes in order to link these to feedbacks (as is done in this study).

So … as usual, rather than mess with ugly observational data, it’s models all the way down. Actually it’s worse, it’s the output of one single solitary model all the way down. Or as a typical adulatory media report of the story says:

Pithan and co-author Thorsten Mauritsen tested air layering and many other Arctic climate feedback effects using sophisticated climate computer models.

Hey, as long they used a sophisticated climate model, and it is reportedly “based on true physics” in the best Hollywood tradition, what’s not to like?

Best to everyone,

w.

The Usual Request: If you disagree with something I say, please quote my exact words so we know what you are referring to. I can defend my own words. I cannot defend some vague claim like “Willis, your logic is wrong”. It may well be … but we’ll never find out unless you quote exactly the logical claims I made that you don’t like.

Data and Code: CERES calculated surface data (in R “save()” format) is here, 110Mbytes. and the CERES measured TOA data is here, 230 Mbytes. CERES Setup.R and CERES Functions.R are needed for the analysis. Finally, the code for this post is Arctic Amplification.R

Also, it’s worth noting that while the CERES top-of-atmosphere data is from measurements, the surface data is calculated from the TOA data using energy balance considerations. Obviously, a global set of observational surface radiation data would be wonderful … but since we haven’t got that, the CERES data is the best we have.