Arctic Albedo Variations

Guest Post by Willis Eschenbach

Anthony has just posted the results from a “Press Session” at the AGU conference. In it the authors make two claims of interest. The first is that there has been a five percent decrease in the summer Arctic albedo since the year 2000:

A decline in the region’s albedo – its reflectivity, in effect – has been a key concern among scientists since the summer Arctic sea ice cover began shrinking in recent decades. As more of the sun’s energy is absorbed by the climate system, it enhances ongoing warming in the region, which is more pronounced than anywhere else on the planet.

Since the year 2000, the rate of absorbed solar radiation in the Arctic in June, July and August has increased by five percent, said Norman Loeb, of NASA’s Langley Research Center, Hampton, Virginia. The measurement is made by NASA’s Clouds and the Earth’s Radiant Energy System (CERES) instruments, which fly on multiple satellites.

The second related claim is as follows:

Kay and colleagues have also analyzed satellite observations of Arctic clouds during this same 15-year period. Kay’s research shows summer cloud amounts and vertical structure are not being affected by summer sea ice loss. While surprising, the observations show that the bright sea ice surface is not automatically replaced by bright clouds. Indeed, sea ice loss, not clouds, explain the increases in absorbed solar radiation measured by CERES.

Since I have the latest CERES data on my computer, I figured I’d see what they were talking about.

Now, it’s not entirely clear from the presentation which dataset they’ve used. Bear in mind that there are two CERES datasets: top-of-atmosphere radiation observations, and surface radiation calculations. Albedo is calculated from the observations, it’s reflected sunlight divided by incoming solar.

On the other hand, they also talk about “absorbed solar radiation” which is only available in the calculated datasets.

So let’s start with the claim that the Arctic albedo has decreased since the year 2000. I assume that they are using the normal definition of “Arctic”, which is above the Arctic Circle at about 66.5° north.

There are several difficulties with albedo near the poles. First, when the total solar input is quite small, the numbers get inaccurate, since it is a ratio and the denominator, the solar input, is near zero. In addition, the numbers are also inaccurate because there’s so little reflected sunlight, which makes both the top and bottom of the ratio quite variable. Finally, it’s difficult to convert the changes in albedo into watts per square metre (W/m2), which is a much more meaningful number.

So let me look at a simpler measurement—how much sunlight is reflected, measured in W/m2. We have three top-of-atmosphere (TOA) observational datasets for that, which are the reflection regardless of the state of cloudiness (called “toa_sw_all”), the reflection from the ground when the sky is clear (called “toa_sw_clr”), and the reflection just from the clouds (called “cre_sw”). Figure 1 shows the first of these, the total sunlight reflected in all conditions:

Figure 1. Total reflection from the Arctic. This includes both cloud and ground reflections. The top panel shows the raw data, with a dotted line showing the average value. The middle panel shows the seasonal component, which is also called the “climatology”. The bottom panel shows the “residual”, which is the difference between the top and middle panels.

I note that indeed the reflections have gotten smaller over the period, meaning that the amount absorbed is larger as the press release stated. The next graph, Figure 2, shows the ground reflections only. Note that as you’d expect, these are less than the total reflections:

Figure 2. “Clear sky” (ground only) reflections from the Arctic. This shows only ground reflections. The top panel shows the raw data, with a dotted line showing the average value. The middle panel shows the seasonal component, which is also called the “climatology”. The bottom panel shows the “residual”, which is the difference between the top and middle panels.

Finally, Figure 3 shows the results from just cloud reflections. Note that rather than decreasing, the cloud reflections are increasing. They are also smaller than the ground reflections.

Figure 3. Cloud reflections from the Arctic. This shows only the effect of the clouds. The top panel shows the raw data, with a dotted line showing the average value. The middle panel shows the seasonal component, which is also called the “climatology”. The bottom panel shows the “residual”, which is the difference between the top and middle panels.

So … there are the three graphs: total reflections, ground reflections, and cloud reflections. So how well does this agree with the claims of the press release?

Now, to start with they’ve done something strange. Rather than look at the changes over the whole year, they’ve only looked at three months of the year, June, July, and August. I disagree strongly with this kind of analysis, for a couple of reasons. The first is because it allows for nearly invisible cherry picking, by simply choosing the months with a particular desired effect. The second is that it makes it hard to determine statistical significance, since there are 12 possible 3-month contiguous chunks that they could choose from … which means that you need to find a much greater effect to claim significance.

So I’m not going to follow that plan. I’m looking at what happens over the whole year, since that’s what really matters. The first point of interest is that the total amount reflected from the Arctic (Figure 1) has indeed decreased over the period at a rate of a quarter of a watt per square metre (-0.25 W/m2) per decade … for a total drop in reflected solar of

 -0.025 W/m2/year * 14 years = 0.35 W/m2

A third of a watt per square metre? All of this hype in the press release is to announce that there’s been a change in Arctic reflections of a third of a watt per square metre in fourteen years??? Be still, my beating heart … that’s a whacking great 1% change in the already small Arctic solar absorption in fourteen years. This is their big news? Now please note, their claim about the change in June, July, and August of a 5% change may indeed be true … but that just emphasizes why that kind of analysis is just cherry picking.

Not much else to say about it once I’ve said that … well, except to say that their claim that “the bright sea ice surface is not automatically replaced by bright clouds” also doesn’t seem to be true. Note that the blue line in the bottom panel of Figure 3, which shows the smoothed changes of cloud reflections, is pretty much a mirror image of the corresponding line in Figure 2 showing the smoothed changes in ground reflections. In fact, the correlation between the unsmoothed cloud and ground residuals is -0.71, with a p-value less than .001. In other words … they’re wrong. The cloud changes do not entirely offset the ground changes, but the bright clouds assuredly move in total opposition to the bright sea ice surface.

Finally, I would note that from 2000 to 2010, the total reflection from the Arctic drops by about one W/m2 (blue line, bottom panel, Figure 1). Then in two years it drops another one W/m2 … and in the next two years it rises by one W/m2. As a result, I’d have to conclude that while these changes may have statistical significance, they may not mean a whole lot …

Regards to all,

w.

PS—If you disagree with someone, please have the courtesy to QUOTE THEIR EXACT WORDS so that we can all understand the nature of your objections.

CODE AND DATA: The R code and functions are here in a 14 Kb zipped folder . The CERES TOA data is here and the surface data is here. WARNING: BIG data files, 200 Mb plus.