Arctic Sea Ice Decline

Reader “fiq” recently asked some questions about sea ice, in particular, whether or not Arctic sea ice decline can rightly be called “staggering.” That’s fine, they were good questions. But when he didn’t find the answers to his liking, he spent “a few minutes looking at the data” and decided to accuse climate modelers of institutionalized incompetence, liken me to a monkey who “sees a short stack of bananas on the next tree and reports to the other monkeys that all the bananas are going extinct,” and refer to my regular readers as “fawning uncritical thinkers,” the kind who “believes you without demanding evidence“.

Regular readers might well take offense at that. Regular readers also know my regular approach.

Let’s look at the data.

---

The question:

Let’s be crystal-clear what the question is:

I was looking at the Daily Sea Ice Extent Time Series chart at NSIDC this morning and thinking about what I learned from a statistics course. For the Arctic the chart depicts a current extent about ~1.5 sigma under the median. As a static data point, isn’t that quite normal? If so, how do we have a staggering decline if we seem to have a perfectly acceptable value in a normal distribution?

Here’s another take on it:

if the actual current data in the series appears to be within statistically normal boundaries why is it being characterized as staggering — using the relevant data (http://nsidc.org/data/seaice_index/ 1980-2010 average, +/- 2SD).

---

First things first

Let’s put that “~1.5 sigma” idea to bed. Here’s some artificial data (yes, this is directly relevant to the question):

No “staggering decline” there! And, none of the data points is more than 1.73 standard deviations from the mean (or median for that matter) so by the method of “fiq” all is within statistically normal boundaries.

But wait — I randomized the time order of the artificial data in that graph. How tricksy of me! Here they are in proper time order:

There’s a clear trend. But it’s still true that none of the data points is more than 1.73 standard deviations from the mean (or median for that matter) so by the method of “fiq” all is within statistically normal boundaries. Me, I’d call it “staggering decline.”

The presence of a trend invalidates the “within statistically normal boundaries” argument. Are we over that “~1.5 sigma” idea now?

---

Extent measured by satellites

Here’s the daily sea ice extent data from NSIDC (National Snow and Ice Data Center):

The two most visible features are a large seasonal cycle (more ice in winter/spring, less in summer/fall) and an overall decline (plainly visible even without the seasonal cycle removed!).

The annual cycle is so much larger than the secular trend that it makes the trend harder to see, although it is still visible. A common (nearly universal, in fact) approach is to remove the annual cycle to transform sea ice extent (or area or volume or whatever) into sea ice extent anomaly. As it’s almost always done, first we choose a baseline period (a “reference period” to define “climatology”), then we compute the average seasonal cycle within the baseline period, then we subtract the average seasonal cycle from the actual data values to define anomaly values. NSIDC doesn’t actually provide the numerical data for their anomalies (which they base on a 1980-2010 baseline), so I computed them myself. I chose the entire time span for my baseline (it’s not going to make any substantive difference, really), which gives us this:

Now the decline is rather plain to see. I’ll delineate the multi-year trend by smoothing the data with a modified lowess smooth. I’ll even add in the likely range of the multi-year trend with dashed red lines above and below the trend line (solid red):

Yes, the trend line is for real. Yes, there are other ways to do it (other choices of “time scale” for the smoothing, and gobs of other smoothing methods), but if you think you can erase the trend by doing so you’re mistaken.

One can even test the reality of the trend to find out whether or not it’s “statistically significant.” It is (99.9999% confidence). As a matter of fact, not only is it trending downward, the rate of decline has increased (99.9% confidence).

Here’s just the (quite real) trend:

Since satellite observations began, Arctic sea ice extent has declined by an area larger than the states of Texas and Alaska combined. For foreign readers, I’ll mention that those are the two largest states of the U.S. and could certainly be described as “big-ass.”

A decline that large is big deal, a huge deal, especially in light of sea ice extent before the satellite era (more about that, soon).

---

September minimum

The sheer size of the decline is not the only big deal. Sea ice varies strongly throughout the year, and change in the annual minimum value (which happens in September) has dramatic consequences, for ocean circulation, for atmospheric circulation (are we enjoying the havoc the jet stream has been wreaking lately?), for climate feedbacks (ice albedo!), for living things (polar bears, human beings), for coastal erosion along shorelines in the Arctic (which native Alaskan town is it that has to move — the whole town — because of erosion caused by summer sea ice loss?), even for commerce (opening the “northwest passage” and “northeast passage” to commercial ship traffic). It’s a big deal.

And what has happened to the annual minimum sea ice extent since satellite observations began? This:

Let’s graph that, not in millions of square kilometers but as the fraction of the 1979 value:

The 2013 minimum was 26% less than the 1979 minimum. Every year since 2007 has been lower than every year prior to 2007. In 2012, the annual minimum was less than half its 1979 value. I’d even say the decline is “staggering,” especially in light what happened before the satellite era (more about that, soon).

---

Volume

Of course ice is 3-dimensional, so also of interest are sea ice thickness and volume. In fact there’s a good argument to be made that sea ice volume is the most relevant metric physically, but I regard that as a matter of opinion — but then, I’m not a physicist.

As far as I know, we only have thickness data for the Arctic (not the Antarctic) and even that is spotty (I suspect, but I don’t know, that there’s more data from the military, specifically submarines, which hasn’t been made public yet). What is clear from the available data is that in addition to covering less area, and less extent, Arctic sea ice has also been getting thinner. Quite a bit. Because of that, volume has declined even faster than area or extent.

As for volume, despite the lack of continuous direct thickness measurements there is very good volume data — the PIOMAS data — from models which incorporate large quantities of observations and make a complete estimate throughout the satellite era, and has been verified with detailed data from the relatively new CryoSat mission. But, volume data throughout the satellite era is also only for the north, I don’t know of any volume data (not even model estimates) for the south.

Anyway, here’s sea ice volume (monthly averages in thousands of cubic kilometers) for the Arctic since 1979 according to PIOMAS:

Again there’s an obvious annual cycle. Again there’s a clearly visible decline. In fact “clearly visible” doesn’t quite capture just how obvious it is.

Here’s the annual average Arctic sea ice volume:

And here it is as a fraction of the 1979 value:

Look closely. The annual average (in spite of the very large annual cycle) last year was 47% less than its 1979 value, and in 2012 it was only half its 1979 value (in spite of the very large annual cycle).

As for the annual minimum, which is extremely important for many reasons (some already mentioned), here ’tis:

And here it is as a fraction of the 1979 value:

Last year’s value was 70% lower than the 1979 value. The 2012 value was 80% lower than the 1979 value. Is 80% decline “staggering”?

---

Before the satellite era

In addition to satellite measurements since about 1979, sea ice extent has been estimated more than a century into the past by careful study of historical records. There’s a surprising abundance of such data because of its importance to the safety of marine navigation. It is hardly as precise as satellite data, and the amount of information declines the further back in time one goes, but what we do have is nonetheless well documented.

Here’s one such estimate: Arctic sea ice extent at its annual maximum and minimum from Kinnard et al. 2008 (GRL 35(2), L02507, DOI: 10.1029/2007GL032507):

It illustrates the unusual nature of the recent decline. Here’s another such estimate, probably the most oft-used data set for this time period, the Walsh & Chapman data, which extends from 1870 through 2006. Here is annual average sea ice extent anomaly, which I’ve aligned with satellite data (shown in red):

This too illustrates the unusual nature of the recent decline, which emphasizes the magnitude of the difference between what we’re seeing now and what has happened in the recent past.

---

Before that

Estimates of Arctic sea ice extent have been extended even further back using proxy data to reconstruct sea ice. Here’s one of the most recent efforts, a proxy reconstruction of late summer Arctic sea ice extent for the last 1450 years by Kinnard et al. 2011 (Nature, 479, 509-512, doi:10.1038/nature10581):

There are clear variations over time, but the precipitous decline in the modern era sticks out like a sore thumb.

This is in accord with a thorough review of Arctic sea ice variations over a wide variety of time scales by Polyak et al. 2010 (Quaternary Science Reviews, 29(15–16), 1757–1778), who say “On suborbital time scales, ice distributions varied in the Holocene, but no evidence exists for large, pan-Arctic fluctuations.” Not, that is, until the last several decades:

The current reduction in Arctic ice cover started in the late 19th century, consistent with the rapidly warming climate, and became very pronounced over the last three decades. This ice loss appears to be unmatched over at least the last few thousand years and unexplainable by any of the known natural variabilities.

Have another look at the 1450-year reconstruction. Staggering?

---

Questions for “fiq”:

1) Of course “staggering” is a subjective term, and no matter how much Arctic sea ice has declined one could argue it’s not that. But given the evidence I’ve provided, I think it’s not at all unfair or misleading for me to apply that adjective. Agree? Yes or no?

2) For some unfathomable reason, this isn’t the first time I’ve been insulted by someone who disagrees with me. I’m a big boy, and I wouldn’t be blogging if I didn’t have a thick enough skin to take it. But my regular readers are not “fawning uncritical thinkers“, and rare indeed is the one who “believes you without demanding evidence“. In fact, when I trip up or get too lazy, they rake me over the coals. So: I think you owe them an apology. Agree? Yes or no?

They’re yes-or-no questions, and I’d appreciate straight answers before you elaborate on your reasons.