Arctic Sea Ice: More than Just the Minimum

The extent of sea ice in the Arctic has dropped dramatically. Here’s the data from NSIDC (the National Snow and Ice Data Center):

The thin black line shows daily values of sea ice extent. Clearly it rises and falls each year, with more ice in winter/early spring and less during summer/early fall. The red dots show the annual maximum values, the blue dots the annual minimum values, and the brown dots show the yearly average values.

The minimum values tend to get more attention than either the maximum or yearly average values. Some people even go so far as to ignore most of those, instead talking only about minimum values from 2007 onward (what’s inside the red box):

We have data for each day over nearly 40 years, we have annual averages for 39 years, and both yearly maxima and minima for 40 years. But for some reason, some people (as in, most climate deniers) only want to talk about 12 minima. Why do you think that is?

I’d like to try something I find interesting. The downward trend in Arctic sea ice was detectable by the end of the 1900s, and we’re interested in whether or not the situation has improved since then, so I’ll take trends estimated from pre-2000 data and extend them up to the present, to see how observations compare to what we would have predicted from the existing linear trend.

We’ll start with the yearly minimum values:

Since 2000, the situation didn’t get better. It got worse.

Yet climate deniers only like to talk about the minimum values since 2007, and they love to declare a “recovery!” in Arctic sea ice. Look at the above graph again: does that look like a recovery to you?

It’s a classic case of cherry-picking — selecting only the data that make you case look good and ignoring the rest. They also don’t generally run valid statistical tests to establish that some real improvement happened. That’s probably because such tests fail.

Here are the yearly maximum values:

There’s definitely no improvement since 2000. It appears to have followed the pre-existing trend pretty closely, and note that the two lowest values are in the last two years, the four lowest values are in the last four years.

Here are the yearly average values:

Again, no improvement, no recovery. The two lowest values are in the last two years, the three lowest values in the last three years. It’s not better than it would have been if it had followed the pre-existing trend; it’s worse.

Arctic sea ice is important for many reasons. For one thing, when ice changes to open water it absorbs more of the sun’s incoming energy, reflecting less back to space, and that warms up the planet even more — it’s one of the classic feedbacks in the climate system. For another thing, removing sea ice allows more heat transfer from ocean to atmosphere in the Arctic, and that dramatically changes wind circulation patterns. It seems to be making the jet stream more wavy, which allows weather systems (including unpleasant ones like heavy rain or heat waves) to be more persistent so bad conditions hang around longer. The effect is profound, and it’s worldwide. Arctic sea ice is something we should talk about more often.

But when climate deniers talk about Arctic sea ice, they paint a rosy picture of “recovery.” It’s a sham, based on blatant cherry-picking and statistical incompetence. It’s also a threat to all of us who are affected by it. Let’s not let the conversation be dominated, or even polluted, by the ignorant and misleading pollution coming from the mouths of climate deniers.

This blog is made possible by readers like you; join others by donating at My Wee Dragon.


6 responses to “Arctic Sea Ice: More than Just the Minimum

  1. Yes, I remember when Judith Curry, who likes to think of herself as a scientist, used the R word (recovery) when the minimum in 2013 was much higher than the record minimum for 2012. This is supposedly a scientist using a single data point to pronounce a recovery. Mind, you, catastrophists (like the mythical Sam Carana) do this kind of thing also, the opposite way. I think both of these extremes are distinctly unhelpful.

    There’s another story to Arctic sea ice, the thickness and the amount of multi-year ice. I understand that both of these variables are looking extremely bad right now (particularly multi-year ice). Another big drop for the record minimum seems to be just a matter of time, a very short time.

    • rhymeswithgoalie

      [Judith Curry] used the R word (recovery) when the minimum in 2013 was much higher than the record minimum for 2012.

      Aye, she conveniently forgot to mention the 2012 Arctic Ocean storm which pushed a lot of the fragmented sea ice to one side and significantly reducing what’s measured as sea ice extent (one such measure is “at least 15% of that grid component has ice”).

      There are, of course, different aspects of sea ice extent. Thin ice makes icebreaker ships unnecessary. Unconnected clumps of ice allow much more heat to pass up to the atmosphere than the same area of contiguous ice sheet. Megafauna (bears and walruses) lose useful floes well before an area is considered ice-free.

      As far as AGW is concerned, I’m more interested in volume than extent, and will take this excuse to post Andy Lee Robinson’s 2018 update of his wonderful sea ice volume visualization.

  2. Why does this remind me of how Republicans are now praising the economy because the stock market prices are high and the unemployment rate is low, conveniently ignoring the exploding deficit, and poor wage growth?

  3. quaesoveritas

    What is a “climate denier”?

    [Response: It’s a common figure of speech referring to climate change deniers. Let’s not have this discussion again.]

  4. Whack a mole here, whack a mole there…

    The flavor of the year this melt season was that ‘the ice extent is above the average for the 2000s’–which was true for much of the later part of the season, thanks in part to a coolish, cloudy July that muted the melt during that time. (You can see the July ‘bend’ in the extent curve, below.)

    You can also see that for the first three months of the year, extent was at records lows for the date, before falling behind the melt pace of the 2016 season, and then, in July, the 2012 season, which still holds the record low minimum.

    What this exercise shows is that there is very considerable variability on monthly time scales, and that that variability strongly affects the rankings of individual minima–which, as pointed out in this post, tends to get the most ‘buzz’, given the human attraction to clear-cut benchmarks (especially for purposes of writing headlines!) And the graph of annual minima in the post* suggests that this variability has become greater in the post-2005 era. Witness, for example, the swing of ~2 million km2 from 2012 to 2013, referenced in the comment above, or the contrary one from 2006 to 2007.

    What’s been haunting me a bit this fall has been the question of what this means for the projected date of the first ‘ice-free’ minimum–often defined as a minimum for which extent is below 1 million km2. Right now, we seem to be in a regime in which the mean minimum value is around 4.5 million km2, and is declining. What I’ve been thinking is that when that value approaches 2 million, we’ll be in a state where any given season has a considerable probability–say, 10%, since most one-year swings are smaller–of reaching the ‘ice-free’ range.

    Right now, it seems to me that we’re in that condition with respect to a ‘shockingly low’ new record–that is, any year now we’re quite likely to hit a minimum around 2.5 million, but probably not ‘ice-free’. Which means that one would like to be able to characterize the post-2005 rate of decline accurately–unfortunately hard to do, with just a decade or so of very noisy data. (Assuming that there has been a regime change, which the data certainly seem to suggest–likely associated with the observed transition from high to low proportions of multi-year ice.)

    *By the way, the data points appear to be misaligned with the year labels on the plot of minima–eyeballing the latest point would lead you to think it was 2019, not 2018, and the then-record lows of 2005 and 2007 appear to be a year later, too. Data coverage began during 1979, so the first minimum (unlike the first maximum) occurred during the former year–not 1980.

    [Response: I plotted the minimum values at the *times* of minimum, which are not the integer part of the year but the latter half of the year. The latest minimum (for 2018) happened closer to 2019.0 than to 2018.0. Perhaps it would have been better to plot minimum values at the integer part of the year, to avoid confusion.

    I recorded those times of minimum (and maximum) to look for a trend in what day of the year the minima/maxima were occurring, but found none.]

  5. Thanks for the clarification. That makes sense, though not a sense I would have (or did!) anticipate. And it’s interesting to know that there is no good evidence for a trend in the timing of minima (particularly since this year was quite late.) Perhaps that’s not surprising, though, in the sense that there are competing effects going on: as extent drops, the ice edge tends to move poleward, making sharper the fall transition from sun to dark.

    I did some playing around on woodfortrees. Since the autumn of 2004, the average linear trend (based on the monthly values wft uses) is something like -0.7 million km2. The mean extent for the last year is 10.33 million, so it would seem that the (naive) linear projection to intercept the zero line would be 15 years, or 2033. (Within the range of model projections, I note, despite the naivete involved.)

    But part of my naivete is that I’m not too clear on just what the zero intercept of the mean annual value actually implies physically. (Ice-free for that part of the year for which typical current values are below the current mean?) So I turn to a more straightforward metric, which is the value at minimum. Currently, we seem to be in the 4.5 million range, as mentioned previously. So, applying the same mean decline–I’m hoping for offsetting biases here–we get 6 years, or 2024. That would qualify, I think, as a pretty aggressive projection.

    For completeness, similar calculations based on the maximum–currently hovering around the 14 million mark–would yield 20 years, or 2038. Again naively, that would seem physically to translate to a year-round ice-free Arctic Ocean, but I think we can rule that projection out on physical grounds–and it’s definitely way out of line with model results. Basically, I think the boundary conditions are so different when we get to prolonged ice-free conditions that extrapolative techniques don’t have much value left.

    But what bedevils me about all this is that, sketchy as the foregoing already was, I really don’t know how to handle the uncertainties. If we assume the linear trend applies, and that a +/-2 million km2 swing is possible in any given year, then it would seem that three years plus or minus would be a pretty safe prediction window. (Quite Maslowskian-sounding!) But I suspect there are all kinds of issues lurking, having to do with arcana like probability density functions and autocorrelation that I understand only in a very basic way.