16 September 2008

Sea Ice Packs

I've already mentioned types of sea ice, but that's only a bare scratch on the surface of the subject of sea ice. Another bit of vocabulary before diving in to today's sea ice: a chunk of sea ice is called a 'floe'. Not a flow, nor a sheet, a floe. Ice sheet is something quite different.

When we get a bunch of floes together, we start to have an ice pack. Three terms come up for describing a region of the ice pack (or maybe the entirety): concentration, area, and extent. Ice pack area makes the most intuitive sense -- add up the area of all ice floes, and that's the area of the ice pack. Concentration and extent are a little more removed. For concentration, draw a curve around some region you're interested in. Then divide the area of sea ice by the total area of the region bounded by your curve. Two common 'curves' used in the science are the footprint of a satellite sensor, and the area of a grid cell. The latter is what you'll see presented on any of the graphics at the sea ice sites I link to. For extent, you then take your grid and for every cell that has more than some concentration (which you'll specify), you add up the area of the entire grid cell. Extent will always be greater than area.

The usual concentration cutoff, and the one to assume if it isn't specified, is 15%. Below this, the ice is not reliably detected by the most commonly-used sensors, and it is a greatly smaller practical problem for ships. Not that ships appreciate bashing in to ice floes, but that at this concentration or lower, it can be manageable to move around them (and get out of the ice pack you were surprised by!).

The most common type of sensor to use for detecting sea ice from space uses passive microwaves. The ice (it turns out) emits microwave energy much more effectively than the ocean around it. This gives it a higher brightness temperature. Between that and some other details, we can get back an estimate of the concentration of sea ice that the satellite was looking at. A word, though, as we're coming out of summer: the method relies on the difference between ice and water. If you have ponds of water sitting on the ice floes, which can happen on thick ice such as the Arctic can have, then your concentration (area) estimate will be biased low. The extent is probably still not too bad. The reason is, by the time you're falling below 15% cover, the thick floes will have been storm-tossed enough that the ponds will have been emptied, or that it's late enough in the season that the melt pond melted its way through the ice floe and there really isn't any ice under the apparent water any more.

In looking at the NSIDC and Cryosphere Today pages on the Arctic melt, one thing to keep in mind is that one uses extent and the other uses area. Their numbers aren't directly comparable. They also differ in how they compute their estimates, in that one uses a longer averaging period than the other. The longer period gives you more confidence about the value (weather over the ice, or ocean, can give you false readings, but it moves pretty fast compared to the ice cover), but will miss some of the details in time.

More to come ... (bwahaha) But, in the mean time, questions you have about sea ice are welcome here.


Anonymous said...

Two questions:

The role of ice thickness and our ability to measure it.

Maslowski's RCM results. I assumed they must have made the rounds, but then in a recent communication posted at DotEarth Marika Holland seemed to be saying she wasn't aware of them. (Specifically she said she didn't know where the 2013 summer sea ice loss projection came from). There's a related question about what the GCMers are doing to adjust to the recent sharp trend, noting Maslowski's claim that his success was due to his RCM being able to properly track the warming currents.

Great blog, BTW.

Robert Grumbine said...

The role of ice thickness in what? Not being disingenuous here; there are many roles played by ice thickness. I'll be going on at more length in another post. In the mean time, one role is that thinner ice is less predictable. See Grumbine, Robert W., The thermodynamic predictability of sea ice, J. Glaciology, 40, 277-282, 1994, for some details on why.

Our ability to measure thickness has generally been far behind our ability to observe extent and area. The first method was to put upward looking sonar on nuclear subs cruising the Arctic. Such data have been collected since the 1950s, and were declassified in the Gore-Chernomyrdin agreement in the 90s. The data were collected with fighting World War III in mind, rather than climate science, so many details aren't as we'd prefer for the science. Nevertheless, it does provide a notion, and that notion is that ice had thinned greatly between the 50-70s and 90s, from over 3 meters to 1.8. Rothrock, Yu, and Maykut, GEOPHYSICAL RESEARCH LETTERS, VOL. 26, NO. 23, PAGES 3469-3472, DECEMBER 1, 1999.

The next realm of getting a grip of Arctic ice thickness was trying to use satellite altimeters designed for open ocean purposes (ERS-1, for instance). The difficulty here is that most of the ice is below the sea surface, and much of what is above the surface is snow rather than ice. Seymour Laxon is one of the people who has worked on that and the one I have talked with about it. This method was limited to ice thicker than about 1 meter, which lets us know something about the multiyear ice in the Arctic, but little about the Antarctic or the half to 3/4ths of the Arctic that is first year ice.

IceSAT, now flying (in fact old for its 3-5 year design life), includes a laser altimeter designed for ice measurements. H. Jay Zwally is one of the names to look for on this. The data are available at http://nsidc.org/data/gla13.html.

That leaves us with our usual sort of choice -- a longer record with worse coverage, a medium record with significant resolution limits, or a record with relatively high resolution (in space, time, and thickness) but very short duration.

Marika's exact wording was:
"(although I’m not entirely clear on how people have come up with a 2013 date)"

That makes sense whether she's read Wieslaw's work (as much of it as I've found myself) or not. And, from my current run through google scholar (not authoritative, but not too bad), it looks like he still doesn't have the full work published in the literature. (Could you provide a link to him using an RCM? The abstracts I found involved his usual high resolution coupled ice-ocean model. RCM would be an atmospheric model, Radiative-Convective Model and not the best thing for a serious prediction of Arctic ice cover)

'entirely clear' is one of those phrases of art, I'm afraid. It translates to normal English better as 'I don't know exactly how the original source arrived at its conclusions.' Marika and I have been at meetings that Wieslaw was presenting at (and vice versa) so we certainly have at least a general idea of what his work is like. The 2013 estimate specifically, though, I only have details by way of a power point presentation. It really says little beyond the fact that if he uses his more or less usual model in more or less predictive ways, he gets seasonally ice free Arctic as early as 2013. I wouldn't say from this that I know exactly where the 2013 date comes from myself. The emphasis above is mine, of course.

Alastair said...

It is very easy to get a date of 2013!

All you have to do is to extrapolate Rothrock et al.'s figures.

This extrapolation assumes a linear trend, which one would expect to be an under estimation since there is a positive feedback from th ice albedo effect.

(I am still busy translating de Saussure so don't have time to expand further:-(

Cheers, Alastair.

Robert Grumbine said...

Do let me know when you finish De Saussure.

But maybe you can take a minute to say how you got the 2013 from Rothrock et al. If I look only at table 2, and not very carefully, I can get that (1.6 m average in 1997, with a trend of 0.1 m/yr thinning). The not very carefully part is to ignore that the signs are mixed, and the net trend involves the small difference between large numbers. Further up the page, last paragraph before Discussion and Conclusions, the authors note (about table 2) "These trends are not well defined by only three cruises, but that are significant at the 80 to 90% level overall and in four of the six regions, and they are netagive in three of those four." As a matter of being consistent with an earlier, better-established, set of observations, we can use such a low significance level. The paper doesn't hang on that point. The real point is table 1.

Table 1 is where Rothrock, Yu, and Maykut are looking at the more climate-oriented variables, as I was suggesting in evaluating a climate trend. Now from this, the averages center on 1967 and 1995, at 3.1 m and 1.8 m respectively. 28 year span, we'll hope that the two figures represent climate numbers (a better chance than drawing a trend line through 3 cruises across 4 years), 1.3 meters thinning. A bit less than 0.05 m/year for the thinning trend.

Using this rate, it's another 36 years to melt away the 1.8 meters average of 1995, so a 2031 disappearance date. This does make the assumption, which is, as you note, dubious, that the melting goes straight line, rather than having any accelerating feedbacks. And we do expect feedbacks.

If there were more and better data on the recent side in this paper, we could argue that the melt rate was accelerating. But ... there aren't more and better data here, and 4 years is almost certainly short for establishing a climate trend in sea ice. With IceSAT, we might be able to reopen the question. (Actually, I'm sure it can be done, if only because it'll give us more data another decade later than the cruises Rothrock, Yu, and Maykut were using.) What the answer will be, I'm not so sure of.

You might be right about the table 2 reading of this paper being the source of 2013. I think it's Maslowski. But maybe it's a combination.

To back up, the 2030 date was, itself, quite the surprise at the time, and remained so until ... well, last fall. Models and extrapolations of the areal coverage trends consistently gave dates of 2050-2060. The daringly pessimistic finally lowered that to 2040-2050 before the melt season began. Only with the melt season well underway did I see Maslowski's 2013 estimate for earliest ice-free.

Alastair said...

It will take more than one minute to produce a well reasoned case that is fully referenced, but you may find this interesting.

As you point out in 1997 the average thickness was 1.6 m and the average rate of melt was 0.1 m per year. At that rate all the ice would melt in 16 years which added to 1997 gives 2013 QED :-)

That paper was not published until December 1999, so the 40% melt that they estimated would have already been exceeded if it had continued at that rate.

On sunday Dr Iain Stewart on the BBC was still saying that nearly half the arctic sea ice had melted, but it is now over ten years after the time when those figures applied! Of course he may have been referring to the change in area. But that only goes to prove that the melt has continued. with the thickness halved and the area halved, then the volume must now be at most a quarter of the 1975 average.

It is understandable that in 1999 Rothrock et al. should have played down the significance of their findings, but since then the apocryphal evidence is that the ice has continued to thin e.g. open water at the pole on a cruise ship that passed no thick ice on its voyage, planes landing on the ice and then sinking, Russian scientists forced to abandon ice stations early, and failure of polar adventurers to reach the pole because of melting ice.

Rothrock et al were not the first to raise the alarm over the ice. McPhee et al had preceded them http://adsabs.harvard.edu/abs/1998GeoRL..25.1729M

And here is an early Maslowski pub. http://www.ametsoc.org/atmospolicy/documents/May032006_Dr.WieslawMaslowski.pdf
from May 2006!

I could go on but it will take more than a minute.

Cheers, Alastair.

Anonymous said...

Thanks for your response, Bob. Actually I would have been a lot more specific with that first question, but I thought you were just asking for future post topics. One specific I'm interested in is whether it seems likely that the minimum ice volume this year is lower than last year's.

Re Maslowski, the Beeb article from last December referred to it as a high-resolution "regional model," which I took to be the same as an RCM. Is that not right?

Anyway, here's the AGU fall meeting abstract in case you haven't seen it:

Understanding Recent Variability in the Arctic Sea Ice Thickness and Volume - Synthesis of Model Results and Observations

Whelan, J (jwhelan@nps.edu), Naval Postgraduate School, Department of Oceanography 833 Dyer Road, Monterey, CA 93943, United States * Maslowski, W (maslowsk@nps.edu), Naval Postgraduate School, Department of Oceanography 833 Dyer Road, Monterey, CA 93943, United States Clement Kinney, J L (jlclemen@nps.edu), Naval Postgraduate School, Department of Oceanography 833 Dyer Road, Monterey, CA 93943, United States Jakacki, J (jjakacki@iopan.gda.pl), Institute of Oceanology, Polish Academy of Sciences 55 Powstancow Warszawy, Sopot, 81-712, Poland

"We examine the diminishing sea ice thickness trend in the Arctic Ocean using results from the NPS 1/12-degree pan-Arctic coupled ice-ocean model. While many previous studies have analyzed changes in ice extent and concentration, this research focuses on ice thickness as it gives a better indication of ice volume variability. The skill of the model is evaluated by comparing its ice thickness output to actual sea ice thickness data gathered during the last three decades. This includes the model comparison against the most recently released collection of Arctic ice draft measurements conducted by U.S. Navy submarines between 1979 and 2000. Our model indicates an accelerated thinning trend in Arctic sea ice during the last decade. This trend is robust and independent of timescales for surface temperature and salinity relaxation. The validation of model output with submarine upward-looking sonar data supports this result. This lends credence to the postulation that the Arctic is likely to be ice-free during the summer in the near future."

A quick search also finds this article, which says there will be a paper out by the end of this year and notes that M. is being close-mouthed about it until then. That would seem to explain why nobody has seen his results.