I’ve got a question for readers: can someone explain to me how the NSIDC sea ice projections were actually made? I can’t get from A to B. Relevant information is in their monthly reports for April 2008 and May 2008.
Remember how much we’ve heard about new ice and why this was expected to lead to 2008 ice meltback surpassing the 2007 record. Let’s try to follow their projection method step by step,
First, we start with March ice, which is the yearly maximum. They say:
Average sea ice extent for March 2008 was 15.2 million square kilometers.
Next we learn:
Figures 4 and 5 indicate that relatively thin, first-year ice now covers 72% of the Arctic Basin, including the region around the North Pole; in 2007, that number was 59%.
By simple multiplication, the area of new ice in March was 10.94 million km^2 (old ice: 4.26 million km^2).
Next we learn:
about 30% of first-year ice typically survives the summer melt season, while 75% of the older ice survives
Their projection using average survival rates is reported as follows:
If we apply the survival rates averaged over all years to current conditions, the end-of-summer extent would be 3.59 million square kilometers.
Can anyone explain this to me? If I multiply opening new ice (10.94 MM sq km) by 30%, I get 3.28 MM sq km; and multiplying opening old ice (4.26 MM sq km) by 75%, I get another 3.19 MM sq km, for a total of 6.48 million sq km – a figure which, by the way, looks like a pretty good estimate to me of where we might end up this year. If you inadvertently divide the surviving old ice by 10, you get something pretty close to 3.59 million sq km ( 3.28 + 3.19/10).
I don’t pretend to be familiar with these calculations and I could easily have been wrongfooted somewhere along the way. So if anyone can figure the calculation out, I’d appreciate it. BTW wouldn’t it be nice to be able to simply look at an archived script and be able to see for yourself what they did?
53 Comments
I thought as you did upon reading their description, but how does that begin to explain the chart which purports to use a year’s data individually to predict 2008. 3.59 msqkm appears to be their ensemble average.
Winds and currents mean more than ‘climate’. They finally admitted that in the fine print last year.
Steve I think you’ll find that the calculation is based on the April data.
Steve,
They applied the rate of Ice Loss based on previous years rates as shown in their Fig 4. Then took the average of those results to make the prediction.
So they calculated the observed changes for both first year and multi-year ice based on observations to arrive at a survivability percentage in each type of ice in each previous year, 1983-2007, and applied it to the current observed extents. This gave them the range from 4.55 – 2.22 Mkm2, they then took the average.
So the numbers they stated and you used were not how they calculated the estimate.
As side comment…
So it would seem to me the bet would be on which year 1983-2007 has the closest observed conditions (outside of extent) to the current year (2008). That would give you the best estimate based on the method illustrated.
Re #5
That would be hard to find I think, this year the Beaufort gyre has contracted considerably (probably connected with the major break up of the Beaufort ice pack this winter: http://ice-glaces.ec.gc.ca/App/WsvPageDsp.cfm?Lang=eng&lnid=43&ScndLvl=no&ID=11892).
This appears to have led to the transpolar drift moving much closer to the N american coast and not through the E siberian and Laptev regions giving the sea ice distribution we see now.
http://iabp.apl.washington.edu/maps_daily_arcticbasin.html
It wouldn’t be surprised to see a major re-distribution of the ice this year, ice free along the N. american side more ice on the asian side.
If you look at the paper I linked to in the other discussion, you’ll see that the r^2 on their predictor is quite bad. It seems reasonable to conclude that a linear model is wrong in this case.
Posit
1) new ice melts faster than old ice
2) more new ice as a fraction of the arctic means lower melt-rates for old ice and new-ice (on average)–due to distributional effects
This would consistent with a linear model having low r^2.
Steve,
I guess we both missed those studied that show that new ice has a smaller heat of fusion than old ice. Once you have the physics right it all makes sense. 🙂
New versus old ice is a contrived excuse mechanism for the alarmists. Once a cooling phase begins all ice accumulations are “new”. I sense a contrived talking point.
Re #8
You also missed the fact that multiyear ice is thicker and has a different albedo.
And you should all know that moshpit guessed 6.5 a couple months ago over
at lucias.
So there.
Steve, I haven’t read all(any) of the responses so far, but I do have these observations:
1. There is an abundance of unrelated numbers here.
2. Sea ice extent in March, 2008 was 15.2 10E6 km^^2. This is not the “Arctic Basin”, it is an absolute number for the area of ice.
3. 72% of the “ARCTIC BASIN” was new year ice. Not % of ice but of the “Arctic Basin”. How big is the Artic basin? Multiply that by 0.72 and you get the amount of “new ice”. Subtract that number from 15.2 10E6 and that is the extent of “old ice”.
4. Next year’s “old ice” will be (0.30 * 0.72 * area of “Arctic basin”) + (0.75 * the result from step (3)).
I suspect that the authors of this report are deliberately confounding the numbers to get more press, thus nore spondulics.
At the risk of making myself look rediculous, I proceed in the general direction. We can extract the following equations, given constant ice:
A = the Arctic Basin area
A * 0.72 >= 15.2 10E6
Old Ice OI = 15.2 10E6 – A * 0.72
also, we are told:
0.216 * A + 0.75 * OI = new OI
Now assuming constant ice coverage:
OI = 0.216 * A / 0.25
I am off to bed now. I bet these equations don’t equate. But, hey, this is climate science, right?
Just to throw my hat in the ring I ran some numbers, I used a last 10 year average (61% Loss) and the 2007 loss rates (71% Loss) and I get a range of 4.437 – 5.982 Mkm2 September Minimum.
Based on the March-September Extent data.
Not very scientific but close enough for a guess.
Putting it here just to record it for posterity.
Why focus on graphs showing sea area with more than 15% sea ice? Why not 30% or 85% or whatever? If the wind blows outward then broken ice will be spread out then this area will fall. If the wind then blows inward that same ice will blow inward and the area will fall but at a different rate. This says nothing about melt (does it not) just distribution of ice over a lesser or greater domain depending on the wind and currents.
From the ALFRED-WEGENER-INSTITUTE in Germany Arctic sea ice very likely will not reach last year’s level.
http://www.awi.de/de/aktuelles_und_presse/pressemitteilungen/detail/item/sea_ice_extend/
Some of the highlights in English:
1. There’s almost a 100% chance the Arctic Ocean will have less ice than it did in 2005, the year with the 2nd lowest recorded ice amount.
2. The chances that the Arctic sea ice will reach last year’s record levels are only about 8%.
3. Exact predictions are not possible because short-term variations in ice extent depend greatly on atmospheric conditions, especially wind, cloud cover and air temperature.
4.To make this year’s estimate, Dr. Rüdiger Gerdes and his team have used atmospheric data from the last 20 years and fed them into an Alfred-Wegener-Institut developed Ocean Sea Ice MODEL.
About the MODEL – in English:
http://www.awi.de/de/forschung/fachbereiche/klimawissenschaften/arctic_sea_ice_in_summer_2008_an_outlook/
It will be interesting to see if this model produces a good result.
Daryl;
AWI model says:
80% probability the minimum ice extent in 2008 will be in the range between 4.16 and 4.70 million km2.
So it looks like you’re a little high…but your low-end number looks like a pretty good guess.
It would be interting to hear Steve’s comments on the AWI’s model methodology.
Just reading what they write, I think the 3.59 million sq km refers only to the FYI (first year ice) — not the total ice.
They said they averaged April Figure 5 data to get an average percent FYI remaining — to apply to this year’s FYI. That is: “…If we apply the survival rates averaged over all years to current conditions, the end-of-summer extent [of FYI] would be 3.59 million square kilometers…”.
Using April Figure 5, and actually averaging the percent FYI remaining — I get a number closer to 32.5 percent remaining (hard to do just eyeballing). Anyhow, apparently the 30 percent mentioned in the May text was merely approximate (the text said “typical”, not average).
Now: 0.325 x 10.94 = 3.56. This is close enough to 3.59 for eyeballed government work. I suspect the actual average percent FYI remaining calculated from figure 5 would be 3.59/10.94 = 0.328 — if one had the actual FYI tabular data used to make the figure (rather than an eyeballed histogram).
FWIW
When the Mojo of your best Totem, The Great Global Average Temperature (The Totem with The Most Mojo Ever in the Entire History of The Planet), begins to wane, you need a replacement. Even if you have to disingenuously invoke its Mojo implicitly through the Mojo of the former Great One.
Referring back to my #20:
I do not actually agree with their methodology (assuming I interpreted it correctly).
Moreover, I think they may be mixing apples and pears — because “FYI fraction %” (Y axis label) may have two meanings in April Figure 5. Anyhow, in the context of their text, the meaning is ambiguous to me. Using one of the meanings, my #20 write up makes sense (but not with the other).
Steve, an hilarious post, if it wasn’t sad to watch such incompetence on those high ranks of science…
#16
While I agree the Arctic ice melt this summer will not likely exceed last year, I have difficulty with the comment that “his team used atmospheric data for the last 20 years.” Chapter 1, Page 3 of the Arctic Climate Assessment Report (the major source for IPCC FAR) has a polar projection map showing changes in observed air temperature for the period 1954 – 2003. Most of the Arctic Ocean has a large white space with the words “No Data”. So the questions become; what data was used to build their model for this region and what 20 year data did they use?
Here is a little treasure trove for people to play with.
#19 Dan Hughes
As Mosh de Pit once (actually more than once) observed: When all else fails, run for the ice!
Cliff
re 24. hehe. benefits of studying rhetoric I suppose. you see the patterns in arguments.
#10
Why not discuss ice in terms of thickness and albedo, then, instead of the irrelevant term, “multiyear ice.” Multiyear ice likely has a great range of thickness and albedo that likely overlaps that of new ice.
The term is meant to distract people from the fact that the amount of arctic ice was greater this winter than last.
Please can someone explain the relationship between the NSIDC numbers and the Cryosphere Today numbers? The former gave 4.13M km^2 for the 2007 minimum, whereas the graph from the latter gave 3.0M km^2. Merely looking at the latter’s anomaly and its current slight downward trend, I predict a 2008 minimum of 3.2M km^2 (anomaly -1.8). But I have no idea how to translate that into an NSIDC number, apart from pro rata, which would be 4.4M km^2.
Rich.
Re #27
NSIDC plots the extent whereas CT plots area.
#Phil 28
OK, I’ll bite! What’s the difference between “extent” and “area”? They both seem to be in km^2.
Re #29
OK Roman:
Ice extent is the total area covered by some amount of ice, including open water between ice floes; ice extent is typically reported in square kilometers.
Ice area doesn’t include the intervening water so is necessarily a smaller area.
For example a 1kmx1km pixel with a concentration of 20% adds 1 sq km to extent but only 0.2 sq km to area.
Phil, in the directory ftp://sidads.colorado.edu/pub/DATASETS/seaice/polar-stereo/nasateam/near-real-time/north, there are files labeled f13 and f15. Do you know what the difference is?
Also do you know how one goes from these data sets to the daily total that one sees published?
Re #31
Sorry Steve for some reason I can’t access that database.
If you mean the extent your calculation seems right to me, if you mean the sea ice area then just multiply by the concentration and sum.
One thought is that CT appears to use 1×1 km squares rather than 2.5×2.5 which would make a difference
Steve, there was a trailing comma in that URL that screwed us up.
I think f13 and f15 refer to the satellite or channel or something. One of the documents in an earlier document I referred to earlier explains f13 etc …
From here.
I suspect it is the same for the datasets you referred to.
Having looked at both the actual values recently and also the previous years basic numbers this is certainly a non-trivial problem if ever there was one. I certainly think extrapolation is something a very brace person would do in the scientific community. OK, I as a punter can say 5m as minimum without too much bye or leave, but scientists should not be spouting off to the media on whether an all time low will be reached. Surely they know that?
How many factors have we got here? Sun, wind, thickness,currents, chemistry, albedo, water temperature to name just a few. Can anyone really calculate and weight all those factors accurately or should they just all play the wait and see game?
# 35, AndyW,
You omitted salinity from your list.
“The exchange of salt between sea ice and the ocean influences ocean circulation across hundreds of kilometres”
(http://nsidc.org/seaice/characreristics/brine_salinity.html)
Now first year sea ice, as one might expect, contains more brine than older ice. The greater extent of first year sea ice this year following last years low will presumably exacerbate this effect as it melts?
I’ve sent an email to S Drobot, the author, seeking a clarification of the methodology.
#26 Gina,
Pure speculation on my part, but wouldn’t using a metric that captures the essential point they’re trying to communicate to a lay audience be the way to go? SMc seems to have no problem using NSIDC’s units in titling his post. The discussion is about how the survival rate of ice formed over the preceding NH Winter vis-a-vis ice formed during earlier winters effects the seasonal ice coverage in the Arctic. If the issue causes you to toss and turn at night, why don’t you just ask them?
They would probably say their correspondences over the paper were confidential.
The bottom line is that 2008 is not on pace with 2007. Period.
And there are data that show that this trend probably won’t accelerate much:
1. SSTs in the Arctic are lower than 1 year ago.
2. Global atmospheric temps are lower.
3. This indicates the melt season will end earlier, by maybe 7-10 days. (Eyeballing the charts, the 2008 melt season appears to have started later too).
4. Last year’s very unusual wind patterns most likely won’t repeat.
5. As in No. 16, the AWI says there’s only an 8% chance 2008 will reach 2007 levels, i.e. very unlikely
My WAG is that we’ll end up 900,000 sqkm over last year – or at 5.2 million sqkm (15% concentration contour). It would be interesting to list all WAGs to see who ends up being right.
Re #40 – **1. SSTs in the Arctic are lower than 1 year ago.**
This is interesting since some have stated that the Arctic waters are warmer. Is it warmer (less cold) below the surface?
I’m certain the record will be broken this summer, and no amount of scientific calculating can convince me otherwise. The ‘Save the Poor Polar Bear Cubs’ commercial I saw last night on cable assured me that Arctic ice is disappearing at such an alarming rate that only sending their organization lots and lots of money will save the poor animals from certain extinction. How this is to be accomplished is not outlined, but I assume it will involve paying a host of ice analysts and polar bear experts a large part of my contribution to make complicated computations using enormous supercomputers that will unequivocally show that old ice + new ice = no ice in September.
My first question was going to be: Have you contacted NSIDC about their methods?, but it looks like Steve has done that (ca. post # 37).
My second – and very long-standing – question is: Does anyone actually have an idea of how much ice we’re dealing with? That is, volume (or mass, if it’s not a constant density) vs. area or extents. Unless you assume consistent depth, it seems that talking about area or extents is pretty meaningless.
I made several ineffectual stabs at this a couple years ago – scrounged around NSIDC & Cryosphere Today; even wrote a few emails. Nada. But then, all these different data sources are still baffling to me.
Steve,
I agree with you that the numbers don’t add up. I also see that they do not put enough data in a press release to be able to recreate what they calculated, but they have given lots of press releases.
The arctic ice extent bottomed near 7 million sq km in 2004, rising to 14.8 (2005), falling to 5.3, rising to 14.5 (2006), falling to 5.7, rising to 14.7,(2007) falling to 4.2, and again rising to 15.2 (2008). [Note, the 2004 data is not absolute minimum, but the sept monthly average.)
Thus the new ice is 7.8 million sq km in 2005 (53%), and 9.2 (63%), 9.0 (61%), and 11.0 (72%) in subsequent yrs. I read 56%, 60%, 59% and 72% off of figure 5 from April 2008, so these numbers are close.
Also from figure 5, I get that the minimum ice extent was 29% new sea ice in 2005, 34% in 2006 and 24% in 2007. This means that the 2005 minimum was 3.3 old and 2.0 new, the 2006 was 3.8 old and 1.9 new, and the 2007 was 3.2 old and 1.0 new.
Old/New sea ice survival rates were 47%/26%(2005), 72%/21%(2006), and 56%/11%(2007). [They quote 13% survival of the new ice in 2007.]
Using those survival rates with the 3.2 old ice and 1.0 million sq km new ice surviving last September, I get predictions of 4.9 (2005 rates), 5.3 (2006 rates), 3.6 (2007 rates), and 6.5 million sq. km using the average survival rates (75%/30% old/new).
The average agrees with your more precise calculation. And 3.6 isn’t the average prediction, but the lowest. The numbers from figure 4, in May 2008 are about 3.2 (2005), 3.4 (2006) and 2.2 (2007), each 61-65% lower than the numbers above. The average prediction is suppressed 55%.
My guess would be that they are doing something more complex than just looking at past survival. One clue might be that they are basing these depressed estimates on a combination of total survival rates and the average decline rate in April of each year, which they mention in the text under the heading Conditions in context next to figure 2 in May. It is reported as greater than the previous year, which had the record minimum extent. This may be the justification they are using to extrapolate to low numbers.
Unfortunately, I haven’t seen this data for any other year in their press releases, so I don’t know if my guess is right.
#38, Whitebeard,
I think most people can conceptualize how thicker ice and whiter ice would melt more slowly than thinner ice and darker ice. Those who don’t will not gain much from the “metric,” “multiyear ice.” “Multiyear ice” mystifies, not clarifies.
I’m wondering how this recent paper( by Domingues et al, 2008) regarding an apparent bias in the Argo buoy measurements of ocean temp fits in to all of this. Lower ocean temps and more ice fits. But they say the figures are wrong. My 2 courses in stats don’t equip me for this.
RE: “You also missed the fact that multiyear ice is thicker and has a different albedo.”
And therefore, ice that gets compressed, and thereby thickened, due to a compression event, becomes “old ice” no matter how old it is.
Now that I’ve had “ice extent” (NSIDC) and “ice area” (Cryosphere Today) explained to me, I feel that the latter is a more fundamental and important measure (though they would obviously have a hig correlation). So why don’t we concentrate on ice area?
As others have said, ice volume would be even more fundamental, but pretty impossible to measure.
Rich.
I sent the above post slightly edited to Sheldon Drobot of NSIDC and asked him to explain the calculations. He replied as follows:
While cheerful enough, I don’t think that I understand it any better after reading the reply, than before. I’ll follow up.
I spent last week in Alaska and just got back. One of the tour operators told us that they got a very late start this season because the sea ice was unusually persistent this year. Last year the ice melted early, but this year was very different.
On a someone what related topic — a bear expert told us that Polar Bears are wondering much further south than typical and are now competing with Brown Bears for territory. She said that the popular theory that diminishing sea ice is causing Polar bears to move south is not supported by the zoology community. They attribute the southern migration to the fact that the berry crop has been quite abundant the past few years causing the population of Polar Bears to swell. It seems that the density of Polar Bears in their native territory is slightly up (including in the sea ice regions), but the total population is up a lot. So the Polar Bears have to move south to avoid competition with other Polar Bears. She then said, “I know that the fact that the Polar bear population is increasing dramatically is disappointing to some of you — but I’m just relating the facts”.
Didn’t see any Sea Ice discussion newer that this, so will post here.
From the WeatherUnderground site, Jeff Master’s blog:
http://www.wunderground.com/blog/JeffMasters/show.html
Latest discussion is on the Arctic ice.
Anbody read the paper he’s referring to?
Re: henry (#51),
Henry, try this.
http://www.sustain.hokudai.ac.jp/2008/11/speech_when_will_summer_arctic_1.php
Since JAXA is running again I decided to look at the differences with previous years. So far 2009 is running ahead of 2005-2008 and behind 2003 and 4. Slightly ahead of the average for the preceding 6 years. If it comes in at the average it will max out at about 14.2.
Steve could we get a new 2009 Sea Ice thread if there’s any interest?