September 1 Eve! Tomorrow, extent measurements start to accumulate into the September average. Today, I’ll show current predictions based on a weighted model involving linear fits predicting September NSDIC based on Cryosphere Today Area and Jaxa Extent area during the period when JAXA extents are available. (That is, since 2001.) Of these two, the JAXA extent gives the best predictor; prediction based on that metric is given a weight of 87%. CT give the second best predictor and contributes 13%. The fit based on JAXA alone is shown in orange; the weighted prediction is given in green:
Our June bets are show in lavender; our late bets are show in purple. My bets are shown in red.
Notice that the lowest June bet is now inside the ±95% confidence intervals based on this more complicated model. That’s happening for four reasons: (1)The uncertainty intervals corresponding to those show in orange here in my previous posts were too narrow — I forgot to include one of the factors in my previous uncertainty calculation. (Doh!), (2) The rate of extent loss has slowed, (3) the prediction based on CT areas is a bit higher than that based on extent and (4) the more complicated method considers the possibility that of the two models the best might be wrong and the less good model might be right. The method of computing the confidence intervals always results in wider confidence intervals than those computed based on the assumption the best model is “right”.
I would get even larger confidence intervals if I included more possible models in my algorithm which I may as the month progresses. But at this point, I think much of the uncertainty is weather dependent and hunting around for a good weather metric that predicts changes in September ice extent might be time consuming and not worth the effort. Still, weather watchers do have a decent chance of beating the curve fit– provided you know how to account for effects of weather in an objective way.
Cue Jaws music!
Only 121 days til Christmas (I could be wrong – it’s a GISS-adjusted figure).
Prediction for Aug. 31st, based on 2008-11 mean x 1.24, was 24,221 km^2 lower than the actual figure.
The prediction for Sept. 1st, was 39521 km^2 higher than the “provisional” actual figure, but that will almost certainly be revised upwards.
The 7 day ma daily change was -53281 km^2/day
Still on target for approx. 3.6 million km^2
Lucia:
I don’t think it would be useful for this year. Weather is strongly affected by surface boundary conditions, and this year is unique in that respect, do I don’t see how statistical methods will yield much light there.
What we do know is almost all of the ice is gone next, except for north of 75°N, most of the ice that remains is higher concentration/multiyear ice, temperatures are dropping rapidly above this latitude, meltponds have nearly frozen, etc.
None of this is really “weather” in the usual sense, you could almost model it as a single column of air above a mixture of ice and water (Blackadar level 1-d physics, e.g. something like this.)
We also know that polar vortexes “happen”, and that these upset the apple cart and could still do so (by pulling more ice south where it gets decimated by warmer air temperatures and above freezing water). Looking at past years for strengths of polar vortexes in September vs August might help in terms of quantifying the uncertainty associated with that.
The JAXA figure for Sept. 1st was revised upwards from 3.741 million km^2 to 3.798 million km ^2, which is 17,667 km ^2 more than the projection based on the 2008-11 mean change *1.24.
The 7 day average change at Sept. 1st was -48,954 km ^2.
The provisional figure of 3.711 million km^2 for Sept. 2nd., is only 30156 km^2 less than the provisional figure for Sept. 1st.
I think we are getting close to the first daily increase in extent using JAXA.
The JAXA figure for Sept. 2nd was revised upwards from 3.711 million km^2 to 3.755 million km ^2, which is 5,638 km ^2 more than the projection based on the 2008-11 mean change *1.24.
The 7 day average change at Sept. 1st was -48,817 km ^2.
The predicted figure for Sept.3rd is 3.715 million km^2, compared to the provisional figure of 3.680 million km^2, a difference of -34825 km^2, but that is well within the recent upward revisions to the daily figures.
The melt ponds are entirely frozen over and covered with snow at this point and high arctic temperature continue to drop rapidly. It’s still a bit wiggleology to declare that we’ve seen a knee point (slowed ice loss), but there’s good physical grounds for expecting it.
You can expect the anomalously warm water surrounding the ice pack to delay the minimum and recovery day. The latest JAXA figure makes it easy to see that the minima are (on average) occurring later in the year.
Even now, when we’re in the period that that’s being averaged over, it is a tough call what the monthly ice average will be, as weather plays a big role in when the annual recovery will actually start.
Carrick,
I’m not sure it is easy to tell from the extent graph that the minima are occurring later.
I thought you might be basing that assertion on the fact that the 2000’s average date was the 15th, while the 1980’s and 1990’s average dates were both the 11th.
However, I have checked the minimum dates since 2002 and there seems to be little evidence of a trend to later minima over this period. September dates of minimums:
2002 9
2003 18
2004 11
2005 22
2006 14
2007 25
2008 9
2009 13
2010 18
2011 9
Average 2002-6 = 14.8
Average 2007-11 = 14.8
Although the averages are identical, the linear trend is slightly down.
I have included 2002, although there was a gap in the data between the 12th and the 21st, so I don’t think we can be sure when the minimum was that year.
Of course, the above period is probably too short to form any conclusions and it would probably be necessary to look at the data from the 80’s and 90’s to be sure.
The JAXA figure for Sept. 3rd was revised upwards from 3.680 million km^2 to 3.747 million km ^2, which is 32,205 km ^2 more than the projection based on the 2008-11 mean change *1.24.
The actual change between the confirmed figures for Sept. 2nd and Sept. 3rd, was -7969 km^2., resulting in a 7 day average change at Sept. 3rd of -45,692 km ^2.
Using the provisional daily figures alone, there was actually an increase of 3437 km ^2 between Sept. 3rd and Sept. 4th.
The projected figure based on the 2008-11 mean *1.24, for Sept.4th is 3.674 million km^2, compared to the JAXA provisional figure of 3.683 million km^2, a difference of -9493 km^2, but once the provisional figure has been revised upwards, I expect the actual to be again higher than the projected figure.
If the upward revision of the Sept. 4th provisional figure is the same as on Sept. 3rd, we could see the first daily increase in JAXA extent, although that doesn’t mean we have passed the minimum, since there could still be more daily falls.
Ray, you can see by fitting to the minimum to a 4th order polynomial that the date of maximum ice extent has shifted later as the ice cap has melted (I use a combination of JAXA + the older GSFC data series.) The reason for fitting to it is to remove weather related noise. Picking individual years without any smoothing is another version of “It’s easy to get a null result using statistics on noisy data.” But that doesn’t mean anything if the method used isn’t robust.
Taking decadal averages achieves the same result of reducing short-period noise. This retains higher frequency signal components than the method I used, and is in general less clunky in terms of picking the optimal smoothing function, but has the disadvantage of introducing serial correlation into the data (hence if you use a 10-year running average, it won’t tell you anything about variations in trends over periods that are less than 10-years).
Trying to look at trends over individual years from a decade by picking the minimum is certainly going to very noisy. With the fitting method I used the trend for date of maximum ice loss is still positive, even in the last decade.
I generally wait a day or so before I look at JAXA numbers, btw, the initial versions for each day are considered provisional. (I wait until I get the next days update before popping corks for the annual ice minimum for example.)
Carrick,
I bow to your superior knowledge on this subject, I am new to the game.
It does appear that on the basis of your method, the dates of minima are getting later, but I did say in my comment that it would be necessary to look at longer-term data than I did, in order to be certain. Of course, that does not preclude the possibility that the minimum date this year will be earlier.
I am aware that the initially published figure for each day is provisional, in fact I have made several references to this in my posts.
By the way, should the scale at the right hand not say “calendar day”, rather than “calendar year”?
Ray, the only thing my method would be useful for is teasing out whether a long-term warming trend is *tending* to extend the date of minimum temperature (and this in turn is useful in really just observing something that the sea temperature anomaly data are already telling you has to happen, that part is “just physics”). Some aspects of what I am calling “weather noise” may actually be deterministic, but they wouldn’t necessarily align year to year (for example forcings from the ENSO).
I’ll fix the scale when I get a chance, thanks for noticing it.
Thanks fossil fuel GHG’s for extending the Northern Arctic Ice melt season. Trending proves tending wrong again…….
The JAXA provisional figure for Sept. 4th was only revised upwards by 43282 km^2, so no increases in the extent, and while the fall of 20,312 km^2 was the second lowest since June 3rd, it was higher than that for the Sept. 3rd. Even so, the 7 day average is down to -38549 km^2.
The actual figure for the 4th was 52774 km^ higher than the projection based on the 2008-11 average x1.24, so there are signs that the decline in the extent is slowing relative to the projection, which would make September mean of over 3.652 million km ^2 more likely.
The provisional figure for the 5th of 3.628 million km^2 suggests a fall of 98,438 km^2 but would still leave the actual figure 4,517 km^2 above the projected figure of 3.624 m km^2, even without the usual upward adjustment.
An increase in the extent based on the confirmed figture for the 5th now seems unlikely but I still think that an increase within the next 7 days is a distinct possibility.
Carrick,
Do you happen to know if the JAXA provisional figure is always lower than the confirmed figure or is that only when the ice is declining?
I think this was the case in August but I have only been keeping a record of the provisional figure since Aug. 31st.
I notice that the confirmed figure for Sept. 4th is even higher than the provisional figure for Sept. 2nd.
In my post #102612, the second paragraph should read:
“The actual figure for the 4th was 52774 km^ higher than the projection based on the 2008-11 average CHANGE x1.24, so there are signs that the decline in the extent is slowing relative to the projection, which would make September mean of over 3.652 million km ^2 more likely.
The JAXA provisional figure for Sept. 5th was revised upwards by 52,969 km^2, resulting in a confirmed figure of 3.681 million km^2, a change in extent of -45,469 km ^2, and a 7 day average change of -33,058 km^2/day.
The actual figure for the 5th was 57487 km^ higher than the projection based on the 2008-11 average change x1.24, so the actual figure is currently running slightly higher than the projected figures, although since Sept. 1st, the projected figure has proven, on average, to be more accurate than JAXA’s own provisional figure and the difference is currently only equivalent to about 1 day’s change.
The provisional figure for the 6th of 3.614 million km^2 suggests a fall of 66,765 km^2 but would leave the actual figure 23,985 km^2 above the projected figure of 3.590 m km^2, without the usual upward adjustment. If the provisional figure for the 6th is adjusted upwards by the same amount as that for the 5th, the confirmed figure for the 6th will be 3.667 million km^2, which is 13,906 km^2 lower than the 5th but 76,954 km^2 more than the projected figure for the 6th.
I therefore suspect that if Lucia were to run her model now, it would predict a September mean above 3.6 million km^2.
Ray– Nope. 3.41. I’m going to post tomorrow. I figure: Friday. I’m looking at something else today.
Oh… when’s UAH out?
Now!
http://www.drroyspencer.com/
0.34c
Never mind, you lose some and you lose some!
Ray:
This is just a SWAG, but I think the error will be opposite to the direction that the ice extent is moving. If there is no movement, the provisional should nearly equal the final figure.
Anyway, that’s my guess.
The JAXA provisional figure for Sept. 6th was revised upwards by 62,187 km^2, resulting in a confirmed figure of 3.676 million km^2, a change in extent of a mere -4,688 km ^2, and a 7 day average change of -28,661 km^2/day.
The actual figure for the 6th was 86,173 km^2 higher than the projection based on the 2008-11 average change x1.24, since June 1st.
The provisional figure for the 7th of 3.602 million km^2 suggests a fall of 74,531 km^2, but even without the usual upward adjustment, that figure would be 66,521 km^2 above the projected figure of 3.535 m km^2. If the provisional figure for the 7th is adjusted upwards by the same amount as that for the 6th, the confirmed figure for the 7th would be 3.664 million km^2, which is 12344 km^2 lower than the 6th but 128,707 km^2 more than the projected figure for the 7th.
The actual daily figure is becoming progressively higher than the projected figures based on the 2008-11 mean x 1.24, suggesting that the estimated September average of 3.65 million km^2 based on the projection, is likely to be too low.
The rate of change which currently most closely reflects the actual figure for the 6th, is now the 2008-11 mean change x 1.227.
Based on that rate of change, the projected mean for September is now 3.733 million km^2, so I am still holding out hope that my original prediction of 3.794 may be in the running for some quatloos!
A lot will depend on how quickly the ice extends after the minimum is reached.
Carrick,
Thanks for your reply. However, I am not so sure.
The average upward revision since Sept. 1st, is now 52,441 km^2,
and shows no sign of falling, despite the fact that daily changes in the confirmed figures are getting lower.
I wish I knew how the provisional figure is arrived at, and I also wish I had been keeping a record for longer!
Ray, thanks. As I said I don’t have any idea how the provisional data is arrived at.
BTW, I’ve modelled the ice loss, my minimum extent day is between September 9th and the 14th and monthly average is between 3.56 to 3.76 Mm2.
We’ll see how badly this does. Weather is still a huge determinant.