September Arctic sea-ice minimum predicted by spring melt-pond fraction

[snowhare]

New letter in Nature Climate Change

tl;dr: Melt pond area in spring predicts September minimum.

[icefest]

Thanks,
Here is the relevant image linking extent and puddles: (for those outside the paywall)

[Steven]

Here's some further information from the paper, about Figure 3 posted above by icefest:

Figure 3 reveals
that there is a highly significant correlation (p-values < 0.005)
between simulated pond fraction and SSM/I September ice extent.
The highest correlation coefficient of R = -0.80 occurs for ice extent
if the pond area is integrated from 1 to 31 May. Extending the
integration time period does not improve the correlation. Although
the pond area integrated from 1 to 31 May contains only around 1%
of the annually integrated pond area (Fig. 1), the melt-pond fraction
in May seems to have the strongest impact on the sea-ice state in
the subsequent September.

The melt pond data in the paper are from a model they developed:

We have developed a new melt-pond model suitable
for forecasting the evolution of melt-ponds¹⁰ and incorporated this
model into the widely applied Los Alamos sea-ice model called CICE (ref. 12).....
The results of our CICE simulation are consistent with
in situ observations^14,15 and pond statistics for the period 2002-2011
based on MODIS satellite data¹⁶.

The superscripts refer to the list of references, link.  The other figures can also be viewed there, e.g. here is the link to Figure 1.

[Neven]

I've just posted a blog post on the ASIB discussing this paper.

[Steven]

The pdf of the paper is now also accessible at:

http://glaciology.weebly.com/uploads/6/6/9/1/6691883/2014-schroeder_nclimate2203.pdf

[crandles]

The scientists say the amount of water ponding on top of the floes as they warm in the spring has been shown to be an excellent indicator.

Using their technique, the Reading University researchers reckon the minimum ice extent this September will be about 5.4 million square km.

It is about the same as at the end of the melt season last year.

http://www.bbc.co.uk/news/science-environment-27870459

[Neven]

Thanks for this, crandles. Very interesting.

Even though this melting season differs from last year's (no persistent cyclones yet, big polynya on Siberian coast, etc), the lack of melt ponding due to clouds and low temps will cause a similar minimum. A perfect test for the theory.

I think it will prove to be right if the current period with high pressure (but again, not a real dipole) doesn't accelerate things. And so far it hasn't, barring some weird oscillations in the CT SIA data.

[Jim Hunt]

An infographic explaining the "melt pond method", via @micheltsamados

[gciriani]

What was the result of the calculation by Reading University, by using this model, for 2013? And are there statistics to calculate the confidence level?

[Steven]

What was the result of the calculation by Reading University, by using this model, for 2013? And are there statistics to calculate the confidence level?

From the paper:

we [...] forecast the September ice extent with an accuracy of 0.50 million km2 by the beginning of June (May pond fraction) and with an accuracy of 0.44 million km2 by the end of June.

For September 2013 we forecast a mean ice extent of 5.55 +/- 0.44 million km2, which is closer to the observed mean value of 5.35 million km2 than any of the 23 [...] predictions presented at the Arctic Sea Ice Outlook webpage in July

Figure 4 in the paper shows the forecasts and hindcasts for 1979-2013.  E.g., Fig. 4c is shown below.  The blue line shows the "forecast" for each year; this is based on the modeled melt pond fraction in the 56-day period from May 1 to June 25 ("pond56").  The black line shows the actual NSIDC September extent for each year.

[viddaloo]

Using their technique, the Reading University researchers reckon the minimum ice extent this September will be about 5.4 million square km.

In which case it will be 7th at PIOMAS. I find that a bit hard to believe, as this year's been ahead of both 2007 and 2009 since August 8th. All 2007, 2009, 2010, 2011, 2012 and 2013 would have lower ice volumes at September Minimum, if Reading's right.

[crandles]

Using their technique, the Reading University researchers reckon the minimum ice extent this September will be about 5.4 million square km.

In which case it will be 7th at PIOMAS. I find that a bit hard to believe, as this year's been ahead of both 2007 and 2009 since August 8th. All 2007, 2009, 2010, 2011, 2012 and 2013 would have lower ice volumes at September Minimum, if Reading's right.

Think there is a mismatch between extent and PIOMAS = volume there.

Per 2nd image in
http://meteomodel.pl/index.php/arcticice
It looks to be 7th at present.

I would say it is looking good for this prediction.

Current NSIDC extent:
014,    08,  27,    5.57198,

[viddaloo]

crandles, I'm sorry, my model has 2014 at 5^th @piomas. If numbers from Reading are right, it would be more like 7^th.

[crandles]

I don't know of any reason why volume can't be 5th and extent 7th especially as that appears to be current situation.

[viddaloo]

I don't know of any reason why volume can't be 5th and extent 7th especially as that appears to be current situation.

I think you misunderstand. 5.4 m km² is 10^th @ijis, not 7^th. I say 5.4 m km² sounds like 7^th @piomas, which would be disappointing, as I currently have it at 5^th. [Reading uses meltponds to predict extent, I use extent to predict volume.]

[crandles]

I hope you accept that their prediction should be judged on what their prediction is attempting to predict. If they predicted extent, they should be judged with extent.

The article says

Using their technique, the Reading University researchers reckon the minimum ice extent this September will be about 5.4 million square km.

It is about the same as at the end of the melt season last year.

NSIDC Sept average for 2013 was 5.35 and
2013,    08,  27,    5.47764

With little further extent loss to go, this year at 5.57198 is marginally above last years 5.47764. So the prediction is looking excellent.

Predictions like SIPN SEARCH try to predict NSIDC monthly average figures not IJIS daily figures.

You might be more interested in volume numbers but I don't see that that gives you licence to say that an excellent looking prediction is 'a bit hard to believe'.

Edit Also

Their extent has diminished from about 7 million square km in the 1990s to less than 5 million square km in five of the past seven years, with a record minimum of 3.6 million square km being set in 2012.

This agrees to 3.63 for 2012 Sept average.

[gciriani]

from the paper

For September 2013 we forecast a mean ice extent of 5.55 +/- 0.44 million km2, which is closer to the observed mean value of 5.35 million km2 than any of the 23 [...] predictions presented at the Arctic Sea Ice Outlook webpage in July

I'm a little bit in a fog, because I'm new here and I think I'm missing something in the terminology. I averaged the September 2013 figures from JAXA (http://www.ijis.iarc.uaf.edu/seaice/extent/plot_v2.csv) and I obtain an average of 4.99 million km2. So obviously the paper refers to something else. What am I missing please?

[Neven]

There are different data set, gciriani. The most well-known being NSIDC and IJIS. For the SEARCH Sea Ice Outlook, or Sea Ice Prediction Network as it is now called, NSIDC September sea ice extent is used (the monthly average). Naturally, there are differences between NSIDC and IJIS.

[viddaloo]

You might be more interested in volume numbers but I don't see that that gives you licence to say that an excellent looking prediction is 'a bit hard to believe'.

You're obviously still very confused, so I simply iterate.

I think you misunderstand. 5.4 m km² is 10th @ijis, not 7th. I say 5.4 m km² sounds like 7th @piomas, which would be disappointing, as I currently have it at 5th. [Reading uses meltponds to predict extent, I use extent to predict volume.]

If you read it the adequate number of times, it is my belief that you will get its simple meaning.

[Steven]

I don't know of any reason why volume can't be 5th and extent 7th especially as that appears to be current situation.

I think you misunderstand. 5.4 m km² is 10^th @ijis, not 7^th. I say 5.4 m km² sounds like 7^th @piomas, which would be disappointing, as I currently have it at 5^th. [Reading uses meltponds to predict extent, I use extent to predict volume.]

The 5.4 million km2 was the Schroeder et al. (University of Reading) prediction for the September 2014 mean ice extent, as measured by NSIDC.  This refers to the NSIDC rather than the IJIS extent, and moreover it refers to the monthly mean extent rather than the daily minimum.

In fact, their prediction for the September 2014 mean ice extent was  5.40 +/- 0.50 million km2, based on modeled melt ponds from May 1 to May 31  (source). 

By the end of June, they updated this prediction to 5.51 +/- 0.44 million km2, based on modeled melt ponds from May 1 to June 25  (source).

[F.Tnioli]

Predicting extent based on melt ponds and the correlation between % of melt ponds at a number of particular dates and annual minimum extent in the past - will always remain an approximate, imho. Furthermore, the error of such an approximation has to increase dramatically in any significant "mode change" event for the Arctic melt season (and i believe, such a change is well started during last 2...5 years, and is revving up). Because substantial part of the melt - is "melt from below" by warm waters (not related to the albedo directly at the site of such a melt).

As such, it is my opinion that method used by Reading scientists already lost some of its usefulness, and i expect that as significance of "melt from below" continue to increase in the observable future - the "Reading method" will continue to be more and more erratic.

Predicting volume based on melt ponds suffers even more uncertainties than predicting extent, since during recent decades we've seen years which had significantly different minimum annual ice volume while having nearly same minimum extent, and vice versa.

Last but not least, personally, i am not happy with the standard definition of the sea ice "extent"/ I understand its meaning and importance for certain calculations, but whenever i am looking for "how much surface in the Arctic is covered by ice" for some very general/basic consideration - i prefer to use ice area numbers. NSIDC's way to call some place being "ice covered" when in reality some 84% of the place is plain and clean open water - is one thing i am laughing at, in particular. In my book, if they can't know for _sure_ whether it's real open water or just thin layer of water on top of a significant ice layer, - then the proper way is to bring in _uncertainty_, instead of saying something like "you know what, we don't care if it's real open water or not - let's just say that if some area has not less than 15% of its surface being ice, then we count whole area being ice". At least, that's what my common sense tells me.

Some further basic info about the difference between extent and area is available at http://nsidc.org/arcticseaicenews/faq/#area_extent .

[crandles]

Predicting extent based on melt ponds and the correlation between % of melt ponds at a number of particular dates and annual minimum extent in the past - will always remain an approximate, imho. Furthermore, the error of such an approximation has to increase dramatically in any significant "mode change" event for the Arctic melt season (and i believe, such a change is well started during last 2...5 years, and is revving up).

I would tend to say it looks one of the best if not the best method we have at the moment. I agree that if the parameters are not tweaked then over time it is likely to start to overestimate extent because the winter thickness may become thin enough to melt out without need for such large melt pond coverage. However parameters will be tweaked to tune to latest situation and I don't see any reason why it couldn't remain a good prediction method.

If "such a change is well started" why is it looking like it will be such a good prediction this year?


If it does start to change so that it become less good a prediction method, that in itself would be useful information IMHO.

the "Reading method" will continue to be more and more erratic.

seems a bit denegrating when the outcome this year looks to show its prediction is excellent.

Why disparage? Can you do better?

[F.Tnioli]

"Why disparade": i didn't "disparade" it. Merely discussed how precise/imprecise the method is and will be in the future in my personal opinion. I dare hope that discussing precision of scientific methods helps to use such methods in a most useful and appropriate way.

"One of the best methods we have": yes, it is. But not because it's so precise - but because other methods are similarly imprecise. Noone's fault. Arctic (and climate in general) are things vast and complex.

"Parameters will be tweaked": suuure. But how well? Presently, past observations help with that much. In a significantly new mode of summer melt, this won't be the case. I've seen how "parameters are tweaked" time and time again when the speed of melt of glaciers and ice caps were "re-estimated" again and again (and/or predictions for how big sea level rise will be by 2100, which is directly related), each time it was said "yeah, this time we got it rather correctly, here's 95% confidence internal, we know what mistake we did last time, now it gotta be correct". Gets old when it's done 3rd time, you know.

"I don't see any reason": ok, no problem. I see, but you don't. It's not impossible that i am "hallucinating" - seeing things which are not existing. I respectfully agree to disagree... sir.

"Why is it such a good prediction this year?" - i think you've seen some of my posts about aerosols, welsbach seeding, my reasons to suspect significantly large geo-engineering in Arctic. I expected much bigger melt this year than what we see now. I can't prove any geo-engineering in the Arctic, of course (and if i could - would i, even in such relatively low-profile but still _public_ place as this one?). But in the same time, i don't think you can prove that intentional and large enough to substantially change annunal minimum ice area/extent geo-engineering project in Arctic is NOT happening since 2013 (continuing this year), eh?

The only one thing i can add - is this: look at CT area numbers during recent weeks. You'll see near-halt of the melt (in terms of sea ice area, of course): year 2014, day 215 = 4,65 MKm2, year 2014 day 229 = 4,53 MKm2. Two weeks of august with average (during this period) daily loss of sea ice area being 0,0085 MKm2 (i.e., mere 8,5k - 8500Km2 / day). But the truly unusual thing about it - is not the rate itself, but continuation of intense melt _after_ such a near-halt. 2013 had similar period of relatively-constant sea ice area during august with continuation of rapid decrease afterwards. This feature is something new; during previous years (1979...2012), i don't remember any single "temporary stop" for some 10...14 days of late-summer melt which then turns into still massive and rapid further decrease of Arctic ice area. And while i can suspect that such a "new behaviour" was caused by highly unusual weather during a single year (say, 2013), - it's much more difficult to seriously consider that we now have 2 "highly unusual weather" years in a row, while during 1979...2012 period we didn't have a single "same kind of highly unusually weather" year at all.

"That in itself would be useful information": oh, i agree. Exactly one of reasons i wrote the previous post, and exactly what i was talking about in the 1st paragraph of this one, you see. You see, right?

"Can you do better": i probably could - with proper data and sufficient computing power and staff, which are things which i do not have at this time. Thus in practice, i can't right now. All i can is to express my opinion about some things which - i hope noone would argue, - influence the usefulness of the method in question. I hope it's not a crime.