September Arctic sea ice extent: 1935-2014

A couple of days ago a paper was published in the journal Revista de Climatología by Diablobanquisa, a Spanish blogger who regularly writes about Arctic sea ice on his blog. In this paper Diablobanquisa, together with his co-author, presents a new time series of September Arctic sea ice extent, including data for the Siberian sector. I think the work looks very solid and gives us an improved view of pre-satellite Arctic sea ice.

Below is the English abstract of the paper, preceded by the graph showing the results of the new time series (updated up to 2015):

A new time series of September
Arctic sea ice extent: 1935-2014

Abstract

Since 1979 satellite-borne passive microwave sensors have provided a continuous and consistent record of Arctic sea ice extent. This record shows a significant downward trend, particularly at September, when Arctic sea ice reaches its annual minimum. Records before 1979 exist, but are not consistent with the satellite record and have limited reliability, specially before 1953. We present a new time series of September Arctic sea ice extent from 1935 to 2014 that includes data for the Siberian sector (AARI operational charts) not used previously in the Arctic wide existing time series (Walsh, HadISST). The new record has been adjusted to be consistent with satellite data. The trend for 1935-2014 is -3.5% decade, while the trend for the satellite era is -13.3% decade. However, the trends since 1935 until early 1980s are positive and statistically significant. The trends turn negative in the 1990s, reaching statistical significance from 2006 onwards. The lowest annual minimum in the pre-satellite era is higher than any annual minimum after 2001.

Median September Arctic sea ice extent for the lowest decade of the pre-satellite era (1935-1944, left) and for the lowest decade of the satellite era (2005-2014, right):

– Download full text (pdf, in Spanish)

– Download extent data values (csv)

– Download gridded data (netcdf)

Methodology and several graphs comparing the results to previous work, can be found over on Diablobanquisa (Spanish version including conclusion here). There's also a separate thread over on the Arctic Sea Ice Forum.

Congratulations and thank you, Diablo + co-author!

Comments

[Diablobanquisa]

Thank you so much for re-blogging this, Neven!

[Diablobanquisa]

It's worth mentioning that Walsh et al. are currently updating their dataset, incorporating AARI and other sources.
This is their last presentation on the topic: http://www.star.nesdis.noaa.gov/star/documents/meetings/Ice2015/dayThree/1_Walsh_J_dayThree.pdf
Although it's still a work in progress, they present some preliminar results on pages 17 and 18.
Their September results look pretty consistent with ours.

[P-maker]

Neven & Diablo,

thank you for bringing this up. Now - for the first time - I see the purpose of logging ship positions way back. The negative imprint of a lat - long position of a sailing ship is clear evidence that open water is not far away, thus sea ice edge may be close by. I stand corrected.

Cheers P

[Colorado Bob]

OT -
Tonight NOVA is airing a new program on the collapse of the krill stocks around Antarctica. I'm sure it will be available tomorrow online for those outside the US :

NOVA
Mystery Beneath the Ice Preview

What’s behind the death of a tiny creature with an outsized role in the Antarctic? Airing January 20, 2016 at 9 pm on PBS

Link

[wayne]

http://eh2r.blogspot.ca/2016/01/northerm-hemisphere-temperature.html

World wide temperature average for the year 2015 was simply significantly higher than the year just past. This was not reflected in sea ice extent as the graph presented above demonstrates. Sea ice extent is very susceptible to dynamical weather, but the long term downward trend seen above is undeniable. However sea ice volume may not be as PIOMAS estimates. My recent observations in darkness have confirmed a remarkable drop in sea ice thickness over a wide area. I'll have more about this at week end.

[AbbottisGone]

So,
1- the negativity starts around the year 2000
2- this means China is significant
3- there has been conjecture that world warII was about hitler not trusting science to be able to deliver on agricultural need and thus standards of living expected by people (whom by assumed definition are not bad) has to be met by other means, thus...
4- Houston, we have a problem!!

Question: am I being over-dramatic?

[Rob Dekker]

Diablobanquisa, thank you for your paper.
I have a question about your 1930's September extent estimates, which seem to be at the core of your conclusion that there was a significant increase in Arctic sea ice between the 30's and the 70's.

I found these estimates (which all hover around 7 M km^2) rather low, so I decided to do a couple of point-checks.

The highest Sept ice extent from the satellite area were in 1980 and 1996, with each something like 7.8 M km^2.

Here is how that situation looked like in (the middle of) August :
http://igloo.atmos.uiuc.edu/cgi-bin/test/print.sh?fm=08&fd=15&fy=1980&sm=08&sd=15&sy=1996

Now, let us compare that to the August 1935 DMI chart :
ftp://sidads.colorado.edu/DATASETS/NOAA/G02203/1935_08.jpg

Noticeable difference is that in Aug 1935 there seems to be more ice around Svalbard, more ice in the Beaufort, more ice extending even into the Bering Strait and more ice in the Greenland sea (even close to Iceland) compared to Aug 1980 and Aug 1996.

So one would expect that Sept 1935 would end up above Sept 1980 and/or 1996, and probably would exceed 8 M km^2.

Yet in your graph, you put 1935 at 6.8 M km^2 or so.

For comparisons with other years in the 30's, it is not much different, which puts your conclusion in at least some serious doubt.

Would you care to comment ?

[Diablobanquisa]

Hi Rob, thank you for your comment.

First of all, I'd like to point out that our work doesn't pretend to be the last word on this, but just our best attempt to build a time series as reliable and as consistent as possible.

In addition, during 1935-1952 we have used a flat climatology that suppresses interannual variability except for the area covered by AARI (Siberian sector). So during that period some years could be higher and other years lower than expressed in our graph.

However, we are pretty confident that the mean extent during 1935-1952 is quite close to our mean extent for that period. And we are also quite confident that there is a significant (although slight) upward trend between 1935 and 60s and 70s.

A relatively low extent during 1935-1952 correlates well with Arctic SAT (the correlation of our time series with SAT is r=-0.84 for the whole 1935-2014 period).

Look also at https://diablobanquisa.files.wordpress.com/2015/12/fig-9.png (black line: our time series; grey band: estimated sea ice extent using a regression of September SIE on Arctic SAT).

The Arctic SAT during 1935-1952 (we use CRUTEM4 data, but it's almost the same using GISS, BEST...) is 0.6ºC above the 1951-1981 mean, while the SAT during 1962-1979 is 0.1ºC below the 1951-1980 mean. This points out to a lower extent during the former period.

Our result comes in overall agreement with the new Walsh dataset (p.17, http://www.star.nesdis.noaa.gov/star/documents/meetings/Ice2015/dayThree/1_Walsh_J_dayThree.pdf) that also shows a slight upward trend from mid 30s to 60s and 70s. Their 1935 value is quite close to ours (p.17, extent: 6.8 mill. sq. km. or so, extracted with http://arohatgi.info/WebPlotDigitizer/app/ ; look also at p.18, they put 1935 with a september sea ice area below 6 mill. sq. km. ; and look also at p.20, where they show their September 1935 map. Here, their 1935 map vs. ours: https://diablobanquisa.files.wordpress.com/2016/01/1935vv.png).

Regarding your map comparison, I'm sure you are aware that we should be very careful when comparing DMI maps to satellite ones. Some reasons for this:

- We don't know if DMI maps are consistent with satellite observations, and we can't test them because there isn't any overlapping period. For instance, we tested HadISST against the satellite data during the overlapping period 1972-1978, and we found that HadISST extent values are 0.83 mill. sq. km higher than satellite ones. (Meier et al. did the same using a different grid, and they found HadISST values to be 0.9 mill. sq. km. higher than satellite ones). We also tested the consistence of including AARI data during the overlapping period 1972-1978, and we found them fully consistent with satellite data.

- As you surely know, the only direct observations on DMI maps are the red symbols, while the white area is just a guess. Besides, we don't know the exact data of each direct observation (We could take them as mid-month values, as you have done, but anyway we don't know the exact data of each one. In the same map, one direct observation could be from August 1st, another one from August 30th, another one from... and the interpolation to mid-september will be affected by this.)

- In addition, you are just eyeballing maps. Even if you are a good observer, and I'm sure you are, eyeballing is a technique with wide error margins. Besides, you have had to eyeball not just the maps, but a interpolation from mid-August to mid-September, what reduces even more the accuracy of the comparison.

Cheers

[Diablobanquisa]

http://www.star.nesdis.noaa.gov/star/documents/meetings/Ice2015/dayThree/1_Walsh_J_dayThree.pdf

https://diablobanquisa.files.wordpress.com/2016/01/1935vv.png

[Diablobanquisa]

Hi again,

I'd like to do a few more comments.

From 1953 onwards, we use a combination of adjusted HadISST and AARI (both tested against satellites during the overlapping period 1972-1978).

So, we needed to extend back the time series from 1952 to 1935 as consistently as possible. We dediced not to use DMI maps neither for this period due to the following reasons:

In addition to the other problems of DMI maps (we didn't know if they were consistent with satellite data, they needed interpolation from August to September...) we found that they had a good geographic coverage only for four years (1935-1938).

From 1939 onwards their coverage is much more reduced. Indeed, between 1940 and 1945 DMI didn't produce any map. From 1946 to 1952, the geographic coverage is low, with observations limited to Greenland, Baffin and surrounding areas. The rest of the Arctic is shown as white. During this period (1939 and 1946-1952), the white area is almost the same every year, and it looks more like an 'area without data' than a guess or an inferred extent. And, if it were a guess (I don't think so), AARI maps would prove it wrong.

DMI maps only offered us four years (1935-1938) with good coverage, so we couldn't rely on them for the whole 1935-1952 period. Besides, we would had needed to solve the other problems (we didn't know if they were consistent with satellite data, they needed interpolation from August to September...) Taking this into account, and in order to keep our time series as consistent as possible, we decided not to use DMI maps neither for the four 'good' years (1935-1938).

Anyway, the new Walsh dataset seems to use DMI maps (as well as AARI, ACSYS and other sources) and its results for the 30s and 40s look reasonably consistent with ours.

Cheers

[Rob Dekker]

Thank you Diablobanquisa for your extensive comments and explanations.

I have many questions and comments in return, but to keep it simple, let me focus on the comparison between your work and HadISST and Walsh' newest presentation, especially regarding the 30's.

You mention that HadISST appears to have a low bias of some 0.9 M km^2, as confirmed by Meier et al. With HadISST logging in at some 9 million km^2 for 1935, the corrected number would thus be in the 8 M km^2 range.

Walsh' chart indeed matches your 6.8 M km^2 for 1935, but it seems to use a different metric than yours. Look at Walsh' chart for the satellite era : The maximum extent in Walsh' chart is 14 M km^2 for the 80's and 90's, while we know from NSIDC that ice extent was 15 M km^2. And Walsh logs the 2007 and 2012 minima also 1 M km^2 lower than the NSIDC extent minima. So either Walsh is not showing extent (but maybe area) or Walsh' chart has a 1 M km^2 low bias.
Either way, to compare your chart with Walsh's, you would need to add about 1 M km^2 to Walsh' chart, which brings 1935 to about 7.8 M km^2, right in line with the compensated HadSST numbers.

Either way your 6.8 M km^2 for 1935 and the the 30's in general seems about 1 M km^2 too low.

[Diablobanquisa]

Hi Rob,

Regarding your first question: the 0.9 bias only applies from 1953 onwards, and not before. HadISST data before 1953 are of a very different nature. The reliability and the consistence of the record drop suddenly before that year, so you shouldn't directly suppose that the same bias correction applies before 1953.
Meier et al. 2012 stated: "the Walsh datasets (and hence the Hadley dataset) contain few observations before 1953 and use mostly climatological averages, so realistic time-varying extent records start in January 1953. Thus, here we employ the
1953–1978 period from Hadley, a 26-yr time series."
We did the same.

Regarding your second question: I don't know which data is using the new Walsh dataset for the satellite era, but I don't think their September values are 1 mill. sq. km. lower than NSIDC's ones.
I've extracted rough values (using this: http://arohatgi.info/WebPlotDigitizer/app/ ) from Walsh (p.17, green line):
September 2013: Walsh, 5.05. NSIDC, 5.35
September 2012: Walsh, 3.33. NSIDC, 3.62
September 2007: Walsh, 4.16. NSIDC, 4.29
September 2006: Walsh, 5.85. NSIDC, 5.91
September 1996: Walsh, 7.72. NSIDC, 7.87
September 1980: Walsh, 7.79. NSIDC, 7.83
September 1979: Walsh, 6.93. NSIDC, 7.19

So the differences range between 0.04 and 0.3 or so, far from the 1 M km2 you pointed out. Anyway, you shouldn't think that the same bias directly applies to all the lenght of the record.
For instance, Meier et al. 2012 present a mean September extent for 1972-1978 of 7.42 M km2. While in the data I've extracted from Walsh, the mean extent for 1972-1978 is 7.77 M km2 or so.
So, Walsh is 0.2 lower than satellites during 1979-2013, but 0.35 higher than satellites during 1972-1978.
(Indeed, in their last presentation, p.21, Walsh et al. state: "Next steps: Use of National Ice Center maps (gridded concentrations) as primary input for 1972-present." So we should still expect some changes in their data.

[Diablobanquisa]

http://www.star.nesdis.noaa.gov/star/documents/meetings/Ice2015/dayThree/1_Walsh_J_dayThree.pdf

p.17: sea ice extent data.
p.18: sea ice area data.

Maybe you were looking at p.18 instead of p.17.

[Diablobanquisa]

I've extracted again and more carefully the extent data values from the new Walsh dataset (p.17, green line). The satellite period looks almost identical to NSIDC data.

I've made some comparisons between our work and the new (and still in progress) Walsh dataset:

- First, the pre-satellite period 1935-1978: https://diablobanquisa.files.wordpress.com/2016/01/vspresat2.png

Walsh values are higher than ours during the whole period, but the trends are very similar. Upward trend from 1935 to 1979.

- Now, the whole period 1935-2014, including also Meier et al.: https://diablobanquisa.files.wordpress.com/2016/01/fullvv.png

From 1979 onwards, the three time series are identical: satellite data.
Before 1979, our values and those of Meier et al. are lower than Walsh.

It's worth remembering that Meier et al. adjusted the pre-1979 values in order to be consistent with (multichannel microwave) satellite data, using (single channel microwave satellite) ESMR data (1972-1978) as a bridge. http://www.the-cryosphere.net/6/1359/2012/tc-6-1359-2012.pdf

Meier et al. stated: "the adjusted Hadley
extent values are clearly more consistent with the XPM time series during 1972–1978. This consistency is obtained by shifting the Hadley estimates down to match the SII estimates and results in lower absolute extent values. The reverse procedure (shifting SII up to match Hadley) could be done if desired and is equally justifiable for consistency of the time series."

Walsh dataset doesn't seem to be adjusted yet. But they stated in their presentation that they will use National Ice Center maps as primary input for 1972-present. NIC values are higher than those derived from microwave satellites, so it seems that they will do a kind of 'reverse procedure' (shifting up the post-1979 period by using operational charts instead of satellite microwave data).

[Diablobanquisa]

(and shifting down 1972-1978)

However, using NIC charts has also some problems, since the time series derived from them are less consistent than those derived from microwave satellite data (it's the same that happens with MASIE: http://neven1.typepad.com/blog/2016/01/a-difference-in-nonsense.html )

So, I prefer using the satellite record (the most consistent time series) and adjusting the pre-satellite era.

[Rob Dekker]

I'm still skeptical of the low numbers for the early part of the series in both your and Walsh graphs, because of two reasons :

1) The August DMI charts which show more ice in more places than August satellite years that ended up well above 7 M km^2 in September, and
2) The discontinuity visible in the Walsh extend graph (page 17) : Prior to 1935 the extend is sharply higher.

But I am fully aware that I have not gathered enough data to contest your or Walsh' findings. Thus I would like to complement you on the work you have done. It's an excellent piece of work, sustained by clear graphs and arguments in your article and your comments alike.

Thank you, Diablobanquisa, for presenting your findings to us with this much detail and reasoning.

[Colorado Bob]

Warm Arctic Storms Aim to Unfreeze the North Pole Again — That’s 55 Degrees (F) Above Normal For January

It’s worth re-stating. The Starks were wrong. Winter isn’t coming. Winter, as we know it, is dying. Dying one tenth of a degree of global oceanic and atmospheric warming at a time. Steadily dying with each ton of heat-trapping greenhouse gasses emitted through our vastly irresponsible and terrifyingly massive burning of fossil fuels.

http://robertscribbler.com/2016/01/24/warm-arctic-storms-aim-to-unfreeze-the-north-pole-again-thats-55-degrees-f-above-normal-for-january/

[Jim Hunt]

Bob^2 is a bit behind the curve on this story. See e.g.

"More Heat Heading for the North Pole"

for a slightly more sober analysis from "Snow White".

She's also been shooting her mouth off about it on Twitter recently!

https://twitter.com/GreatWhiteCon/status/690961876113604609

Oh, and on Facebook too:

https://www.facebook.com/GreatWhiteCon/

[Neven]

Thank you, Diablobanquisa, for presenting your findings to us with this much detail and reasoning.

Thanks from me too, DB.

Warm Arctic Storms Aim to Unfreeze the North Pole Again

Anomalously warm, okay, but I don't believe it will get anywhere near 'unfreezing'.

[Jim Hunt]

Anomalously warm, okay, but I don't believe it will get anywhere near 'unfreezing'.

That makes two of us Neven, or three if you include "Snow White".

If I can persuade Typepad to display an image, here's the current forecast (not anomaly) for Monday night, which as near as the North Pole gets to "unfreezing" in the short term:

Green is above zero. Blue is not. Unridged ice up there will be well on the way to 2 meters thick by now. QED?

[Colorado Bob]

I always love these debates where X points out an extreme event , and Y says sorry you were 3 yards short. The point is it's above freezing North of Greenland in January.

In the dark.

[Colorado Bob]

So last night I re-watched the NOVA program :

Mystery Beneath the Ice

This is about the crash of the krill populations around Antarctica. By 50%, and in some places 90%. And they explain why this happening.

For the last few years we have heard the deniers harp on the slight growth on sea ice around Antarctica as proof the world is not warming. As anyone who who has studied this is aware, it's the fresh water layer flowing off the ice cap that is freezing. Fresh water freezes a 32 F degrees, salt water freezes at 26 F degrees. So the ocean surface is freezing at much higher temperature than in the past. Because there is an ever higher melt rate of fresh water sitting on top of the ocean.

But as this was happening , something else, far more worrisome was taking place. The melt season is 90 days longer. It melts sooner, it freezes later. The Antarctic peninsula has warmed by 11 F degrees in the last 50 years.

Nobody told the krill.

All this change has slammed this key species in the food web. The Arctic gets lot's of focus, but the changes at the bottom of the world are more more worrisome. I wish everyone would watch this program. It's a very important piece of the new puzzle we find ourselves in.

[Colorado Bob]

To add on -

Ryan in New England / January 25, 2016

Well said, Bob. This was a very enlightening episode, and the images beneath the ice were breathtaking. As Bob points out, the melt season is 90 days longer, but only now are experts realizing how important the ice is to the krill. The krill feed on phytoplankton, which don’t grow during the sunlight free depths of winter. Even the experts were uncertain how the krill sustained themselves through the winter. So they went there and drilled through the ice and dove into the icy depths to find out.

It turns out the krill actually graze on the bio-film found frozen on the underside of the sea ice. The images are beautiful. Countless krill grazing along the surface of the ice like herds on the plains of Africa. It just goes to show how little we understand the fundamentals of systems we are destroying. It’s an example of how the small changes we’ve already seen (small compared to what’s coming) can have profound effects.

Like Bob has warned us before, it’s the little things that will cause our demise. And we are killing the little things.

Colorado Bob / January 25, 2016

Ryan –
Thanks , I needed that. An excellent follow on .

To see those baby krill feeding on an ever more thinning grazing area, it reminded me of the caribou. Their numbers are crashing because ice coats the land, and not snow. It reminded me of moose where ticks get the upper hand.

It’s these changes where we see nothing , and nature is reeling.

[Colorado Bob]

Neven

Many , many people gain so much from your work. And the people who add so much here.

Thanks for it all .

It’s these changes where we see nothing , and nature is reeling.

[Jim Hunt]

There you go again Bob. My point is that in actual fact it's not "above freezing North of Greenland in January. In the dark."

It is however currently "above freezing" in Svalbard, and (just!) above -25°C in the North of Greenland:

http://www.wunderground.com/q/zmw:00000.1.04312?sp=04312

Let's settle for Neven's "anomalously warm" shall we? In January. In the dark.

[Neven]

Thanks, Bob, and thanks for the man useful links.

The point is it's above freezing North of Greenland in January.

In the dark.

Again, even if this were 100% true, this might not be so unusual. We don't need spectacular events every two weeks to make a point. It's the slow, ongoing decline that's the point. The Arctic has been (increasingly) anomalously warm for several years now. Sooner or later, this will be reflected in Arctic sea ice cover.

These events are pretty interesting in themselves (not all of them), but in the end it's all about the effect they're having on what the ice does during the melting season.

[John Christensen]

I would agree with Neven; the anomalous events do not appear to be that far off compared to the historic record, at least when looking at DMI 80N.

The spike in December peaked for 80N at 257K, which is -16C. The DMI 80N record from 1958-2002 seems to show 13 similar or more significant temperature peaks during winter months - although the incredible spike in early January 2000 looks like an error - maybe Y2K related..;-).

In January 2016 the 80N temp is hovering at -19 to -26C, compared to the normal temp of -29 to -31C.

However, you cannot really understand the temperature readings in a meaningful way, unless you consider the weather that caused the temperature.

The weather in the past weeks seems to have been influenced in a major way by the clashing of the pool of warmer than normal water south of Nova Scotia and Newfoundland with the pool of cooler than normal water south of Iceland and Greenland. This natural breeding ground of winter storms has therefore been more active than usual, causing stronger than usual storms to enter the Denmark Strait pounding both Iceland and south east Greenland with much more snow than usual, and also causing equally higher than usual foehn in south west Greenland.
And finally, due to more lows than usual climbing up the east coast of Greenland, winds around Svalbard and the Fram Strait, have been primarily from the south, causing warm, moist air to get near the Pole and cause the consistently high 80N temp readings.

However, with the AO turning positive again, it seems the next low lingering in the Denmark Strait will move east with the jet stream, and a high will develop over the Arctic causing the gyre to get started and the 80N temp to move closer to normal.

Makes me wonder if the ice during peak winter months prefers the 'not bird nor fish' style of weather that also seems preferable in peak summer months..

[Kris]

Neven wrote:

It's the slow, ongoing decline that's the point. The Arctic has been (increasingly) anomalously warm for several years now.

Yes indeed. But it looks to me we are forgetting about the direct impact.

Do allow me to remind you all of the basics of the Jet Stream. A "ribbon band", created by the difference between cold and warm air. The greater the difference between cold and warm air, the stronger the jet stream, and in our case, the better the Polar Jet will be able to keep the cold air into the Arctic.

As we all know, in the past few years the anomalously warm Arctic air weakened considerably the Polar Jet, triggering this Polar Jet to meander in giant meanders. Resulting in large pockets of cold air moving to the South.

Last Winter the East North America had been struck for weeks by this phenomena. This Winter, until last week, the cold air moved nearly always into the central Atlantic - to be warmed up there and then to be moved back again into the Arctic, with even greater warmings up as a consequence.

It even happened this year the Polar Jet and the Subtropical Jet merged, at the longitude Florida --> Portugal!

[Neven]

Kris, sure, and I try to stress as often as I can that Arctic Amplification and what happens to Arctic sea ice cover has consequences, one of which is changes to the jet stream. Some occasions offer the opportunity for that, others don't. At least, that's how it works for me (and sometimes I'm just lazy ;-) ).

What I find more interesting right now, is what is happening to global sea ice area. It's approaching the period in which its minimum occurs, and it is already 4th lowest compared to years since 2005 (only 2006, 2007 and 2011 are lower). I'm not sure if a new record is possible, but it would be funny, as climate risk deniers use it as an argument that all is well (like saying that more obese than undernourished people means there is no world hunger).

Other than that it's not a very useful metric. But it's a race, and races can be (statistical) fun.

[Jim Hunt]

Neven - See also this ASIF poll on how low Antarctic sea ice extent will go this SH summer:

http://forum.arctic-sea-ice.net/index.php/topic,1467.0.html

Some more sad news from Antarctic:

http://www.bbc.co.uk/news/uk-35398552

The BBC has just reported that:

Explorer Henry Worsley has died after suffering exhaustion and dehydration during a solo attempt to cross Antarctica.

The former Army officer, 55, was 71 days into his effort to become the first person to cross the Antarctic unaided.

His wife Joanna announced the news, saying she felt "heartbroken sadness".

[John Christensen]

Kris said:

As we all know, in the past few years the anomalously warm Arctic air weakened considerably the Polar Jet, triggering this Polar Jet to meander in giant meanders. Resulting in large pockets of cold air moving to the South.

Do you have something to substantiate this hypothesis Kris?

As you seem to be cognizant of the works of the jet stream, you will also know that when the AO index is positive, then the jet stream moves further north and strengthens, while when the AO index is negative, the jet stream moves further south and slows down, following a more wobbly path.

Do you have anything substantial showing that:
- The jet stream has slowed down under AO positive or negative conditions compared to 'normal' jet stream speed?
- Or that negative AO (causing slower, meandering jet stream) has become more frequent than 'normal' occurrence of negative AO index?

[John Christensen]

And from Kris also:

As we all know, in the past few years the anomalously warm Arctic air weakened considerably the Polar Jet, triggering this Polar Jet to meander in giant meanders. Resulting in large pockets of cold air moving to the South.

Last Winter the East North America had been struck for weeks by this phenomena.

Well, yes the jet stream caused the cold in East North America in 2015, but it was caused by the negative AO index.

Considering the history of the AO and jet stream during winter months, it is interesting to note that the AO was primarily negative from 1950 to 1987 and then turned positive until 2011.
And yes; in 2015 it was negative, so nothing unusual about that:

http://www.cpc.ncep.noaa.gov/products/precip/CWlink/daily_ao_index/JFM_season_ao_index.shtml

When you compare the temperature in the state of NY in 2015 to the longer record, you also note that while it was cold in 2015, the anomaly was not at all extreme:

http://www.ncdc.noaa.gov/cag/time-series/us/30/0/tavg/1/1/1895-2015?base_prd=true&firstbaseyear=1901&lastbaseyear=2000

In other words, the jet stream has been meandering many times in the past, and much more forcefully than in the past few years.
The relative dominance of positive AO during winter from 1987-2011 could have created a false impression that Arctic temperatures are impacting the jet stream.

[Kris]

John Christensen wrote:

Do you have something to substantiate this hypothesis Kris?

My worthless person hasn't anything at all. OTOH, the Britts have, so do have a search onto the netweather website.

Incidentally, do you have something to substantiate the AO is all that important!???

[Kevin O'Neill]

John - Regarding the changing jet stream: Jennifer Frances has done a lot of work in this area.

There's also a new paper showing much the same results: Arctic Air Masses in a Warming World

"In the model, a pattern with negative θe850 anomalies over the central Arctic becomes less frequent in the future. There is also an increase in the frequency of patterns associated with an amplified ridge (trough) with positive (negative) θe850 anomalies over western (eastern) North America. We hypothesize that the increase in frequency of such patterns is the result of enhanced forcing of baroclinic waves owing to reduced sea ice over the western Arctic."

[John Christensen]

From Kris:

Incidentally, do you have something to substantiate the AO is all that important!???

That is a very curious comment, since the jet stream is defined by the AO..

Think about that for a minute.

[John Christensen]

Hi Kevin,

IMO Jennifer Francis is looking at the tail of the elephant.

[Leslie Graham]

Turns out Colorado Bob was right and the temperature DID briefly rise above freezing almost all the way to the north pole.
Check out Reanalyzer at 0300 Monday 25th Jan.
Not that I'm claiming the first recorded time the temperature has ever risen above freezing at the north pole in January is important or anything but I think it IS important to be accurate and note that it did indeed happen.

[Kevin O'Neill]

JC writes: "IMO Jennifer Francis is looking at the tail of the elephant."

Increasing CO2 leads to a warmer globe. This is amplified in the arctic. A warmer arctic leads to less sea ice. Less sea ice leads to changes in synoptic weather; including a jet stream that moves slower and with increased waviness.

If you mean the jet stream is the tail, yes she's looking at the tail, but not because she's ignoring the rest of the elephant.

Otherwise I have no idea what you mean.

[Neven]

Turns out Colorado Bob was right and the temperature DID briefly rise above freezing almost all the way to the north pole.

Indeed, the anomalous heat is reaching closer than I thought it would go:

[Rob Dekker]

Maybe we should carry this discussion to the "North Pole temperature anomaly (big)" thread, but this latest influx of heat apparently reaching the NP was predicted both by ClimateReanalyser
http://pamola.um.maine.edu/fcst_frames/GFS-025deg/ARC-LEA/T2/01.png
and by DMI
http://ocean.dmi.dk/anim/index.uk.php
for the past 5 days.

Interestingly enough, this latest anomaly does not yet seem to show up in the buoy data.
Legendary buoy 300234062785480 located between Svalbard and the NP is still hovering around -15 C :
http://iabp.apl.washington.edu/raw_plots.php?bid=300234062785480#top
and buoy 300234062788470 just behind the NP is still logging in at -30 C :
http://iabp.apl.washington.edu/raw_plots.php?bid=300234062788470#top

Also DMI's "north-of-80" temp tool is not yet showing exceptional temps
http://ocean.dmi.dk/arctic/meant80n.uk.php

Let us check in a day or two, because it seems that this latest wave of heat from the Atlantic is just a bit late.

After all, Climate ReAnalyser and DMI are forecast tools.

[P-maker]

The North Pole is under attack from melting temperatures and the Tropics are under attack from freezing cold air masses. Although this song was written after the Spanish Civil war and this particular version was recorded after the Breivik shootings, the message is evident. We are “surrounded by enemies” and this blog is our main weapon against it! Please take a deep breath and enjoy: http://www.magcartz.com/v/y-xv56p4AlI/til-ungdommen-sunget-af-sissel-kyrkjeb%C3%B8-norge-mindes-de-77.aspx

[Rob Dekker]

Kevin O'Neill said

If you mean the jet stream is the tail, yes she's looking at the tail, but not because she's ignoring the rest of the elephant.

Otherwise I have no idea what you mean.

Thank you Kevin. Very well said.

[Rob Dekker]

Diablobanquisa, since you state "we decided not to use DMI maps" I am still wondering exactly which data you used to reconstruct the 1935 (and the 30's in general) sea ice extent.

From your block I read :

En tercer lugar, extendemos la serie hasta 1935 combinando los datos de AARI con una climatología seleccionada en base a la correlación entre temperatura y banquisa

But what does that mean exactly for 1935 or any year in the 30's ?

[Neven]

Rob, if I've understood correctly, they extrapolate the correlation between SAT and sea ice cover (during the satellite era, I assume), constrained by historical data for the Siberian regions as provided by AARI. Is that correct, Diablo?

[Rob Dekker]

Also, I see you calculated confidence intervals for the series back to 1944 :
https://diablobanquisa.files.wordpress.com/2015/12/fig-7.png
but not before that.
Why not ?

[AbbottisGone]

Global sea ice must of course be a prominent statistic for the general public to accept as changing before the world at large feels the need to rearrange its priorities... The explanation of the Southern Hemisphere obviously hampers that.

This is why I say the clearest and quickest indication to the mass public can only be the signals being given by multi-year sea-ice! Any idea how those indicators are going?

I make this point: deniers are scared of multi-year sea-ice statistics and I claim that as a fact worth knowing! Pity no one talks about them!!

[Jim Hunt]

Leslie/Rob - Forgive me for being pedantic, not least because I'm currently debating such issues with some "skeptical" types in the Twittosphere, but your CCI charts show above freezing temperatures NOT reaching the North Pole.

FYI I've even been debating "coverage bias" in the Arctic with the likes of NASA and BEST recently:

https://twitter.com/jim_hunt/status/690221204041502720

Quoting "Gavin Schmidt":

We use interpolation from land-based weather stations. Comparisons of this to Arctic Buoy records shows a pretty good relationship.

As Rob points out, those images are forecasts not measurements. It will indeed be interesting to see what the few thermometers in the region reveal, but there is no doubt that Longyearbyen was "above freezing" for quite a while, although that is no longer the case.

I can state with very high confidence that the North Pole hasn't just "unfrozen". How about you?

[John Christensen]

Kevin said:

Increasing CO2 leads to a warmer globe. This is amplified in the arctic. A warmer arctic leads to less sea ice. Less sea ice leads to changes in synoptic weather; including a jet stream that moves slower and with increased waviness.

If you mean the jet stream is the tail, yes she's looking at the tail, but not because she's ignoring the rest of the elephant.

Kevin, seriously, this is non-sense.

This is what James Hansen has to say about the AO and the jet stream:

"The degree to which Arctic air penetrates into middle latitudes is related to the AO index, which is defined by surface atmospheric pressure patterns. When the AO index is positive, surface pressure is low in the polar region. This helps the middle latitude jet stream to blow strongly and consistently from west to east, thus keeping cold Arctic air locked in the polar region. When the AO index is negative, there tends to be high pressure in the polar region, weaker zonal winds, and greater movement of frigid polar air into middle latitudes."
(Hansen, James; Reto Ruedy; Makiko Sato; Ken Lo (2009). "If It’s That Warm, How Come It’s So Damned Cold?")

The cold spell in 2010 explained by the NCDC:
"Cold arctic air gripped western Europe in the first three weeks of December. Two major snowstorms, icy conditions, and frigid temperatures wreaked havoc across much of the region...The harsh winter weather was attributed to a negative Arctic Oscillation, which is a climate pattern that influences weather in the Northern Hemisphere. A very persistent, strong ridge of high pressure, or 'blocking system', near Greenland allowed cold Arctic air to slide south into Europe. Europe was not the only region in the Northern Hemisphere affected by the Arctic Oscillation. A large snow storm and frigid temperatures affected much of the Midwest United States on December 10–13...."
(http://www.ncdc.noaa.gov/sotc/2010/12)

If I should provide a hypothesis of how the AO (and therefore the jet stream) could be impacted in a warmer world with less Arctic sea ice, then the increase in open water in the Arctic should cause an increase in fall and winter low pressure in the Arctic.
This would lead to a positive AO index becoming more prominent, which again would move the jet stream further north and keep the jet stream strong.

Dr. Jennifer Francis to my knowledge is not tackling the challenge of whether we will see slow, meandering jet stream paths in AO positive conditions, or if we will see AO negative conditions becoming more prominent, as they were prior to 1987.
But if AO negative becomes more prominent, as it was until 30 years ago, what then caused it to change: Ocean circulations or global warming?

Dr. Jennifer Francis is looking at a limited geographic area over a limited period of time, and her conclusion therefore seems to miss the broader influence of the AO in our NH atmospheric patterns.

[Neven]

If we're going to discuss this, maybe we should move it to some other thread, or the Forum?

[P-maker]

John,

” the increase in open water in the Arctic should cause an increase in fall and winter low pressure in the Arctic. “

Could you please substantiate your assumption here (or in the thread proposed by Neven) before you move on.

I seem to recollect a situation a few years ago, when autumn weather was dominated by persistent lows in three distinct areas – Barents/Kara, Baffin Bay and Beaufort Sea. Although these areas are not part of the Arctic Ocean proper, this still may constitute an new pattern yet to be seen emerge more permanently, as sea ice disappears.

[Leslie Graham]

Jim
I'm not familiar with CCI charts so can't comment, but I did notice that on 'Reanylizer' the north pole was showing above freezing temperatures at 0300 on Monday.
I had presumed that 'after the fact' charts were actual temperatures and not still projections. There are reasons I hesitated to post; I'm not an expert, I'm fairly sure 'unfreezing' isn't even a word and I'm not going to die in a trench defending any cause
However, yes - I do feel you are being just a tad pedantic about it. It doesn't really matter if the North Pole was still a half degree below freezing after all. The point I was trying to make was that these temperatures are, in my limited experience, absolutely extraordinary and likely unprecedented.
UCAR’s North Pole temperature data record, for example, shows not a single day above freezing at the North Pole during January since records began in 1948.
I just thought this was worth a mention.
This might also be off topic but it's currently colder in Vietnam than it is on the northern Siberian Arctic coast.

[John Christensen]

You are right, sorry Neven.

The best thread may be this one, which I guess I killed with all the AO numbers a while back:

http://neven1.typepad.com/blog/2015/08/a-wetter-and-warmer-arctic.html

P-maker,

Please see my comments there.

[Diablobanquisa]

Neven wrote: 'Rob, if I've understood correctly, they extrapolate the correlation between SAT and sea ice cover (during the satellite era, I assume), constrained by historical data for the Siberian regions as provided by AARI. Is that correct, Diablo?'

Yes, Neven, it's absolutely correct.

Rob Dekker wrote: 'Also, I see you calculated confidence intervals for the series back to 1944 :
https://diablobanquisa.files.wordpress.com/2015/12/fig-7.png
but not before that.
Why not ?'

We started with a 10 year trend, because trends for shorter periods are very noisy.

That graph shows the evolution of the trend as years are added to the end of a time series starting in 1935. The first point of the black line shows the trend of a 10-yr time series (1935–1944), and subsequent points extend the time series by one year each up to a trend over the full 80-yr 1935–2014 time series.

It's the same that Meier et al. (2012) showed on their Fig.3.

In our graph, the green, blue and red lines (limited by the grey area) show the 99% confidence level, each line according to a different method. (according to the first method, an upward trend is significant when both green lines are above zero, and a downward trend is significant when both green lines are below 0; the same applies to blue and red lines with the other two methods)

[Rob Dekker]

Diablobanquisa, thank you for your patience with me.
I understand the confidence intervals graph now. It is for the trend, not the confidence in the data itself.

But the actual determination of the early data is still a mystery to me :

Neven wrote: 'Rob, if I've understood correctly, they extrapolate the correlation between SAT and sea ice cover (during the satellite era, I assume), constrained by historical data for the Siberian regions as provided by AARI. Is that correct, Diablo?'
Yes, Neven, it's absolutely correct.

I'm probably just slow, but how exactly does that work ?
Could you possibly explain how you obtain an example year (such as 1935) ?

[Rob Dekker]

Apologies for being so skeptical about the 30's extent numbers.
But here is another reason why : You post the September 1935-1944 median ice extent numbers in this graph :
http://neven1.typepad.com/.a/6a0133f03a1e37970b01bb08ae82ac970d-pi
Note that it looks like the North-West Passage is virtually open.
And this is the "median", where your earlier years show even less ice.

If that were really the case, why did the first NW Passage (after Amundsen's 1906 partial traversal) not happen until 1942 by Larsen ?And it took him 2 years to get through.

[John Christensen]

Hi Rob,

Considering the length of the NW passage from the Baffin Bay to the Chuckchi Sea, the brief period of low ice extent late Aug/early Sep, the area that must have been covered with 20-60% ice concentration, and lastly the speed of the ship at the time, it would be impossible to make the journey in one season, even if the steady ice pack seemed to leave the passage open at the peak of the minimum.

[Diablobanquisa]

Rob Dekker wrote: I understand the confidence intervals graph now. It is for the trend, not the confidence in the data itself.

Exactly.

how exactly does that work ?
Could you possibly explain how you obtain an example year (such as 1935) ?

We calculated the correlation between SAT and September SIE during the satellite era (1979-2014): r=-0.91.

We calculated the correlation between SAT and our September SIE during the period 1953-2014: r=-0.88.

We smoothed with a 18 year filter both SAT and September SIE time series and calculated the correlation between them during 1953-2014: I'd have to find the exact number, but it was almost -1.

Taking into account this strong negative correlation, and the physics behind it, we thought that SAT could be used as a proxy for September SIE.

Then, we calculated the SAT anomaly for the 18 year lenght 1935-1952 period: 0.58ºC above the 1951-1980 mean.

We looked for the 18 year period during 1953-2014 with the closest SAT anomaly: it is the 1985-2002 period, with a SAT anomaly of 0.59ºC above the 1951-1980 mean.

We calculated the median sea ice extent from the gridded data for the 1985-2002 period (6.994.000 km2).

We used this median sea ice extent for 1935-1952, incorporating into it the AARI data for each year of the period. For instance, for 1935 we used these two AARI charts: ftp://sidads.colorado.edu/pub/DATASETS/NOAA/G02176/pngs/1935/e19350902_tc.v0.png and ftp://sidads.colorado.edu/pub/DATASETS/NOAA/G02176/pngs/1935/w19350916_tc.v0.png.

Interestingly, the incorporation of AARI data barely changes the mean extent for the 1935-1952 period (from 6.994.000 to 7.039.000 km2, a difference of +45.000 km2). So, the median sea ice edge derived from AARI maps for 1935-1952 matches very closely the sea ice edge of the 1985-2002 climatology.

We tested the use of other climatologies.
For instance, we looked for the coldest 18 year period during 1953-2014: it is 1963-1980, with a SAT anomaly of -0.13ºC below the 1951-1980 mean. We calculated the median sea ice extent for this period: 7.707.000 km2, and used it for the 1935-1952 period. The incorporation of AARI data into this climatology resulted in a change of -340.000 km2. From 7.707.000 to 7.367.000 km2.

We tested also a warmer period: 1995-2012, with a median sea ice extent of 5.967.000 km2. The incorporation of AARI data resulted in a change of +430.000 km2, from 5.967.000 to 6.397.000 km2.

So, if we use a higher climatology, AARI data push it down. If we use a lower climatology, AARI data push it up. If we use the selected climatology (on the basis of the correlation between SAT and SIE), AARI data barely change it. This strenghthens the confidence on the selected climatology.

In addition, we also performed the regression equation of September SIE on summer SAT (on the basis of their relationship during the satellite period) and used it for estimating September SIE for 1935-2014. This estimation is shown by the grey area on Fig.9: https://diablobanquisa.files.wordpress.com/2015/12/fig-9.png The black line is our time series, and the blue line is HadISST.

Rob wrote: You post the September 1935-1944 median ice extent numbers in this graph :
http://neven1.typepad.com/.a/6a0133f03a1e37970b01bb08ae82ac970d-pi
Note that it looks like the North-West Passage is virtually open.

Look at this high resolution version of the same graphic: https://diablobanquisa.files.wordpress.com/2016/01/fig-6-eng.png

I'd say Northwest Passage is closed. Parry Channel is fully closed. Even Amundsen's route presents some ice covered areas. Indeed, the ice extent in the Nortwest Passage seems similar to the mean of the 1971-2000 period, and it isn't so different from what explorers found during the XIX Century: http://onlinelibrary.wiley.com/doi/10.1029/2003EO400003/epdf "The extent of summer sea ice during the 19th century (at the Northwest Passage), insofar as it is shown in patterns of navigability inferred from ship tracks, the direct observations of explorers, and a number of native accounts, is remarkably similar to present ice climatology" Present ice climatology means 1971-2000 average, when the Passage was closed (obviously, the sea ice extent during the last decade is much lower).

Cheers

[Rob Dekker]

Ah. Now I understand.
In my words : You created your own gridded climatology for the 1935-1952 period, by using 1985-2002 sat images since that period matches the temperature range of 1935-1952.

On that brand new gridded climatology map you then applied the gridded AARI observations for the Kara and the Siberian Sea, and that is how you obtained your 1935-1952 extent numbers.

Doublecheck : So if you would not have applied AARI we would have seen a flat line at 6.994.000 km2 in your series, from 1935-1952.

Did I get that right ?
If so, I think that is very creative but it raises a couple of comments and a few questions. Let me do the questions first :

1) Why did you only apply AARI observations ? Why not other observations that Walsh and Chapman included in their 2001 paper ?

2) Which gridded climatology did you use for the 1953-1978 period ?
And how do you argue that these two periods, with two different gridded climatologies, are still "consistent" ?

And may I add that the correlation numbers between temp and SIE are indeed remarkable. You may be onto something there.

[Rob Dekker]

One more question, this one about the temp - SIE correlation :

You choose April - Sept HADCRUT4, 70 - 90 deg as your temperature reference. Could you tell us a bit more about that choice ?
Did April - Sept give the best correlation with Sept SIE ?

I'm curious, because SIPN is looking for better ways to predict Sept SIE based on data earlier in the year.

[Diablobanquisa]

Hi Rob,

Yes, you got it right (although AARI covers Kara, Laptev, ESS and some years parts of Chukchi and Barents).

The only 'observations' that Walsh and Chapman used in their 2001 paper for 1935-1952 are August DMI maps and Kelly grids (Kelly grids are just the digitized white area of DMI maps). So, there wasn't too much to use... (discontinuous and sparse data, with large temporal and geographic gaps, without knowledge of their internal consistence, whithout knowledge of their consistence with satellite data, that are needed to interpolate from August to September...)

On the other hand, AARI maps are derived from direct and systematic observations, mostly from aircrafts, without gaps, used for operational purposes (shipping along the Northern Sea Route), that we tested against satellites during a overlapping period... So, AARI data are much more complete, reliable and consistent that DMI maps.

We didn't use any climatology for the 1953-1978 period. We didn't need it because there are enough data. The data for the years of this period is a combination of AARI with adjusted HadISST. No climatologies involved.

Why do I argue the 1935-1952 is still consistent in our time series? Because AARI is tested to be consistent, and because the relationship between SAT and SIE is very high during 1953-2014. Why should it be significantly different during 1935-1952? (there are obvious physics linking SIE and SAT).

Yes, we choose April - Sept CRUTEM4, 70 - 90 deg because it gives the best correlation with September SIE.
We took CRUTEM4, GISS and BEST. We calculated SAT anomalies from these three datasets for different months and periods of months and for different latitudinal bands. We got thousands of SAT anomaly time series. We calculated the correlation of each of them with our time series during 1953-2014. The highest correlation (r=-0.88) was found to be with April - Sept CRUTEM4, 70 - 90 deg.

(It would be interesting to use DMI and ACSYS August data for the 1935-1952 period, but there are a lot of issues to solve in order to keep them consistent with the rest of the time series, maybe for our next version).

(In their new product, Walsh et al. are merging AARI, ACSYS and DMI for that period, filling the gaps with an 'analog approach' instead of a climatology).

(As we saw, although absolute values are different, the trends for 1935-1978 derived from our time series and from the new Walsh dataset, seem pretty similar and consistent)

[Neven]

Diablo, something else: Are Walsh and Meier, etc., aware of your paper? And do you consider publishing in a perhaps higher profile English journal? Or maybe The Cryosphere?

[Diablobanquisa]

Hi Neven,

Well, while we were working on the article, we sent a couple of questions to Walt Meier, that he very kindly answered. Recently, I wrote him to tell him that the paper has been published. So yes, he is aware of our work.
Not Walsh or others, unless they read your blog...

We are fully amateur, so this work was conceived as a zero cost project. Considering that our english skills aren't enough to write a paper ourselves, we would had had to pay for a translation. Besides, we wanted to publish it in an open access journal. Unfortunately, most open access journals aren't free for authors...

Maybe for the next one, we will consider again if it's worth translating and trying to publish it in an english and maybe higher profile journal.

Thank you so much, Neven.

[Rob Dekker]

Thanks Diablobanquisa for your continued replies.
Let me first answer your question:
You asked:

Why should it be significantly different during 1935-1952? (there are obvious physics linking SIE and SAT).

Yes, there are obvious physics linking SIE and SAT. And your regression analysis over the satellite era confirms that (with correlation of r=0.9).
But the problem for the 1953-1978 period is that you are working with sparse data sets, both for temperature observations and SIE observations. And that can break the correlation to the point where even the trend is not clear.
For example, can you please determine the correlation between your SAT data (HATCRUT4 April-Sep 70+ deg) and your final SEI numbers over the same period 1953-1978 ?
I suspect that you get much lower correlation than during any time during the sat era (where you obtained 0.9).

I still do not understand how you obtained your 1953-1978 numbers.
You write :

The data for the years of this period is a combination of AARI with adjusted HadISST. No climatologies involved.

How did that work exactly ?
On this graph :
https://diablobanquisa.files.wordpress.com/2016/01/image001.png
HadISST gridded hangs around 9.9 M km2 for 1953, and
(non-gridded) Meier et al logs in at about 7.9 M km^2 for 1953, while
your estimate is about 7.2 M km2.

Did you obtain that number (7.2) only by doing AARI adjustment on the HadISST gridded product ? Or did you do something more ?

[Rob Dekker]

I'm sorry, Diablobanquisa. In my prior post I ask for the correlation between SAT and your SIE numbers, for the 1953-1978 range, but I meant to say the 1935-1953 range.
That is the period that you used your own climatology, based on SAT, and my point is the SAT-SIE link breaks over that period with sparse data.

If the SAT-SIE correlation over that 1935-1953 period is weak, you cannot draw conclusions about trends between that period and later periods.

[Diablobanquisa]

Did you obtain that number (7.2) only by doing AARI adjustment on the HadISST gridded product ?

Yes, only by incorporating AARI data into adjusted HadISST:

Firstly, we adjust HadISST gridded data to match the satellite data (original HadISST for 1953 is 8.71, and adjusted HadISST is 7.88)

Secondly, we incorporate AARI data into adjusted HadISST. And it gives 7.20 for September 1953.

We do the same for every year during 1953-1978.

The correlation between our SIE dataset and SAT during 1953-1978 is r=-0.57 (p value 0.0006)
The correlation using HadISST instead of our timeseries is lower: r=-0.40 ( p value 0.0260)
So, the incorporation of AARI data improves the correlation with SAT.

To answer your question about SAT/SIE correlation just during 1935-1952: r=-0.54 (p value, 0.0444)
The correlation using HadISST instead of our timeseries is lower: r=-0.40 ( p value, 0.3586).

The lower correlation during 1935-1952 doesn't come as a surprise, because the reliability of the interannual variability in our time series for the 1935-1952 period is lower than from 1953 onwards, because during the former we use a flat climatology for the areas not covered by AARI.

But the problem for the 1953-1978 period is that you are working with sparse data sets, both for temperature observations and SIE observations.

I think you were refering to 1935-1952.
Although both are sparse, I'd say that the instrumental temperature record is much more complete, consistent and reliable that the sea ice datasets for that period.

Even if the correlation during that period is not as high as during the satellite era, I'd say it's due to a lack of fully realistic interannual variability in our sea ice data for that period, as I explained above.

If the SAT-SIE correlation over that 1935-1953 period is weak, you cannot draw conclusions about trends between that period and later periods.

I don't think so. If you trust the instrumental temperature record for that period, and taking into account the strong correlation between SIE and SAT on decadal time scales, you should conclude that the average extent during 1935-1952 must be quite close to ours (unless you have arguments to explain why the relationship between actual SIE and actual SAT on decadal time scales should be significantly different before 1953).
So, I think it's all about trusting the instrumental temperature record before 1953, or not.

In conclusion:

- is our September SIE time series perfect? Absolutely not.

- is our September SIE time series better than the previously existing ones? (HadISST and the 'old' Walsh; remember that the 'new' Walsh is a work still in progress and not published yet).

I'd say yes.

We focused on September and we focused on consistence.

HadISST and the 'old' Walsh don't use any direct observation for September during 1935-1952. Their September data during that period is fully extrapolated, interpolated, inferred... from:
- Some sparse direct observations during August (DMI maps - red areas, 1935-1939 and 1946-1952)
- Inferred sea ice edge for August (DMI maps - white areas, 1935-1938)
- Something guessed as an inferred sea ice edge (DMI maps, white areas, 1939 and 1946-1952) that doesn't look as an inferred edge, but as an 'area without data' and that fully disagrees with the available direct observations from other sources.
- It's worth remembering that the full 1940-1945 period is built without any data (DMI didn't produce any map for those years).
- Remember also that the consistence of DMI maps with other records is not tested.

And, what we use for September 1935-1952?:

- A continuous dataset of direct and systematic observations during September for the whole Siberian sector (and their consistence is tested).
- A climatology to fill the gaps, derived from a very strong correlation between SAT and SIE on yearly (-0.91) and decadal (-0.96 smoothed with a 18 year filter) time scales during 1953-2014.

Now, let me ask you a couple of questions: - why do you think that 1935-1939 values must be significantly higher than ours?
- why do you think that the trend from mid 30s to 60s and 70s can't be positive? I don't understand the reasons why you look to be so sure about these two questions...

Cheers

[Diablobanquisa]

Instead of...
- A climatology to fill the gaps, derived from a very strong correlation between SAT and SIE on yearly (-0.91) and decadal (-0.96 smoothed with a 18 year filter) time scales during 1953-2014.

...I should have said:
- A climatology to fill the gaps, derived from a very strong correlation between SAT and SIE on decadal time scales during 1953-2014 (-0.96 between the two time series smoothed with a 18 year filter ; remember that we were using SAT as a proxy of SIE for a 18 year period, 1935-1952).

[Neven]

Maybe for the next one, we will consider again if it's worth translating and trying to publish it in an english and maybe higher profile journal.

1. Your English is really good, and you obviously speak 'science English'.
2. I work as a translator (albeit mostly for audiovisual translations), but if you need a proofreader, let me know.

I believe The Cryosphere (which you probably know) is very well-read by cryospheric scientists. I think your paper could fit in well there.

[Diablobanquisa]

Thank you so much, Neven, it's very kind of you.

[Diablobanquisa]

Hi again Rob,

Regarding your doubts about how we obtained our 1953-1978 numbers, I'd like to explain it again:

- Firstly, we adjusted HadISST September 1953-1978 extent values almost exactly as Meier et al. did. (we use a different grid, so we found the bias to be of 0.83 instead of 0.9)

- Secondly, we apply this adjustment to HadISST gridded data. How do we do this?

Meier et al. stated: "The adjustments to match Hadley with the adjusted XPM (and thus SII) are larger, particularly during summer (Table 2). This is because, as mentioned above, Hadley is more conservative in detecting thin ice and not underestimating
melt effects. In the original Hadley fields, passive microwave summer concentrations and extents are adjusted upward to account for that bias. Here, we effectively do the reverse. We ignore the passive microwave bias – focusing on consistency for trends and variability – and adjust the Hadley downward to be consistent."

So, the bias is very likely due to low concentration areas in the Marginal Ice Zone.

Taking this into account, we adjusted HadISST gridded data for every September during the 1953-1978 period by removing the grid cells with lowest concentrations until we reach the target extent (0.83). See Fig. 1: https://diablobanquisa.files.wordpress.com/2016/01/fig-1-eng-v3.png

The resultant time series is almost identical to Meier et al. However, we also adjusted the gridded data instead of adjusting only the extent values.

- Thirdly, we incorporated AARI gridded data into the adjusted HadISST gridded data. (Both transformed in extent data, to avoid the lack of consistency on sea ice concentration). See Fig.2: https://diablobanquisa.files.wordpress.com/2016/01/fig-2-eng-v3.png

And that's all.

(During 1952-1971, and specially during the 50s, HadISST has sparse data for the Siberian Sector, and the incorporation of AARI data into HadISST has been recommended by Mahoney et al. 2008.)

The incorporation of AARI data during the target period 1972-1978 barely changes its mean extent, from 7.42 to 7.40 (-0.02). So, AARI data result pretty consistent during the target period.

The incorporation of AARI data during 1953-1971 improves the reliability of the record in the Siberian sector, and results in a change of -0.16 on the mean extent of the period, from 7.76 to 7.60.

Some years the incorporation of AARI data reduces extent (1953, 54, 55, 56, 59, 60, 61...) and other years it results on an extent increase (1957, 58, 62, 63, 68...). We could say that it corrects the sea ice edge in the Siberian sector, where HadISST has its lowest density of direct observations.

As a result of the incorporation of AARI data, the correlation between SAT and SIE during 1953-1978 increases from -0.42 to -0.57.

Cheers

[Rob Dekker]

Thank you Diablo, for your extensive explanations.
I think you did a great job in trying to reconstruct ice extent in the Arctic back to 1935.
I have some concerns with your methods, summarized as follows :

1) You use a different 'baseline' for the 1935-1953 period than for your 1953-1976 period, which causes an artificial discrepancy :

Imagine you did not do any AARI adjustment at all.
In that case, your series will show a flat line from 1935-1953 at 6.994 M km2 and then suddenly jump to the Meier et al 'adjusted' value of 7.8 M km2. That is an 800 k km2 discrepancy which is the cause for your conclusion that there is a "significant" uptrend between 1935 and 1978. While in fact that uptrend has nothing to do with ice observations, and is only caused by your choice of 'baseline' (climatology) over that period.

2) By creating that climatology using temperature data, you are polluting the signal with temperature info, and by adjusting HadISST on concentration, you are even overruling any ice observations with your temperature signal.

Result is that your ice extent numbers have more to do with the known temperature graph that you used as an input than with ice extent observations.

That all said, it is mighty hard to choose a good climatology in an era that has sparse data, so I do not envy you.

It would be best if we could incorporate more ice observations, rather than relying on an assumption of SAT-SIE links.
I understand that Walsh is working on that and I also would recommend that you reconsider some observations that you currently left out :

- It turns out that the DMI 1935 map has a decent overlap with AARI charts from August and September, and is pretty consistent in the overlap. That puts some confidence that maybe the other DMI observations are also reliable.
It would be interesting to incorporate these DMI findings into your work.
- There are more other observations over the 1935-1978 period, for example the ice edge data from the Norwegians. That data shows a mild dip around the 1940's but not significant.

I would be happy to help out find more data sets that could further restraint the (arbitrary) choice of a baseline gridded product, and even help out finding better methods to homogenize different data sets.

[Diablobanquisa]

OK Rob, thank you for your opinions.

Just a final comment:
You think that the upward trend from 1935 to 60s and 70s is an artifact created by our methodology. Walsh et al. are using a fully different methodology for their new dataset (indeed, their methodology is closer to the one you are proposing). However, their trend from 1935 to 60s and 70s is upwards as well, and quite similar to ours, as we saw above: https://diablobanquisa.files.wordpress.com/2016/01/vspresat2.png

Anyway, their results are still provisional.

Thanks again for all your comments and suggestions.

[Diablobanquisa]

Rob, only another comment to clarify one of your latest points:

by adjusting HadISST on concentration, you are even overruling any ice observations with your temperature signal.

The adjustment applied to HadISST has nothing to do with temperature, it's just the bias found between HadISST and satellite sea ice data during the overlapping period (1972-1978).

[Rob Dekker]

Diablo,
You are right. For the 1953-1978 period, HadISST "scalar" adjustments (as per Meier 2012) were done based on the overlapping period 1972-1978 of satellite observations.
Yet, HadISST over the full 1953-1978 period were based on non-satellite ice observations, mostly from the Atlantic and Canadian/Alaska side of the Arctic. Notably few observations from the Siberian side.
So when you adjust that record on the "gridded" form based only on ice-concentration, you may have erased some real ice observations, and thus you AARI adjustments may have been double-counted.

You think that the upward trend from 1935 to 60s and 70s is an artifact created by our methodology.

Do you agree ?

[Diablobanquisa]

Obviously, I don't agree.

Regarding the HadISST adjustment: the bias during the overlapping period 1972-1978 isn't caused by a lack of observations in some areas, but by a systematic bias in concentration values. That's well established:

Walsh and Chapman 2001: "The solid curve NIC data for 1978^94) shows larger ice-covered areas than does the time
series utilizing passive-microwave SMMR/SSMI) data for the same period. The differences range from approximately
5% in winter +Fig.1a) to 15-20% in summer +Fig.1b).The differences, which arise from the smaller ice concentrations in the passive-microwave dataset, are largest in regions and time periods when melt creates a wet surface."

Rayner et al. 2003: " For example, satellite-borne passive microwave retrievals of sea ice concentration are not consistent with historical charts based on in situ observations, aerial reconnaissance and infrared satellite images. (...) A bigger problem, however, is that thin ice is not identified as such by the microwave retrievals: instead it is returned as a mixture of thick ice and open water [Emery et al., 1994]. Also, ponds resulting from summer melting on top of the ice often cause the microwave instrument to return a 10–30% lower than actual concentration of sea ice [Comiso and Kwok, 1996]: this particularly affects the Arctic in summer,

Meier et al. 2012: " However, they are not consistent with the passive microwave satellite record and any quantitative estimates of trends or variability across the 1978–1979 boundary are limited by uncertainties resulting from the inconsistent data sources. (...) The adjustments to match Hadley with the adjusted XPM(and thus SII) are larger, particularly during summer (Table 2). This is because, as mentioned above, Hadley is more conservative in detecting thin ice and not underestimating melt effects. In the original Hadley fields, passive microwave summer concentrations and extents are adjusted upward to account for that bias. Here, we effectively do the reverse. We ignore the passive microwave bias – focusing on consistency for trends and variability – and adjust the Hadley downward to be consistent.

So, if concentration drops from 100% to 75%, that hasn't any effect on extent numbers. However, if concentration drops from 25% to 0%, obviously there is an effect on extent numbers. So, the effect of the concentration bias on extent numbers is necessarily located at lowest concentration areas.

During 1972-1978, HadISST has a good coverage for the whole Arctic.

However, in previous decades the coverage isn't so good, particularly at the siberian seas during the 50s. The incorporation of AARI data fills these gaps.

I don't think we are doing a 'double-counting'. If concentration adjustment has removed a gridcell and AARI has ice on that cell, the ice will return. If AARI hasn't ice on that cell, the adjustment was right.

Three questions for you:

- Why are you so sure that the extent values during 1935-1939 must be necessarily much higher than ours?

- Why are you so sure that the upward trend since 1935 to 60s can't be right?

- If the upward trend is an artifact of our methodology, why a fully different methodology (new Walsh) obtains almost the same upward trend?

[Diablobanquisa]

Walsh and Chapman 2001: "The primary sources of the post-1972 data are the hemispheric fields of sea-ice concentration from the U.S. National Ice Center (NIC), whose weekly grids (derived primarily from satellite data) span the period 1972-94" ftp://pscftp.apl.washington.edu/incoming/PolarFridays/2-walsh_2001.pdf

Rayner et al. 2003 (HadISST paper): " in HadISST1, Northern Hemisphere Walsh fields for 1901–1978 were used as the main data source for that period. (...) U.S. National Ice Center (NIC) analyses [Knight, 1984]. The charts were based on U.S. Navy, Canadian and Danish aerial reconnaissance data and from retrievals from advanced very high resolution radiometer (AVHRR), passive microwave, and other satellite instruments. (...) NIC charts were used mostly to calibrate the summer passive microwave data, although they are indirectly included through the use of the Walsh data set." http://www.metoffice.gov.uk/hadobs/hadisst/HadISST_paper.pdf

Walsh and Johnson, 1978: "For some months in the earlier part of the study period, no data were available for certain areas. In particular, there were no data for the Siberian sector for several years in the 1950s."
http://journals.ametsoc.org/doi/pdf/10.1175/1520-0485%281979%29009%3C0580%3AAAOASI%3E2.0.CO%3B2

Walsh, 1978: " Some data for the Soviet Arctic are contained in the charts of the US Fleet Weather Facility and the British Meteorological Office, but reliable data for the years prior to the late 1960s need to be provided by the Soviets" https://nsidc.org/pubs/documents/gd/GD-2_web.pdf (p. 49-51)

[Rob Dekker]

Hi Diablo,

Starting with that last note by Walsh, I think the AARI data you analyzed is providing a wealth of data, specifically about the Siberian side of the Arctic, which was not well covered by the Walsh and Chapman 2001 data that was at the core of the HadISST data set.

Since you adjusted that data set using AARI (after compensating for Meier et al 2012's adjustment), I have no issue with your 1953-and-later assessment of the Arctic. Your work helped in fine tuning Meier et al's work in reconstructing Arctic sea ice extent over that period.

You ask :

Why are you so sure that the extent values during 1935-1939 must be necessarily much higher than ours?

No necessarily. But I argue that there is a 800 k km2 discontinuity between 1952 and 1953 in your data set, even WITHOUT any (AARI or other) ice observations.

Why are you so sure that the upward trend since 1935 to 60s can't be right?

I did not say any of that. I'm just saying that you cannot declare a "statistically significant" uptrend in ice extent between the 30's and the 70's if you introduced it yourself by choosing a "climatology" from 1935-1952 that is 800 k km2 below the 1953 numbers you use.

The choice of climatology before 1953 should not matter if there are enough ice observations before 1953.

So you could have chosen a climatology that levels out at 7.8 M km2 at 1953, so it nicely blends in with the adjusted gridded HadISST around that period. If you make the AARI adjustments on THAT climatology, you could show how much the Siberian side of the Arctic creates a 'trend' between the 30's and the 70's.

OR you could have stuck with the 7 M km2 climatology you choose for 1935-1952 and made AARI adjustments on that stretching on to 1978. That would show how AARI shows a trend as well.

But you can't do both, because then the trend is pre-determined by your input (the temperature graph in your case).

If the upward trend is an artifact of our methodology, why a fully different methodology (new Walsh) obtains almost the same upward trend?

Walsh' findings are preliminary, and please note that there are still some odd irregularities in his graph as well around the 20s.

Other researchers such as Vinje 2000, Polyakov et al 2003, Johanessen et al 2004, and Divine & Dick 2004, all show a steady decline that started in the 19th century, with possibly a 'hold' or a mild dip in the 30's, but it would be hard to call that "significant", followed by accellerated trends in the 90's and 2000's that we all agree on.

Thanks for your work Diablo. And I promise that I'll give future publications the same scrutiny that I gave yours. Be it from Walsh or anyone else :o)

[Diablobanquisa]

Hi Rob, thank you for your replies and your comments.

Just a point:
So you could have chosen a climatology that levels out at 7.8 M km2 at 1953, so it nicely blends in with the adjusted gridded HadISST around that period. If you make the AARI adjustments on THAT climatology, you could show how much the Siberian side of the Arctic creates a 'trend' between the 30's and the 70's.

Indeed, we did a test using a 7.71 M km2 climatology for 1935-1952. After the incorporation of AARI data, the upward trend from 1935 to 60s and 70s was still there, although certainly much slighter. Red line: our work; blue line: our work using a 7.71 climatology for 1935-1952: https://diablobanquisa.files.wordpress.com/2016/02/rrtyy.png

However, I think that the choice of the 7.71 climatology is arbitrary, because it doesn't rely on anything (well, on an unchecked assumption that the mean extent during 1935-1952 should be similar to the mean extent during 1953-1978; indeed, even the mean extent during 1953-1978 isn't as high: it's 7.55 after the incorporation of AARI data into adjusted HadISST).

Our choice of the 7.0 climatology relies on the very strong correlation on decadal time scales between Arctic SAT and Arctic SIE. In addition, the mean sea ice edge derived from AARI data for 1935-1952 matches very closely the sea ice edge of the 7.0 climatology. On the contrary, the mean edge derived from AARI doesn't match the edge of the 7.71 climatology. So, at least in the Siberian sector, actual observations for 1935-1952 are much closer to the 7.0 climatology than to the 7.7 one.

Anyway, I understand your caveats about this issue.

Thank you for all your comments and suggestions. I hope that, as you stated, you'll give future publications the same scrutiny, I'll be reading ;-)

Thanks again, Rob.

[Diablobanquisa]

Rob, I forgot to mention that I'm glad to see that you have no issue with our work from 1953 onwards.

I acknowledge that our values before 1953 are more uncertain and rely notably on the correlation SAT-SIE, so I understand your skepticism about them.

Thanks again for all your feedback.

Cheers

[P-maker]

Rob & Diablo

Thank you for giving us this insight, Some folks may have learned how difficult it is to get the details right. I recall discussions with Nick Rayner and other Met Office staff back then in the 90ies how to interpret old Danish Met Office ice charts. I think progress has been made, but I also encourage you to know, that Neven's blog on sea ice is the place to be, when the going gets rough!

[Diablobanquisa]

Thank you, P-maker!

Rob, regarding this:

"It would be best if we could incorporate more ice observations, rather than relying on an assumption of SAT-SIE links. I understand that Walsh is working on that and I also would recommend that you reconsider some observations that you currently left out"

Even if I still think that SAT-SIE link works, if we eventually work on a new version we'll take into account this suggestion. Particularly, I think we should try to incorporate ACSYS August sea ice edge data (https://nsidc.org/data/docs/noaa/g02169_nordic_sea_ice/ ) as well as the direct observations from August DMI charts ( http://nsidc.org/data/docs/noaa/g02203-dmi/ ).
I don't know if we'll be able, but at least we'll attempt to do it.

[Rob Dekker]

Hi Diablo,
ACSYS and DMI both have good coverage over the Nordic seas over the 1935 - 1953 period, so it would be a great addition to your work.

Mahoney et al 2008 describes a method (using longitudinal slices) that could work very well for ACSYS, since that data set deals with latitudinal ice edges.
http://seaice.alaska.edu/gi/publications/mahoney/Mahoney_2008_JGR_20thC_RSI.pdf

Incidentally, that paper also analyzed the very same AARI data set, so it is interesting to compare with your findings.

Mahoney find a statistically significant DECREASE in summer ice extent in summer during the 1935 - 1953 period, while your graph suggest no trend or maybe a slight uptrend.

Do you know why that is ?

[Diablobanquisa]

Hi Rob,

I was aware of that, and it's a good point.

As you stated, they show an overall downward trend from 1936 to 1952 whereas our trend for that period is essentially flat (indeed, very slighly uptrend, as you pointed out).

First of all: Mahoney et al's methodology looks really good, and they are true experts, while we are just amateurs.

Anyway: why the difference? I'd say it might be due to a combination of:

- Their first summer point corresponds to 1936 instead of 1935.
- They present a 3 month mean (JAS), while we present a snapshot around mid-September.
- They left out western Kara sea (just in that area, there is more ice during 1946-1951 than during 1935-1945).
- Their methodology could lead to an extent overestimation when data are sparse (by assuming the same edge for a whole sea).
- Our methodology could lead to an extent underestimation when data are sparse (by relying on SAT-SIE climatology).
(data are more sparse at the first years of the period).

I extracted their summer values for 1936-1952 from their Fig.6 and plotted them against ours. The two time series certainly look very different until 1940, but from 1941 onwards they correlate pretty well (1941-1952, r=0.81). Both time series also agree that the lowest extent during the 1936-1952 period corresponds to 1952.
Red, ours ; blue, Mahoney et al.: https://diablobanquisa.files.wordpress.com/2016/02/mahoney.png

( https://diablobanquisa.files.wordpress.com/2016/02/extracting.png )

(A bit off topic: this statement from Mahoney et al. encouraged me to start our work... "Rayner et al. [2003] acknowledge that the AARI charts, which were not available in digital form at the time, were not incorporated into the HadISST data set, but we strongly encourage their assimilation in future updates.". However, the recently presented HadISST2 dataset hasn't incorporated AARI data, and it still relies mostly on the 'old' Walsh: Titchner HA, Rayner NA (2014): The Met Office Hadley Centre sea ice and sea surface temperature data set, version 2: 1. Sea ice concentrations. J. Geophys. Res. Atmos., 119:2864-2889. (if you'd like to read the full text, let me know. Their data aren't available for download yet: http://www.metoffice.gov.uk/hadobs/hadisst2/ ) )

[Diablobanquisa]

Obviously, the most likely reason to explain the discrepancy is a failure of our methodology, particularly during the earlier years of the record (1936-1940), when data are sparse. (I'd say the second reason is the western Kara issue, particularly around 1949, when sea ice extent is at its highest level in that area).

In 1952 the coverage of AARI charts is excellent (we used these: ftp://sidads.colorado.edu/pub/DATASETS/NOAA/G02176/pngs/1952/E19520910_tc.v0.png and ftp://sidads.colorado.edu/pub/DATASETS/NOAA/G02176/pngs/1952/W19520915_tc.v0.png ). So, our error margins are much smaller, since we don't rely on our 'SAT-SIE climatology' for anywhere at the Siberian sector.

Assuming that both time series level at 1952, this would be (roughly) the effect of using Mahoney extents (blue) instead of ours (red):
Pre-satellite era: https://diablobanquisa.files.wordpress.com/2016/02/presamah.png
Whole time series: https://diablobanquisa.files.wordpress.com/2016/02/fullmah.png

[Diablobanquisa]

Instead of 'failure' I should have said 'feature', I didn't mean that our methodology is wrong.

[Diablobanquisa]

I have been looking at Mahoney et al. SIE numbers here: ftp://sidads.colorado.edu/pub/DATASETS/NOAA/G02182/RArctic_ip_area.csv

( http://nsidc.org/data/docs/noaa/g02182_seaice_edge_extent_russia/index.html )

Indeed, they have a lot of 'missing' Septembers during the earlier years of the record.

For instance, for 1935 they give a September number for Kara sea only. However, the two maps we used for 1935 ( ftp://sidads.colorado.edu/pub/DATASETS/NOAA/G02176/pngs/1935/e19350902_tc.v0.png and ftp://sidads.colorado.edu/pub/DATASETS/NOAA/G02176/pngs/1935/w19350916_tc.v0.png ) allowed us to integrate some data from Chukchi, ESS, Laptev and Kara. (We use the data as given in the chart: cells that are open water are integrated on our gridded data as open water ; cells with ice, are integrated as ice. Although I tell 'chart', obviously what we integrate is the gridded data corresponding to each chart, for instance: ftp://sidads.colorado.edu/pub/DATASETS/NOAA/G02176/binary/1935/e19350902.v0.bin ).

For 1936 they give a September number for Barents, Kara and ESS. However, from the two charts we used for September 1936, we integrated some data from the five seas (Barents, Kara, Laptev, ESS and Chukchi).

For 1937 they only give a September number for Kara and Laptev, while we integrated some data from Kara, Laptev, ESS and Chukchi ...

... Until around 1943, their September coverage doesn't become more or less complete and continuous.

So, I think that this issue could also explain the difference between the two time series (a difference that depends mostly on the earlier years of the record).

[Rob Dekker]

Thanks Diablo,

I've looked through the extent files from Mahoney et al, and yes, it seems clear that the greatest challenge is how to find the best methodology to estimate the missing data in these temporal/spacial sparse data sets of the 30's and 40's and equally important : to estimate an error margin.

In that respect, I have to admit that at this point, I do not understand exactly how you, nor Mahoney have done this for the pre-1953 period.

[Rob Dekker]

And we have not even included the other side of the Arctic :
The Nordic seas, Baffin Bay, the Canadian Arctic, and the Beaufort.

Now I understand why Walsh is taking his time coming up with a pan-Arctic estimate before 1953.

[Diablobanquisa]

Florence Fetterer has let me know that the 'new Walsh' is up.
They have just published this: https://nsidc.org/data/g10010

Documentation: https://nsidc.org/data/docs/noaa/g10010-sea-ice-1850-onward/G10010_SIBT1850.pdf

Gridded data: ftp://sidads.colorado.edu/pub/DATASETS/NOAA/G10010/

I haven't look at the data yet.

[Rob Dekker]

And the report is here :
https://nsidc.org/data/docs/noaa/g10010-sea-ice-1850-onward/G10010_SIBT1850.pdf

I have not looked at the data and methods yet either, but figure 1 shows ice extent from 1850-present.
Seems to collaborate your findings of an uptrend in September from the early 50's to the end of the 60's, but put in a wider context of a 'flat' and higher extent before that.

[Rob Dekker]

I'm sorry, that should read as a dip from the 30's to the 60's as the most profound finding in September extent, preceded by a higher extent earlier.
Other months are also interesting.

August is remarkably variable pre-1950, while October, November and December are remarkably flat.

Time to look into the details of the methods they used.

[Diablobanquisa]

At first glance, their September values for 1935-1978 look rather similar to those they presented in their previous presentation and we were looking at.

[Bill Fothergill]

@ Diablo

Thank you for this illuminating article.

You make reference to the Walsh & Johnson paper first published in the Journal of Physical Oceanography. Buried in there - just below Fig 5 - is a little statistic I love to use when one gets subjected to bollocks about "Arctic sea ice was as low 60 or 70 years ago". The overall growth than Walsh and Johnson clocked over the 1953-77 period of their study was a miserly 3.14 thousand sq kms p.a.

Something that might be tangentially related to your work is the summary by the US Army of the materiel shipments to the Soviet Union during WWII. http://www.history.army.mil/books/wwii/persian/appendix-a.htm#t1

Here in the UK, most people are only aware of the horrible Arctic convoys round the top of Norway to Murmansk and Archangelski, but these accounted for just 23% of the equipment shifted. About 47% was shipped from the US West Coast to the Russian Far East, with a only a rather minimal 2 or 3% getting delivered further than the Bering Straits. (Although this was going close to Japan, there was no declaration of war between the Soviets and Japan until the closing stages of the conflict.)

The point is that, had there been any chance of getting stuff through the Bering Straits, Joe Stalin would have had this materiel delivered to Archangelski and Murmanski, rather than having to be transported overland. In other words, the Northern Sea Route wasn't open with any regularity or reliability during the war. (Yep, I know that, whilst Hitler and Stalin were still buddies, the Komet made the journey West-East, but that was just about it for the NSR.)

[Rob Dekker]

Accurding to Walsh, his data set is best viewed with Panoply.
http://www.giss.nasa.gov/tools/panoply/
Even though I just installed Java 8, I still can't get Panoply to work on my Mac.
Does anyone else have this issue ?

[Diablobanquisa]

Thank you very much, Bill.

Rob, I'm sorry but I can't help you, I use Panoply without any problem, but I run it under Windows.

[Rob Dekker]

I have Panoply running on a Windows box, and I am able to visualize Walsh' data.
I have some initial findings to share, and I think you will like it.
Neven, is there a way in which I can upload pictures with TypePad ?
Or do I need my own separate site before I can show images here ?

[Neven]

Yes, you need to link to images from somewhere else using the img tags. And make sure they're no wider than 400-450 pixels because otherwise the rest gets cut off.

You can also send images to me, I'll upload them to Picasa and then link to them here. I believe we have e-mailed before, so you probably still have my mail address, but there's a button in the menu at the top of this blog as well.

[Jim Hunt]

Rob,

I have PanoplyJ running fine under Scientific Linux.

I have a MacBook too, but currently it is running Windows 7 with a dead screen and I daren't reboot it just at the moment!

[Diablobanquisa]

September ASIE 1935-2014: new Walsh vs. ours (it's a preliminary analysis): https://diablobanquisa.files.wordpress.com/2016/02/new-walsh-vs-ours.png

September ASIE 1935-2014: new Walsh, ours, Meier et al. 2012: https://diablobanquisa.files.wordpress.com/2016/02/new-walsh-ours-meier.png

September 1973, 1974 and 1976: new Walsh, ours, and ESMR (https://nsidc.org/data/docs/daac/nsidc0009_esmr_seaice.gd.html) charts:

https://diablobanquisa.files.wordpress.com/2016/02/1973vs.png

https://diablobanquisa.files.wordpress.com/2016/02/1974vs.png

https://diablobanquisa.files.wordpress.com/2016/02/1976vs.png

[Rob Dekker]

Thanks guys. I have a process running Panoply on a Windows box (at work) and Photobucket for posting images. So I'm good for now.

After I exposed a "low" bias for the 1935-1953 period in Diablo's paper, due to his choice of climatology, I promised him that I would give Walsh an equally critical look.

Here is the start of that, and up-front I would like to let you know that I don't know where the evidence will lead, so this is really the very beginning of an exploration of the data that Walsh analyzed and presented.

So I started with spot checks.
I choose to check 1935, 1952 and 1963 for starters,
1935 because it is the start of Diablo's timeseries, and any decadal trend should manifest itself at this extteme.
1952 because it is an anomalously low extent in both Diablo's and Walsh' timeseries, and
1963 because Walsh' series seems anomalously high w.r.t. Diablo's and Meier et al.

So let's start with 1935.
Here, I start with August 1935, because we have good observations there, not just from the Russian side of the Arctic (AARI) that Diablo used, but also from DMI :
ftp://sidads.colorado.edu/DATASETS/NOAA/G02203/1935_08.jpg
which also covers the Beaufort, the Greenland Sea and the Barents.

Here is the Aug 1935 ice concentration map from the new Walsh' data set :

The first thing to note is that the light blue range matches quite nicely with the AARI data in the Russian Arctic, and also with the DMI data (linked above).

Specifically the way in which the ice boundary in the Greenland sea almost touches Iceland, as well as the ice south of Svalbard and the line of the ice boundary north of Novaya Zemlya closely matches with the DMI observations. And the Russian Arctic matches closely with BOTH the AARI images from August as well as that DMI image.
Moreover, ice in the Beaufort is close to the coast in the DMI image as well as in this Walsh' image.

So this Walsh image appears to be consistent with ice observations in Aug 1935, as long as we take the light blue ice boundary as the ice edge.

However, and here is the first major issue :
The ice cover in the picture above is something like 10 - 11 M Km^2.
Yet, in Walsh' "extent" graph Aug 1935 comes in at about 9 M km^2 :
https://nsidc.org/data/docs/noaa/g10010-sea-ice-1850-onward/G10010_SIBT1850.pdf

And to make matters worse, Sep 1935 clocks in at 7 M km^2 or so.
Loosing 2 M km^2 between Aug and Sept is already quite unprecedented. But loosing 3 - 4 M Km^2 between Aug and Sept ( the difference between the ice observations from DMI and AARI from Aug 1935 and the reported 7 M km^2 Sept 1935 extent from Walsh' figure 1) is IMPOSSIBLE.

So something is not right here in 1935.

[Diablobanquisa]

Thank you for your analysis, Rob, I'm looking forward to see the following updates.

A preliminary comment: during 1935-1952, I wouldn't expect August and September charts to be consistent. Walsh et al. are using very different sources and approaches for each one of those two months.

The first thing to note is that the light blue range matches quite nicely with the AARI data in the Russian Arctic, and also with the DMI data (linked above).

Specifically the way in which the ice boundary in the Greenland sea almost touches Iceland, as well as the ice south of Svalbard and the line of the ice boundary north of Novaya Zemlya closely matches with the DMI observations.

I agree, although it doesn't come as a surprise as their sources for that August map are AARI and DMI (and this chart from ACSYS: ftp://sidads.colorado.edu/pub/DATASETS/NOAA/G02169/ice_edge_positions/browse/aug/iep19350826.jpg )

However, and here is the first major issue :
The ice cover in the picture above is something like 10 - 11 M Km^2.
Yet, in Walsh' "extent" graph Aug 1935 comes in at about 9 M km^2

I haven't still been able to calculate the extent values from the gridded data (I'd need some help from my coauthor, who is the smart guy that wrotes most of the code, but unfortunately he is too busy right now) although I tend to assume that Walsh et al. have done it right: I assume the extent derived from that map should be 9 M km^2 or so.

Another point: the Walsh chart matches closely the DMI one (its red coloured direct observations)... except for Baffin, Foxe and Labrador coast. In these three areas, Walsh's chart shows a larger sea ice extent than derived from DMI direct observations.

According to DMI, Baffin Bay is "probably open water" whereas Walsh's map shows a lot of (low concentration) extra ice (200.000 km^2 or so?)

DMI map doesn't present any direct observation at Foxe Basin, but Walsh's map shows it completely ice covered (low concentration, but almost 200.000 km^2).

Labrador coast: DMI chart shows icebergs (not a surprise) but no sea ice, while Walsh's chart shows the area as sea ice covered from Southern Baffin Island to Labrador coast and along it (200.000 km^2 or so?)

(Looking at seaice_source, it seems that they have derived those areas from "Kelly ice extent grids". Those grids were mainly derived from the white areas on DMI charts, although in this case the extent is even larger than DMI's white area, so I dont't understand how Kelly got it. Kelly 1979 https://nsidc.org/data/docs/noaa/g02203-dmi/docs/Kelly_GD5_1979.pdf ).

So, at Baffin, Foxe and Labrador, Walsh's map could be showing 600.000 km^2 or so of not observed extra ice. This would reduce the actual loss from August to September from 2.0 M km2 to 1.4 M km2 or so. Not that impossible.

In addition: are the Beaufort direct observations from August 31th, 15th or 1st? It's not the same in order to estimate the required ice loss...

Well, and CAA and Chukchi are also derived from Kelly grids-DMI white areas, so no direct observations there.

In conclusion, I think that August and September ice charts in the new Walsh dataset during 1935-1952 aren't consistent between them, but that doesn't imply that August maps are 'right' and September maps are 'wrong', neither the contrary. It just shows the difficulty of getting consistent and reliable sea ice data before the satellites and particularly before 1953.

(In this case(1935) I'd say that it seems that Walsh et al. are likely overestimating August extent -Baffin, Foxe, Labrador, CAA, Chukchi- and maybe slightly underestimating September extent - Greenland-Barents-.)

Cheers