Casual prediction: lowest yet May arctic sea ice extent

[Paddy]

Based on the NSIDC's Charctic graph so far, I reckon this is the lowest May average extent we've seen.  (Not quite the lowest area, though).

EDIT: The question is, though, what does this mean, if anything, for the rest of the melting season?

[DavidR]

If we don't see a rise over the next few days it  will be the lowest , certainly since 2006. What  it  means for the remainder of the season is simply that there is a smaller area with a high albedo. and therefore probably slightly warmer water going into the remainder of the melt  season.

We are currently  150K km^2 below the record if  we stay at that anomaly for the next few days we are likely to stay at  record lows for the next month.

[Tor Bejnar]

From the NSIDC's Charctic graph, 2006 was lower than 2015 for the first 16 days of May.  Without crunching the daily numbers, it looks like 2015 was lower, on average, during the 2nd half of May than 2006 was lower than 2015 during the 1st half of May.  Regardless of how close the numbers work out to be, 2015's probable 'momentum' going into June (per DavidR's post, but don't discount the last 3 days of May yet!), probably doesn't mean much as the 'ahead-ness' is largely due to early Hudson Bay melt that has a fleetingly small affect on the sea ice melt in the CAB or its peripheral seas.

[ChrisReynolds]

Paddy.

NSIDC Extent bottom ten extents for 1 May to 28 May, 1979 to 2015:

2015   12.510
2006   12.576
2004   12.599
2011   12.751
2014   12.754
2007   12.840
2010   12.959
2005   12.981
1995   13.023
2002   13.033

We are lowest, probably by enough of a margin for this to hold over the next three days (here I have not done the numbers), but the average over the next few days would have to claw back 0.066 on an average for the last 28 days. So lowest May looks a good bet.

Here's a plot of the regional breakdown.



The maths is clear to those who do such things, but don't worry about the details (unless you want me to explain). Basically read that graph as - the black lines show the range between the highest and lowest for 28 May from 1979 to 2015, the red triangle is 28 May 2015.

Hudson and Bering are low, they're outside the Arctic Ocean. But the Arctic Ocean itself is the lowest on record for 28 May, while for the 1 to 28 May average it is the 3rd lowest extent since 1979.

Barents and Kara are low, which does raise prospects for a good melt from the Atlantic side. For the 28 May Chukchi is the (EDIT) 'second' lowest on record.

However, June has seen large losses in extent and area in years that have seen record or near record lows since 2007 (except 2013). So I'd wait until late June and see what June losses are like - anything over 3 million kmsq (AREA) is a reasonable (though not certain) indicator of a good melt season.

[ktonine]

Significant changes began happening in 2004 and were dramatized by 2007.

Trying to discern what *changed* is difficult.  One of the interesting items I've noticed is that the Greenland Sea average September extent is (post-2007) negatively correlated with all of the other regions.  This was not the case prior to 2007:

Region. . Post-2007      Pre-2007

Beaufort   -0.5348       -0.0050
Chukchi   -0.7217        0.4265
ESS         -0.8132         0.2353
Laptev      -0.0221      -0.0186
Kara          -0.8212        0.0223
Barents     -0.2498        0.3475
Central       -0.3764        0.4705
CAA           -0.6224       -0.0887

The only region that has not changed in relation to the Greenland Sea is the Laptev. 

The intuitive explanation is that the ice has become more mobile and in low extent years more of it is ending up in the Greenland Sea.  And while the absolute amount remaining in the Greenland Sea in any given year doesn't seem like a large value it probably represents an amount of ice several times larger since the Greenland Sea is essentially a killing ground for ice.

I've only looked at the correlations for September average extent.  It might be interesting to see at what time of year this pattern first becomes established.

[DavidR]

However, June has seen large losses in extent and area in years that have seen record or near record lows since 2007 (except 2013). So I'd wait until late June and see what June losses are like - anything over 3 million kmsq is a reasonable (though not certain) indicator of a good melt season.

The transition to  large June melts seems to be a fixture, although only 1 year has had a melt  greater than 3 M km^2; 2012 obviously. All of the past five years (2010-2014) have seen June NSIDC extent drop by more than 2.55 M km^2. Prior to 2010 not one year had a decline that large. 

1999 came closest with 2.527 and set a record September low that year. It was the year after the last big El Nino, and both 1998 and 1999 had June melting  more than 500 K km^2 higher than the previous highest figure.  This does not look good for this year or next year.   

Currently we would need a smaller melt over the next 34 days, than in any of the past 5 years,  to not be at record low levels on July 1st. If we get  3M km^2, we will be well below the previous July 1st record and a record or near record September minimum would seem inevitable.

[ChrisReynolds]

Sorry about that David,

I've just checked and the spreadsheet page I have on June melt uses area, took me a while to figure that out, area carries a stronger signal than does extent - the offfending page is now noted to be using area!

In descending order of Sept extent, June area loss using Wipneus's regional NSIDC extent/area dataset.

Year / June Area loss / Sept NSIDC Extent
2012   3.18   3.62 - OK
2007   3.19   4.29 - OK
2011   3.00   4.61 - OK
2008   2.55   4.72
2010   3.06   4.92 - prospects killed by MYI
2014   2.32   5.28
2013   2.97   5.35
2009   2.28   5.38
2005   2.74   5.56
2006   2.54   5.91

[P-maker]

Ktonine:

”Greenland Sea average September extent is (post-2007) negatively correlated with all of the other regions.”

Thank you indeed for this post. This was exactly my concern (stated less elegantly in a different thread a few days ago).

I know you work with small numbers, but still I think you make a convincing case, that the Greenland Sea will be the “graveyard” of the Arctic Ocean.

Chris,

The cat may have been born with a fat tail – actually two fat tails - coming out of the Fram and Nares Straits respectively. However, if these fat tails are cut off due to much higher mobility in the Central Arctic Ocean and other peripheral seas – as suggested by Ktonine’s numbers – we will only see the skinny remnants of the tail in the Greenland Sea statistics.

Your most resent ranked list of low numbers also made it clear, that extent and area are getting closer to each other as the numbers get smaller.

[Shared Humanity]

Significant changes began happening in 2004 and were dramatized by 2007.

Trying to discern what *changed* is difficult.  One of the interesting items I've noticed is that the Greenland Sea average September extent is (post-2007) negatively correlated with all of the other regions.  This was not the case prior to 2007:

Region. . Post-2007      Pre-2007

Beaufort   -0.5348       -0.0050
Chukchi   -0.7217        0.4265
ESS         -0.8132         0.2353
Laptev      -0.0221      -0.0186
Kara          -0.8212        0.0223
Barents     -0.2498        0.3475
Central       -0.3764        0.4705
CAA           -0.6224       -0.0887

The only region that has not changed in relation to the Greenland Sea is the Laptev. 

The intuitive explanation is that the ice has become more mobile and in low extent years more of it is ending up in the Greenland Sea.  And while the absolute amount remaining in the Greenland Sea in any given year doesn't seem like a large value it probably represents an amount of ice several times larger since the Greenland Sea is essentially a killing ground for ice.

I've only looked at the correlations for September average extent.  It might be interesting to see at what time of year this pattern first becomes established.

We know the ice is more mobile then it was 20 years ago and this negative correlation would suggest that this new mobility is a factor in the melt of a specific season. It's not as simple as saying the Greenland Sea will be the killing ground for ice. While substantial amounts of ice exits out of the Arctic via the Fram throughout the year, mobility of sea ice is playing out all through the CAB.

I feel that the models are not capturing the dramatically increased mechanical processes that are occurring in the Arctic. This increasingly fractured and mobile ice pack is being subjected to all kinds of mixing, grinding, etc. The lowest extent on record occurred during the GAC of 2012. The following year saw a significant rebound in minimum extent and was remarkable as a  low  energy season....I described it as "torpor" when it was occurring. I am not suggesting that stormy weather will always result in big melts but these kinds of seasons have their impact on the quality of the ice even if it is not always obvious when we look at SIE and SIA.

[ChrisReynolds]

P Maker,

Yet the volume export from Fram remains relatively level as increased area export is opposed by thinner ice export. I think of Nares as a thin tail.

The negative correlation after 2007 implies that as extent in the Arctic Ocean goes down export goes up. This may be a result of the coincidence of unusual Greenland High in years of low extent, as the Greenland high drives southward winds in the Greenland sea.

[epiphyte]

P Maker,

Yet the volume export from Fram remains relatively level as increased area export is opposed by thinner ice export. I think of Nares as a thin tail.

The negative correlation after 2007 implies that as extent in the Arctic Ocean goes down export goes up. This may be a result of the coincidence of unusual Greenland High in years of low extent, as the Greenland high drives southward winds in the Greenland sea.

I think of Nares export not as a thin tail, but as yet another example where the volume , area , extent numbers tell a different story than the one they used to tell. IMO the fact that the exported ice is thinner, and therefore less voluminous, than in the past doesn't make it less significant to the overall picture, but rather the opposite.

 - for one thing, if the ice that is exported is thin and mobile, and the CAB area overall is not changing, and mild-normal freezing conditions exist over the region, (as has been the case in April/May) then there is a large amount of even thinner ice (i.e. only a few weeks old and a few inches thick) scattered throughout the pack.

- for another, if the exported ice is thin, then it takes relatively little energy to melt it, which means that exporting a given area will have less effect on Greenland Sea surface temps than it used to, which will push the boundary of the melting region north, even in the face of south/westward export.

I've been following the movement of individual floes approaching the Fram from the North/East for a few weeks now - and what I'm seeing looks consistent with the above.... A large area of ice has been moving continually SW and melting quickly and completely, yet the melting boundary region is remaining static or moving North, with sea surface temps behind the boundary region showing high positive anomalies, and currents consistently pushing that warm water Northwards. (I posted a pic of this a couple of weeks ago.)

All of this feeds into my general distrust of the PIOMAS numbers w.r.t. the overall state of the younger, more mobile ice. I can see stuff melting out at the edges, (notwithstanding surface freezing conditions), I can see leads opening up and freezing over in Kara and the CAB N of Svalbard as the ice becomes less compact, so even though the area is remaining the same, the volume/unit area must be shrinking.

From above on any given day it appears that nothing much is going on, and the numbers reflect a slow start on the Atlantic side... But I think it's quite possible that the whole region is continuously thinning out, and that the mechanism by which this is happening is one that has only come into play in the past five years, a period which post-dates the formulation and validation of PIOMAS.

Last year the late-season rally in the melt took many by surprise - but isn't it just what one would expect if the dynamics of the overall pack become more fluid as thin ice becomes more ubiquitous?

IMO the current situation in the Beaufort also exemplifies this - The mobility of the ice has led to large expanses which are liable to become open water at the slightest provocation. This year the early season weather has been warmer there - and the the effect is correspondingly dramatic. But consider that now, as never before, this could happen just about anywhere in the arctic given the right weather for just a few days.

Of late, many have observed that the May/June weather plays a critical, and largely unpredictable role in any given melt season. Would any question that this too is more true than used to be the case? Thirty years ago, the exact timing of a cyclone occurring two weeks either side of June 1st might have had little significance to the overall area at the end of the season. Right now, a few inches of well-timed snow might tip the balance between end-of-season melt/not melt over millions of square miles of area

Interesting times indeed.

[ChrisReynolds]

Epiphyte,

My reason for calling Nares a thin tail is that it is so much narrower than the Fram Strait, so the potential for export is even less.

In HYCOM there is a clear depletion of thick ice around Nares, but within a short distance towards Banks island the thickest ice remains a constant distance band from the coast of the Canadian Arctic Archipelago.

I've yet to be persuaded that the PIOMAS model is failing, what I see in the numbers is physically consistent with the changes since 2007.

[epiphyte]

@ChrisReynolds - My bad! Total brain failure - I had read "Nares" and understood "Fram"... Hopefully I won't start driving on the wrong side of the road next.

[ChrisReynolds]

 

[Paddy]

Paddy.

NSIDC Extent bottom ten extents for 1 May to 28 May, 1979 to 2015:

2015   12.510
2006   12.576
2004   12.599
2011   12.751
2014   12.754
2007   12.840
2010   12.959
2005   12.981
1995   13.023
2002   13.033

We are lowest, probably by enough of a margin for this to hold over the next three days (here I have not done the numbers), but the average over the next few days would have to claw back 0.066 on an average for the last 28 days. So lowest May looks a good bet.

Here's a plot of the regional breakdown.



The maths is clear to those who do such things, but don't worry about the details (unless you want me to explain). Basically read that graph as - the black lines show the range between the highest and lowest for 28 May from 1979 to 2015, the red triangle is 28 May 2015.

Hudson and Bering are low, they're outside the Arctic Ocean. But the Arctic Ocean itself is the lowest on record for 28 May, while for the 1 to 28 May average it is the 3rd lowest extent since 1979.

Barents and Kara are low, which does raise prospects for a good melt from the Atlantic side. For the 28 May Chukchi is the (EDIT) 'second' lowest on record.

However, June has seen large losses in extent and area in years that have seen record or near record lows since 2007 (except 2013). So I'd wait until late June and see what June losses are like - anything over 3 million kmsq (AREA) is a reasonable (though not certain) indicator of a good melt season.

Thank you for this analysis - helps make sense of things a lot.  Out of curiosity, do you have figures for the whole of May now?  It'd be interesting to see how different of a record is being set.

[ChrisReynolds]

Paddy,

I am in the process of writing my May Status blog post right now. Check my blog shortly after the PIOMAS data is out.
http://dosbat.blogspot.co.uk/

[ChrisReynolds]

Based on the NSIDC's Charctic graph so far, I reckon this is the lowest May average extent we've seen.  (Not quite the lowest area, though).

EDIT: The question is, though, what does this mean, if anything, for the rest of the melting season?

NSIDC Sea Ice Index
ftp://sidads.colorado.edu/DATASETS/NOAA/G02135/
For May
ftp://sidads.colorado.edu/DATASETS/NOAA/G02135/May/N_05_area.txt

Bottom ten lowest May extents.

Year / Extent
2004   12.58
2006   12.61
2015   12.65
2014   12.77
2011   12.81
2007   12.88
1989   12.98
2003   12.99
2005   12.99
1995   13.04

My blog post on May sea ice state is here:
http://dosbat.blogspot.co.uk/2015/06/may-2015-status-four-dipoles.html
PIOMAS will follow.

And the lines and triangles range plot for the regions is updated for 1 June atb the end of that post.

[Paddy]

Cheers, thank you!

Odd how the overall May extent here is higher than the 1st to 28th May extent in the earlier post despite the last few days having continued loss below 12 million km².  Are these from the same data?

[ChrisReynolds]

The NSIDC extent monthlyaverage is calculated from a grid box average for the region. The monthly average is from the average of the extent for all days in the month. The monthly average may be lowest, I didn't get round to calculating last night. Can do later when I'm back home, if nobody else does it.

[DavidR]

Cheers, thank you!

Odd how the overall May extent here is higher than the 1st to 28th May extent in the earlier post despite the last few days having continued loss below 12 million km².  Are these from the same data?

Based on the daily data,
   ftp://sidads.colorado.edu/DATASETS/NOAA/G02135/north/daily/data/ ,
I  have 2015 at 12.485, 2006 at 12.517, and 2004 at 12.598.  The figures vary  slightly if we use the 5 day average but not enough to  change the order.
I'm not sure what  Chris is basing his calculations on.

[ChrisReynolds]

David,

The data is here, as linked to above.
ftp://sidads.colorado.edu/DATASETS/NOAA/G02135/May/N_05_area.txt

year mo    data_type region extent   area
1979  5      Goddard      N  14.04  11.03
1980  5      Goddard      N  14.03  10.88
1981  5      Goddard      N  13.88  10.92
1982  5      Goddard      N  14.15  11.09
1983  5      Goddard      N  13.52  10.79
1984  5      Goddard      N  13.66  10.80
1985  5      Goddard      N  14.22  11.18
1986  5      Goddard      N  13.50  10.62
1987  5      Goddard      N  13.80  10.90
1988  5      Goddard      N  13.68  11.35
1989  5      Goddard      N  12.98  11.29
1990  5      Goddard      N  13.29  10.83
1991  5      Goddard      N  13.51  11.43
1992  5      Goddard      N  13.25  11.36
1993  5      Goddard      N  13.53  11.34
1994  5      Goddard      N  13.73  11.59
1995  5      Goddard      N  13.04  10.76
1996  5      Goddard      N  13.05  11.21
1997  5      Goddard      N  13.31  11.16
1998  5      Goddard      N  13.80  11.34
1999  5      Goddard      N  13.85  11.60
2000  5      Goddard      N  13.18  11.17
2001  5      Goddard      N  13.71  11.32
2002  5      Goddard      N  13.11  11.10
2003  5      Goddard      N  12.99  10.82
2004  5      Goddard      N  12.58  10.85
2005  5      Goddard      N  12.99  10.82
2006  5      Goddard      N  12.61  10.38
2007  5      Goddard      N  12.88  10.70
2008  5      Goddard      N  13.18  11.19
2009  5      Goddard      N  13.40  11.42
2010  5      Goddard      N  13.11  10.81
2011  5      Goddard      N  12.81  10.67
2012  5      Goddard      N  13.11  10.99
2013  5      Goddard      N  13.08  11.20
2014  5      Goddard      N  12.77  10.99
2015  5      NRTSI-G      N  12.65  10.78

 Important Notes:

 1) The "extent" column includes the area near the pole not imaged by the
 sensor. It is assumed to be entirely ice covered with at least 15%
 concentration. However, the "area" column excludes the area not imaged by
 the sensor. This area is 1.19 million square kilometers for SMMR (November
 1978 through June 1987), 0.31 million square kilometers for SSM/I (July
 1987 through December 2013), and 0.029 million square kilometers for SSMIS
 (January 2008 to present). Therefore, there is a discontinuity in the
 "area" data values in this file at the June/July 1987 boundary and at the
 December 2007/January 2008 boundary.

 2) When insufficient satellite data is available to process the area and
 extent values, the data is flagged as -9999. For more information see
 Processing Steps, Monthly Mean Ice Concentration in the documentation -
 http://nsidc.org/data/docs/noaa/g02135_seaice_index/index.html#processingstep.

[Paddy]

Thank you!

[Juan C. García]

NSIDC:

"2.2 Mean Concentration Fields and Median Ice Edge Position
On monthly extent images, ice ends and water begins where the concentration estimates of grid cells in the gridded average, or mean, concentration field for that month drop below 15 percent."
http://nsidc.org/data/docs/noaa/g02135_seaice_index/

NSIDC uses a method to calculate the Monthly Average, that it is different to what we usually think as an average. For NSIDC, if a grid cell has ice in 15% of the days or more, then the grid cell has ice for all the month. In May we have 31 days and 15% of 31 days is 4.65 days.
Rounding to 5 days, if a grid has 5 days with 100% ice, it means that the grid cell will have 100% ice, in order to calculate de NSIDC SIE for the month. If another grid cell has 50% of ice for 10 days in the month, that grid cell also enters with 100% ice to calculate the average.
Given this rules to calculate NSIDC Monthly Average, in months that there is SIE melting, the first 5 or 10 days of the month are the ones that "turn on" most of the grid cells that will defined the NSIDC Monthly Average.
So it is not important what happened on the last 15 days of May 2015, the May NSIDC Monthly Average will reflect what happen on the first days of May.
On the other hand, on the first days of June we are having daily records that are below to any other year. On a usual month, that will be enough to make us have a NSIDC June 2015 record. The only thing that could change that is if something very unusual happens, like a cyclon moving the ice to a region that was free of ice at the beginning of the month. If that happens and a grid cell has ice on 15% of the days, then that grid cell will also contribute to establish the NSIDC monthly average.
For me, it was interesting that I was able to predict a June record on NSIDC monthly Average, at the ending of May:

It is very interesting the form in which the NSIDC SIE has fallen apart on the last two weeks. It hasn't been seen in any other year. Knowing that the NSIDC June average is defined the first 7 days of June, we are going to have a huge record on NSIDC June SIE average, when it will be announced on July.

[Tor Bejnar]

Juan,
I wondered what you meant (on May 29) when you wrote "NSIDC June average is defined the first 7 days of June".  Now I know.  Thanks for the explanation.

[oren]

Thanks for the very clear explanation. What a strange averaging method, especially for something like sea ice that is sharply trending over the month.

[Juan C. García]

Juan,
I wondered what you meant (on May 29) when you wrote "NSIDC June average is defined the first 7 days of June".  Now I know.  Thanks for the explanation.

You are welcome Tor Bejnar.
Regarding the 2015 May average, I do not expect that we will have a new record on 2015, but it could be. The true is that the ice that is moving on the second part of the month also plays some role.
If we see only the NSIDC Charctic Graph and we concentrate on the begining of May, then 2004 should be above 2006 May Average. But 2006 May average was above 2004 May Average.
My bet is that on May 2015 we will not break the record, but we will do it on June 2015.
Let's wait and see.

[Paddy]

Yep, my casual prediction was off.  Partly because of not understanding/remembering that the NSIDC monthly average extent wasn't calculated from the average of the daily extents.

[ChrisReynolds]

Thanks for the very clear explanation. What a strange averaging method, especially for something like sea ice that is sharply trending over the month.

Yes, that was a much better explanation than my rushed and ultimately poor attempt up thread.

[Juan C. García]

Thank you, Oren, Paddy and ChrisReynolds for your comments.

NSIDC has already published the May figures and "May 2015 averaged 12.65 million square kilometers (4.88 million square miles), the third lowest May ice extent in the satellite record."

But what it is interesting is if we can forecast the breaking record that we should have on June 2015, that is competing with 2010, 2011 and 2012, being 2010 the smallest figure with 10.87 million square kilometers.

The 2015 NSIDC June Average will depend on ice concentration (Charctic shows extent but no concentration) and ice movement. But lets ignore those two components.

On June 3th, the difference between 2010 and 2015 is 200,000 km2. So lets forecast 10.67 million km2 +/- 100,000 km2, as the 2015 NSIDC June Average.

[Juan C. García]

At this moment it is difficult for me to understand why the NSIDC 2012 June Average is below the 2011 June Average. But anyway, I already made my forecast... 

[Juan C. García]

On June 3th, the difference between 2010 and 2015 is 200,000 km2. So lets forecast 10.67 million km2 +/- 100,000 km2, as the 2015 NSIDC June Average.

I find hard to believe that I forecasted that NSIDC 2015 June average will be less than NSIDC 2010 June average!
In less than two weeks we will see if it is true.