Monday, 6 April 2015

March 2015 Status

Data is now in for March 2015, the early maximum is confirmed, temperatures over autumn/winter continue the pattern of recent years, and the volume increase in PIOMAS remains. This year takes volume back to 2010 levels, but the impact of 2007 remains clear. My entry for the 2015 Sea Ice Prediction Network (SIPN) is given at the end of this post.

NSIDC Extent continues to show a relatively flat maximum, however I am not convinced that this means anything strong for the coming melt season. Indeed the early peak suggests that a new record is not likely, as discussed previously. However it does no more than suggest, and as I have no idea what the mechanism might be there, I will not factor that into considerations of the coming melt season.

The record of 14.54km^2 on February 2015 (NSIDC) looks increasingly certain. The following graphic shows sector extents, the sectors as defined in this post. Being contained within the Arctic Basin the Siberian and Alaskan sectors are flat (almost). The Pacific Sector (Okhotsk & Bering) remains on a bumpy plateau, but the Atlantic (Laptev, Barents, Greenland) and Canadian (Baffin, St Lawrence, and Hudson Bay) are both falling away from their peaks.

As usual, to make the results comparable with DMI I have calculated average surface temperatures north of 80degN for September to March. These are presented as anomalies (differences) from the NASA GISS baseline of 1951 to 1980.

Winter 2015 confirms the visual impression of a rapid rise of temperature starting in about 1995 and ending about 2006. This has been followed by a period of far lesser rise in temperature. But before passing that over, the rise of over 4 degrees C is staggering. Looking at the steepest rate of warming, I have taken the averages of two ten year periods, 1985 to 1994 (1.0degC) and 2006 to 2015 (4.7degC) this is an increase of 3.7degC, which happened in ten years.

Subtracting the 1951 to 1980 average from the 2015 period shows that the warming is spread over the Arctic Ocean towards Siberia and around the Atlantic Ocean ice edge.

Regular readers of this blog will recognise the following type of profile. Here, the same difference of periods is used as the above plot, but a vertical cross section is used.

So the astonishing warming seen between those two periods is shown to be a surface hugging feature with warming higher in the atmosphere being around a quarter of that at the surface. This is potentially indicative of two processes, one is heat flux through ice increasing as ice thins, the other is the decline of the surface temperature inversion typical of Arctic winters over the pack. With regards the latter effect, it is difficult to disentangle the decline of the inversion as this would itself be disrupted by an increase of heat flux as the ice thins. Either way, it seems to me, given the geographical distribution of warming seen in the preceding map, that the cause of the warming is sea ice decline.

I have read opinions that if the rate of winter warming keeps up we will see virtually sea ice free conditions early next decade. If we see the previous rate of warming resume then this is feasible. However my reading of the situation is that as the ice has thinned it has hit a situation of diminishing returns, Once the ice tends to a first year state the warming rate reduces. So further rapid winter warming due to ice loss seems to me unlikely to be as fast as it was between the 1990s and 2000s.

PIOMAS data is now available, as usual here is a map plot of thickness from the gridded thickness data.

This shows a thickening over the 2m threshold into Beaufort, as shown in my previous post on the matter, the Drift Age Model shows a substantial transport of multi year ice into this region. I suspect this means higher survival prospects in Beaufort, Chukchi, and possibly the East Siberian Sea in the coming summer.

Because people may find it interesting, I post an animated gif of September 2014 to March 2015 below.

And now for the usual plot of regional volume, which remains high in the Central Arctic region, but not convincingly high elsewhere. Note that the black trace for Central Arctic is on a scale four times as great as that for the rest of the regions.

Using the PIOMAS main series of volume from the Polar Science Centre's website, the volume as of 31 March 2015 is 24.036k km^3, this is 1.397k km^3 above that for the same date last year. Volume is virtually the same as that for 31 March 2010, 24.053k km^3. At present I am using the monthly gridded PIOMAS data (heff files), for these March volume is 23.207k km^3, which is 1.419k km^3 above that of March 2013. The Central Arctic is 1.337k km^3 above the volume in March 2013, so accounts for most of the increase in volume.

Looking at thickness/volume breakdowns, the increase in the Central Arctic for grid boxes reporting over 4m thick is 2.082k km^3, which is offset by reductions in thinner categories of ice. The increase for grid boxes reporting over 4m thick for the whole PIOMAS model domain is 2.094k km^3. So the picture seen throughout the autumn and winter is maintained: The volume increase is mainly due to the thickest ice in the Central Arctic.

Breaking the March PIOMAS gridded data down into five thickness bands shows that virtually all of the increase of volume is from grid boxes over 4m thick. One must go back to 2002 to find a greater volume of ice in that thickest category for March. However it is worth pointing out that for the whole PIOMAS domain, in 2008 volume for grid boxes over 4m thick was 2.188k km^3, but by March 2009 that had dropped to 0.437k km^3.

Despite the increased volume much of the pack remains relatively thin compared to past decades, this leads to a faster rate of thickening through the autumn and winter. This is shown in the volume gain from the previous year's minimum to 31 March. This year's volume gain was similar to last year and low for a post 2007 year, but remains hight compared to previous years. 2014's autumn winter gain behaviour was very much a post 2007 behaviour.

The March thickness distribution for 2015 largely follows the 2007 to 2014 pattern, apart from the substantial volume increase for grid boxes over 4m thick.

The volume peak is in April, taking the volume growths from 31 March for 1979 to 2014, and applying these small gains to 31 March 2015 volume (both for the whole PIOMAS domain) yields a series of simple projections of volume for the coming peak. This would mean a daily peak volume for April between 24.05 and 24.75k km^3. The average projection is 24.42k km^3, which would mean volume this April would be around 1k km^3 above that of March 2013 and 2014, and would keep volume around the 2010 level.

So near the end of the volume cycle for 2014/2015 the impact of 2010 seems to have been largely erased, but the impact of the 2007 volume increase remains largely intact. General conditions regards the PIOMAS data, and HYCOM and Drift Age Model images, suggest to me that there is little prospect of a new record or even extent near that of 2007 or 2011 in September 2015. Of course weather conditions similarly exceptional to 2007 could mean record losses, however I don't consider it sound to prognosticate on the basis of the inherently unpredictable.

Such unpredictable factors merely add to the 'noise' in the predictive model used, or need to be accepted as outliers for which prediction fails. In making my 2015 prediction for average September sea ice extent I will be taking the latter approach. To get in early for this year's SIPN (of course I can't actually enter it until June), here is my prediction for the 2015 September sea ice extent:

5.20 million kmsq +/-0.65 million kmsq.

The method uses March PIOMAS volume for the Arctic Ocean, and is totally statistical with no heuristic adjustment this year. Based on the hindcast from 1979 to 2014 the method is 81% successful, I have decided to accept a level of failure to predict hindcast years.

Here is the hindcast.


Nightvid said...

If the volume increase from March 2014 to March 2015 is due to the very thick ice near the CAA/Greenland, what physical reason to we have for expecting this to hold up the minimum extent in September? Since that area will stay ice no matter what, seems the impact should be limited compared to the areas that may or may not melt (such as the "SIberian central arctic", for lack of a better term).

Chris Reynolds said...

The PIOMAS effective thickness is from a sub grid thickness distribution. So even though a boax might report only 2.25m it might contain a significant part of ice many metres thick. This further level of detail is in files known as 'gice'.

Now, I have started to look at gice but work keeps getting in the way, so I haven't really got to grips with it. My suspicion is that it will show thicker mechanically deformed ice in Beaufort/Chukchi much as HYCOM and the DAM suggest this influx. However in the grid box effective thickness this manifests merely as a slight (25cm) increase in thickness. It doesn't take a great deal of multi year ice within first year ice to keep extent up because extent is only >15% (or about 1/7 cover).

As we've discussed before; I don't think extent as low as last year is guaranteed in Laptev, because last year's Laptev low was because of special conditions. I don't think Beaufort will be very low because of the MYI export. That also affects Chukchi, and may affect the ESS. The degree to which those latter two regions are impacted depends on the amount of export from Beaufort to those regions in the next few months.

Last year I ignored that other evidence from HYCOM and DAM, despite a commenter here saying I was unwise to do so. That was a major reason for my prediction last year being so low. This year I have no heuristic adjustment but have used roughly the same method. The numbers come out as they do, and I'm presenting those as a prediction. The hindcast is there for anyone wanting to get a feel for whether the method is any good.

David Rennie said...

I'm looking forward to your April analysis of the maximum, particularly the graphs showing the volume of ice in each 10 cm band so I can compare it with last year.

I am wondering why the total March volume shown in each of the 5 bands in the graph above do not add up to the total volume for March.

Chris Reynolds said...


Sorry about those graphs, I've uploaded correct versions of the 5 thickness band graph and the PIOMAS thickness distribution graph.

There is definitely an occasional problem with Excel. The graphs look OK in Excel but on copy and paste the previous view of the graph using an earlier month's data pastes. As this happens only occasionally it's hard to catch every time.

I'll post both this year and last for the thickness distributions when I come to April's post. Probably a difference plot will be of use too.

Carmiac said...

I've been thinking about how variable individual meting seasons are, and the answer generaly comes down to "weather". Is there a better way to quantify that? I'm specifically thinking of your FDD calculations, and wondering if you could do the inverse, Melting Degree Days and see what there is to see?

Chris Reynolds said...

Hi Carmiac,

Good question.

The FDD based calculation of ice growth is flawed because it only represents the growth of ice due to the heat flux through ice alone. A major issue is that my calculations don't include snow. However it does track sea ice growth fairly well for such a simple model, accounting for most of the growth in a model like PIOMAS. In winter volume growth seems to be mainly dependent on ice thickness, with weather being a secondary player.

I've found that the summer melt is not such a simple process. As PIOMAS is run its concentration of ice is continually adjusted by observed sea ice concentration. This seems to have less of an impact in winter mainly because most of the volume growth happens in land-locked seas in the Arctic Ocean, where concentration is high.

In summer the observed concentration seems to play more of a role in adjusting PIOMAS volume, especially after June. In short I have found what you suggest, that weather is a dominant factor. Even if the physics of volume loss in the summer were simple enough for me to model (they aren't), the unpredictability of weather makes the problem of prediction very hard.

There is another factor that makes use of a temperature based model difficult. Just as ice cubes cool a drink, as the ice starts to melt it keeps surface temperatures close to zero. This is why summer warming over recent decades is far less than autumn/winter/spring warming. So temperature isn't a good factor to use when looking at ice melt, it isn't going to show a strong past relationship with ice loss during the summer. What is more important are temperatures above the ice, which have a role in infra-red (IR) radiation from atmosphere to ice. Also cloud is crucial but is not simple. Cloud around June tends to block sunlight and reduce melt, while cloud around August tends to increase back-radiation of IR which dominates over reduced sun because the sun is at such a low angle (clouds help scatter light, where direct rays would be highly reflected), and the power from the sun is so low.

Anyway, I have calculated Melting Degree Days for ocean area north of 70degN. Calculation is between days 121 and 234.

1990 624.0769385
1991 613.3838791
1992 602.516682
1993 606.1220559
1994 601.5122688
1995 590.3993561
1996 546.2837775
1997 575.3468549
1998 587.4648855
1999 596.7308991
2000 576.2508283
2001 578.965969
2002 591.7001793
2003 549.9003996
2004 601.2558726
2005 533.4567557
2006 563.5716835
2007 589.5539167
2008 590.8724466
2009 601.0035764
2010 591.8651312
2011 598.6934373
2012 576.2727878
2013 663.4655465
2014 560.9031095

Check out 2013, according to this 2013 was 'warmer' than 2012 or 2014. Yet we know it was a cold summer. I've not had the time to dig deeper, but I suspect it was simply a longer period with surface temperatures above 0degC.

Carmiac said...

Thanks for that data and the explanation. It looks like there is a trend of about -3.5MDD/decade, but the noise is highly dominant. If we ignore 2013 as an outlier, it becomes almost -10MDD/decade. Which is odd? I would have expected that it would be going up with time.

Actually, maybe that makes sense if these are surface temperatures. The increase in melt would be keeping the surface cooler.

Chris Reynolds said...

Sorry, I should have been more clear. Those are derived from T2m from NCEP/NCAR Reanalysis 2. T2m is the product designed to be surface temperature 2m above the surface.

The galling thing about summer is we know more energy is going into melting ice. But it's hard to quantify where from, although PIOMAS can give clues. Summertime T2m over the ocean should rise in the coming decades once the ocean becomes ice free in summer.