Monday, 25 May 2015

The Slow Transition: The Central Arctic

For the Arctic Ocean to become seasonally sea ice free the Central Arctic region must become sea ice free. What is happening in PIOMAS in the Central Arctic, and what is likely to happen in future?

As always the definition of the Central Arctic I use is the region shown by Cryosphere Today, the central green coloured region labelled Arctic Basin. I have previously shown the fast transition of the peripheral seas of the Arctic Ocean to a virtually sea ice free state in late summer. Being first year ice dominated in recent years these seas give a picture of how a seasonally ice free Arctic Ocean will behave. There is a clear increase in September virtually ice free state within the peripheral seas of the Arctic Ocean basin (Beaufort, Chukchi, East Siberian, Laptev seas), as I outline here, and this process will entrench as winter peak thickness continues to decline and the region continues to warm.

An alternate view of this fast transition can be gained by looking at Arctic Ocean and Central Arctic September volume, presented in this way it is clear that in recent years almost all ice volume at the end of summer resides in the Central Arctic, and that this is a recent development, by no means typifying past behaviour of the Arctic Ocean's sea ice.

So the process of transition of the regional seas of the Arctic Ocean to a seasonal ice pack is approaching completion, what remains for the whole Arctic Ocean to transition is the Central Arctic.

Another perspective is to look at April volume (the month of peak volume for the whole Arctic) and volume losses from April to September (melt season losses). By definition, when melt season losses equal April volume the result is virtual ice free conditions in late summer, in the East Siberian Sea (ESS) this state has persisted since 2007, a behaviour common to other peripheral seas.

Using the same perspective we can look at the Central Arctic, the largest single region, and the region that needs to become ice free to allow the Arctic Ocean to transition to a seasonally ice free state.

The trend fits have no underlying physical significance, they are merely illustrative quadratic fits. But in the case of the Central Arctic April volume and melt season losses are some way off meeting. The simple quadratic fits used in the above graph are merely to indicate a possible result of what may happen if things continue as those fits suggest. However throughout this series of posts I have repeatedly pointed out the problems with such curve fitting and the necessity for considering the underlying physical processes. I do not expect these curves to continue to converge so sharply.

With regards the April volume decline, as I have shown on many occasions, the volume loss in PIOMAS has all come from thicker ice, with thinner ice actually increasing in volume. In the Arctic Ocean there are two categories of sea ice, first year ice grows mainly thermodynamically (see my previous two posts), older multi year ice (which has survived at least a year) thickens to the point where thermodynamic growth is less important and mechanical deformation takes over as the dominant way the ice thickens. Ice in April that is thinner than 2.2m is probably representative of first year ice dominance, grid boxes thicker than 2.2m probably reflect substantial inclusion of thicker, older, multi-year ice.

This decline of thicker ice is in line with findings based on submarine upward looking sonar transects which show a preferential thinning of thicker ice. This thinning cannot continue indefinitely, levels of multi year ice are already very low, and they cannot drop to zero except in the case of a fully seasonal Arctic Ocean ice pack.

I showed in an earlier post how a simple thermodynamic model of sea ice growth, combined with the use of Freezing Degree Days as an indicator of freeze season (September to April) cold indicates that without massive winter warming, thickness growth over winter will be held up for ice growing thermodynamically. Turning now to the April volume of sea ice in PIOMAS I have calculated for a 1.5cm per year decline in April thickness of thermodynamically grown ice, as calculated in this post. This is then applied to April 2014 Central Arctic Ice thickness to produce a suggested trajectory of thinning. In fact my  work on Freezing Degree Days (FDDs) and the simple model of sea ice growth, and later work on PIOMAS Gice, implies that 1.5cm per year reduction in thermodynamic growth is probably too strong a loss, but this is a discussion of principle, not a prediction.

Allowing for the negative feedback from the thickness-growth feedback arresting the decline in winter volume, while keeping the same quadratic fit to melt season losses pushes the point at which the curves meet back by a few years, so the summer ice survives for longer.

Now to look at the quadratic fit to the melt season volume losses, the blue curve of increasing losses. I suspect that what is going on here is an increase of open water formation efficiency caused by the movement of greater volume (or area) of ice into thinner ice categories. From this viewpoint the volume loss events of 2007 and 2010 play a role in dropping the thickness of regions of the Central Arctic, with a slightly rising linear trend best fitting the years before those step decreases in thickness in the outer edges of the Central Arctic.The following graph should read 'melt season' for 'summer'.

The whole process is happening fast amidst substantial variability due to weather, so I am unable to produce a mathematical backing to this suspicion. Despite this, if the linear increase from 1978 to 2006 is the 'background' trend of increasing melt season loss, and the thinning following 2007 and then 2010 acts to shift that line up without affecting the slope, then it would delay the meeting of melt loss and April volume even further.

So I am not convinced by extrapolating the trends of melt season loss and April volume which implies an ice free Arctic by the very early 2020s.

Gice is the thickness distribution used by the PIOMAS model within each grid box. The following graphic shows the timeseries of April volumes from 11 thickness categories (12 available, but 0m is excluded as by definition it has no volume).

This shows that volume for the 4.23m thick band and thicker bands is declining. In contrast to this, the volumes for thinner ice are not declining, 2.61m is relatively level compared to thicker ice, while 1.46m and 0.71m are increasing in volume. So why is this thinner ice not declining as all thinner ice categories are? It might be supposed that the thinning is affecting all thicknesses and dropping the thickness profile towards zero, this is not what is happening,

In my previous post about the BCE region I showed how despite the decline of thicker ice there is a wall stopping the decline of volume in thinner ice. The same is happening within the Central Arctic Region. As with that previous post I have drawn the 'wall' as a fuzzy band because it will thin as the Arctic warms.

So the thinning within the Central Arctic is not just an issue of moving the thickness profile to a profile of the same shape, but lower average thickness. Instead we see the thicker ice collapsing, but the thinner ice being maintained, the profile of thickness is being crushed up against an invisible wall, and this invisible wall is due to thermodynamic growth of ice. Winter cold opposes the thinning by regrowing new ice and as the decline of thicker ice is progressing faster than winter warming the thick ice collapses while thinner ice does not.

This is why the gice thickness of 4.23m and above is declining while the ice of 2.61m remains roughly steady, and the thinner ice volume increases.

A crucial test of any hypothesis is prediction, when using a physically derived model one need not await further years of data, one can hindcast (running the model on past data). A physical model that is not derived through fitting to data is blind as to whether data is for the past or future.

In my post The Simplest Model of Sea Ice Growth I showed the derivation of a simple model using basic physics. In my post on the thickness growth feedback I used the simple model to calculate sea ice volume gain for the Central Arctic, and compared this to actual evolution of volume gain in the PIOMAS model. The agreement was very good.

The simple model of winter ice growth is able to closely match the winter growth of ice seen in PIOMAS. As also shown in that post using the model, without massive winter warming, there is no immediate prospect of a rapid crash in winter volume in the Central Arctic.

On the subject of winter warming, it is worth looking at NCEP/NCAR reanalysis surface temperatures north of 70degN for November to December. With reference to the plots below and above, it is clear that even as winter temperature has increased, so has the above autumn/winter volume growth.

So despite the massive winter warming of over 3 degrees during recent decades, the transition to thinner ice in the Central Arctic has still been able to increase ice growth substantially, when one might reasonably supposed ice production has decreased. There is no shortage of cold to grow ice over winter in the Central Arctic. Furthermore, once the seas around the Central Arctic become ice free earlier in the summer and, as in 2012, the ice edge begins to penetrate the Central Arctic regularly, there is even more potential for increased volume gain because of the relationship between thickening and initial thickness.

The above plot is from the simple model of sea ice, what it shows is that as initial sea ice thickness declines,  the amount of thickening over winter icreases. Various levels of winter cold, indicated by Freezing Degree Days are shown on the plot to give a qualitative feel for the impact of winter warming. Paying attention to the lowest light blue line (FDD is 0), where there is no ice growth over winter, at 2.5m September thickness the spread of the curves is only 0.8m but for a September thickness of 0m the spread is 2.2m. This behaviour means that right down to an ice free state in September with a winter having and FDD of between 3000 and 4000m the September thickness maintains at around 2m. For comparison FDDs for recent winters are around 3800 for the region north of 80degN. The model is a simple representation of the process of freezing, it does not include snow or ocean heat flux, both of which will make real ice (or ice modelled in a more complex model) thinner.

If thickening increases then so does ice volume production over winter, and the more winter volume is held up the more volume there is to melt during the summer, so the earlier the ice edge needs to enter the Central Arctic to bring about virtually ice free conditions.

The Central Arctic continues to thin, but there is little prospect of a rapid collapse of winter peak thickness. Rather what will happen is that with continued human driven global warming the atmosphere and ocean heat fluxes in winter will increase and winter warming will reduce the peak thickness. As winter thickness declines increasing loss of summer extent within the Central Arctic becomes feasible and the prospect of a virtually ice free Arctic in September will eventually become a reality.


Anonymous said...

It's disappointing to see an analysis made in May 2015 that ignores the most recent data available; i.e. April 2015 PIOMAS data. You've deliberately omitted it from your plots and the FDD plot actually shows a previously predicted drop in 2015 to ~9000 instead of the actual increase to 12000. I understand that you want to emphasize the long term trend instead of the recent rebound that you attribute to the weather; but, these omissions aren't good science or engineering practice.

It's much better, and more credible, to serve up all the facts and present an imperfect analysis than to filter out the unsavory data to fit a favorite theory.

Chris Reynolds said...

Anon, I disagree.

The bulk of this post was written in January, I've just been wanting to find an appropriate moment to fit this in. I have covered the volume increase extensively. In this post from September I state a 'massive volume gain'.

In my April status posts I covered the volume increase extensively. Earlier in the winter I have cautioned that the volume increase would fully carry over to 2015.

You are correct that I am concerned with long term fundamental process, that was one reason to omit the April 2015 data. Furthermore I do not have monthly gice data for April 2015. Thus the correct decision was not to invest extra time in recalculating everything to include data which would confuse an already complex argument.

In actual fact I could have included the April 2015 data without comment, that would have seemed to support my theory (it is a volume increase following the 2012 crash!), so your allegation that I am omitting data to favour my hypothesis is utter crap.