Saturday, 7 June 2014

PIOMAS May 2014. Part 2

As promised here is a more detailed breakdown of the PIOMAS May data release.

Using the main PIOMAS series here are the anomalies (baseline 1980 to 1999), on schedule the spring volume loss anomaly is occurring. This can be seen at the end of the red trace below, years from 2000 are included to show how unusual this behaviour is.

This means that every year from 2010 onwards shows anomalously large volume losses in spring, these have been present despite the upgrade from version 2 to version 2.1, and even the muted melt year of 2013 showed the same behaviour. This supports my previous opinion (see here) that the enhanced spring melt is due to ice conditions, not due to weather. Its emergence follows the 2010 volume loss event, see here, and as with 2007 this seems to have caused a change in seasonal behaviour in the ice pack.

This shift is not just apparent in PIOMAS data. The plot below shows NSIDC Extent June losses for 1979 to 2013, after 2010 there is an increase in June extent losses, despite varying weather conditions.

So along with the 2007 loss event which changed the seasonal cycle, within four years we have another game changing event. Am I the only one who finds that pretty amazing?

Now turning to the PIOMAS gridded data I'll go through the usual sample regions in order.

Whole PIOMAS Domain.

The thickness profiles, as usual, are visualisations of the volume (vertical axis) as a function of grid box thickness (horizontal axis). First the differences between 2014 and 2013 or 2012. The peak lower down followed by negative values for thicker ice shows that this year the bulk of the pack is thinner than in 2012 and 2013. However at the high end, for grid boxes 4m and thicker, there is present a large volume of thicker ice when compared to the last two years, this is the residual volume from the 2013 volume gain and is restricted to the centre of the pack off Greenland and the Canadian Arctic Archipelago (CAA).

Comparing May 2014 to other recent years for the whole PIOMAS domain shows that whilst there is a peak of volume in the thickest category the peak in the distribution is similar to 2011 and 2013. Despite being thicker 2012 turned out to be a record breaking minimum year, which shows that thickness/volume is not everything on the year to year timescale.

On the subject of 2012, in each of these regions I'm going to give the thinning profile for 2012 to introduce the thinning profile for 2014, which I will be following through this melt season. The graph below shows profiles of thickness for each month, from April to September.

As in the above graphic 2014's thinning profile shows little change, but as will be seen in the following regions this is not generally the case. Note the unallocated colour bars in the legend of that graph, these are waiting for future months data to be added.

Beaufort Sea.

The current state in the Beaufort Sea is seen to be different from 2012 and 2013 in a near identical manner, we see the positive differences below 2m thick and negative above indicating the thinner profile of ice in Beaufort this year. Note that there is no increase in thick ice volume in 2014, this is not a factor in PIOMAS data in Beaufort this year.

Comparing with post 2010 thickness distributions and current conditions look to be about the thinnest. Indeed at an average thickness of 1.719m Beaufort is currently the thinnest in the PIOMAS data, with 2008 being the next thinnest at 1.722m.

In 2012 there was very little thinning in Beaufort between April and May.

However in 2014 a thinning has occurred, this might be due to movement of ice away from the Alaskan coast rather than extensive melt as such because temperatures over much of the pack in May were still sub zero (off the Alaskan coast around 0 to -0.3 (NCEP/NCAR)).

East Siberian Sea (ESS).

Similar to Beaufort the ESS shows a thinning compared to both 2012 and 2013. Note that, as with Beaufort,  there is no increase in thick ice volume in 2014.

2014 is seen to be most similar in thickness profile to 2011, however the peak of volume is far less. At 1.568k km^3 the ESS May volume was the lowest in the dataset, 2012 and 2013 being around 2k km^3.

Looking back again to 2012, as with Beaufort between April and May there was insignificant thinning, what there was seems to appear as a movement down the thickness bands losing a sight increase in volume around 2.7m thick in in April and translating that into a raising of the volume at around 2.4m.

This year there has been a noticeable drop in thickness from around 2.1m to 1.8m at peak, again a translation into higher volume at 1.8m in May has occurred, this may be due in large part to export and open water because it is associated with a drop in volume from 1.732k km^3, to 1.568k km^3 between April and May (off the Siberian coast around -0.3 to -0.6 (NCEP/NCAR)).

Central Arctic.

In the Central Arctic the increase of ice 4m thick and greater makes an appearance, comparison of 2012, 2013 and 2014 shows a more complex picture than in Beaufort and the ESS. Around 2m we see the same change of sign of differences caused by ice around 2m thick in 2014 being thinner than in 2013, in 2012 there was marginally less ice around 2m thick. Note that the title is incorrect, it should be Central Arctic, not ESS.

The overall thickness distributions for recent years shows that while we have the increased volume of the thickest ice the peak of volume around 2m and above is reduced due to increased extent of thicker ice.

Again I turn to 2012, here the thinning profile throughout that melt season is rather complex, although by July the thinning is marked and changes the nature of the distributions. This is similar behaviour to the other regions in the Arctic Ocean and show the importance of insolation after the solstice falling on wet ice (lower albedo) preconditioned by events in the first half of the melt season.

Looking at the first two months of the 2014 season, note that the increase of thickest ice volume (4m and greater) is almost identical to that shown for the whole PIOMAS domain, the thickest ice and its resultant increase in volume remains contained in the Central Arctic so will play no role in the coming melt season until the ice edge enters the Central Arctic. However later in the season we may find that the periphery of the pack in to the Central region presents no more an obstacle to melt than in 2011 and 2012.


Due to Wipneus raising the point, thickness calculations have changed, so I've stopped using concentration in thickness calculations and am able to update to the current month. In closing here are a couple of graphs, for the Arctic Ocean, thickness by decade and then for post 2010 years.

Thickness (aka volume) may explain the downward trend in area and extent, but it's not a great predictor of what happens in any one year.


Anonymous said...

I've been thinking about the unusual reduction in the volume anomaly this winter and propose that the PIOMAS systemic bias errors may be a contributing factor.

The recent results clearly show a substantially larger portion of the ice field has become 3.5 to 4 meters thick, and
The PIOMAS validation charts demonstrate that the model underestimates thick ice in comparison to the measured thickness for ice > ~3.5 meters.

So, the recent change to the distribution of ice thickness has included a bias error for the thick ice, thereby underreporting the overall volume. The perceived reduction in overall volume is because of the change in the model error from overestimating thin ice to underestimating thick ice.

The "adjusted" volume plot on the validation page only extents to 2010 when the predominance of thin ice suggested a negative adjustment from the prediction. This winter the ice field shifted to a larger region of thick MYI ice as the winds pushed it against the CAA instead of toward the Fram Strait. The increased thickness of that substantially thicker and larger ice field has shifted the model error away from overestimating the volume toward underestimating it.

I suspect that the reported loss of the 2013 volume increase is influenced as much by model error as it is any moderate temperatures for a couple weeks in February.

Chris Reynolds said...

Hi Anon,

Schweiger 2011 (Uncertainty in Modeled Arctic Sea Ice Volume) states that PIOMAS has a tendency to "underestimate thick and overestimate thin ice". As the pack has tended to become first year ice and the extent of thicker multi-year ice has decreased the bias is to be expected to be as the adjusted volume anomaly suggests, that PIOMAS underestimates sea ice volume (at least during the winter).

As can be seen from the 'post 2010 thickness distributions' graphic for the three regions considered, the thick ice is mainly confined to the central Arctic. It may be that there is more thick ice in the Central Arctic region, but this region plays little role in the summer melt, and the long term history of the preferential thinning of thick ice suggests this ice has a limited future.

I seriously doubt that model error accounts for the evolution of ice volume this past winter. The model is subject to NCEP/NCAR forcing, the response to forcing does not evolve significantly with time within the seasonal cycle. It is true that as volume moves into thicker thickness categories it will be more subject to the thickness error tendency in PIOMAS. However this doesn't account for the observed behaviour of the volume 'pulse'.

This graphic is from an earlier post. It shows the volume difference for the dates on the x axis and twelve months earlier, i.e. 2013/14 minus 2012/13.

Chris Reynolds said...

Sorry, should have closed that off..

Considering that graphic of difference from 12 months earlier, note the precipitous drop in volume in February. During February there was a PNA induced warming aloft - see the earlier post for more detail.

Anonymous said...

I'm not making an assessment of how this might effect the summer melt toward minimum.

The point is that the volume of thick ice (>3 m) has increased in 2014, it is large relative to the last 4 or 5 years. If the validation plot indicating a negative bias error is correct, then the reported volumes for those regions is about 15 - 20 % low. Estimating from your thickness plots, that suggests that PIOMASS is under reporting the volume by 300 to 500 km^3. Therefore, all of the volume gain during 2013 was NOT lost, the model error changed. The low biased result for 2014 is actually greater than the high biased result for 2013.

Chris Reynolds said...

Sorry, looking towards the minimum is rather on my mind throughout the melt season.

I do get what you're saying, and yes there may be a slight under-reporting of volume in the central Arctic. But to be clear, the first link at the end of my 18 June 2014 10:25 post does demonstrate that what happened to remove much of the volume difference was a discrete event. How this can be due to a continual bias doesn't make sense to me. But yes, it is likely that thickest ice volume is being under-reported.