Arctic Dipole: A dominant role.

Zhang et al (2008) identify a recently emerging Dipole Anomaly in the Arctic, they call this the Arctic Rapid Change Pattern, or ARP, throughout this post I will however use the commonly accepted term; Arctic Dipole (AD). They use a running window EOF technique to examine small time periods over which the AD has changed. This means that instead of Wu et al's approach of applying the EOF to a long period, Zhang et al apply the EOF to short (5 year) runs over the period 1886 to 2006, this allows them to analyse the spatial changes with time. They don't refer to the previously discussed Wu et al paper.

The following is figure 1 from the paper, this conveys their key findings graphically.

Figure 1 of Zhang et al, 2008, "Recent radical shifts of atmospheric circulations and rapid changes in Arctic climate system."

In the above figure the panels show the first EOF (dominant mode of atmospheric variability) for six consecutive periods, the percentage above each panel is the amount of variance in the period that the dominant mode accounts for. The first 5 panels show changes in the Arctic Oscillation (AO) over the period 1987 to 2002. The final shows the emergence of the AD as a dominant mode.

It is notable that the centre of action of the AO shifts northeastwards, having originally been over the Icelandic Sea. The shift is stated to be related the observed poleward shift of storm tracks and intensification of Arctic storm activity. I'm still trying to reconcile this finding with the apparently counter-intuitive finding that atmospheric heat flux into the Arctic has decreased since the 1980s (Smedsrud et al). In the previous post I mentioned that the AO and Fram Strait ice flux only show a correlation after the 1970s. In view of that it would be intersting to see what changes there were in the AO prior to the 1980s, such changes in the centre of action may explain the change in AO/Fram-ice-flux correlation.

Zhang et al find that the AO robustly explains most of the variance (over 20%) over most of the period that they consider, previously I noted that Wu et al 2006 stated the AO accounted for around 60% of variance. However from 2002 onwards the AD becomes the dominant mode, explaining 21% of the variance. It's worth noting here that Zhang et al's use of EOFs over small windows as opposed to Wu et al's EOF over a long period should (as I understand it) bias the amount of variance accounted for in Zhang et al downwards. When a longer dataset is considered short periods of random fluctuations are in a sense averaged out, so the findings of Zhang should as I understand it be considered more conservative than those of Wu et al (If you're reading this and can correct my best guess please feel free to enlighten me).

Zhang et al propose that this shift to the AD, and the AD pattern itself, marks an unprecedented climate change signature. Wu et al found that the AD, although less dominant and explaining only 13% of variance, was visible as the second leading mode in the Arctic's climate: So I would question the novelty of the AD because it was present in the 20th century. However the shift to a dominant AD does seem to be a new and important signature of climate change. It's worth noting here that the AD pattern of Zhang et al apparently differs from the pattern described in Wu et al, however this may be because Wu et al consider a longer period than the time windows used by Zhang et al, and Zhang et al are considering the AD's move to dominance, Wu et al's AD was in the second not the first leading mode.

Zhang et al examine the correlation between the AD, temperature and sea-ice changes. They find that when the AD is negative it draws air from the Atlantic into the Atlantic sector of the ice-pack, this notably causes a negative correlation between temperature at Longyearbyen (Svalbard) and the AD. They also carry out a lagged correlation test on sea ice anomalies in the summer with the preceding winter's AD index and find a significant decrease in sea-ice in the Barents, Kara and Laptev seas with a preceding winter's negative AD index this fits with the findings of Wu et al as discussed in the preceding AD post. Zhang et al's figure 2 also shows that with a negative AD there is an impact of cooling and increase of sea-ice in the Pacific sector over the period 2002 to 2007. In contrast Wang et al show that the greatest impact upon sea-ice is with the positive AD mode, this is supported by a later paper which I'll cover in my next post.

Furthermore Zhang et al present a timeseries of the amplitude of the AD from 2001 to 2007. This shows that the AD had a downward trend from Winter 2001/2 until winter 2005/6, when it began an upward trend into a positive index leading up to the crash of 2007. The AD seems to have been the key player in the both preconditioning and the anomalous high and Fram Strait export of of 2007. Other research supports this contention, so this raises the probability of further crashes like 2007. However we are in the early throes of a new atmospheric regime in the Arctic, so it cannot yet be said how often and how persistent such extreme positive AD phases will be. Furthermore as this year is certainly showing, the sea ice, being thinner than in the past, is more responsive to atmospheric impacts.

So we now seem to have a plausible physical reason to home in on the the year 2002 as a feasible breakpoint in assessing Arctic trends, as we now see the emergence to dominance of an old, formerly secondary, atmospheric player. It's been obvious that the summer sea ice extent loss is accelerating and often people fit non linear curves to the decline. But I thought it interesting to see what happens when I fit linear trends using 2002 as a breakpoint. The Arctic Dipole should change the rate of loss as it becomes dominant because it affects ocean, and atmosphere heat fluxes and directly impacts sea ice transport.

Trends in September Average sea-ice extent. Blue line, NSIDC September average sea-ice extent. Red line, pre and post 2002 trends. Data from NSIDC.

1979 to 2001 is period 1, 2002 to 2010 is period 2. From Excel (Linest and Fdist functions): Period 1 has a slope of -0.046M km^2, r2 of 0.33, probability that the trend is by chance 0.0038. Period 1 has a slope of -0.177M km^2, r2 of 0.53, probability that the trend is by chance 0.0054. The loss rate in period 2 is almost 4 times the loss rate in period 1.

NB - See comments regarding the following paragraph.
Recently over at Neven's Sea Ice Blog, Seke Rob, one of the posters, has revealed a stunning graphic of variation of days at certain levels of extent through the melt season. You can view the image here, I've deliberately not done it as a graphic in this post because it needs higher resolution than my blog will comfortably accomodate. Firstly the days of the year are shown as a day count on the X axis, the Y axis are years, the graph shows the progression of each year. The coloured bars for each year are the number of days in a certain band of sea-ice extent. So starting from the bottom green (less than 12 million kmsq) is around day 160 - 175, 20 years on, 1999 and it's moved on to between days 145 and 160. But move up a bit further and you'll see a jump in 2003 a jump that's sustained for the following years. Is there a connection between this and the findings of Zhang et al?

In a previous post I discussed Hakkinen et al's findings of an increase in sea-ice drift due to increase in the speed of the transpolar drift and the action of increased storms. Their figure 1 shows a long term trend of increase in speeds, however the summer speeds in particular have an apparent jump in speeds, again this is after 2002, although other evidence does not show a trend in the AD after 2002, despite it's rise to dominance.

I've already touched on the impact of the AD on the sea-ice, but there is one very useful paper that covers this excellently and I'll address it in my next post in this series.


Lars Smedsrud et al, 2008, "Recent and Future Changes of the Arctic sea-ice cover."
http://www.folk.uib.no/gbsag/Smedsrud_etal_2008.pdf
Bingyi Wu et al, 2005, "Dipole Anomaly in the Winter Arctic Atmosphere and its Association with Sea Ice Motion."
http://www.atmos.washington.edu/~dennis/WuWangWalsh_Arctic_06.pdf
Xiangdong Zhang et al, 2008, "Recent radical shifts of atmospheric circulations and rapid changes in Arctic climate system."