Cold Winters: The Snow Advance Index.

Recently (well, back in November) Dr Judah Cohen published a new paper on the linkage between Siberian snow cover and Northern Hemisphere winters. The paper By Judah Cohen and Justin Jones is "A new index for more accurate winter predictions." There's a link to it at the end of this post. I've previously commented on the possibility of increased incidence of cold winters due to low solar activity and Arctic sea-ice. I've seen those as biassing for colder winters, with Cohen's previous work on Siberian Snowfall showing a trigger for specific cold winter events. However Cohen and Jones' new index makes me wonder what room there is for the other factors.


The major factor dominating Northern Hemisphere in winter is the Arctic Oscillation (AO). For us in the UK (and Europe) the useful rule of thumb is that when the AO is negative winter weather is typified by blocking high conditions, clear skies and colder than average weather. When the AO is positive the UK's weather is typified by low pressure systems from the Atlantic. We've seen this recently; in the last few weeks the UK has been battered by low pressure systems, strong winds, and days marred by wintery showers. That's typical high index AO weather, and the AO has been high (NOAA). In contrast, during the winter of 2009/10 the AO index dropped to record lows in February and was low throughout the winter.

Cohen & Jones follow up previous work investigating the correlation between October Siberian snowfall and the state of the AO in December, January and February. However in this paper they use a new index of Siberian snowfall: The Snow Advance Index (SAI). What Cohen & Jones do is they move from using Siberian snow extent to the rate of change of Siberian snowcover. They use snowcover in Eurasia south of 60degN and by moving to the rate of change of snow cover find a better correlation with the winter AO. Their figure 2c shows the best correlation using the highest quality data (daily data are only available for this period).

Figure 2c of Cohen & Jones. October SAI - blue. Winter (Dec-Feb) AO - black.

Here the correlation is 0.844, using weekly snow cover extent and October extent the correlations are lower. It would be more convenient if we had more events like 2009/10 in the series, however in my opinion Cohen has already previously demonstrated a role for Siberian snowfall in the winter using NCEP/NCAR reanalysis to examine what was going on in the atmosphere. As the SAI and AO index (NOAA) used in Cohen and Jones are indepenent of the NCAR/NCEP reanalysis the two lines of evidence give independent corroboration.

What I find challenging about Cohen and Jones is this:

Previously I've discussed how Lockwood et al found a link between low solar activity and cold European winters using the Central England Temperature series (first link in this post). How then am I to resolve Cohen and Jones with this line of evidence? Lockwood et al leave open the issue of mechanism, although they do discuss candidate mechanisms. Cohen & Jones and Cohen et al 2010 suggest that the winter of 2009/10 was instigated at the surface, could the reduced solar activity linkage be through impacts on clouds in line with Svensmark's hypothesis? It would be good to see if the SAI/AO correlation held in the Little Ice Age, but I can't think of a proxy that would work for this purpose.

Furthermore as previously discussed (second link this post) James Overland has been arguing that the winter of 2009/10 was probably a consequence of the loss of Arctic sea-ice. I've found the idea of reduced sea-ice causing impacts on winter weather persuasive, particularly due to the observation by Francis et al of persistent impacts to the Atlantic geopotential height thickness. Am I now to assume that this actually has a near negligible impact on European weather? If loss of sea-ice were having an impact I'd expect to see a reduction in correlation in Cohen & Jones as time moves on. However on balance their figure 2 doesn't suggest the sort of deviation I had initially expected might be there. Even with the longer timeseries used in the paper there doesn't seem to be a perceptible divergence between SAI or SCE and winter AO, so I'm left wondering if the Arctic sea-ice - cold winters paradigm has been overstated.

If Cohen and Jones are correct and the correlation is as high as 0.84 (84 71% of variance) then these other factors acting directly are left with 16%. There is of course the indirect effect, such as increased atmospheric humidity due to open water in the Arctic. However even this indirect effect doesn't seem to be supported as Rutger's Snowlab doesn't suggest a trend in October or November anomaly.

I think Cohen and Jones are onto something, and as a result being able to predict such winters as 2009/10 as early as November will be a substantial benefit allowing preparation to reduce the impact of such winters.

But as for the Cold Winters-Solar/Arctic linkages, at present I don't know. I remain unwilling to throw the baby out with the bathwater and dismiss these ideas. However I'm less persuaded than I was as to the magnitude and mechanism of the effect. Lockwood (et al) does seem to have found a real link between low solar activity and cold winters in the Central England Temperature series (ref), and this seems to be associated with increased incidence of blocking highs during winter at such times (ref), also the long term behaviour of Arctic sea-ice (ref) seems to suggest a support for Lockwood's theory. In terms of the impact of reduced sea-ice there is no doubt that this is having an impact upon the Arctic atmosphere: The Arctic Dipole Anomaly has risen to dominance (ref, ref), and through it spring sea-ice concentration has impacts on summer rain in China (ref), and may have wider impacts in the Northern Hemisphere (ref). Amidst this evidence there still seems reason to expect a winter impact of reduced solar activity and reduced Arctic sea-ice.



Cohen & Jones, 2011, "A new index for more accurate winter predictions."
http://web.mit.edu/jlcohen/www/papers/CohenandJones_GRL11.pdf