Northern Hemisphere Blocking.

Over at Neven's Sea Ice Blog, Steve Bloom posted a link to some data on Northern Hemisphere blocking, I've got to work on the data Steve linked to, but haven't really got to the bottom of the matter. This data holds some exciting detail. I've not reached a firm conclusion on this, I need to read some more papers on blocking. But this post summarises where I've got to.



Atmospheric Blocking is caused by static high pressure systems that block the path of low pressure systems, which are carried from west to east by the Jetstream, it is the blocking of such flow that defines a blocking event. Semi-permanent highs such as the Azores High are not blocking in nature. These events are associated with events such as the 2003 European Heatwave, the 2010 Russian Heatwave, and the 2009/10 severe winter. So it is apparent that changes in blocking events could have implications of a practical, rather than purely scientific, nature.

The blocking data is from Dr Anthony R Lupo & Colleagues, available here. Temperature and geopotential height data are from NCEP/NCAR, available here, grids have been area weighted. All data used is for the Northern Hemisphere. Data were originally downloaded in late December 2012, this affects the end date of some calculations as December 2012 falls in the calendar year 2012, 2011 is used as the end year in some calculations. I normally link to papers in a reference list at the end of each post, I've not done that for this post because I can't claim to have properly read and considered the papers linked to in the text.

First place to start is to look at time series, the data is provided as timeseries of individual blocking events, the variables that seemed of most use initially were length of blocking event (days) and intensity of blocking event. Blocking intensity is as defined in Wiedenmann et al (2002). Geopotential height (GPH) is the height of a given pressure level in the atmospheric column, Wiedemann et al use the height of the 500mbar pressure level, taking the peak GPH at 500mb within the blocking pattern they scale this proportionally to what they term a representative contour. The representative contour (a single number) is derived by tracking across the line of latitude at which GPH is maximum, the result is that the BI gives a measure of the GPH gradients in the region of the blocking high.

A lot of these graphics are big, just click on the graphics to enlarge.

Fig 1. Total Blocking Days per Month.

Fig 2, Average Blocking Intensity per Month.

It's clear that something is afoot; since around 2000 there's been changes in both total blocking days and average intensity per month, notable is that months without blocking events become almost non existent after 2000.

Concentrating on the total blocking days per month, what change has caused this post 2007 shift? To avoid incomprehensibly spaghettified graphs I've broken down the data into seasons, using meteorological seasons:

Winter - December, January, February. DJF
Spring - March, April, May. MAM
Summer - June, July, August. JJA
Autumn - September, October, November. SON

These also line up with seasons of Arctic sea ice and consequent Arctic climate, which will become relevant further on in this post.

There are two factors that could cause a change in the total of blocking days each month; a change in the average length of blocking events, and/or a change in the number of blocking events.

Firstly the number of blocking events per month, which is more directly comparable with the first two graphs.

Fig 3, Number of blocking events, by month.

Now seasonal breakdowns.

Fig 4, Number of blocking days, by season.

Fig 5, Average length of blocking events, by season.

Fig 6, Number of blocking events, by season.

The increase in blocking days seems to be mainly due to an increase in the number of events.

The scale of the change can be seen in figure 4, but can also be appreciated from a histogram of distributions of total monthly days. This is calculated by calculating the distributions of the total number of days of blocking per month for two periods; 1968 to 1999, and 2000 to 2012. Then normalising the resultant distributions to allow comparison by expressing each total as a fraction of 1, where 1 is scaled to the total number of days.

Fig 7, Normalised frequency distribution of total blocking days per month.

Dr Lupo's data also includes the starting longitude of each event. So we can look at an aspect of the geographical distribution of both duration and intensity of each event.

Fig 8, Time and longitude, with width of each data point relating duration.

Fig 9, Time and longitude, with width of each data point relating intensity.

In both I've plotted negative longitude above the axis, positive below. This relates to the Americas being above the axis, Greenwich (UK) at zero, Europe and Eurasia to Asia below the axis, which seems to me to be the most intuitive way of presenting. Limitations of Excel prevented me from rotating those to up-down.

What's odd to me is the clear decrease in intensity at the same time as an increase in duration after 2003. These changes are apparent from earlier graphs but they seem to be concentrated across Europe into Eurasia, the 2003 peak is clear, but the general increase in duration and decrease in intensity is also clearly not a transitory phenomenon around 2003 but continues to the present, mainly affecting Europe, but also apparent in the American sector. Also, what happened to intensity in 2001 and 2002? This seems to be reflected in figure 4 as well.

Blocking intensity itself is strongly seasonal, being greatest in the winter (when polar to mid latitude temperature gradient is strongest) so I've calculated the monthly average for the period 1951-1980 (GISS baseline) and thence calculated anomaly from the average. Here is a time series of seasonal anomalies of intensity.

Fig 10, Timeseries of seasonal average intensity anomalies (baseline 1951 to 1980).

Blocking intensity has decreased in all seasons after 2000, with a tendency towards small generally positive anomalies.

***

So what is causing the post 2002 change in blocking duration and intensity. My first thought was the work of Francis & Vavrus, "Evidence linking Arctic amplification to extreme weather in mid latitudes" 2012. They identified changes in the Jetstream due to the differential warming of mid latitudes and the Arctic, changes that lead to increased latitudinal amplitude of Rossby waves on the Jetstream, associated with slower moving weather patterns. This could give rise to increased duration of blocking (which is small but plausible - fig 5), increased number of blocking events (strong - fig 6) as the conditions giving rise to one event persist giving rise to further events, and possibly a reduction in blocking intensity (strong - fig 8) due to gradients being lessened with larger Rossby waves. Notably figs 7 & 8 do not suggest (to me) that any strong change in wavenumber or phase of Rossby waves is a cause of the post 2002 changes, by which I mean the longitudinal distribution remains largely the same before and after 2002. However that conclusion could be wrong with blocking events in the given regions being tied to geography, notably continental landmasses adjoining the Atlantic.

I've used NCEP/NCAR to generate monthly time series of average temperature over the following bands:

Arctic = 90degN to 65degN
Mid Latitude = 65degN to 25degN
Northern Hemisphere = 90degN to 25degN
Difference in Temperature (DIT) = Mid Latitude minus Arctic

Fig 11, Seasonal averages of temperature difference between mid latitudes and the Arctic (DIT).

The downward deviation into this century is due to the accelerated warming of the Arctic, the lack of such a substantial downward deviation in JJA is due to the melting of sea ice, energy that would go into sensible warming in the Arctic goes into melting ice. Comparison of figs 4 & 6 with fig 11 seems to suggest that what is going on is not solely related to the DIT change. However it is worth looking at scatter plots of temperature difference and blocking days, per season, and per year.

Fig 12a.

Fig 12b. - error, second period should read 2003 to 2012.

Fig 12c.

Fig 12d.

Fig 12e.

Figs 12a to e, Scatterplots of temperature difference (DIT) against number of blocking days.

Correlation does not imply causation without a viable mechanism, in which case it can be taken as supportive together with other evidence. The differences between the post 2003 period and the period before reflects the massive Arctic warming in tandem with the increase in blocking events. So is there any reason to suspect temperature may have a role? As mentioned above figure 10, the seasonal cycle shows the greatest number of blocking events in winter, when the temperature difference is greatest, however what I've been looking at is a decrease in the DIT and suspecting a link to an increase in blocking. That noted, what I've seen of the literature suggests a role for Northern Hemisphere temperature in blocking days, for example this white paper from Mokhov et al. They reference earlier work, Lupo 1997, which shows using a model that warming leads to an increase in events which may be weaker and more persistent.

Comparing a control experiment (CO2 330ppm) and a doubled CO2 experiment, Lupo et al 1997 finds that:

...in general, blocking anticyclones were more persistent but weaker than their counterparts in the control experiment. All other characteristics of the overall sample such as; frequency of occurrence, size, preferred genesis locations, and annual variations in size and intensity, were generally similar to the control experiment.

The major difference between this statement and what is seen post 2000 is that the number of events per month (frequency of occurrence) has gone up substantially, as seen in fig 6.

To see whether the temperature and the total annual blocking days are possibly related requires that the trend be removed. Without doing this any correlation will be stymied by concurrent trends; during this century both temperatures and blocking days have increased together. Using annual data I've removed the trend by taking inter-annual differences, both in temperature and total number of blocking days. Then I've calculated correlations over different periods between annual blocking days and the four different temperature indices.

Fig 13, Correlations for the four different temperature indices against annual day count of blocking. Emboldened and italicised are statistically significant at the 95% level, ref.

For northern hemisphere temperatures (NH) in fig 13 it is only the post 2000 period that is statistically significant. The Arctic is statistically significant for the whole period from 1970, and for the post 2000 period. However I view it as meaningful that the temperature difference between Arctic and mid latitudes (DIT) is statistically significant for all three periods. Note that this is a negative correlation which fits with the relationship seen in the source data (not de-trended) that as the DIT reduces blocking increases.

At present I still think it very plausible that Arctic warming is playing a role, and the way it could play a role seems to me to be via the change in temperature gradient between mid latitudes and the pole, the gradient that drives atmospheric circulation. It is interesting that there seems to be no connection with the new Greenland centred summer circulation pattern (Summer Arctic Dipole). But Zhang's found that the Arctic Dipole, termed the Arctic Rapid Change Pattern in that work, arose from around 2003, ref; is there a connection? And what role does Cohen's finding of a lack of winter boreal warming due to low AO index play? Figure 4 is interesting in this respect with a reduction back to normal levels of blocking in Summer and Autumn, but maintained high levels in Winter and Spring.

I really need to do some reading and think about this matter.



Thanks to Steve Bloom for putting me onto this data, and to the originators of the data, Dr Anthony R Lupo & Colleagues, for producing the dataset and making it publicly available.

Note 1, I've also looked at changes in the difference between GPH for mid latitudes and Arctic. There are no substantial changes in GPH difference (calculated as anomaly due to large seasonal cycle, and consequently no significant behaviour for scatter plots, DJF, MAM, JJA, SON, Annual.

Note 2, Some cursory checking against ENSO, AO, PNA, AMO, did not suggest any of these play a role.