Greenland Blocking Index 1851–2015: a regional climate change signal

We present an extended monthly and seasonal Greenland Blocking Index (GBI) from January 1851 to December 2015, which more than doubles the length of the existing published GBI series. We achieve this by homogenizing the Twentieth Century Reanalysis version 2c‐based GBI and splicing it with the NCEP/NCAR Reanalysis‐based GBI. For the whole time period, there are significant decreases in GBI in autumn, October and November, and no significant monthly, seasonal or annual increases. More recently, since 1981 there are significant GBI increases in all seasons and annually, with the strongest monthly increases in July and August. A recent clustering of high GBI values is evident in summer, when 7 of the top 11 values in the last 165 years – including the two latest years 2014 and 2015 – occurred since 2007. Also, 2010 is the highest GBI year in the annual, spring, winter and December series but 2011 is the record low GBI value in the spring and April series. Moreover, since 1851 there have been significant increases in GBI variability in May and especially December. December has also shown a significant clustering of extreme high and low GBI values since 2001, mirroring a similar, recently identified phenomenon in the December North Atlantic Oscillation index, suggesting a related driving mechanism. We discuss changes in hemispheric circulation that are associated with high compared with low GBI conditions. Our GBI time series should be useful for climatologists and other scientists interested in aspects and impacts of Arctic variability and change.

H a n n a , E d w a r d , C r o p p er, Tho m a s E., H a ll, Ric h a r d J. a n d C a p p el e n, Joh n 2 0 1 6. G r e e nl a n d Blocki n g In d ex 1 8 5 1-2 0 1 5: a r e gio n al cli m a t e c h a n g e si g n al.
In t e r n a tio n al Jou r n al of Cli m a t olo gy 3 6 (1 5) , p p . 4 8 4 7-4 8 6 1. 1 0. 1 0 0 2/joc. 4 6 7 3 file P u blis h e r s p a g e : h t t p:// dx. doi.o r g/ 1 0. 1 0 0 2/joc. 4 6 7 3 < h t t p:// dx. doi.o r g/ 1 0. 1 0 0 2/joc. 4 6 7 3 > Pl e a s e n o t e: C h a n g e s m a d e a s a r e s ul t of p u blis hi n g p r o c e s s e s s u c h a s c o py-e di ti n g, fo r m a t ti n g a n d Thi s v e r sio n is b ei n g m a d e a v ail a bl e in a c c o r d a n c e wit h p u blis h e r p olici e s. S e e h t t p://o r c a . cf. a c. u k/ p olici e s. h t ml fo r u s a g e p olici e s. Co py ri g h t a n d m o r al ri g h t s fo r p u blic a tio n s m a d e a v ail a bl e in ORCA a r e r e t ai n e d by t h e c o py ri g h t h ol d e r s .

Introduction
High-pressure blocking over Greenland has been of great interest to weather forecasters and climatologists for many years. This blocking has traditionally been discussed in quite descriptive terms but more recently a Greenland Blocking Index (GBI) has been defined by Fang (2004) and popularized by Hanna et al. (2013Hanna et al. ( , 2014Hanna et al. ( , 2015, where GBI is the mean 500 hPa geopotential height for the 60-80 ∘ N, 20-80 ∘ Wr e g i o n ( Figure 1). A basic GBI dataset has recently been made available on the US National Oceanographic and Atmospheric Administration (NOAA)'s Earth System Research Laboratory Physical Sciences Division at: http://www.esrl.noaa.gov/psd/data/timeseries/daily/GBI/ but, in common with the above-referenced work, it only extends back to 1948. The GBI represents, and clearly depicts changes in, blocking across the entire Greenland region. A comparison with the blocking index of Davini et al. (2012), which is based on a considerably smaller subset of the GBI region (62.5-72.5 ∘ N, 20-70 ∘ W), shows very strong agreement with the GBI: the two series are correlated at ≥0.98 based on monthly data from NCEP/NCAR Reanalysis for 1948 to near present. Similarly, the Absolute Geopotential Height blocking measure of Scherrer et al. (2006) shows a preferential blocking region centred over Greenland, which is effectively captured through the GBI. Instantaneous anticyclonic blocking over both Greenland and Siberia has also been observed by Davini et al. (2012Davini et al. ( , 2013, who related it to Northern Hemisphere jet-stream flow.  discuss changes in both of these regional blocking features, which appear to be related to a more amplified northern polar jet stream during the last 10-20 years of Arctic Amplification (Francis and Vavrus, 2015).
Here we extend the GBI back to 1851 using the newly available Twentieth Century Reanalysis version 2c (20CRv2c; Compo et al., 2011Compo et al., , 2015 and are thus able to place previously published studies on observed GBI changes in the last 67 years in a much longer-term (165-year) climatic context. A great advantage of 20CRv2c data is the provision of data back to the mid-19th century, purely reliant on surface measurements, which avoids homogeneity issues in the satellite records caused for example by changes in satellite technology. However, recent work Villarini, 2012, 2014;Wang et al., 2013) has highlighted homogeneity issues in the 20CR data, some of which are likely to be non-climatic in origin, that we attempt to remove through homogeneity testing.   are depicted by a circle surrounding the location points.

Methods and datasets
Previously published GBI studies were based on the NCEP/NCAR Reanalysis (which spans from 1948 to near present; Kalnay et al., 2010), so here we used this dataset as our GBI 'baseline' and use 20CRv2c data to extend it backwards in time. This provided a continuous monthly GBI record from January 1851 to December 2015. It was not possible to simply join the 20CR and NCEP-NCAR-based GBI series together as the mean and standard deviation of the two series were not consistent (Table 1). Therefore, for each calendar month, we merged the two GBI time series in a two-stage procedure. The first step was to homogenize 20CRv2c-based GBI by applying the Bai-Perron structural change point test (Bai and Peron, 2003). This test is able to detect multiple break points, which could arise either through climatic shifts or artificial shifts as an artefact of data assimilation, through changes in the spatial density and number of observations, or through a change in the data-collection method. A breakpoint is determined as non-climatic if it coincides with an estimated inhomogeneity in the time series of the 56-member ensemble standard deviation spread from 20CRv2c, which can act as metadata for the identification of inhomogeneities Villarini, 2012, 2014). Coincidence of breakpoints was established if there was any overlap of 95% confidence intervals of the breakpoints in GBI and ensemble spread data for the relevant field. Such breakpoints were corrected for by adding the mean difference between the parts of the time series after and before the breakpoint onto the early part of the series before the breakpoint, and this was done for all necessary breakpoints (there can be more than one breakpoint in a given monthly series). In practice, only 5 months needed correcting: April (1942 breakpoint), May (1942), June (1916), July (1876and 1917 and August (1874 and1918). The results were checked using the Pettitt test, Standard Normal homogeneity test and Buishand range test. While these tests only identify a single breakpoint they can also be applied to subsections of the data to identify further breaks. The corrected time series were tested again for homogeneity and found to be homogenous. We also used all available long-term reference mean sea-level pressure (MSLP) time series from Greenland coastal synoptic stations (Cappelen, 2015; Figure 1) to help interpret the GBI homogeneity test results. This is a reasonable procedure because SLP from in/around Greenland can be thought to be partly reflective of the 500 hPa geopotential height (GBI) conditions. Second, for each monthly GBI time series, we splice 20CRv2c GBI with NCEP/NCAR Reanalysis GBI by multiplying the former with a coefficient so that the variances match for the overlap period; the value of the coefficient was iteratively determined. Next, we subtracted the mean of the 20CRv2c values rescaled in the previous step and added the mean of the NCEP/NCAR GBI values,  (Table 1). Following homogenization and splicing of the combined 20CRv2c and NCEP/NCAR Reanalysis GBI 1851-2015 monthly time series, we calculated seasonal GBI series for the standard 3-month meteorological seasons (MAM = spring, JJA = summer, SON = autumn and DJF = winter, where winter refers to the year of the January); seasonal values are based on means of the GBI monthly anomalies normalized to 1951-2000.
We used standard descriptive statistics and correlation analysis to analyse GBI monthly and seasonal variability and trends over the full period and various sub-periods (some of these being standard climatological normal time periods, as well as longer and more recent sub-periods). Correlation is based on linearly detrended data. We use a 5-year running standard deviation metric to analyse temporal changes in GBI variance. Statistical significance is tested using a t-test applied to temporal correlation analysis (after Abramowitz and Stegun, 1965), using the standard significance threshold of p ≤ 0.05. Standard wavelet analysis was also used as a supporting analysis tool on raw and filtered (7-point binomial) data (Grinsted et al., 2004;Ve l e d a et al., 2012). We also analysed extreme high and low GBI values for each month, to establish whether or not there were any unusual changes in the occurrence of extreme GBI years, with a particular focus on the most recent 15 years of record during the period of strong Arctic Amplification. We used the hypergeometric distribution test to establish the statistical significance of runs of extreme GBI values for particular years and seasons, following a similar procedure by Hanna et al. (2015). This was done based on two time periods 2001-2015 and 2007-2015 during the present century and recent period of strong Arctic Amplification and late summer Arctic sea-ice decline (Cohen et al., 2014;Francis and Vavrus, 2015).
We evaluate the strength and changes of correlation between monthly/seasonal GBI and (1) North Atlantic Oscillation (NAO) indices and (2) an updated Composite Greenland Temperature 3 (monthly and seasonal means of surface air temperatures from five weather stations in coastal western and southern Greenland; Hanna et al., 2012Hanna et al., , 2014, and discuss whether these comparisons yield possible insights into (1) the connection of Greenland blocking with wider-scale changes in atmospheric circulation over the North Atlantic, extending down to mid-latitudes, and (2) local-regional scale temperature changes around Greenland. For defining NAO we use a newly developed and extended 1850-2015 Azores-Iceland NAO index , both because this encompasses our GBI timespan and also as it uses the traditional/preferred southern station node (Azores) whose pressure record is most strongly anti-correlated with that of Iceland (Jones et al., 2003) and is relatively thermally stable throughout the year; alternative southern stations on mainland Iberia are rather off a south-north Atlantic axis and suffer from seasonal continental heating/cooling effects. We also use the Hurrell principal component (PC)-based NAO index because, although it only extends back to 1899, it is more widely representative of MSLP changes across the North Atlantic Hurrell, 2015). Finally, for the three standard summer months (June-August) we employ the Summer NAO (SNAO) index (Folland et al., 2009;Bladé et al., 2012) which is defined as the first empirical orthogonal function (EOF) of the observed summertime extratropical MSLP over the 40-70 ∘ N, 90 ∘ W-30 ∘ E region, and has been demonstrated to better capture NAO patterns and variations and their effects during summer in Northwest Europe and the UK. SNAO is derived from NCEP/NCAR reanalysis and downloaded from climexp.knmi.nl. The Hurrell PC NAO is similarly based on the PC time series of the leading EOF of monthly MSLP but over the much larger region of 20-80 ∘ N, 90 ∘ W-40 ∘ E (Hurrell, 2015). The final substantive results section uses composite analysis of high minus low GBI months and seasons to unravel the relationship between physical climate fields at the regional to hemispheric level. A concluding discussion section places our results in a broader-scale contemporary climate-change context.  Table 2 show seasonal and monthly GBI variability since 1851. Multidecadal peaks and troughs are evident in all four seasonal GBI series ( Figure 2). Wavelet plots of GBI series for all 12 calendar months do not show many common sustained features in the raw GBI data ( Figure S1). Perhaps the most noteworthy feature is an apparently significant approximately 7-to 8-year period which shows up since about the year 2000 in the December GBI wavelet transform, and also appears in the December NAO wavelet transform ; Figure S2). However, once the GBI data have been filtered, periods of around 8 years become apparent in most months ( Figure 3): although filtering time series naturally emphasizes longer-term periods, this dominant feature is robust to whether a 5-, 7-or 11-point binomial filter is used (not shown) and also appears in the filtered NAO data wavelets ( Figure S3). Cross-wavelet transform plots show a tendency for the NAO to lead GBI, which is especially evident for winter months in the last few decades ( Figure S4).

GBI variability and trends
All four seasonal GBI plots show: a peak in the late 19th century around 1880 (although the exact timing of this varies by season), with GBI having risen from a relatively low base in summer and autumn; a trough around 1910 or 1920; another peak or increasing trend around 1930-1940; a generally decreasing -or for autumn rather flat -trend until the 1980s or very early 1990s; and an increase since then ( Figure 2). The recent increase is most marked in summer and apparent for winter but displays a mixed signal for autumn and spring -with high values present throughout the first decade of the 20th century but combined with an apparent short-term negative trend during the past decade. Trends in monthly and seasonal GBI for the full period (1851-2015) are all insignificant except for decreases in October, November and autumn (Table 2). Table 2 also splits the full period into sub-periods for more detailed analysis of monthly and seasonal GBI trends and variability. This shows several significant increases in GBI for 1851-1900 -in May and June and for the year as a whole -whereas there were no significant decreases. For 1901For -1950 are no significant changes. For Table 2. Monthly and seasonal GBI variability (expressed through standard deviation -sigma) and trends for standard climatological normal and other periods. Parameter 1851Parameter -2015Parameter 1851Parameter -1900Parameter 1901Parameter -1950Parameter 1951Parameter -2000Parameter 1981Parameter -2010Parameter 1991Parameter -2015Parameter 2001Parameter -2015  1951-2000, February, March and winter show significant decreases. We also focus on three more recent periods of more contemporary climate change. Over 1981-2010, all seasons, annual, March, July, August and October had significant GBI increases. For 1991-2015, these GBI increases were sustained for summer, July and August (the sustained summer increase over the last 25 years is clearly seen in Figure 2). There were no significant trends for the shorter recent 2001-2015 period except for an increase in July. Most monthly and all seasonal time GBI series do not show significant overall (1853-2013) trends in variability; the exceptions are May and December, both with significantly increasing variability (Table 3). Figure 4 shows changes in GBI variability for the seasons and December, with a striking upward trend clearly visible for the latter.

Changing frequency of extreme GBI values
We next analyse low and high GBI values for each month and season (Table 4). 2010 features as the highest December, winter, spring and annual values in the entire GBI series since 1851 (followed by the record low spring and April values in 2011). 2010 also appears in the highest 10 values in January, February, April, May, August and autumn. Some five of the ten highest summer GBI values occurred between 2007 and 2012, with 2015 and 2014 having the 10th-and 11th highest GBI values in the 165-year summer GBI series. Years since 2001 feature in both the lowest ten and the highest ten December GBI values (there are five extreme years since 2001), which is a symptom of the significantly increasing GBI variability in that month as discussed above.
To test the significance of this recent clustering of extreme GBI values, we use the hypergeometric test and focus on the following seasons and months in which more than two of the highest ten and/or lowest ten extreme values for that month or season occurred since 2000: specifically summer (JJA), June and December (Table 4). For summer, where six of the ten (and 7 of the 11) highest years in the entire record occurred since 2007, this is found to be highly significant in both samples (i.e. for both of the above time periods): this is especially so for 2007-2015 (p = 0.0001) but also for 2001-2015 (p = 0.00370). These  1900 1920 1940 1960 1980 2000 1860 1880 1900 1920 1940 1960 1980 2000 1860 1880 1900 1920 1940 1960 1980 2000 1860 1880 1900 1920 1940 1960 1980 2000 32 16 8 4 1860 1880 1900 1920 1940 1960 1980 2000 1860 1880 1900 1920 1940 1960 1980 2000 1860 1880 1900 1920 1940 1960 1980 2000    results for summer are supported by a similar analysis for June, for which four of the ten highest GBI values occurred since 2007, with the second highest June GBI in 2012 (Table 4). According to the hypergeometric test the chance probability of getting this clustering of extreme June GBI values during 2007-2015 is 0.0135, which we accept as significant at the standard p ≤ 0.05 level. The summer and June GBI series are normally distributed, which engenders high confidence in these results. A similar recent clustering of high summer GBI values is not seen in the August monthly series, and the occurrence since 2007 of three of the ten highest July GBI values is not quite statistically significant; however, all of the recent extreme high summer GBI years also have extreme high June and/or July GBI values, so this does appear to be predominantly an early-to mid-summer phenomenon. Turning our attention to December, we note that 5 (3) of the 20 most extreme (i.e. the ten highest and ten lowest) GBI years occurred since 2001 (2007). The hypergeometric test gives an associated probability of p = 0.0213 (0.08), which is significant (just insignificant) for the respective period. The December GBI values are not normally distributed but, because each value in the hypergeometric test is either true or false, the base distribution of the data is not expected to seriously affect the reliability of the results as long as the time series are not strongly autocorrelated (which could enhance clustering of extreme values). The GBI monthly series for June and December are hardly autocorrelated -respective correlation coefficients of the series correlated against themselves displaced by 1 year are 0.18 and −0.12 -and the summer GBI series is similarly only weakly autocorrelated (r = 0.24). Therefore, we judge that the GBI has had a significant number of extreme (both high and low) years in December since 2001, and that June and summer have had significant clusterings of high GBI years since 2007 and also since 2001 in the case of summer.

Relationship between GBI and other climatic indices
Tables 5-7 contain detrended correlation coefficients between GBI and respectively NAO and CGT3 for the same periods as in Table 2. The GBI-NAO correlations are mainly significantly negative. For the station-based NAO  for the full (1852-2014) time period, GBI-NAO correlations are strongest in January to March (all below −0.8) and weakest in July at −0.37 (Table 5). For the two most recent time periods since 1991, correlations for December are below −0.9 and as low as −0.96 for 2001-2014. The month with the least strong correlations (although still significant for the overall period) is July. The time variations in correlation coefficients reflect a mix of non-stationary relationships between the GBI and NAO indices, random fluctuations and/or variations related to data quality. Using the Hurrell PC-based NAO index (Hurrell, 2015), correlation coefficients are generally below −0.7 except for July and August in the earliest  period, when they are slightly higher (i.e. less strong; Table 6 (Table 7). There is no clear or consistent seasonal cycle in these correlations but they hold up well for the earlier periods (Table 7). Strong correlations have been found between winter temperatures across southwestern Greenland and high blocking activity in the Greenland/North Atlantic region by Rimbu and Lohmann (2011). The same authors also identified a typical pattern of an out-of-phase relationship between blocking frequency and the NAO -mirroring the strongly negative NAO-GBI relationship that we have found here. However, the authors also note the potential for the occurrence of synoptic conditions where North Atlantic blocking can also be associated with a positive NAO, although the centre of the block would typically be shifted more towards the UK than centred over Greenland.

GBI composite analysis
Northern Hemisphere plots of various meteorological parameters (geopotential height at 850, 700, 500 and 300 hPa), 850 hPa (near-surface) temperature, precipitation, 500 hPa wind, omega and sensible and latent heat fluxes) are examined for the ten highest minus the ten lowest GBI years for the main seasons, and two months June and December highlighted above as being of special interest. Only the plots for winter and summer are shown in the main text (Figures 5 and 6) but for completeness the other seasons' and June and December plots are presented in Figures S5-S8. The impact of GBI changes on geopotential height (GPH) is clearly greatest in winter (when the circulation is strongest and most variable), and the focus of maximum GPH difference is greatest just off of southeast Greenland, compared with southwest Greenland in the summer difference plots. In terms of broader hemispheric circulation features, high GBI is associated with moderate troughing over central Europe in winter but only weaker troughing displaced north and west to the UK and westwards regions in summer. Also with high winter GBI, low heights are focused in the north-east Pacific off the west Canadian coast.
Temperature anomalies under a high GBI regime are high over Greenland and the Labrador Sea, partly because greater GPH heights are thermodynamically linked to higher temperatures (e.g. Overland et al., 2015, supported by the strong positive correlations in our Table 7) but in winter this anomaly extends further west to encompass Hudson Bay and the surrounding regions. The accompanying negative winter temperature anomaly over Siberia, Scandinavia and the UK is equally striking. However, in summer the temperature high is much more restricted and in Greenland is only really high along the west coast and extending out over the Labrador Sea; there is also a pronounced warm anomaly around the north Russia coast and a cold anomaly over northern Canada, in contrast to winter.
Precipitation anomalies are mainly positive over Greenland under high GBI because milder, moister southerly winds are then advected over the Ice Sheet. This situation reverses for southern Greenland in summer, which has a distinct dry anomaly. This reflects the seasonal northward migration and general weakening of the northern polar jet stream between winter and summer, with negative precipitation anomalies (linked with the jet and a negative NAO) under a positive GBI migrating north towards Greenland. This seasonal change is accompanied by opposite precipitation anomalies over much of the UK (most evident and predominantly wet in summer but tending towards dry in winter) linked with high GBI. Hanna et al. (2008) and the UK Met Office (2012) document the exceptionally wet 2007 and 2012 UK summers, and Overland et al. (2012) discuss how the recent run of wet UK summers may be linked to recent jet stream anomalies, a high GBI and positive Arctic Dipole. However, it is clear from its geographical position that the UK is very sensitive to slight changes in GBI and associated precipitation anomalies, especially in winter.
The collective winter plots in particular show clear elements of the Scandinavian or Eurasia-1 and East Atlantic circulation patterns, for example with the southern temperature anomaly displaced eastwards to give a distinct southeast-northwest temperature dipole (Barnston and Livezey, 1987).
The 500 hPa wind anomaly plots show an easterly anomaly over much of the mid-North Atlantic that is again related to a weak jet stream under high GBI, and again this is most evident in winter. In summer, the predominant westerly circulation over the UK does not weaken under high GBI and actually an anomalous anticlockwise/cyclonic circulation is then apparent. Sensible Table 4. Lowest 10 and highest 10 GBI years for each month and season (based on January 1851-December 2015).

Month/Season
Month 1899-2014 2001-2014 1991-2014 1981-2010 1951-2000 1901-1950 Jan  Tables 5 and 6 are highlighted in bold. All coefficients are significant at p ≤ 0.05 except for those highlighted in italics. and latent heat flux (SH and LH) anomalies -which are plotted here as positive upwards from the surface -are negative south of Greenland and positive around the UK and Scandinavia (SH) and southwest of the UK (LH) in winter, with much weaker anomalies in summer. This reflects much lower (higher) sea-air winter heat release around Greenland (UK) under conditions of positive GBI and higher (lower) near-surface temperatures. Finally, omega anomalies are more consistent between winter and summer, being mainly positive in east and extreme southwest Greenland (indicating descending air motion), with negative omega (marking positive vorticity advection and ascending air motion, e.g. Holton, 1992) to be found further north in west Greenland.
To summarize this section, extreme GBI events are clearly linked both thermodynamically and dynamically, with pronounced meteorological anomalies via polar jet-stream changes both downstream and upstream of the Greenland block, and have major repercussions for weather conditions experienced at mid-latitudes. Thus, the GBI temporal changes noted here have a much wider Northern Hemisphere climatological significance well beyond the Greenland region.

Discussion/Summary
The new homogenized/extended GBI time series here should be of wide interest and use to climatological researchers. Our key results from analysing this new time series are that GBI has significantly increased in summer and become significantly more variable with an unusual run of recent extreme values (since 2001) in December. These are in line with previous findings , which are mainly related to the strongly associated NAO (Tables 5 and 6), but are here based on a much longer, homogenized GBI time series dating back to 1851. In this section, we explore possible reasons for these changes and what this tells us about current climate-change processes. Enhanced Greenland blocking may arise from Temperature anomaly (°C) Figure 5. Difference in various meteorological parameters for the mean of the ten highest GBI years minus the mean of the ten lowest GBI years for winter (DJF).  Figure 6. Difference in various meteorological parameters for the mean of the ten highest GBI years minus the mean of the ten lowest GBI years for summer (JJA). Table 7. Detrended correlation coefficients between monthly and seasonal GBI and Composite Greenland Temperature 3 (CGT3; Hanna et al., 2012Hanna et al., , 2014, for standard climatological normal and other periods. CGT3 is a mean of the five western and southern Greenland weather stations shown in Figure 1. 1895-2014 2001-2014 1991-2014 1981-2010 1951-2000 1901-1950  thermodynamic and dynamic influences, although in reality both work together and the balance and interplay of these effects is hard to decipher. However, Rajewicz and Marshall (2014) attribute around 38-49% of Greenland summer air temperature and melt extent variability to circulation anomalies, with lower tropospheric background warming explaining 13-27%. This background warming is easy to explain, as Greenland has been one of the fastest-warming regions of the Northern Hemisphere in the last 10-25 years (e.g. Hanna et al., 2012Hanna et al., , 2014Cohen et al., 2014;Overland et al., 2014). As Greenland warms, warm air expands and raises geopotential heights, which on average increases the intensity and frequency of high pressure: here we see this effect in summer/June, in line with previous observational results (e.g. Francis and Vavrus, 2012;Overland et al., 2012;Hanna et al., 2013Hanna et al., , 2014. At the same time, Arctic Amplification of global warming has recently been linked to a sometimes more wavy, meridional and slower-moving jet stream and more negative NAO (e.g. Cohen et al., 2014;Francis and Vavrus, 2015;. This slower-moving jet is associated with stronger blocking features over both Greenland and Siberia, although working out the cause and effect of such jet stream and blocking changes is far from straightforward and requires significant additional dynamical insights ; moreover, the apparent waviness is contentious and may be metric-dependent (e.g. Barnes, 2013;Screen and Simmonds, 2013). Multidecadal variations of the GBI are -especially in winter -partly related to similar, 10-to 20-year variations in NAO (not shown), and to longer-period variations in the Atlantic Multidecadal Oscillation (AMO). Hakkinen et al. (2011) found that winters with more frequent blocking between Greenland and western Europe are related to a warmer, saltier subpolar North Atlantic ocean and a positive phase of the AMO -for example during the 1960s, late 1970s and early 2000s -and these are times of high and/or increasing GBI (Figure 2(a)). A 20-year periodicity in blocking frequency over the Scandinavian region has also been linked to solar variability (Rimbu et al., 2014), although our GBI wavelet analysis shows no evidence of a sustained cycle of this length: this is perhaps unsurprising given the difference in regional foci. Although here we only do a formal statistical comparison between GBI and NAO variations, changes in GBI are clearly related to wider North Atlantic blocking and circulation changes, including the East Atlantic and Scandinavian patterns, as revealed through earlier work (Scherrer et al., 2006;Hakkinen et al., 2011;Rimbu et al., 2014).

Month/Season
The greater variability in December GBI and extreme GBI events in recent Decembers is linked to more positive GBI on average, in line with the negative trend in the NAO since 2000  and the more variable December NAO and Arctic Oscillation (AO) over the last few decades Overland and Wang, 2015). The causes of this are at present unclear, and may be related to Arctic sea ice loss, particularly in the Barents-Kara Sea region (e.g. Liptak and Strong, 2014;, or to influences from the tropics (Trenberth et al., 2014) or to decadal fluctuations in the North Pacific and Atlantic and atmospheric internal variability (Perlwitz et al., 2015). Sea-ice loss may lead to a destabilization of early winter jet stream arising from particularly large autumn sea-ice losses and increased ocean-atmosphere heat fluxes (the ice-ocean heat flux feedback) in recent years . This may have led to a greater likelihood in some years of extreme negative AO and NAO December values and concomitant high GBI values (since the GBI and AO/NAO act mainly in antiphase; Hanna et al., 2015) and more meridional/less zonal jet-stream flow. The tropical influence is from enhanced tropical convection exciting the poleward propagation of Rossby waves, which may result in more persistent climate anomalies in the extra-tropics (Trenberth et al., 2014). Ding et al. (2014) identify tropical forcing as a major contributor to recent observed warming over Greenland, via Rossby-wave propagation, associated with the recent negative trend in the winter NAO, and by implication the positive trend in GBI. A recent study suggests that tropical forcing accounted for the positive NAO trend at the end of the twentieth century, which has recently been overshadowed by the impacts of accelerating sea-ice loss (Feldstein and Lee, 2014). The combination of these possible forcing effects through Arctic Amplification, which is traditionally thought to be greatest in winter due to a lagged response to summer sea-ice losses (Deser et al., 2010;Screen and Simmonds, 2010) and limited ice-/snow-melt during the cold season, may have recently led to more frequent destabilization of the winter jet and polar vortex. However, as this does not happen every year, this may have the effect of making the winter atmospheric circulation in mid-high northern latitudes more variable on an interannual basis. This perhaps reflects interplay between the tropical, Arctic and mid-latitude influences outlined above, together with internal atmospheric variability. Future modelling and dynamical process studies should focus on defining the relative influences of the various climatic forcings and feedbacks discussed above on Greenland Blocking changes. However, whatever the causes of these recent changes, our GBI index effectively captures them and enables them to be placed in a longer-term climatic context.

Statement of data availability
The monthly and seasonal GBI series in this paper are available from http://www.sheffield.ac.uk/geography/ staff/hanna_edward/gbi The 20CRv2c data were accessed in January 2016. We plan to revisit and update our GBI online archive at least annually, if not more frequently.

Supporting Information
The following supporting information is available as part of the online article: Figure S1. GBI wavelet transforms for all calendar months, based on our new/extended (1851-2015) GBI time series and raw (unfiltered) GBI values.  Figure S3. NAO wavelet transforms for all calendar months, based on the Cropper et al. (2015) NAO time series and (7-point binomial-filtered) NAOI values. Figure S4. Cross-wavelet transforms for all calendar months, between our new/extended (1851-2015) GBI time series and the Cropper et al. (2015) NAO series, each of which was filtered with a 7-point binomial filter beforehand. Left-pointing arrows indicate the NAO timeseries leads the GBI and upward (downward) arrows indicate an (in) out of phase relationship. Figure S5. Difference in various meteorological parameters for the mean of the ten highest GBI years minus the mean of the ten lowest GBI years for spring (MAM). Figure S6. Difference in various meteorological parameters for the mean of the ten highest GBI years minus the mean of the ten lowest GBI years for autumn (SON). Figure S7. Difference in various meteorological parameters for the mean of the ten highest GBI years minus the mean of the ten lowest GBI years for the month of June. Figure S8. Difference in various meteorological parameters for the mean of the ten highest GBI years minus the mean of the ten lowest GBI years for the month of December.