Frequency, persistence and intensity of high and low AAI events
Seasonal means of daily AAI values for the Arctic, land and sea are shown in Fig. 2, while the equivalent for the Arctic sectors is Figure S1. All regions and sectors show an overall positive trend of increasing AAI values, although in some instances these are relatively slight and not statistically significant (e.g. the Beaufort–Chukchi seas, Figure S1). Trends for the Arctic, land and sea regions are significant. The daily land AAI has a strong, significant correlation with the Arctic AAI (0.93), but correlates more weakly, but still significantly, with the sea AAI (0.39). The sea AAI correlation with the Arctic AAI is significant but moderate (0.70, DJF). Figure 3 shows that winter 11-year moving window seasonal trends are significant for years centred on 2001–2004 for the Arctic, Barents-Kara and East Siberian-Laptev sectors and the land and sea regions, and significant trends in the Beaufort-Chukchi Seas extend from 2002 to 2006 (p < 0.1); however, the significant trends arise earlier for the Canadian Archipelago (1994–1999). Interestingly there are few significant trends in the period after this.
Frequency and persistence of AA events
Figure 4 shows the frequency and persistence of high and low AAI events for winter for the Arctic, land and sea regions, identified using Markov chain analysis. Figure S2 shows the same for the Arctic sectors. There are significant decreases in the frequency of low AAI events for the Arctic, land and sea regions and significant decreases in low AAI event persistence in the Arctic as a whole and over the sea (Fig. 4a, b). The Arctic, land and sea regions all have significant increases in frequency of high AAI events, while significant increases in persistence of high AAI events are found for the Arctic and over the sea (Fig. 4c, d). In the Arctic sectors, significant negative trends in frequency (persistence) of low AAI events occur for the Barents-Kara and Greenland (East Siberian-Laptev) Seas (Figure S2a,b). Positive significant trends of high AAI events occur only for frequency (Barents-Kara, East Siberian-Laptev, Greenland; Figure S2c). The Beaufort-Chukchi seas and Canadian Archipelago present an interesting contrast. There are no significant trends in either persistence or frequency of high and low AAI events. However, interannual variability in frequency of low AAI events is much higher than for the other sectors, while interannual variability in persistence of low AAI events is very low (Figure S2). Low AAI events for these regions are characterised by a high frequency but short persistence, events being often restricted to single days. The pattern for high AAI events is similar to those from other sectors.
Intensity of AA events
AAI event intensity is presented in Fig. 5 for the Arctic, land and sea regions, and in Figure S3 for the other sectors. There are significant positive trends over land for both low and high AAI events, that is both tails of the distribution are showing a significant shift towards more positive values, so extreme cold events are becoming less intense while extreme warm events become more intense. The only other significant trend is a positive trend of high AAI values over the sea. Here the high AAI events become distinct and consistently higher than those for the Arctic and land regions around 2000 (Fig. 5a). Since 2012, there has been a decrease in both high and low AAI events, which is most marked over land. The intensity time series for the Arctic sectors show significant trends for GRE for both high and low events (Figure S3). ESL shows positive trends for both high and low AA events, while there are further positive trends for low events in BK and BC (Figure S3a).
Trends in persistence, frequency and intensity
All time series of frequency, persistence and intensity are very noisy, showing greater interannual variability than the magnitude of the overall trend, and a linear trend does not always appear to be the best fit. For example, in the Arctic as a whole, the frequency of high AAI events appears to have little trend prior to 2000, then shows a marked increase to around 2010 before flattening off again (Fig. 4c). To further analyse trends of persistence and frequency and intensity, 11-year moving window trends were calculated and are shown in Fig. 6. Significant trends for low (high) AAI events mostly show decreases (increases) in frequency, persistence and intensity. The most significant winter trends for the whole Arctic, land, sea, Barents-Kara and East Siberian-Laptev Seas are concentrated between 2001 and 2009, while other Arctic sectors show very few or no significant trends. There are some outliers for the Arctic, Barents-Kara and East Siberian-Laptev Seas in the 1980s and early 1990s, including a number of 11-year trends of opposite sign to those which would be expected from the overall time series. For example, in the 1980s the Arctic shows significant increases (decreases) in the persistence of low (high) AAI events. For the Arctic, land and sea, significant decreases in persistence, frequency and intensity of low AAI events occur earlier than the significant increases in high AAI events (1998–2004 cf. 2000–2008, Fig. 6), while for the Arctic sectors, they are more synchronous (Figure S4).
Spatial variability of Z500 and 2mT during high and low AAI events
Composite plots of Z500 and 2mT for the 100 highest AAI days minus the 100 lowest AAI days for the 40 years are presented for the Arctic, land and sea regions (Fig. 7) and for the sectoral AAIs (Figure S5). These show some notable differences between regions in the locations of temperature and Z500 maxima and minima. For all regions and sectors the composite differences in Z500 are co-located with areas of significant high (low) 2mT differences. Where the two sets of anomalies agree one can assume that Arctic surface heat anomalies are probably more strongly related to and driven by dynamical (circulation) changes, while near-surface temperature anomalies that are unrelated to thickness anomalies may have resulted from more localised/shallower heat sources such as sea-ice removal increasing turbulent heat flux from the ocean. Here comments concern high AAI days: the converse is true for low AAI days.
For the Arctic, significant areas of increased Z500 heights occur over Northern Siberia and centred on the north coast of Alaska, with a region of significant height increase over the central North Atlantic Ocean (Fig. 7a), while longitudinally extensive regions of significantly lower heights extend around the mid-latitudes. Significant positive temperature anomalies occur over the central Arctic and the northern coasts of Siberia and North America, with cold anomalies stretching east-west over a broad swath of central Eurasia and extending southeast from southwest Canada across much of the central USA (Fig. 7d). This is the well-established Warm Arctic-Cold Continents pattern (Overland et al. 2011; Chen et al. 2018) and generally corresponds with mid-tropospheric height anomalies (Fig. 7a). The land-based AAI events show a very similar pattern to those of the Arctic as a whole (Fig. 7b, e), indicating that the more extreme Arctic AAI days, both high and low, occur over land (54% shared high days, 80% shared low days), rather than over the sea (32% shared high and low days). The sea-based AAI shows a monopole of high Z500 anomalies centred over the Kara Sea, but extending eastwards to the Bering Strait (Fig. 7c). The 2mT anomaly is centred slightly to the North of the Z500 centre, closer to Svalbard (Fig. 7f). Low heights over the UK and western Scandinavia coupled with high heights over north Siberia, and a similar pattern of low heights south of Bering Strait and increased heights over Alaska, indicate the transfer of warm southerly air masses right up into the central Arctic (Fig. 7a–c). Temperature and GPH differences over Greenland, although significant, are more modest than those found over Siberia and Alaska in winter.
Figure S5 show patterns for the locations of Z500 and 2mT anomalies for the Arctic sector AAIs. Z500 anomalies occur at largely within the sector concerned, as do 2mT anomalies although these are often located where warmer air flow anticyclonically around the Z500 anomaly, indicating higher surface pressure. For the Canadian Archipelago, 2mT anomalies, while significant, are weaker than for other sectors, which, due to the location of the sector away from the Arctic gateways, means more limited advection of warm air.
Energy fluxes for high AAI events
Figure 8 shows lead-lag composites for high AAI events, illustrating changes in energy fluxes before and after the maximum AAI day. THF are consistently positive (downward) for the East Siberian-Laptev and Beaufort-Chukchi sectors (Fig. 8a), peaking two days before the AAI maximum (AAmax). THF for other regions are negative (upwards), the upwards flux being greatest at maximum leads and lags, when the vertical temperature and moisture gradients between surface and atmosphere are greatest. This upwards flux diminishes at AAmax as the vertical temperature gradient is reduced. Further reductions in upwards flux will arise as the air overhead contains more moisture, reducing the upward latent heat flux (not shown). The distinction here between positive and negative THF regions is between areas with high sea ice concentration (SIC), which reduces or prevents upwards turbulent heat fluxes from the ocean surface (East Siberian-Laptev and Beaufort-Chukchi sectors), and those with a higher proportion of exposed ocean, where upwards THF are less restricted in larger areas of the sector.
DLR is much greater for the Greenland and Barents-Kara Seas (Fig. 8b), associated with greater quantities of water vapour in the atmosphere and the passage of extratropical cyclones into the regions from lower latitudes, although the amplitude of the increase at AAmax is less than that in the Beaufort-Chukchi and East Siberian-Laptev Seas. As with THF, the minimum amplitude is for changes in the Canadian Archipelago. This is likely related to limited cyclone activity. Figure S5g shows that for AAI high events in this sector, the usual route of storms into the Arctic via Baffin Bay is blocked by anticyclonic anomalies.
NWVT is greatest over the Greenland Sea, this sector showing the greatest magnitude of increase at AAmax (Fig. 8c). The Beaufort-Chukchi Seas also shows a marked increase peaking just prior to AAmax. These two regions are in close proximity to the oceanic gateways into the Arctic (Greenland-Norwegian Seas and Bering Strait respectively), which will provide access for moisture transport into the Arctic by cyclones. In the East Siberia-Laptev seas, NWVT is actually slightly negative (southwards), from − 5 to + 5 days.
NHT is highest for the Greenland Sea (Fig. 8d), showing large increases peaking at AAmax, likely associated with Greenland blocking and synoptic-scale storms. For other regions, a more complex picture emerges. For the East Siberian-Laptev Seas, NHT declines to a minimum at AAmax, while for the Canadian Archipelago and the Barents-Kara Seas it is consistently negative. In the Beaufort-Chukchi Seas there is peak in NHT at d-1, after which values become negative.
These regional differences in energy and moisture flux are reflected in SIC responses at AAmax. The Beaufort-Chukchi and East Siberian-Laptev Seas demonstrate a similar SIC response; initial growth at longer lead times, reflecting the seasonal cycle of ice growth, followed by a sharp decline in SIC from d-2, with a sea-ice minimum at AAmax as a response primarily to increased DLR (Fig. 8b), followed by a steady recovery in SIC (Fig. 8f). The Greenland and Barents-Kara Seas on the other hand show a more gradual but extended SIC decline from d-6, and the minimum is not reached until 2 days after AAmax (Fig. 8e). Here, while DLR plays an important role in SIC reduction, contributions from NHT and NWVT for the Greenland Sea are particularly important. SIC in the Canadian Archipelago, on the other hand, as with the energy and moisture flux terms, behaves completely differently. Over the whole period there is a steady increase in SIC reflecting the growth phase of the annual cycle. However, from d-3 to d + 1, this growth is halted, when AA max occurs, although actual reduction in SIC is minimal. After this, the growth cycle resumes. This is the one region where moisture intrusion is lower; with the exception of along the east coast of Greenland, net moisture transport over the archipelago is often southwards (Woods et al. 2013) and the peak in DLR is smaller than for all other regions. 2 mT temperature anomalies are modest compared with other regions (Figure S5g), resulting in a cessation of SIC increase rather than the reduction seen elsewhere.
Sources of AA in winter
From the above, differences between Arctic sectors are apparent, which may indicate different sources of forcing for high AAI events. Examining Z500 anomalies and θ on 2PVU can help to explain these differences (Fig. 9).
For high AAI events in the Barents-Kara sector, over lead times from d-6 to d0 a positive Z500 anomaly develops over the Kara Sea coast of Siberia, on the eastern side of the storm track into the Arctic, at the same time as the development of anticyclonic Rossby wave-breaking (AWB) over northwest Siberia and Scandinavia (Fig. 9a). East Siberian-Laptev Sea high AAI events show positive Z500 anomalies along the northern coast of Siberia, which coalesce over the Taymyr Peninsula at d0. Z500 anomalies reveal evidence of a wavetrain over the North Pacific and North America. AWB takes place further south and west, over western Europe, but there is also a suggestion of cyclonic wave breaking (CWB) developing over the eastern Pacific (Fig. 9b). For the Beaufort-Chukchi seas, the development of increased Z500 anomalies over Alaska is accompanied by a wavetrain with low heights over the western US and Canada and increased heights over the eastern US. This is accompanied by CWB along the western seaboard of North America (Fig. 9c). For the Canadian Archipelago, positive Z500 anomalies grow over Baffin Bay and eastern Greenland from d-6 to d0 accompanied by clear CWB over the north Atlantic and up into Baffin Bay (Fig. 9d). Finally, for high AAI events in the Greenland Sea, positive Z500 anomalies develop over the Barents Sea and shift westwards, over the Norwegian Seas by d0, with AWB over the North Atlantic (Fig. 9e).