On 2 August 2012 a dramatic storm formed over Siberia, moved into the Arctic, and died in the Canadian Arctic Archipelago on 14 August. During its lifetime its central pressure dropped to 966 hPa, leading it to be dubbed ‘The Great Arctic Cyclone of August 2012’. This cyclone occurred during a period when the sea ice extent was on the way to reaching a new satellite-era low, and its intense behavior was related to baroclinicity and a tropopause polar vortex. The pressure of the storm was the lowest of all Arctic August storms over our record starting in 1979, and the system was also the most extreme when a combination of key cyclone properties was considered. Even though, climatologically, summer is a ‘quiet’ time in the Arctic, when compared withall Arctic storms across the period it came in as the 13th most extreme storm, warranting the attribution of ‘Great’.
 On 2 August 2012 a storm formed over northern Siberia and subsequently made its way into the Arctic basin where it intensified and achieved a central pressure of 966 hPa on 6 August. The storm lasted for almost 13 days before eventually decaying in the Canadian Archipelago. This pressure minimum and cyclone longevity are very atypical of Arctic storms, particularly in August. The system attracted much scientific and media attention, and was dubbed ‘The Great Arctic Cyclone of August 2012’. Part of this interest was due to the fact that the Arctic sea ice extent had recently plummeted to new record lows, raising questions as to cause and effect and what dramatic sea ice reductions might mean for future Arctic storm activity. Here we undertake analyses to assess key potential physical mechanisms responsible for this August storm (hereafter referred to as ‘AS12’), including baroclinicity, surface energy fluxes associated with diminished sea ice coverage, and the influence of an upper-level tropopause polar vortex (TPV) which had developed some two weeks earlier. Finally, we investigate a broad range of aspects of AS12 to determine the extent to which it can be considered extreme, in the context of both August and annual storm behaviour.
2. Data and Methods
 Our analysis is undertaken with the ‘new’ generation National Centers for Environmental Prediction (NCEP) Climate Forecast System Reanalysis (CFSR). One of the significant advantages of CFSR for this work is that it includes atmosphere-ocean coupling and assimilates sea ice concentration data [Saha et al., 2010]. The atmospheric component is run at T382L64, and the dataset covers the period 1 January 1979–31 December 2010. Atmospheric data as well as surface turbulent heat fluxes and sea ice concentration (SIC) have been downloaded at 0.5° resolution at 6-hourly intervals. Beginning on 1 January 2011, CFSR was extended by the CFS Version 2 (CFSV2) operational model, a model identical to that used to create CFSR. We downloaded this data at 0.5° resolution starting from 1 January 2011, and at 1.0° from May 2011.
 Mean sea level pressure (P) and z300 cyclone analyses are undertaken with the scheme described in Simmonds and Keay  and Keable et al. . The algorithm has shown considerable skill in the early identification of cyclones and the temporal integrity of their tracks [Mesquita et al., 2009], and hence is ideally suited to this investigation where we track long cyclone trajectories. It is also very valuable for our present purpose in that it determines a range of influential morphological aspects of each cyclone it detects. The four key ones we use here are the central pressure, intensity (as measured by ∇2P in the vicinity of the center of the cyclone) and subsequently denoted by DsqP, radius (R) (taken as the weighted mean distance from the cyclone center to the points at which ∇2P is zero around the ‘edge’ of a cyclone), and depth (D) which is the difference between the mean pressure at the edge of the cyclone and that at the center. Synoptically, each of these characteristics reflect related but different aspects of cyclonic structure and influence. These metrics in totopresent a comprehensive and multi-faceted picture of cyclone structure, a crucial aspect of our investigation.
 Part of our diagnosis makes use of the Eady growth rate (EGR), a measure of baroclinicity and hence of the ability of the environment to foster cyclonic systems. It is defined as [Simmonds and Lim, 2009]
where Nis the Brunt-Väisälä frequency (where , g the acceleration due to gravity, z the vertical coordinate, and θ the potential temperature), f is the Coriolis parameter and u (z) is the vertical profile of the horizontal wind vector. In our investigation we calculate EGR at the 500 hPa level (EGR500). (The vertical derivatives in the EGRexpression were approximated with centered finite differences using vector wind and potential temperature data at the 400, 500, and 600 hPa levels). One of our aims is to diagnose conditions at the tropopause and in the upper troposphere leading up to, and during, the storm. A number of works have shown that analyses undertaken at the 500 hPa level are very insightful in diagnosing the evolution and interaction between dramatic features at the tropopause, in the upper-troposphere and at the surface [e.g.,Bosart et al., 1996]. As indicated above, there is reason to believe that a TPV influenced the evolution of AS12. TPVs are characterised, in part, by the downward intrusion of stratospheric air to about 500 hPa and beyond, and hence have the potential to influence surface development [Cavallo and Hakim, 2012]. See also in this context the comprehensive analysis of Kew et al. .
3. The August 2012 Storm
 The storm was first identified by the cyclone algorithm at 00UTC 2 August over Siberia near 59°N, 98°E (Figure 1). It then tracked east northeast, and crossed into the Arctic basin (defined here as the region north of 70°N) in the East Siberian Sea at 18UTC 4 August. AS12 subsequently made its way into the central Arctic and reached the minimum central pressure of 966.38 hPa at 18UTC 6 August. The contour plot in Figure 1 shows the Pdistribution at this time. Subsequent evolution of the storm saw its central pressure steadily increase as it travelled east and then south and was last identified at 18UTC 14 August near 76°N, 260°E in the Canadian Arctic Archipelago (CAA). The storm therefore lasted almost 13 days (51 6-hourly synoptic times), of which 10½ days (41 synoptic times) were spent in the Arctic basin.
 In Figure 2 (top) we display the evolution of the central pressure over the life of the storm. When first identified AS12 had a central pressure of 1001 hPa and, while at no stage can its reduction be regarded as ‘explosive’ [Lim and Simmonds, 2002], it was monotonic and continued for 4¾ days. The Figure reveals an acceleration of the deepening rate after the cyclone passed from the continent to over the East Siberian Sea.
 As discussed above, central pressure presents only a limited perspective of cyclone structure and significance. To provide a comprehensive perspective of the characteristics of AS12 we also present in Figure 2 (top) the storm's intensity, radius and depth over its lifetime. These variables exhibit fairly modest values when the cyclone is over the continent. However when it moves out over the East Siberian Sea the intensity increases from 1.7 hPa (deg. lat.)−2 to 4.9 hPa (deg. lat.)−2 in 30 hours, and peaks 12 hours before the minimum central pressure is reached. From then on the intensity undergoes an almost monotonic decrease. By contrast, the radius of the system exhibits only a modest increase during this period. Then, starting at 06UTC 7 August (12 hours after the pressure minimum) the radius increases markedly from 4.2 to 6.6 deg. lat. in half a day. Thereafter the size of AS12 slowly decreases, a structure similar to that exhibited by midlatitude storms in the NH summer [Simmonds, 2000; Rudeva and Gulev, 2007]. The time evolution of the cyclone Depth reflects characteristics of both of these parameters. Similar to intensity, the Depth reached a maximum shortly before the central pressure minimum. However, it achieves its absolute maximum of 18.2 hPa at 18UTC 7 August, a full day after the pressure minimum.
 To understand the structure and evolution of AS12 we diagnose the key aspects of the environment through which the storm progressed. The factors presented here are EGR500, the surface sensible (SH) and latent heat fluxes (LH), their sum (TH) (all measured positive in the upward direction), and SIC. At each synoptic time the mean value of these parameters are calculated over the area of the cyclone, and their time series are shown in Figure 2 (bottom). EGR500 shows sizeable and increasing values while the storm was over Siberia, indicating that baroclinicity was involved with its initiation and development. On the storm reaching the coast this parameter increased further, and maintained high values for the 36 hour period over which the central pressure decrease had accelerated. Subsequently the values of EGR500 decreased but still assumed significant magnitude. While over Siberia AS12 experienced quite large sensible heat fluxes in the early afternoon (local time) when the surface is warm, and modest values outside this time. It is noteworthy that after crossing into the Arctic basin the sensible and latent heat fluxes in the confines of the storm were negative during the period of greatest pressure decrease. Shortly after, the latent heat flux over the cyclone domain became positive (i.e., upward) with typical values of 20 Wm−2. Offsetting this was the fact that the SH remained negative for most of the cyclone's life, with small positive values being diagnosed late in its evolution over the CAA (where a diurnal cycle is apparent with, as above, peaks in the early afternoon (local time)). Accordingly, the total heat flux, while positive for most of the time, exhibited quite small values. This observation leads one to conclude that the turbulent heat fluxes played little or no role in the storm's progress, a perspective supported by the fact that the period of greatest deepening corresponded to a time of sensible and latent heat loss from the atmosphere.
 Finally, in Figure 2 (bottom) we show the time series of the SIC averaged over the domain of the cyclone (the initial small values and rapid rise in SIC reflects the fact that not all of the cyclone was over the Arctic Ocean at this stage). For much of the middle part of the cyclone's life it found itself over regions where the SIC ranged between 0.65 and 0.8. We note that SIC values fell into the lower part of this range between 00UTC on the 7 and 9 August, the period over which the cyclone achieved its greatest size and depth. We comment here that there is no indication of a change in heat fluxes or EGR500 during this period of low concentration. This absence of a signal should come as no surprise given that August is the month when the total surface energy flux is small, being in the process of changing from negative to positive [Serreze et al., 2007; Simmonds and Keay, 2009]. Hence it is a time when one expects the turbulent heat fluxes to be small and to have relatively little sensitivity to sea ice conditions, in marked contrast to strong associations in winter [e.g., Murray and Simmonds, 1995]. This leads to the view that it was the enhanced influence of the cyclone which contributed to the reduction in ice area, rather than low sea ice area being responsible for releasing energy to maintain the system. This perspective is strongly consistent with analyses we have conducted (not shown) that indicate AS12 caused the dispersion and separation of a significant amount of ice, while its removal left the main pack more exposed to wind and waves associated with AS12, facilitating the further decay of the main pack. (See also the studies of Overeem et al.  and Asplin et al.  on the influence of storms on the ice.)
4. An Upper Level Vortex
 In addition to a role played by baroclinicity, dynamics suggest that an upper-level feature was probably associated with this dramatic surface cyclone. It is well established that upper-level disturbances can play a major role in surface cyclogenesis and development [Holton, 2004]. The Arctic region is particularly favourable for the occurrence of long-lived TPVs, with relatively weak vertical and horizontal wind shear providing an environment conducive for such longevity [Cavallo and Hakim, 2012]. This persistence presents an enhanced potential for the interaction between upper-level and surface disturbances.
 To verify the presence and character of an associated TPV during the evolution of AS12 we applied the cyclone tracking scheme at the 300 hPa level. We identified the genesis of a TPV near Svalbard (near 12°E, 77°N) at 06UTC 17 July 2012. This feature spent time over the Greenland and Norwegian Seas before moving off to the east and skirting to the north of Europe and Siberia. The vortex found itself over Severnaya Zemlya (near 101°E, 79°N) with a central height of 8787 gpm at 00UTC 2 August, this being the first time the surface system was identified. The z300 pattern at this time is presented in Figure 3a, and the genesis point of AS12 is indicated by a star. The 300 hPa vortex, which at this stage has been in existence 16 days, extends a strong trough to the south over the area in which the surface system was born. Two days later the tropopause system has moved further east, and it can be seen that it has strongly influenced the translation of AS12 (Figure 3b). Figure 3cshows that after a further 48 hours (00UTC 6 August) AS12 is located very close to the upper-level system; the two systems display even greater vertical organisation shortly after this time, and that relation persists right through to the lysis of AS12 over the CAA. We are at present directly quantifying the influence of the TPV and lower-level processes by considering the vertical profiles of vorticity at the center of AS12. This is being done with powerful techniques which have already been developed [e.g.,Lim and Simmonds, 2007; Čampa and Wernli, 2012]. However, at present one can say that the TPV clearly played a strong role in the evolution of the storm.
5. Just How Unusual Was This Storm?
 Finally we use our range of cyclone attributes to quantify the extent to which AS12 can be regarded an extreme cyclone. To place its very low central pressure in context we show in Figure 4a the distribution of the central pressures (using 5 hPa bins) of all 1618 August Arctic cyclones in our database (1979–2012). (To qualify for inclusion a cyclone track must spend at least one synoptic time in the Arctic basin. The central pressure ascribed to each storm is the lowest of all the synoptic times when the track was actually in the Arctic.) The plot shows that AS12 was at the tail of the distribution and, at 966.38 hPa, was the lowest in our record, beating the previous deepest (966.94 hPa) (for a storm at 06UTC 7 August 1995) by 0.56 hPa. The next lowest central pressure, 969.23 hPa, was associated with a cyclone at 06UTC 22 August 1991, followed by the fourth lowest storm central pressure in the earlier part of that month 00UTC 7 August 1991 (970.47 hPa).
Figures 4b–4d show that the other three storm structural parameters fell well down at the extreme end of their distributions. DsqP, R and D assumed values of 4.91 hPa (deg. lat.)−2, 6.62 deg. lat., and 18.21 hPa, respectively, representing rank positions of 17, 12, and 13 over all the August cyclones. Each of these ranks is in the lowest 1 percentile of their distributions, except for DsqP (1.05 percentile). To obtain an overallmulti-faceted estimate of the extremity of the storm, we have formed a score by summing for each August storm its ranking on the above four measures (e.g., for AS12 this score is 1 + 17 + 12 + 13 = 55), and then ranked those scores. AS12 ranks third on such an overall measure, behind the 1995 and the 7 August 1991 storms mentioned above. There was no August storm which ranked above the present storm onall four indices taken individually.
 In our cyclone analysis we also determine the lifetime of each system, expressed in terms of how many (6 hourly) synoptic times the system was found in the Arctic basin. Figure 4e shows the August climatology of these, with about one half of all systems spending between 1 and 3 days in the basin. (Our central pressure and ‘tracking points’ distributions are consistent with the summer results of Sepp and Jaagus [2011, Table 1], who used the cyclone database described by Gulev et al. .) AS12 at the tail of the distribution (41 synoptic times) ranks at number 12 on this characteristic. When the ranking of this parameter is also included in the score calculation AS12 resumes its prime position as the most extreme August system.
 It is also insightful to place this storm in the context of all (i.e., over the entire year) Arctic storms for the period of analysis. Summer is climatologically a less energetic time in the Arctic and, e.g., the frequency distribution of central pressures is much less broad than for the winter season [Simmonds et al., 2008]. Over the record there are 19625 storms which spent some time in the Arctic basin. On the score derived for the four cyclone parameters AS12 ranked 47 across all of these cyclones (0.24 percentile); when the track lifetime was included in the score the rank rose to a remarkable 13 (0.06 percentile).
6. Summary and Concluding Remarks
 We have analysed a very dramatic storm which was formed over Siberia on 2 August 2012, moved into the Arctic basin, and finally finished its days in the CAA at 18UTC 14 August, a remarkable 12¾ days after its formation. During its lifetime the central pressure of the storm dropped to 966.38 hPa, while the structural attributes of DsqP, R, and D attained their extrema of 4.91 hPa (deg. lat.)−2, 6.62 deg. lat., and 18.21 hPa, respectively. (We comment that the central pressure was very similar to the 966.84 hPa determined when the cyclone scheme was applied to the MERRA reanalysis [Rienecker et al., 2011].) This remarkable storm made its appearance during a period when the extent of Arctic sea ice was on the way to reaching a new all-time low over the satellite era. Given the importance of the Arctic region, it was of significance to diagnose the structure and evolution of AS12 and, in particular, ascertain whether the highly anomalous sea ice coverage played a role in its development.
 Our analysis indicates that the storm behaviour was strongly influenced by baroclinicity and by the presence of a tropopause polar vortex. Baroclinicity was seen to be a key factor associated with the rapid decrease in central pressure when AS12 crossed into the East Siberian Sea. The surface system formed in a deep southward trough associated with the TPV. After four days the centers of the TPV and AS12 became vertically aligned, and remained so till the latter's demise over the CAA. There is little evidence to suggest that the large negative anomalies of sea ice extent had any substantial impact on the cyclone. By contrast, the storm greatly impacted on the ice distribution (NSIDC Arctic Sea Ice News and Analysis, 2012, available from the NSIDC Web site, http://nsidc.org/arcticseaicenews/2012/10/poles-apart-a-record-breaking-summer-and-winter/).
 A range of potential mechanisms responsible for the record-breaking sea ice extent in 2012 have been suggested. A key candidate is that the significant reduction in areal ice coverage and the significant thinning of the ice left the remaining Arctic ice cover in 2012 more vulnerable to the storm than it would have been in earlier decades (see discussion on the effects of ‘preconditioning’ inScreen et al. ). As to the Arctic storm itself, our analysis shows it to be a truly remarkable one analysed over the 1979–2012 period covered by CFSR/CFSV2. Using our multiple-index approach (based on cyclone properties and longevity) we conclude that AS12 was the most extreme August Arctic cyclone (out of a population of 1618). When all Arctic cyclones were considered (which included the more vigorous winter systems) AS12 ranked in position 13 out of a compilation of 19625 storms. This storm truly deserves the title of ‘The Great Arctic Cyclone of August 2012’.
 The authors are grateful to NCEP for making the CFSR and CFSV2 datasets readily available. Parts of this research were made possible by a grant from the Australian Research Council.
 The Editor thanks Heini Wernli and an anonymous reviewer for their assistance in evaluating this paper.