The Contribution of Adam A. Scaife was written in the course of his employment at the Met Office, UK and is published with the permission of the Controller of HMSO and the Queen's Printer for Scotland.
Cold air outbreaks (CAOs) are departures of cold air masses into warmer regions. Over land, these events can lead to deaths and damage (Mercer, 2003; Barnett et al, 2005; Pinto et al, 2007). Over the ocean, CAOs are important for a number of reasons: they give rise to mesoscale weather phenomena such as polar lows (Bracegirdle and Gray, 2008), they lead to enhanced heat and momentum fluxes from the ocean to the air (Renfrew and Moore, 1999) and may therefore influence the ocean circulation (Pickart et al, 2003), and they cause rapid formation of sea ice in marginal ice zones (Skogseth et al, 2004). In recent years it has emerged that anomalies in the stratospheric circulation can be associated with tropospheric CAOs (Thompson et al, 2002; Cai and Ren, 2007; Scaife et al, 2008).
Normally, the extratropical stratosphere is characterised by a strong westerly circumpolar flow. In winter, planetary waves of tropospheric origin propagate continuously into the stratosphere (Charney and Drazin, 1961), where they break and exert a drag on the zonal flow (McIntyre and Palmer, 1983; Polvani and Waugh, 2004). This violates the geostrophic balance and induces a poleward drift of air masses. At high latitudes, the air converges, sinks and warms adiabatically. If there is severe wave-breaking, the stratospheric zonal flow reverses, giving rise to stratospheric sudden warmings (SSWs: Matsuno, 1971), which may last for days to weeks (Limpasuvan and Hartmann, 1999). After their first appearance in the upper stratosphere, circulation anomalies are occasionally found at successively lower levels (Matsuno, 1970; Lorenz and Hartmann, 2003). After reaching the tropopause, the anomalies may impact the troposphere through an interaction with synoptic-scale eddies (Song and Robinson, 2004), or more directly through induced meridional circulations. As a result, a negative Northern Annular Mode (NAM: Thompson and Wallace, 2001) and North Atlantic Oscillation (NAO: Hurrell et al, 2003) pattern may occur near the surface some weeks after the first warming signal in the upper stratosphere (Baldwin and Dunkerton, 2001; Baldwin et al, 2003; Limpasuvan et al, 2004).
Negative NAM and NAO regimes in the troposphere have a profound influence on the weather in large and widespread regions of the Northern Hemisphere (NH) (Kenyon and Hegerl, 2008). Atlantic and Pacific storm tracks shift latitudinally (Hurrell and Van Loon, 1997; Baldwin and Dunkerton, 2001), Greenland and Newfoundland warm (Thompson et al, 2002), and the frequency and severity of CAOs increase over large parts of east Asia (Chen et al, 2005; Jeong and Ho, 2005), northern Eurasia (Scaife et al, 2008) and eastern North America (Thompson and Wallace, 2001; Walsh et al, 2001; Cellitti et al, 2006). Over the ocean, negative phases of the NAO, and positive height anomalies over Greenland in particular, are associated with marine CAOs over the Nordic Seas (Kolstad et al, 2009).
Motivated by the link that has been observed between anomalous stratospheric events and the tropospheric climate, we aim to provide a detailed description of tropospheric cold anomalies in relation to such events. Thompson et al(2002) investigated the mean temperature response during the first 60 days after the onset dates of stratospheric anomalous vortex conditions. Here, we extend their work by assessing the temperature development and changes in the probability of CAOs at different stages of stratospheric weak vortex events. We find that the tropospheric temperature development goes through several distinct and well-defined stages of stratospheric weak vortex events and we identify CAOs over both continental and oceanic regions. These results are corroborated by data from 300-year time slices of 13 coupled model simulations.
2. Data and methods
Daily mean fields from the National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) re-analysis (hereafter referred to as NNR) data (Kalnay et al, 1996) were used throughout the study. The analysis period was from the autumn/winter of 1958 to the winter/spring of 2009.
Monthly mean data from 13 models in the World Climate Research Programme's (WCRP's) Coupled Model Intercomparison Project phase 3 (CMIP3) multi-model dataset were also used. An analysis of the stratospheric variability of these models is found in Cordero and Forster (2006). The models used are listed in Table I (note that we used a more recent BCCR-BCM model simulation, as described by Otterå et al(2009)). These were the only models that included 300-year time slices of pre-industrial control simulations, with no anthropogenic or natural forcing. The time slices were chosen arbitrarily from the years that were available for download.
Table I. The official CMIP3 designations of the models that were used in this study.
The most commonly used measure of stratospheric variability is the NAM index. However, as both the spatial structure and the temporal variability of the NAM differed greatly across the models, we defined a Vortex Strength Index (VSI) as −ZP, where, , Z′ = Z − Z̄, Z is the geopotential height, Z̄ is its climatological mean, φ is the latitude, and the sum was performed on all grid points north of 65°N. The reason for the minus sign is that the vortex is weak when the pressure is high and vice versa. Anomalies were formed by removing the date-wise climatological mean for each grid point. To ensure that the climatology was smooth, we applied a 31-day running-mean filter. The highest pressure level (below 10 hPa, where some of the models appeared to be unreliable) for which data was available from all the models was 50 hPa, so this was used as the stratospheric reference level for both the models and the NNR. Baldwin and Thompson (2009) have shown that a VSI in this form is practically identical to the daily zonal-mean NAM index. The monthly VSI was computed from the model data in a similar way, although the climatological monthly means were not smoothed.
The analysis of this paper focuses on the temporal development of the signals of anomalous temperature and geopotential height. Both parameters were evaluated at a pressure level of 850 hPa and their date-wise climatological means and anomalies were found in the same manner as for the area-averaged 50 hPa geopotential height anomalies described above, although for each grid point separately.
Our analysis is centred on composites of days and months for which the stratospheric vortex is weak. We define Weak vortex days (WVDs) in the NNR as the days for which the daily VSI falls below its overall wintertime (December–March) 10th percentile. An alternative to this method would be to remove the seasonal cycle (by using the date-wise climatological 10th percentile as a threshold instead), but this would have forced the WVDs to be equally distributed among the winter months. Cold days are defined as days with an 850 hPa temperature below its date-wise climatological 10th percentile. When identifying cold days we did remove the seasonal cycle, as the purpose of defining cold days is to assess whether a given day is colder than ‘normal’. Weak vortex months (WVMs) and Cold months in the models are defined with respect to the overall 10th percentiles of the monthly mean anomalies.
To assess the sensitivity of the results to the choice of method, the analysis was also done by compositing with respect to a set of SSW central dates, as defined and identified by Charlton and Polvani (2007) (hereafter referred to as CP07). Six SSWs have occurred since CP07, yielding a total of 31 SSWs since 1958. The new central dates, as derived from the NNR, are listed in Table II
Table II. SSW central dates since the ones identified by Charlton and Polvani (2007), as derived from the NNR.
18 Jan 2003
7 Jan 2004
21 Jan 2006
24 Feb 2007
22 Feb 2008
24 Jan 2009
3.1. Weak vortex events
In Figure 1(a), a matrix of VSI values for each day in the analysis period is shown. The values were grouped with respect to deciles. The blue days are the WVDs as defined above. The SSW central dates are shown using crosses. As mentioned earlier, due to the way they were computed, the density of WVDs is higher in midwinter than in early and late winter. Figure 1(a) shows that this complies well with the seasonal distribution of the SSW central dates. An advantage of CP07's approach is that all their events are independent, and the study of lead/lag processes is therefore free of the effects of artificial smoothing. The composite zonal-mean of the zonal wind at 10 hPa and 60°N relative to the SSW central dates is shown in Figure 1(b). The rapid weakening and gradual recovery of the polar vortex is clear. However, a disadvantage of CP07's approach, or indeed any approach that selects a specific reference date for each event, is that one must be certain that the correct date has been chosen in each case. Otherwise, the temporal signal may be distorted. The relative scarcity of observed SSWs adds weight to this issue. The VSI-based approach, in which each WVD is regarded as a separate ‘event’, leads to runs of days and a smoothing of the temporal signal. However, it is simple and sensitive to only one a priori choice: the selection of a threshold value for WVDs. The composite 50 hPa polar cap geopotential height anomalies relative to WVDs are shown in Figure 1(c). The symmetrical evolution of the height anomalies about the WVDs is a result of the smoothing introduced by the VSI-based approach. The symmetry also shows that the WVD approach is biased towards the middle date of longer events. We also note that in composites of WVDs, persistent events are given more weight than transient ones.
3.2. Tropospheric signature
In this section, we analyse temporal developments in the troposphere throughout the life cycles of stratospheric weak vortex events using both CP07's approach and the VSI-based approach, with an emphasis on cold anomalies. To simplify the notation, we build loosely on the terminology of Limpasuvan et al(2004). They examined the evolution of wave activity fluxes and atmospheric pressure fields in several sub-periods of SSW life cycles. We define the following phases of weak vortex events: Precursor (45–31 days before the central dates and WVDs), Onset (30–16 days before same), Growth (15–1 days before same), Peak (0–14 days after same), Mature (15–29 days after same), Decline (30–44 days after same) and Decay (45–59 days after same). Note that the developments that are seen in Fig. 9 of Limpasuvan et al(2004) are not necessarily directly comparable to the developments in our time intervals.
In Figure 2, the development of the 850 hPa geopotential height and temperature anomalies relative to both the SSW central dates (Figure 2(a)) and WVDs (Figure 2(b)) are shown. In the early stages (Precursor, Onset and Growth), positive height anomalies centred over northwest Eurasia and negative anomalies near the Bering Strait are found. This corresponds to a pattern that has been found to favour stratospheric warmings through an enhancement of upward-propagating tropospheric wave-number-one waves (Kuroda and Kodera, 1999; Garfinkel et al, 2010). It is therefore thought to be a tropospheric precursor of warmings aloft. Cold anomalies arise over north-eastern Asia through anomalous northerly wind components, resulting in southward advection of cold, Arctic air masses. Cold anomalies are also found over Europe, where there are anomalous easterlies. These winds relate to the anomalous ridge over northwest Eurasia. To our knowledge, these cold anomalies, which appear too early to be affected by downward propagation of the negative NAM-like signals after SSWs, have not been documented previously in relation to weak vortex events. Bueh and Nakamura (2007) document similar temperature patterns in response to the so-called Scandinavian Pattern, which resembles the anomalous height pattern found in Figure 2(b) (Precursor stage).
By the Growth and Peak stages in the WVD framework (Figure 2(b)), an NAO-like anomaly has appeared. This leads to an anomalous northerly flow and cold anomalies in northern Europe. This is consistent with an increased frequency of marine CAOs in the Nordic Seas region under negative NAO conditions (Kolstad et al, 2009). As the pressure anomalies are contained primarily in the Atlantic sector by this time, the cold anomalies in Asia diminish in magnitude. At the same time, cold anomalies appear on the east coast of North America. Corresponding warm anomalies over Canada and the Mediterranean/North Africa complete the quadrupole pattern of temperature anomalies that are associated with the NAO (Stephenson and Pavan, 2003). In the Mature, Decline and Decay stages, the NAO pattern gradually weakens, and the most prominent cold anomalies are found in Asia and Europe. This is consistent with findings from previous studies (Baldwin and Dunkerton, 2001; Thompson et al, 2002; Chen et al, 2005).
To summarise, Figure 2 shows temporally and geographically distinct cold anomalies throughout the life cycle of weak vortex events. The cold anomalies were found using the CP07 method (Figure 2(a)) and using the WVD method (Figure 2(b)), although the exact timing and relative amplitudes differ slightly.
3.3. Relative frequency of stratosphere-related cold air outbreaks
Figure 2 was based on changes to the mean temperature field, with no regard to its extreme values. We now define the quantity α as the fractional change in the number of cold days (days for which the 850 hPa temperature anomaly falls below its date-wise 10th percentile) with respect to climatology. If α = 1.5, cold days are 50% more likely than normal. As this parameter is only concerned with cold extremes, the evolution of α, which is shown in Figure 3, provides a useful complement to Figure 2.
In the early stages of weak vortex events, i.e. during the Precursor and Onset stages, α > 1.5 most consistently in Asia and Europe. By the Growth stage, α > 1.5 off the coast of North America in the WVD framework. In the later Mature and Decline stages, the strongest cold anomalies are again confined to Asia and Europe. It is only in Asia that α > 1.75 during all the periods shown. The large fractional changes in Figure 3 show that the frequency of cold days is affected strongly around the time of sudden stratospheric warmings.
3.4. Robustness of the results
In Figure 4, the same analysis that was used to produce Figures 2(b) and 3(b) was applied to the two halves of the analysis period. Note that we averaged over longer time periods than in Figures 2 and 3. In general, the features shown previously for both the mean and extreme events are robust to this halving of the data period. High pressure is observed over northwest Eurasia prior to the weak vortex events, and a subsequent negative NAO-like pattern is seen. Similarly, cold anomalies are found over Asia, Europe and near the east coast of North America. The perhaps largest difference between the two periods is found over the Pacific in the Mature/Decline phases. In the first part of the period, an anomalous low over the Aleutian Islands (Figure 4(a)) is associated with advection of cold, continental air out over the Pacific (Figure 4(c)). The large Pacific high anomaly in the latter part of the period (Figure 4(b)) is associated with a 75% increase in the number of cold days along the west coast of North America (Figure 4(d)).
As an additional test of robustness, we examined monthly mean data from 13 coupled climate models. This part of the analysis is done using the WVM framework. Some of the CMIP3 models have low model tops and many of them underestimate the stratospheric variance (Cordero and Forster, 2006). The models will therefore be used to evaluate the temporal and spatial variability rather than the exact amplitudes of the anomalies. Based on the symmetry of Figure 1(c), it is natural to define that WVMs correspond to the Growth and Peak phases of weak vortex events. Figure 5(a) shows the 850 hPa geopotential height and temperature anomalies during WVMs (Growth/Peak phase), as well as during the preceding (Precursor/Onset phase) and the succeeding (Mature/Decline phase) months. The initial cold anomaly in Asia, the westward shift of the northwest Eurasia warm anomaly and the appearance of cold anomalies in Europe and off the east coast of North America are all seen in the model ensemble around weak vortex months. The temperature developments are consistent with the anomalous northerly and easterly flow that is associated with the pressure anomalies, whereas the overall westward progression of the temperature pattern indicates further potential for seasonal predictability.
In Figure 5(b), the changes to the probability of cold months in the different stages are shown. The following cold anomalies are associated with a higher than 50% increase in the number of cold months: (1) The cold anomaly over Asia in the Precursor/Onset and Growth/Peak phases, (2) the cold anomaly over northern Europe in the Growth/Peak and Mature/Decline phases, and (3) the cold anomaly over north-eastern North America in the Growth/Peak phases. Qualitatively, the features in Figure 5 are in good agreement with Figures 2 and 3, although the magnitudes of the anomalies are generally weaker. This is at least partly due to the much larger sample size of the model data.
4. Concluding remarks
The relationship between stratospheric weak vortex events and tropospheric developments, and cold air outbreaks (CAOs) in particular, were investigated using 51 winters of re-analysis data and a set of coupled climate models. We found large increases in the frequency of cold air outbreaks (Figure 3) that coincide geographically with the regions of mean temperature change (Figure 2). The probability of CAOs was found to increase: (1) by 75% or more in some regions of northern Asia throughout the life cycle of weak vortex events (from the Precursor phase to the Decay phase), (2) by 50% or more in some regions of Europe (from the Onset phase to the Decline phase), and (3) by 50% or more in the Peak phase off the east coast of North America.
Changes in the frequency of cold air outbreaks associated with the stratosphere are therefore large compared to the climatological incidence of CAOs. Such substantial changes make this signal important for the long-range forecasting of the likelihood of CAOs. If the signal is predictable, then there will be an associated predictability of CAOs. However, if it is unpredictable, then it represents an important limit on the long-range predictability of CAOs.
A potential obstacle to the predictability of CAOs based on the state of the stratospheric vortex is the fact that many of the cold anomalies seen in Figure 3 occurred before the SSW central dates and WVDs. The early CAOs in Europe and Asia were associated with the perhaps clearest precursor of stratospheric weak vortex events, a high-pressure anomaly centred over the northwestern edge of Eurasia in the Precursor, Onset and Growth phases. Although its location changed with time, this positive height anomaly persisted for all the phases and was confined to high latitudes in the Atlantic sector. More work is therefore needed to address the chain of cause and effect and to investigate tropospheric precursors of weak vortex events, adding to existing studies of troposphere–stratosphere interactions (Kuroda and Kodera, 1999; Chen et al, 2003; Polvani and Waugh, 2004; Reichler et al, 2005; Scaife et al, 2005; Cohen et al, 2007; Martius et al, 2009; Mukougawa et al, 2009; Garfinkel et al, 2010).
We did not directly address the issue of cause and effect of CAOs in this paper, but interestingly, we found a hemisphere-wide pattern of lower-tropospheric temperature signals both before and after weak vortex events. In general, such temperature signals are associated with pressure anomaly dipoles in the form of anomalous ridges upstream (such as the precursory high anomaly over northwest Eurasia) and anomalous troughs downstream of the cold anomalies. Such patterns lead to changes to the flow, and the resulting temperature advections may well act as positive feedback mechanisms, as documented for the negative phase of the surface NAM (Thompson and Wallace, 2000). The association between pressure anomaly dipoles and CAOs is known from previous studies (Konrad, 1996; Walsh et al, 2001; Chen et al, 2005; Takaya and Nakamura, 2005; Cellitti et al, 2006; Kolstad et al, 2009). It is quite possible that some of the regional CAOs identified in this paper are at least partly set up or sustained by cold air advection, as part of the chain of events outlined by Konrad (1996).
Given the strong stratospheric link to many CAOs, it could be that attention needs to be paid to the simulation of the stratosphere in climate models. However, parts of our analysis were repeated with an ensemble of 13 coupled climate models. Somewhat surprisingly, considering that many of these models have low model tops and poorly resolved stratospheres (Cordero and Forster, 2006), the model results corroborated the relationships between the weak vortex events and the cold anomalies listed above. This may indicate that the main aspects of the tropospheric temperature developments during the life cycle of the stratospheric weak vortex events are associated with internal processes in the troposphere and lower stratosphere, as suggested by Polvani and Waugh (2004).
We wish to thank the editors and two anonymous reviewers for contributing to an improved paper. We also acknowledge the modelling groups, the Program for Climate Model Diagnosis and Intercomparison (PCMDI) and the WCRP's Working Group on Coupled Modelling (WGCM) for making available the WCRP CMIP3 multi-model dataset, as well as NOAA/OAR/ESRL PSD for providing the NCEP/NCAR re-analysis data. Erik Kolstad's work was funded by the Norwegian Research Council through its International Polar Year programme and the project IPY-THORPEX (grant number 175992/S30). Tarjei Breiteig's work was supported by the COMPAS project (grant number 165424), also funded by the Norwegian Research Council. Adam Scaife was supported by the Joint DECC and Defra Integrated Climate Programme—DECC/Defra (GA01101). This is publication no. A259 from the Bjerknes Centre for Climate Research.