Climate Dynamics

, Volume 30, Issue 7–8, pp 871–885 | Cite as

Marine cold-air outbreaks in the future: an assessment of IPCC AR4 model results for the Northern Hemisphere



For many locations around the globe some of the most severe weather is associated with outbreaks of cold air over relatively warm oceans, referred to here as marine cold-air outbreaks (MCAOs). Drawing on empirical evidence, an MCAO indicator is defined here as the difference between the skin potential temperature, which over open ocean is the sea surface potential temperature, and the potential temperature at 700 hPa. Rare MCAOs are defined as the 95th percentile of this indicator. Climate model data that have been provided as part of the Intergovernmental Panel on Climate Change (IPCC) Assessment Report Four (AR4) were used to assess the models’ projections for the twenty-first century and their ability to represent the observed climatology of MCAOs. The ensemble average of the models broadly captures the observed spatial distribution of the strength of MCAOs. However, there are some significant differences between the models and observations, which are mainly associated with simulated biases of the underlying sea ice, such as excessive sea-ice extent over the Barents Sea in most of the models. The future changes of the strength of MCAOs vary significantly across the Northern Hemisphere. The largest projected weakening of MCAOs is over the Labrador Sea. Over the Nordic seas the main region of strong MCAOs will move north and weaken slightly as it moves away from the warm tongue of the Gulf Stream in the Norwegian Sea. Over the Sea of Japan there is projected to be only a small weakening of MCAOs. The implications of the results for mesoscale weather systems that are associated with MCAOs, namely polar lows and arctic fronts, are discussed.

1 Introduction

A number of regions around the globe are prone to dangerous weather conditions associated with outbreaks of cold and stably stratified polar and continental air masses over relatively warm ocean surfaces. Such outbreaks, characterised by roll clouds and small-scale fronts at small fetches (i.e., close to the coast or ice edge) and deeper convection further from the coast, are referred to here as marine cold-air outbreaks (MCAOs). The strong atmospheric convection that is characteristic of MCAOs can act to enhance the intensification rate of weather systems and also contributes significantly to climatological precipitation totals.

It is problematic that the current generation of climate models do not resolve mesoscale weather systems, such as arctic fronts and polar lows, that are associated with MCAOs (e.g., Rasmussen and Turner 2003), thus omitting important weather phenomena from projections of future climate change. For instance one common feature of MCAOs, roll clouds, can have a profound impact on flux modelling of the planetary boundary layer and remain a key problem for climate models (Liu et al. 2006). Indeed this issue is not limited to low-resolution climate models, for example Pagowski and Moore (2001) found that a mesoscale model “grossly over-estimated” heat fluxes. On the positive side MCAOs, which are large-scale phenomena, are represented in climate models. We define here an indicator (or index) of MCAOs. The indicator is based on the difference between the sea surface temperature of open water and the air temperature at 700 hPa, drawing on the empirical work by Kolstad (2006) and Bracegirdle and Gray (2007). A somewhat comparable approach was recently used by Camargo et al. (2007) to assess climate models’ ability to represent the potential for tropical cyclone development.

Changes in a range of factors that influence the strength of MCAOs have been manifested during recent years and projected in future global warming scenarios. Observation-based data have shown recent sea-ice retreat (Serreze et al. 2002; Zhang et al. 2003) and a poleward shift of the storm tracks (McCabe et al. 2001; Zhang et al. 2004). Future projections show these trends continuing (Yin 2005; Arzel et al. 2006; Bengtsson et al. 2006; Zhang and Walsh 2006) along with a change of large-scale atmospheric teleconnection patterns [e.g., the positive trend of the North Atlantic Oscillation (NAO); Stephenson et al. 2006], a weakening of the AMOC (Atlantic Meridional Overturning Circulation; Gregory et al. 2005; Schmittner et al. 2005) and increases of atmospheric static stability (Frierson 2006). These changes are not uniform across the globe and will affect the strength and frequency of MCAOs to varying degrees in different regions. Some studies have found that in some regions trends of extreme cold-air outbreaks do not necessarily follow local trends of mean temperature: a phenomenon known as the ‘climate paradox’ (Vavrus et al. 2006). The response of MCAOs to climate forcing therefore requires a fully coupled climate model to take into account a wide range of factors.

The importance of cold-air outbreaks to severe weather in the Arctic can be illustrated by considering their role in the dynamics of two related weather phenomena: arctic fronts and polar lows.

Arctic fronts are shallow features that define a boundary between cold and highly stable Arctic air masses and unstable, modified air near the warm sea surface during major cold-air outbreaks (Shapiro et al. 1989; Drüe and Heinemann 2001). Model simulations indicate that the corresponding large ocean-air heat fluxes and release of latent heat strengthen frontogenesis and enhance a low-level jet (e.g., Grønås and Skeie 1999). Arctic fronts usually develop near the edge of the sea-ice, and ‘ice breeze’ mechanisms can be important for their initial development (Langland et al. 1989; Brümmer 1996). Associated near-surface winds can sometimes reach hurricane force (Økland 1998; Grønås and Skeie 1999) and have most likely caused many accidents at sea near the marginal ice zone (Grønås and Skeie 1999).

Our second example of potentially severe MCAO weather is polar lows, which are intense small-scale cyclones that commonly occur in cold-air outbreaks over high-latitude oceans (Businger 1985; Lystad 1986; Shapiro and Fedor 1989; Rasmussen and Turner 2003). Their development can be conceptualised in two stages, where baroclinic development, often along arctic fronts, is followed by a period of intensification dominated by latent heat release (Nordeng 1990; Kristjansson 1990). During the second stage of intensification polar lows sometimes possess a structure similar to hurricanes, with a warm core and a clear central eye surrounded by spiral cloud bands (Nordeng and Rasmussen 1992). The strong dependence of the intensification rate of polar lows on latent heat release and surface fluxes has been demonstrated in a range of modelling experiments, from high-resolution idealized simulations with atmospheric convection explicitly resolved (e.g., Craig and Gray 1996, who used a horizontal grid spacing of 5 km) to lower-resolution simulations of case studies using weather forecasting models with convection parameterized (Claud et al. 2004, who used a horizontal grid spacing of 25 km).

MCAOs are not only important for mesoscale weather systems. During MCAOs, the heat loss from the ocean to the atmosphere leads to evaporation, cooling and freezing, which again leads to positive salinity anomalies and ocean convection. For instance, Greenland Sea deep water formation is a result of the water masses being retained long enough for significant cooling to take place (Hakkinen 1995; Ronski and Budeus 2005). Changes in the strength and/or prevalence of MCAOs potentially have global implications since the rate of North Atlantic Deep Water production is an important contributor to the strength of the AMOC (Griffies and Tziperman 1995; Delworth and Greatbatch 2000). Climate-modelling experiments that simulate a weakening or shut down of the AMOC show large changes of temperature and sea level in both hemispheres (Stocker 2002; Levermann et al. 2005).

The key aims of this paper are to establish the climatological characteristics of MCAOs, and thereby indirectly the likelihood of corresponding severe weather, to evaluate the ability of climate models in simulating MCAOs and to identify and assess the projected future changes. The climate model data used from this study were taken from the Intergovernmental Panel on Climate Change (IPCC) Assessment Report Four (AR4) archive. In the next section an MCAO indicator is defined, followed by a description of the methods and the model data which are used. Climatological properties of the indicator and model trends in the Northern Hemisphere are then presented. A discussion of the results follows before a few concluding remarks.

2 A marine cold-air outbreak indicator

There is no widely accepted quantitative definition of a ‘cold-air outbreak’. Vavrus et al. (2006) defined cold-air outbreaks as the occurrence of two or more days during which the local mean daily surface air temperature is at least two standard deviations below the local wintertime mean temperature. However, another weather event that the term is often used to describe is the advection of cold air over a relatively warm ocean, referred to here as a marine cold-air outbreak (MCAO). MCAOs can be readily identified on satellite imagery due to distinctive cloud characteristics such as roll clouds and cellular convection.

Most definitions used in studies of cold-air outbreaks are designed to detect cold events over densely populated land regions (Boyle 1986; Konrad and Colucci 1989; Vavrus et al. 2006). As a result these definitions are not good indicators of the severity of MCAOs since they are based on atmospheric temperature anomalies relative to local climatologies, which over the ocean do not in general reflect the local air–sea temperature difference. For instance, over the Sea of Japan, which is known to be particularly prone to MCAOs (Dorman et al. 2004; Ninomiya et al. 2006) the frequency of cold-air outbreaks as defined by Vavrus et al. (2006) is low. Unlike Vavrus et al. (2006), Dorman et al. (2004) do not use an anomaly approach, but instead define a ‘Very Cold Siberian Air Outbreak’ over the Sea of Japan as occurring when the 0°C isotherm is south of 40°N for more than 24 h. However, this definition is rather specific to the region of the Sea of Japan and is not a reliable indicator of the occurrence of severe MCAOs in other regions due to large variations of both SST and atmospheric temperature with latitude.

The approach adopted in this study draws from an analysis conducted by Bracegirdle and Gray (2007), which assessed the effectiveness of various parameters for identifying polar lows that occur in MCAOs over the Nordic seas. They found empirically that the thermodynamic disequilibrium between the ocean and atmosphere (TDOA), as measured by the difference between the sea surface temperature (SST) and the wet-bulb potential temperature at 700 hPa (θw,700), is a more reliable indicator of MCAOs than atmospheric temperature alone. It should be noted that the only atmospheric levels available for their analysis were 900, 700 and 500 hPa. As pointed out by Bracegirdle and Gray (2007), the most likely reason for the TDOA defined using thermodynamic variables at 900 hPa being relatively ineffective is that in cold-air outbreaks large surface fluxes of heat and moisture from the ocean to the atmosphere act to rapidly modify the boundary layer and the cold air-mass at 900 hPa can rapidly lose its identity. The 700 hPa level is generally above the boundary layer and therefore less likely to be modified rapidly by surface fluxes. A possible reason for the TDOA defined using thermodynamic variables at 500 hPa being less effective is that in some cases conditions at 500 hPa may not correspond to the presence of an MCAO in the lower troposphere. A similar approach to that used by Bracegirdle and Gray (2007) is in fact already in practical use as Norwegian meteorologists routinely assess the difference between the SST and the temperature at the 500 hPa level as part of a polar low advance warning service (Gunnar Noer 2005, personal communication).

The MCAO indicator defined here was based on the empirical evidence of Bracegirdle and Gray (2007), but with a few minor adaptations to the parameter they chose. Firstly, we used the ‘skin temperature’ (SKT), which is the temperature on the surface and corresponds to the SST of open water and the soil (or ice) temperature elsewhere. This is an ideal parameter for our purposes, as the temperature of the ocean surface under the ice is irrelevant to this study. We also used the ‘dry’ potential temperature at 700 hPa (θ700) rather than θw,700, a choice that made a negligible difference to discriminant performance in Bracegirdle and Gray (2007). The basis for choosing the dry value here is that at warmer lower latitudes, not analysed by Bracegirdle and Gray (2007), the greater moisture capacity of the atmosphere means that the wet-bulb temperature can deviate significantly from the actual temperature. For the same reason the use of wet-bulb values at both the surface and 700 hPa was not found to be an effective indicator of MCAOs. Furthermore, to take into account the sea-level pressure, we calculated the potential temperature of the skin temperature (θSKT) as opposed to using the skin temperature value itself. We also divided the vertical temperature difference by the pressure difference between the surface and the 700 hPa level. These last adaptations were motivated by the large range of climatological sea-level pressure values (and the spatial variations of the typical height of the 700 hPa surface) that occur across different parts of the globe.

The MCAO indicator is defined as
$$ \frac{\Delta \theta}{\Delta p} = \frac{\theta_{\rm SKT} - \theta_{700}}{{\rm SLP} - p_{700}}, $$
where SLP is the sea level pressure, p700 is 700 hPa, and the potential temperatures have been defined above. The unit is K/bar (1 bar = 105 Pa).

3 Data and methods

The MCAO indicator defined above is transient in nature, mostly due to rapid fluctuations of air temperature and surface pressure. The variation of the skin (ocean) temperature is slower, which is probably why skin temperature (and sea surface temperature) is only available as monthly averages in the IPCC AR4 models. Only the models which provide daily temperature fields at the 700 hPa pressure level could be used in the study. Unfortunately, this lead to the exclusion of UKMO-HADCM3, UKMO-HADGEM1, NCAR-CCSM3 and others. The models which were used, along with their average horizontal resolution in the study area and the name of their realisations under the two scenarios, are listed in Table 1. More details about the models and their full names can be found at All the model data were collected from
Table 1

Abbreviations of the models that were used in the analysis, their country of origin, the name of the runs for each scenario, and their horizontal resolution

Model name




Lon × Lat





2.8 × 2.8





2.8 × 2.8





2.8 × 2.8



run1, run2


1.9 × 1.9



run1, run4

run2, run4

1.9 × 1.9





2.5 × 2.0





2.5 × 2.0





4.0 × 3.0





1.1 × 1.1



run1, run2


3.8 × 2.5





1.1 × 1.1





2.8 × 2.8





2.8 × 2.8


Europe (ECMWF)



2.5 × 2.5

The 2.5°  × 2.5° ECMWF ERA-40 reanalysis (Uppala et al. 2005) is used as the observational reference throughout the paper. These data are available on T159 (roughly 1°  × 1°) resolution, but the coarse version was chosen for consistency with most of the climate models, which are generally run at resolutions lower than T159 (see Table 1). The pre-1997 cold bias over the central Arctic Ocean in ERA-40 that was identified by Bromwich et al. (2007) is not important over regions of open ocean, which are the focus of this study. To compile the model ensemble and to be able to compare the results from the models with ERA-40, the model data were interpolated onto the ERA-40 grid.

Because we wanted to use skin temperature along with daily atmospheric temperature fields, it was necessary to use an interpolation scheme to artificially increase the temporal resolution from monthly to daily. For each day the skin temperature at each grid point was set to be the monthly mean value for the corresponding month. Then the daily values were replaced with 31-day running-mean values at each grid point. This is equivalent to interpolating monthly mean values back onto individual dates, an approach used by NOAA’s Climate Prediction Center (CPC) to calculate long-term daily mean NAO index values. At the mid-point of each month, the value is thus the monthly mean itself, whereas at either end of each month, it is influenced about equally by the mean values from the two closest months. As the ocean temperature varies on a slower time scale than that of the atmosphere, this is assumed to yield fairly realistic daily mean fields.

To verify the validity of this approach, we compared artificial daily mean ERA-40 skin temperature values calculated using the method described above with true daily mean fields. In Fig. 1a, the true daily mean values of the potential temperature based on the skin temperature and the sea level pressure for the open-water grid point at longitude 0° and latitude 70°N is shown along with the artificial values, which were computed using monthly mean skin temperatures and daily mean sea level pressure. The true and artificial daily mean skin temperatures are also shown. The period shown was arbitrarily chosen and runs from 1 November 1991 through the following 150 days. Even the high-frequency variability of the skin temperature is captured well by our method; the correlation coefficient between the true and artificial daily mean skin temperatures for this grid point is 0.99 for the period 1991–2000. Pressure fluctuations dominate the variability of the potential temperature, and the correlation coefficient between the true and artificial daily mean potential temperatures is also 0.99. In Fig. 1b, the correlation coefficients between true and artificial daily mean skin temperature are plotted for each grid point in the same period (all significant at the 0.05-level). In regions with variable sea-ice cover, the correlation is lower than over open ocean, but the coefficients are always above 0.85. We conclude that our approach yields a viable daily mean skin temperature field.
Fig. 1

a Time series for the extended winter from 1 November 1991 and the following 150 days of ERA-40 fields for the grid point at longitude 0° and latitude 70°N: potential temperature calculated from daily mean skin temperature and daily mean sea level pressure (SLP); potential temperature calculated from the artificial daily mean skin temperature (see text for details) and daily mean SLP; daily mean skin temperature; and artificial daily mean skin temperature; b correlation coefficients between true daily mean and artificial daily mean ERA-40 skin temperature for the period from 15 January 1991 to 15 December 2000 (all significant at the 0.05-level)

To quantify the change of the MCAO indicator during a period of specified climate forcing, we used the 20C3M scenario as the reference state of the model system, representing the current climate. The moderate SRES A1B scenario, in which the CO2-equivalent concentration peaks at 720 ppm in 2100 (Houghton et al. 2001), was chosen to represent the future. Due to the varying availability of the data, the reference periods were 1981–2000 and 2081–2100, respectively. Twenty years would normally be considered insufficient to avoid variance due to internal variability in both models and the real atmosphere. Under the assumption that such unwanted effects are smoothed out, this is an incentive to use a model ensemble. To assess the sensitivity to choice of reference period, the fields described above were calculated for both the 1961–1980 and the 1981–2000 ERA-40 periods (see the discussion in the next section). In an attempt to reduce the impact of this internal variability when comparing the model data with ERA-40, we used the 40-year period from 1961 to 2000 as the ERA-40 reference period throughout the paper.

A key aim of this paper is to assess projected changes of the strength of MCAOs. We define events for which the 0.95 quantiles (the 95th percentile) of the MCAO indicator is exceeded as rare MCAOs at that location. On average, for a given grid point, this happens roughly 18 days each year. Assuming that most of the MCAOs occur during the extended winter from late autumn to mid-spring, say in the course of 6 months, the threshold is exceeded three times a month or once every 10 days or so. However, MCAOs tend to persist for more than 1 day, so a certain amount of clustering can be expected. If an average strong MCAO event lasts for 3 days, they will only occur once per month during the extended wintertime when the 0.95 quantile is used as a threshold. It seems that this approach is not very sensitive to the choice of threshold. The 0.99 quantiles, for which an exceedance will occur less than 4 days a year on average, and can thus be said to define exceptional MCAOs, has a spatial distribution which is roughly similar to the 0.95 quantile.

4 ERA-40 climatology

In Fig. 2, the 1961–2000 ERA-40 0.95 quantile of the MCAO indicator Δ θ/Δ p in K/bar is shown, along with the 0.95 quantiles for the two periods 1961–1980 and 1981–2000 with the 0.95 quantiles for the entire period subtracted.
Fig. 2

The MCAO indicator 0.95 quantiles based on ERA-40; a for the reference period 1961–2000; b the difference between the quantiles for the period 1961–1980 and the reference period; c the difference between the quantiles for the period 1981–2000 and the reference period. The unit is K/bar

The most pronounced differences between the two periods is due to the relative absence of sea-ice in the Greenland Sea in the latter period, which has been the subject of many studies (e.g., Deser et al. 2000; Johannessen et al. 2004). In the regions with retreating sea-ice in the Nordic Seas after 1980, the MCAO indicator quantiles are higher because of the new presence of open ocean. This negative sea-ice trend coincides with a positive NAO trend, and significant negative correlations between the sea-ice extent and the NAO index have been found for the Greenland Sea (Deser et al. 2000) and the Barents Sea (Sorteberg and Kvingedal 2006). To the south and southwest of Iceland, the high NAO index values after 1980 lead to higher MCAO indicator values in the Labrador and Irminger Seas. In the Pacific Ocean, the same positive trend is observed, except for a distinct decrease of MCAO indicator quantiles to the north of Japan after 1980.

Overall, three regions stand out with the strongest MCAOs. They are the extreme Northeast Atlantic (over the Norwegian and Barents seas), the extreme Northwest Atlantic (around the southern tip of Greenland over the Labrador Sea and extending to Iceland) and the Northwest Pacific (in particular around Japan and over the Sea of Okhotsk).

Warm ocean currents and southwesterly winds bring warm waters to high latitudes in the Northeast Atlantic, leading to ice-free conditions all year in large parts of the Greenland, Norwegian and Barents Seas. Quantile values exceed 10 K/bar in the Norwegian Sea region. These large values are consistent with other analyses of MCAOs and associated mesoscale weather systems in that region (Lystad 1986; Noer and Ovhed 2003; Kolstad 2006).

A tongue of moderate MCAO indicator quantiles (>−10 K/bar) extends to the southwest of the Norwegian Sea across the North Atlantic to the eastern coast of the USA. Here, the Gulf Stream brings very warm water to relatively high latitudes. In the summer, the water in this region is sometimes warm enough to sustain tropical cyclones (e.g., Jagger and Elsner 2006). During winter, outbreaks of cold continental air have been known to produce powerful low-pressure systems that have much in common with polar lows (Grossman and Betts 1990; Businger et al. 2005).

There is also a tongue of warm water extending into the Labrador Sea, where exceptionally large quantiles (up to 10 K/bar) are found. It known that advection of cold air from Arctic Canada produces strong MCAOs in this region (Mailhot et al. 1996; Renfrew et al. 1999; Pagowski and Moore 2001; Moore and Renfrew 2002). The large heat fluxes from the ocean to the atmosphere that are associated with MCAOs contribute to the Labrador Sea being an important region for North Atlantic Deep Water formation (Wood et al. 1999; Pickart et al. 2002).

The largest MCAO indicator quantiles in the Northwest Pacific region occur around Japan and in the Sea of Okhotsk, where they lie between 0 and 10 K/bar. A range of scenarios producing cold-air outbreaks over the Japan Sea were identified by Dorman et al. (2004), who attributed the most extreme cold-air outbreaks to outbreaks of very cold Siberian air during periods of an expanded Siberian High. The Northwest Pacific is a region where both polar lows (Fu et al. 2004) and explosive extratropical cyclones (Yoshida and Asuma 2004) are known to form.

The northern North Pacific is another region where polar lows have been observed, both over the Bering Sea (Businger 1987; Businger and Baik 1991; Bresch et al. 1997) and over the Gulf of Alaska (Bond and Shapiro 1991; Douglas et al. 1991). However, the MCAO indicator values are fairly low (<0 K/bar), which is consistent with analyses of polar lows that have found atmospheric convection to be more important in the intensification of Atlantic systems than those that form over the Pacific (Sardie and Warner 1983).

5 Results

5.1 Model bias and spread for twentieth century climatology

Figure 3 shows the average 20C3M (1981–2000) 13-member model ensemble mean 0.95 quantile of the MCAO indicator (a), the ensemble bias with respect to ERA-40 (the ensemble mean minus the 1961–2000 ERA-40 average in the top panel of Fig. 2) (b), and the inter-model standard deviation (c).
Fig. 3

a The 13-member model ensemble mean of the MCAO indicator 0.95 quantiles for the period 1981–2000 (20C3M scenario); b the ensemble mean minus the ERA-40 quantiles as calculated for the reference period 1961–2000. The unit is K/bar; c the inter-model standard deviation

The bias of each individual model with respect to ERA-40 are shown in Fig. 4. Most of the models struggle with the sea-ice distribution in the Greenland and/or Barents Seas. This is a region where the annual sea-ice distribution varies substantially (Deser et al. 2000; Sorteberg and Kvingedal 2006). The Barents Sea, which in reality is largely free of ice even at the height of winter, is perpetually frozen in some models, the most extreme of which is CSIRO-Mk3.0.
Fig. 4

The MCAO indicator 1981–2000 (20C3M scenario) 0.95 quantile for each individual model minus the ERA-40 1961–2000 0.95 quantile. The zero contour (no bias) is drawn with a solid black curve. The unit is K/bar

In terms of the model ensemble average (Fig. 3), the models produce MCAO indicator values that are too weak in regions close to sea-ice. Negative differences exceeding 20 K/bar in the eastern Barents Sea, 10 K/bar in the Labrador and Kara Seas and Baffin Bay, and around 5 K/bar in the Greenland Sea, north of Siberia (the East Siberian and Chukchi Seas) and in the Sea of Okhotsk are found (Fig. 3b). The inter-model standard deviation (Fig. 3c) is considerable in most of these locations, with values ranging from higher than 20 K/bar in the Greenland and Barents Seas,  >10 K/bar in the Labrador Sea and north of Siberia, but well below 10 K/bar in the Sea of Okhotsk.

Over the open ocean the models generally overestimate the quantiles of the MCAO indicator. Positive differences of more than 10 K/bar are found between Iceland and Norway, adjacent to the eastern USA and to the east and west of Japan, while differences of around 5 K/bar are found along the Gulf Stream past the British Isles, near the Aleutian Islands and in the Beaufort Sea. High levels of model scatter (>10 K/bar) are found in the Beaufort Sea and in the Sea of Japan.

The biases above depend partly on the models’ representation of ocean currents and sea-ice. In Fig. 5, the 0.95 quantiles of the ERA-40 skin temperature, the model ensemble bias with respect to ERA-40, as well as the inter-model standard deviation are shown.
Fig. 5

a ERA-40 skin temperature 0.95 quantiles for the reference period 1961–2000; b the 13-member model ensemble mean of the skin temperature 0.95 quantiles for the period 1981–2000 (20C3M scenario) minus the ERA-40 1961–2000 0.95 quantile; c the inter-model standard deviation. The unit is K

It is clear from Fig. 5b that there is too much ice in the Barents Sea, with a negative bias of around 3 K. Biases of comparable magnitude are also found in the Seas of Okhotsk and Japan, off the eastern coast of USA and near the date line around 40°N. The model scatter is very large in these regions, ranging from 2 to 4 K.

In some regions of open water, such as the Norwegian Sea and near Japan, there is a positive MCAO indicator bias and a negative SKT bias. This is because the air masses are too cold. In Fig. 6, the 0.05 quantile of the ERA-40 700 hPa potential temperature is shown along with the model ensemble bias and the inter-model standard deviation. (Note that the 0.05 quantile is used here to reflect extremes on the low end of the temperature range.) It is interesting to note that the ensemble bias is negative everywhere except over Greenland and the northeast American continent.
Fig. 6

As Fig. 5, but for the 700 hPa potential temperature 0.05 quantiles

5.2 Future projections

Changes to the MCAO indicator projected for the twenty-first century under the SRESA1B scenario are presented as an ensemble average in Fig. 7 and for each model separately in Fig. 8. Since the bias of the model ensemble is large in most locations, close attention will be paid to the robustness of projected changes by taking account of the inter-model spread of the projections. To assess the relative contribution to the changes from skin temperature and 700 hPa potential temperature, changes to these variables are shown separately. The projected changes of the skin temperature and 700-hPa temperature quantiles are shown in Figs. 9 and 10, respectively.
Fig. 7

a The 13-member model ensemble mean of the MCAO indicator 0.95 quantiles for the period 2081–2100 (A1B scenario) minus the mean of the 1981–2000 (20C3M scenario) 0.95 quantiles; b the inter-model standard deviation of the differences. The unit is K/bar

Fig. 8

The MCAO indicator 2081–2100 (A1B scenario) 0.95 quantile for each individual model minus the 1981–2000 (20C3M scenario) 0.95 quantile. The zero contour (no change) is drawn with a solid black curve. The unit is K/bar

Fig. 9

As Fig. 7, but for the skin temperature 0.95 quantiles. The unit is K

Fig. 10

As Fig. 7, but for the 700 hPa potential temperature 0.05 quantiles. The unit is K

At a glance, the MCAO indicator has a negative trend over open ocean and a substantial positive trend over the sea-ice-covered Arctic region (Fig. 7). The 0.95 quantile of the ocean surface temperature increases uniformly (Fig. 9), although not by much to the south of Greenland (<1 K). On the whole the 0.05 quantile of the temperature at 700 hPa increases more rapidly, with the smallest positive changes on the order of 2–3 K to the south of Iceland. However, there is a considerable regional variation of the projected changes.

Rapid sea-ice retreat along the southern rim of the Arctic Ocean leads to a large increase of the strength of MCAOs (Fig. 7). This is most prominent to the north of the Bering Strait and in the northern Barents Sea, where changes of around 20 K/bar are projected. Due the poor representation of the initial sea-ice distribution in the latter region, the Barents Sea region has the largest inter-model scatter. The standard deviation in the Barents Sea and near Spitsbergen amounts to 10–20 K/bar. The changes here should therefore be treated with caution and a major improvement in models is required to reproduce current observed conditions and future changes more accurately. Much of the uncertainty in this region stems from large errors at the start of the twenty-first century and all the models project a decrease of the sea-ice from those initial conditions. To the north of the Bering Strait, the inter-model standard deviation is around 10 K/bar.

We note that the individual model results indicate that negative changes are found in regions of open water near the models’ twentieth century sea ice edge, while large positive changes are found where the sea-ice disappears in the twenty-first century. This appears to be a consistent signal, and is evident near the Bering Strait, in the extreme Northwest Atlantic (including Hudson Bay, where large positive changes are projected with low inter-model scatter) and in the northwest Pacific as well.

The changes over the open ocean are largely negative. The most dramatic decreases are projected for the Norwegian Sea and at the entrance to the Labrador Sea south of Greenland, where the negative changes amount to more than 6 K/bar (Fig. 7). This minimum is collocated with the smallest ocean warming in the Northern Hemisphere (Fig. 9). This contributes to a large decrease of the strength of MCAOs south of Greenland as the atmosphere warms more than the sea surface (Fig. 10). There is a large model spread in the Labrador Sea due to the twentieth century sea-ice extending too far south in some of the models.

Further south, off the eastern seaboard of the USA, the results also show decreases of around 4–6 K/bar. Despite the strong warming projected over the northeast American continent, these changes are quite small due to the simultaneous strong warming of the SKT along the coastline that is simulated in the AR4 models.

The changes to the MCAO indicator in the Pacific region are smaller than those found in the Atlantic, and the model scatter of the projections is more moderate. Over the Sea of Okhotsk there is hardly any change to the MCAO indicator quantiles. In the northern Bering Sea large positive changes are found (up to 6 K/bar), although this is most likely due to erroneous sea-ice representation in the twentieth century. In the Gulf of Alaska and near Japan, negative changes are found. As most of these changes are accompanied by low inter-model standard deviation (<4 K/bar), they must be considered more robust than some of the changes in parts of the North Atlantic.

6 Discussion

In this study the current representation and future changes of marine cold-air outbreaks (MCAOs) have been assessed. As a result of the diversity of factors which influence them, the projected changes of MCAOs vary greatly between different regions. This will contribute to regional differences in future changes of weather associated with MCAOs.

The changes to the strength of MCAOs have important implications for changes in the activity of unresolved mesoscale weather systems. For the Atlantic region, there are two main regions of large decrease, the Labrador Sea and Norwegian Sea. Over the marginal ice zone to the north of the Nordic seas large increases are projected.

The changes in the Labrador Sea region over the twenty-first century are dominated by the influence of the weakening AMOC (Schmittner et al. 2005), which is likely to be the main reason for an anomalous lack of a warming skin temperature (SKT) trend south of Greenland and the relatively weak warming of the southern Norwegian Sea. This weakens MCAOs in the region over the twenty-first century because the SKT remains nearly constant whilst the atmospheric temperature associated with MCAOs increases. The magnitude of the decrease of the 95th percentile of MCAOs is approximately 8°C/bar, which will act to weaken associated polar lows and arctic fronts.

A robust signal over the marginal ice zone in the Nordic seas is that the retreat of sea-ice produces drastic increases of the strength of MCAOs in regions of recently exposed ocean and decreases immediately south of the current marginal ice zone. Only three models start out with ‘reasonable’ sea-ice. These are: MPI-ECHAM5, which in fact has too little sea-ice at the end of the 20C3M; MRI-CGCM2.3.2a, with almost no bias; and GFDL-CM2.1, which has too much ice in the southeastern part of the Barents Sea and too little ice near Spitsbergen. They all project decreases of up to 10 K/bar in the region where they start out with open ocean, and substantial increases where they start out with sea-ice. This seems to be a consistent signal in most of the models, even when the initial sea-ice edge is misplaced with respect to late twentieth century climatology.

However, due to the wide range of projected changes and initial conditions there is large uncertainty over the regional detail of the changes in the Nordic seas. Although much of the weather associated with the ice edge, such as arctic fronts, will migrate northwards along with the ice edge, two important observations about the projected changes were noted. Firstly, as the ice edge moves further to the north of the Nordic Seas away from the warming influence of the Gulf Stream, peak MCAO values will, as well as migrating, decrease in strength. This will act to reduce the intensification rate of mesoscale weather phenomena in the region of maximum MCAOs as it moves north. However, the second observation is that for polar lows, even if intensification rates are reduced, the longer life-time over the expanse of open ocean may contribute to an increased intensity of the lows before they make landfall (Emanuel and Rotunno 1989).

In the Pacific region the most prominent location for MCAOs is along the western boundary. This shows only small decreases in strength over the twenty-first century compared to the decreases projected over parts of the Atlantic. This is mainly due to large SKT increases, particularly in the region of the Sea of Japan. Although the slight decrease in the strength of MCAOs suggests a reduction of the severity of associated mesoscale weather systems, the coincident warmer conditions may to some extent cancel this out due to the possibility of more latent heat release. Therefore the parameters assessed in this study will likely contribute little to changes of mesoscale weather systems in the Japan Sea region over the twenty-first century.

Future changes to severe weather associated with MCAOs are dependent on many factors apart from air–sea temperature difference and therefore this study should only be viewed as a first step in assessing their sensitivity to climate forcing. Other factors important to the development and maintenance of mesoscale weather phenomena include baroclinicity, forcing from upper level troughs and the ocean mixed layer depth. Many of these factors depend on the storm track, which is projected to strengthen over the British Isles in the future (Bengtsson et al. 2006). This is consistent with the findings of Yin (2005), where a poleward migration of the storm track was found. It is not clear how such changes in the mean flow can be related to cold-air outbreaks, which are not necessarily tied to the modes of variability that are associated with storm track variability. Only through resolving mesoscale weather phenomena explicitly will their sensitivity to climate forcing be comprehensively established.

As mentioned in the introduction, MCAOs are important for air–sea interaction. Large fluxes of heat and moisture from the ocean to the atmosphere on the order of 600 W/m2 can occur during MCAOs (Renfrew and Moore 1999). [In a case study of an arctic front in the Norwegian Sea, Grønås and Skeie (1999) found that the simulated sensible heat flux exceeded 1,200 W/m2, but Pagowski and Moore (2001) have shown that heat fluxes in mesoscale models can be exaggerated.] Changes to MCAOs over the Labrador Sea will likely have the greatest impact on ocean circulation since this is a key location for North Atlantic Deep Water formation, which contributes to the AMOC. To first order weaker MCAOs will lead to decreased heat fluxes from the ocean to the atmosphere, which will result in less North Atlantic Deep Water formation and a further weakening of the AMOC. However, weaker mesoscale convective systems associated with the weaker MCAOs will produce less precipitation and act to decrease the freshwater input and may therefore act to strengthen the AMOC. With a large modelling error associated with both of these factors (Moore and Renfrew 2002; Ninomiya et al. 2006), the role of MCAO changes in modifying ocean circulation is still uncertain and remains for future work.

7 Conclusions

The model ensemble has large biases with respect to a 40-year ERA-40 climatology of MCAOs. The most pronounced negative biases are due to a poor representation of sea-ice in the Arctic, particularly over the Barents and Greenland Seas, but also over the Northwest Atlantic. Over open ocean, the models generally overestimate the strength of MCAOs, mainly because the ensemble model atmosphere is too cold at 700 hPa.

In terms of future changes, a robust signal is that the strength of the MCAOs in what is now the marginal ice zone decreases substantially. At the same time, the regions with strong MCAOs move polewards with the retreating sea-ice. Over the open ocean an overall decrease of the MCAO indicator values is found because the atmospheric warming is more pronounced than that of the sea surface. However, there are large regional variations. The largest weakening of MCAOs is around the southern tip of Greenland, where the ocean warming is small as a result of the simulated weakening of the AMOC. Changes over the Pacific are smaller due to a more rapid ocean surface warming, in particular over the Seas of Japan and Okhotsk.



We would like to thank two anonymous reviewers for their thorough and useful comments. Erik Kolstad thanks Burghard Brümmer and Stephen Mobbs for fruitful discussions. Alan Condron is thanked for taking part in the initial planning of the paper. This is publication no. A179 from the Bjerknes Centre for Climate Research. We also acknowledge the modelling groups for making their simulations available for analysis, the Program for Climate Model Diagnosis and Inter-Comparison (PCMDI) for collecting and archiving the CMIP3 model output, and the WCRP’s Working Group on Coupled Modelling (WGCM) for organizing the model data analysis activity. The WCRP CMIP3 multi-model dataset is supported by the Office of Science, U.S. Department of Energy.


  1. Arzel O, Fichefet T, Goosse H (2006) Sea ice evolution over the 20th and 21st centuries as simulated by current AOGCMs. Ocean Modelling 12:401–415CrossRefGoogle Scholar
  2. Bengtsson L, Hodges K, Roeckner E (2006) Storm tracks and climate change. J Clim 19:3518–3543CrossRefGoogle Scholar
  3. Bond N, Shapiro M (1991) Polar lows over the Gulf of Alaska in conditions of reverse shear. Mon Weather Rev 119:551–572CrossRefGoogle Scholar
  4. Boyle JS (1986) Synoptic conditions in midlatitudes accompanying cold surges for the months of December 1974 and 1978. Part I: Monthly mean fields and individual events. Mon Weather Rev 114:903–918CrossRefGoogle Scholar
  5. Bracegirdle TJ, Gray SL (2007) A new climatology of the dynamical forcing of polar lows. Int J Climatol (in press)Google Scholar
  6. Bresch JF, Reed RJ, Albright MD (1997) A polar-low development over the Bering Sea: nalysis, numerical simulation, and sensitivity experiments. Mon Weather Rev 125:3109–3130CrossRefGoogle Scholar
  7. Bromwich DH, Fogt RL, Hodges KI, Walsh JE (2007) A tropospheric assessment of the ERA-40, NCEP, and JRA-25 global reanalyses in the polar regions. J Geophys Res 112:D10111. doi:10.1029/2006JD007859 CrossRefGoogle Scholar
  8. Brümmer B (1996) Boundary-layer modification in wintertime cold-air outbreaks from the arctic sea ice. Bound Layer Meteorol 80:109–125CrossRefGoogle Scholar
  9. Businger S (1985) The synoptic climatology of polar low outbreaks. Tellus 37A:419–432Google Scholar
  10. Businger S (1987) The synoptic climatology of polar low outbreaks over the Gulf of Alaska and the Bering Sea. Tellus 39A:307–325CrossRefGoogle Scholar
  11. Businger S, Baik JJ (1991) An arctic hurricane over the Bering Sea. Mon Weather Rev 119:2293–2322CrossRefGoogle Scholar
  12. Businger S, Graziano T, Kaplan M, Rozumalski R (2005) Cold-air cyclogenesis along the Gulf-Stream front: investigation of diabatic impacts on cyclone development, frontal structure, and track. Meteorol Atmos Phys 88:65–90. doi:10.1007/s00703-003-0050-y CrossRefGoogle Scholar
  13. Camargo SJ, Sobel AH, Barnston AG, Emanuel KA (2007) Tropical cyclone genesis potential index in climate models. Tellus 59A:428–443Google Scholar
  14. Claud C, Heinemann G, Raustein E, McMurdie L (2004) Polar low le Cygne: satellite observations and numerical simulations. Quart J Roy Meteorol Soc 130:1075–1102CrossRefGoogle Scholar
  15. Craig G, Gray S (1996) CISK or WISHE as the mechanism for tropical cyclone intensification. J Atmos Sci 53:3528–3540CrossRefGoogle Scholar
  16. Delworth T, Greatbatch R (2000) Multidecadal thermohaline circulation variability driven by atmospheric surface flux forcing. J Clim 13:1481–1495CrossRefGoogle Scholar
  17. Deser C, Walsh J, Timlin M (2000) Arctic sea ice variability in the context of recent atmospheric circulation trends. J Clim 13:617–633CrossRefGoogle Scholar
  18. Dorman CE, Beardsley RC, Dashko NA, Friehe CA, Kheilf D, Cho K, Limeburner R, Varlamov SM (2004) Winter marine atmospheric conditions over the Japan Sea. J Geophys Res 109:C12011CrossRefGoogle Scholar
  19. Douglas M, Fedor L, Shapiro M (1991) Polar low structure over the northern Gulf of Alaska based on research aircraft observations. Mon Weather Rev 119:32–54CrossRefGoogle Scholar
  20. Drüe C, Heinemann G (2001) Airborne investigation of arctic boundary-layer fronts over the marginal ice zone of the Davis Strait. Bound Layer Meteorol 101:261–292CrossRefGoogle Scholar
  21. Emanuel K, Rotunno R (1989) Polar lows as arctic hurricanes. Tellus 41A:1–17Google Scholar
  22. Frierson DMW (2006) Robust increases in midlatitude static stability in simulations of global warming. Geophys Res Lett 33:L24816. doi:10.1029/2006GL027504 CrossRefGoogle Scholar
  23. Fu G, Niino H, Kimura R, Kato T (2004) A polar low over the Japan Sea on 21 January 1997. Part I: Observational analysis. Mon Weather Rev 132:1537–1551CrossRefGoogle Scholar
  24. Gregory JM, Dixon KW, Stouffer RJ, Weaver AJ, Driesschaert E, Eby M, Fichefet T, Hasumi H, Hu A, Jungclaus JH, Kamenkovich IV, Levermann A, Montoya M, Murakami S, Nawrath S, Oka A, Sokolov AP, Thorpe RB (2005) A model intercomparison of changes in the Atlantic thermohaline circulation in response to increasing atmospheric CO2 concentration. Geophys Res Lett 32:L12703CrossRefGoogle Scholar
  25. Griffies SM, Tziperman E (1995) A linear thermohaline oscillator driven by stochastic atmospheric forcing. J Clim 8:2440–2453CrossRefGoogle Scholar
  26. Grønås S, Skeie P (1999) A case study of strong winds at an Arctic front. Tellus 51A:865–879Google Scholar
  27. Grossman RL, Betts AK (1990) Air–sea interaction during an extreme cold air outbreak from the eastern coast of the United States. Mon Weather Rev 118:324–342CrossRefGoogle Scholar
  28. Hakkinen S (1995) Simulated interannual variability of the Greenland sea deep-water formation and its connection to surface forcing. J Geophys Res 100:4761–4770CrossRefGoogle Scholar
  29. Houghton J, Ding Y, Griggs D, Noguer M, van der Linden P, Xiaosu D (2001) Climate change 2001: the scientific basis: contributions of working group I to the third assessment report of the intergovernmental panel on climate change. Cambridge University Press, CambridgeGoogle Scholar
  30. Jagger TH, Elsner JB (2006) Climatology models for extreme hurricane winds near the United States. J Clim 19:3220–3236CrossRefGoogle Scholar
  31. Johannessen OM, Bengtsson L, Miles MW, Kuzmina SI, Semenov VA, Alekseev GV, Nagurnyi AP, Zakharov VF, Bobylev LP, Pettersson LH, Hasselmann K, Cattle AP (2004) Arctic climate change: observed and modelled temperature and sea-ice variability. Tellus 56A:328–341Google Scholar
  32. Kolstad EW (2006) A new climatology of favourable conditions for reverse-shear polar lows. Tellus 58A:344–354Google Scholar
  33. Konrad CE, Colucci SJ (1989) An examination of extreme cold air outbreaks over eastern North America. Mon Weather Rev 117:2687–2700CrossRefGoogle Scholar
  34. Kristjansson J (1990) Model simulations of an intense meso-beta scale cyclone. The role of condensation parameterization. Tellus 42A:78–91Google Scholar
  35. Langland RH, Tag PM, Fett RW (1989) An ice breeze mechanism for boundary-layer jets. Bound Layer Meteorol 48:177–195CrossRefGoogle Scholar
  36. Levermann A, Griesel A, Hofmann M, Montoya M, Rahmstorf S (2005) Dynamic sea level changes following changes in the thermohaline circulation. Clim Dyn 24:347–354CrossRefGoogle Scholar
  37. Liu A, Moore G, Tsuboki K, Renfrew I (2006) The effect of the sea-ice zone on the development of boundary-layer roll clouds during cold air outbreaks. Bound Layer Meteorol 118:557–581CrossRefGoogle Scholar
  38. Lystad M (1986) Polar lows project; final report: polar lows in the Norwegian, Greenland and Barents Sea. Technical report, The Norwegian Meteorological Institute (DNMI), Oslo, NorwayGoogle Scholar
  39. Mailhot J, Hanley D, Bilodeau B, Hertzman O (1996) A numerical case study of a polar low in the Labrador Sea. Tellus 48A:383–402Google Scholar
  40. McCabe GJ, Clark MP, Serreze MC (2001) Trends in Northern Hemisphere surface cyclone frequency and intensity. J Clim 14:2763–2768CrossRefGoogle Scholar
  41. Moore G, Renfrew I (2002) An assessment of the surface turbulent heat fluxes from the NCEP-NCAR reanalysis over the western boundary currents. J Clim 15:2020–2037CrossRefGoogle Scholar
  42. Ninomiya K, Nishimura T, Suzuki T, Matsumura S (2006) Polar-air outbreak and air-mass transformation over the east coast of Asia as simulated by an AGCM. J Meteorol Soc Japan 84:47–68CrossRefGoogle Scholar
  43. Noer G, Ovhed M (2003) Forecasting of polar lows in the Norwegian and the Barents Sea. In: Proceedings of the 9th meeting of the EGS Polar Lows Working Group, Cambridge, UKGoogle Scholar
  44. Nordeng T (1990) A model-based diagnostic study of the development and maintenance mechanisms of two polar lows. Tellus 42A:92–108Google Scholar
  45. Nordeng T, Rasmussen E (1992) A most beautiful polar low—a case-study of a polar low development in the Bear-Island region. Tellus 44A:81–99Google Scholar
  46. Økland H (1998) Modification of frontal circulations by surface heat flux. Tellus 50A:211–218Google Scholar
  47. Pagowski M, Moore G (2001) A numerical study of an extreme cold-air outbreak over the Labrador Sea: sea ice, air–sea interaction, and development of polar lows. Mon Weather Rev 129:47–72CrossRefGoogle Scholar
  48. Pickart RS, Torres DJ, Clarke RA (2002) Hydrography of the Labrador Sea during active convection. J Phys Oceanogr 32:428–457CrossRefGoogle Scholar
  49. Rasmussen E, Turner J (2003) Polar lows: mesoscale weather systems at high latitudes. Cambridge University Press, CambridgeGoogle Scholar
  50. Renfrew IA, Moore G (1999) An extreme cold-air outbreak over the Labrador Sea: roll vortices and air–sea interaction. Mon Weather Rev 127:2379–2394CrossRefGoogle Scholar
  51. Renfrew I, Moore G, Holt T, Chang S, Guest P (1999) Mesoscale forecasting during a field program: meteorological support of the Labrador Sea deep convection experiment. Bull Am Meteorol Soc 80:605–620CrossRefGoogle Scholar
  52. Ronski S, Budeus G (2005) Time series of winter convection in the Greenland Sea. J Geophys Res 110:C04015. doi:10.1029/2004JC002318 CrossRefGoogle Scholar
  53. Sardie JM, Warner TT (1983) On the mechanism for the development of polar lows. J Atmos Sci 40:869–881CrossRefGoogle Scholar
  54. Schmittner A, Latif M, Schneider B (2005) Model projections of the North Atlantic thermohaline circulation for the 21st century assessed by observations. Geophys Res Lett 32:L23710CrossRefGoogle Scholar
  55. Serreze MC, Maslanik JA, Scambos TA, Fetterer F, J Stroeve KK, Fowler C, Drobot S, Barry RG, Haran TM (2002) A record minimum arctic sea ice extent and area in 2002. Geophys Res Lett 30:2002GL016406Google Scholar
  56. Shapiro M, Fedor L (1989) A case study of an ice edge boundary layer front and polar low development over the Norwegian and Barents Seas. In: Polar and Arctic Lows, A Deepak, Hampton, VA, USAGoogle Scholar
  57. Shapiro M, Hampel T, Fedor L (1989) Research aircraft observations of an arctic front over the Barents Sea. In: Polar and Arctic Lows, A Deepak, Hampton, VA, USAGoogle Scholar
  58. Sorteberg A, Kvingedal B (2006) Atmospheric forcing on the Barents Sea winter ice extent. J Clim 19:4772–4784CrossRefGoogle Scholar
  59. Stephenson DB, Pavan V, Collins M, Junge MM, Quadrelli R, Groups PCM (2006) North atlantic oscillation response to transient greenhouse gas forcing and the impact on European winter climate: a CMIP2 multi-model assessment. Clim Dyn 27:401–420. doi:10.1007/s00382-006-0140-x CrossRefGoogle Scholar
  60. Stocker TF (2002) North-south connections. Science 297:1814–1815CrossRefGoogle Scholar
  61. Uppala SM, Kallberg PW, Simmons AJ, Andrae U, Bechtold VD, Fiorino M, Gibson JK, Haseler J, Hernandez A, Kelly GA, Li X, Onogi K, Saarinen S, Sokka N, Allan RP, Andersson E, Arpe K, Balmaseda MA, Beljaars ACM, Van De Berg L, Bidlot J, Bormann N, Caires S, Chevallier F, Dethof A, Dragosavac M, Fisher M, Fuentes M, Hagemann S, Holm E, Hoskins BJ, Isaksen L, Janssen PAEM, Jenne R, McNally AP, Mahfouf JF, Morcrette JJ, Rayner NA, Saunders RW, Simon P, Sterl A, Trenberth KE, Untch A, Vasiljevic D, Viterbo P, Woollen J (2005) The ERA-40 re-analysis. Q J R Meteorol Soc 131:2961–3012CrossRefGoogle Scholar
  62. Vavrus S, Walsh JE, Chapman WL, Portis D (2006) The behavior of extreme cold air outbreaks under greenhouse warming. Int J Climatol 26:1133–1147CrossRefGoogle Scholar
  63. Wood R, Keen A, Mitchell J, Gregory J (1999) Changing spatial structure of the thermohaline circulation in response to atmospheric CO2 forcing in a climate model. Nature 399:572–575CrossRefGoogle Scholar
  64. Yin JH (2005) A consistent poleward shift of the storm tracks in simulations of 21st century climate. Geophys Res Lett 32:L18701CrossRefGoogle Scholar
  65. Yoshida A, Asuma Y (2004) Structures and environment of explosively developing extratropical cyclones in the northwestern Pacific region. Mon Weather Rev 132:1121–1142CrossRefGoogle Scholar
  66. Zhang XD, Walsh JE (2006) Toward a seasonally ice-covered Arctic Ocean: scenarios from the IPCC AR4 model simulations. J Clim 19:1730–1747CrossRefGoogle Scholar
  67. Zhang XD, Ikeda M, Walsh JE (2003) Arctic sea ice and freshwater changes driven by the atmospheric leading mode in a coupled sea ice–ocean model. J Clim 16:2159–2177CrossRefGoogle Scholar
  68. Zhang XD, Walsh JE, Zhang J, Bhatt US, Ikeda M (2004) Climatology and interannual variability of arctic cyclone activity: 1948–2002. J Clim 17:2300–2317CrossRefGoogle Scholar

Copyright information

© Springer-Verlag 2007

Authors and Affiliations

  1. 1.Bjerknes Centre for Climate ResearchBergenNorway
  2. 2.British Antarctic SurveyCambridgeUK

Personalised recommendations