SEARCH

SEARCH BY CITATION

Keywords:

  • seasonal forecasts;
  • climate forecasts;
  • ECMWF model

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Experimental design and analysis methods
  5. 3. Sensitivity of the atmosphere to the 2007/2008 ice anomaly in atmospheric simulations
  6. 4. Coupled versus uncoupled response
  7. 5. Ice forcing versus SST forcing
  8. 6. Implications and conclusions
  9. References

The impact on the atmospheric circulation of the unprecedented Arctic sea-ice anomalies during the summers 2007 and 2008 is evaluated using the atmospheric model of the ECMWF operational seasonal forecasting system. Results show that the ice anomaly had a significant impact on the atmospheric circulation over the Euro-Atlantic Sector, characterized by a high pressure over the Arctic (Greenland) and low pressure centres over Western Europe and Northwest America. The impact is similar for the two consecutive years, and it is consistent with the observed atmospheric anomalies. Results also show that the impact of the ice is strongly dependent on the underlying sea surface temperature. Results from partial coupling experiments indicate that the sea surface temperature over the Northwest Atlantic strongly affects the mean state of atmospheric circulation over the Euro-Atlantic sector (first-order impact), and conditions the response of the atmosphere to a given ice anomaly (second-order impact). The implications of these results for seasonal and long-term predictions are discussed. Copyright © 2010 Royal Meteorological Society


1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Experimental design and analysis methods
  5. 3. Sensitivity of the atmosphere to the 2007/2008 ice anomaly in atmospheric simulations
  6. 4. Coupled versus uncoupled response
  7. 5. Ice forcing versus SST forcing
  8. 6. Implications and conclusions
  9. References

Arctic ice extent reached unprecedented minima during the summers of 2007 and 2008 (Comiso et al., 2008; Stroeve et al., 2008), causing concern about the possible acceleration of the long-term trend of declining Arctic sea ice. Several observational and modelling studies have addressed the reasons for this long-term decline (Rigor and Wallace, 2004; Serreze and Francis, 2006; Ogi and Wallace, 2007, among others). There is growing consensus that the anomalous Artic ice extent during the 2007 and 2008 summers were largely the result of atmospheric forcing, although the nature of the forcing (dynamical or thermodynamic) may vary (Deser and Teng, 2008, and references therein provide detailed discussion). Evidence exists that anomalous atmospheric conditions had been the driving force for the dramatic anomalies in summer 2007 (Slingo and Sutton, 2007; L'Heureux et al., 2008; Kay et al., 2008; Schweiger et al., 2008; Zhang et al., 2008), together with a preconditioning resulting from warmer ocean conditions (Polyakov et al., 2007).

A complementary question is whether the 2007 and 2008 ice anomalies had any impact on the atmospheric circulation. The answer to this question is especially relevant in the context of a warming climate, when large decreases in the summer ice are expected to become more common, and may also be particularly important for the design of a seasonal forecasting system.

This paper evaluates the impact of the 2007 and 2008 observed ice anomalies in the ECMWF model used for the operational seasonal forecasts. The experiments aim at answering three main questions:

  • (i)
    Did the Arctic ice anomalies in the summer 2007 and 2008 influence the atmospheric circulation?
  • (ii)
    Can we trust state-of-the art coupled climate models to represent the impact of sea-ice anomalies on the atmospheric circulation?
  • (iii)
    How does the atmospheric response to a given ice anomaly depend on the underlying SST?

The impact of ice anomalies on the atmospheric circulation has been addressed in previous modelling studies, but a clear picture still fails to emerge. Deser et al.(2004) and Magnusdottir et al.(2004) used the Community Climate Model (CCM3) to investigate the equilibrium response of the atmospheric wintertime circulation to the forcing resulting from sea surface temperature (SST) and sea-ice extent patterns. These forcing patterns were representative of the observed trends in the second half of the twentieth century. It was found that the atmospheric circulation responded linearly (nonlinearly) to the amplitude (polarity) of the forcing. The response to the observed ice anomaly resembled the negative phase of the North Atlantic Oscillation (NAO) and/or Arctic Oscillation (AO), and this was stronger when the polarity of the forcing SST pattern was reversed. These findings contrast with those of Singarayer et al.(2006), who ran the Hadley Centre Atmospheric Model (HadAM3) with climatological SSTs and observed sea-ice concentrations from 1978 to 2000 to investigate the impact of sea ice on the atmospheric trends. Their results showed that the response of the atmosphere projected positively on the NAO. Alexander et al.(2004; AL04 in what follows), using an ensemble of integrations of the CCM3.6 model, investigated the impact of the observed 1982–1983 and 1995–1996 winter ice anomalies on the atmospheric circulation. Their results showed a modest but significant response of the atmospheric circulation (about 20 m at 500 mb), which they interpreted as a positive (negative) feedback in the North Pacific (Atlantic) sectors. More recently, Bhatt et al.(2008; BH08 in what follows) investigated the impact of the observed ice anomalies during the summer of 1995 (the lowest ice anomaly prior to 2007 and 2008) using the same model version and similar experimental set-up to that of AL04. They found that the ice anomalies caused higher sea-level pressures (SLPs) and upper-level heights in the North Pacific, accompanied by increased (decreased) precipitation north (south) of the Pacific storm track. Francis et al.(2009), in an observational study, concluded that the summer anomalies in the ice cover are related to the atmospheric circulation of the following autumn and winter. All of these studies point to a substantial influence of sea-ice variability on Northern Hemisphere (NH) circulation, although the variety of experimental designs makes the results difficult to compare. The comparison of results is particularly difficult if the atmospheric response is indeed nonlinear, as found by Deser et al.(2004).

This paper is organized as follows: section 2 describes the experimental design and analysis methods. Section 3 discusses the results of the experiments where the atmosphere model, forced by the observed SST for 2007 and 2008, was run with observed and climatological Arctic ice cover. Section 4 shows the response of the coupled ocean–atmosphere model to the ice anomalies, and discusses the influence of the midlatitude SST errors on the results. Section 5 presents the sensitivity of the atmosphere model to the 2007 ice anomalies under a wide range of SST anomalies. The implications of the results are discussed in section 6.

2. Experimental design and analysis methods

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Experimental design and analysis methods
  5. 3. Sensitivity of the atmosphere to the 2007/2008 ice anomaly in atmospheric simulations
  6. 4. Coupled versus uncoupled response
  7. 5. Ice forcing versus SST forcing
  8. 6. Implications and conclusions
  9. References

All the experiments discussed in this paper have been conducted with the atmospheric component of the operational ECMWF seasonal forecasting system S3. This uses a (TL)159 spectral truncation (approximately 125 km horizontal resolution), and 62 levels in the vertical, the highest of which reaches 5 hPa. More details on S3 specifications and performance are given in Anderson et al.(2007) and Molteni et al.(2007).

Table I presents a summary of the experiments conducted to evaluate the impact of the 2007 and 2008 observed ice anomalies in the NH atmospheric circulation. Pairs of experiments with prescribed observed ice (OI) and with climatological ice (CI) conditions were conducted under a three different surface boundary settings, where the SST were

  • (i)
    prescribed from observations of SST (OS),
  • (ii)
    predicted by the ocean–atmosphere coupled model (CS), and
  • (iii)
    resulting from a hybrid partial coupling experiment (PS) as described below.
Table I. Summary of experiments conducted to assess the impact of the 2007 and 2008 ice anomaly in the atmospheric circulation. Each experiment consists of 40 ensemble members integrated from May to August of 2007 and 2008.
ExperimentIce forcingSST boundary forcing
OI_OSObserved IcePrescribed Observed SST (daily)
CI_OSClimatological Ice 
OI_CSObserved IceFrom Ocean–Atmosphere coupled model (CGCM)
CI_CSClimatological Ice 
OI_PSObserved IceFrom CGCM but for the Northwest Atlantic (30–60°N, 80–30°W)
CI_PSClimatological Ice 

The experiments, each comprising two initial dates (2007 and 2008), were integrated for the period May to August, and consisted of 40-member ensembles for each date.

The first set of experiments (OS) was conduced by forcing the atmospheric model with observed daily SSTs for 2007 and 2008. One of the experiments (OI_OS) was forced by daily values of the analyzed ice cover, while the other (CI_OS) is forced by daily values of the climatological ice cover (as used in the ECMWF seasonal forecasts). In the areas free of ice, the corresponding value of the observed SST was prescribed. Comparison between these two experiments gives information about the impact of the ice anomalies on the atmospheric circulation. They also represent an upper limit for the sea-ice-induced predictability at seasonal time-scales, since they assume perfect predictability of both SST and ice conditions.

The second pair of experiments were conducted in coupled mode (CS, for Coupled SST), i.e. the SST evolution was predicted by the ocean–atmosphere coupled model (the same coupled model as is used for the operational ECMWF seasonal forecasting system S3). As before, one experiment (OI_CS) was conducted by prescribing the observed ice extent for 2007 and 2008, while the other was conducted using climatological ice cover (CI_CS). In the areas free of ice, the SST value was provided by the ocean model. Comparison between this pair of coupled seasonal forecast experiments can assess the impact of perfectly predicted ice conditions in models where the SST evolution is not perfect.

A third set of experiments (PS) was conducted in partial coupling mode, where the ocean and atmosphere were freely coupled everywhere except for a region by the Northwest Atlantic (30–60°N, 8–30°W), where the ocean model was strongly relaxed to the observed SST. These experiments were conducted to explore the role that the western boundary currents have in the atmospheric circulation and, in particular, in the atmospheric response to anomalous ice forcing.

The daily values of ice cover and SST were derived from the OI_v2 dataset (Reynolds et al., 2002) by daily interpolation of weekly values. The climatological values of ice cover are those used in the S3 seasonal forecasting system, which cover the period 1981–2001 and were ultimately derived from the ERA-40 sea-ice climatology (Uppala et al., 2005). The observed values of ice cover were prescribed in the ocean model gridpoints as binary values (ice or no ice): an ocean model grid point will be considered completely free of ice unless the observed ice concentration exceeded the 55% threshold value, where it would be consider 100% ice.

The ensemble of forecasts was created by applying perturbations to the SST during the first month of the integrations. The SST perturbations were those used to generate the ensemble in the S3 seasonal forecasting system, with size and spatial patterns representative of the uncertainty in SST analysis (Vialard et al., 2005). Each SST perturbation is applied with a plus or minus sign, so as to guarantee symmetry in the final ensemble.

Table II presents the experiments conducted under a different set-up, designed to test if the atmospheric response to a given ice anomaly depends on the SST forcing. In other words, these experiments were designed to test the linearity of the atmospheric response to the ice anomaly. Two additional experiments (I07_VS) and (CI_VS) were conducted. Each experiment consisted of 20 sets of five ensemble members each (amounting to a total ensemble size of 100), where the atmospheric model was integrated forward for four months (May to August) with prescribed daily values of observed SST. For each of the 20 sets, the SST and atmospheric initial conditions were taken from individual years of the 1987–2006 period. Each year was sampled five times, adding small perturbations to the SST as in the experiments in Table I. In experiment I07_VS, the atmospheric model was forced by the ice conditions of May–September 2007, and in experiment CI_VS climatological ice conditions were used. The degree to which the differences between experiments I07_VS and CI_VS depart from the differences between experiment OI_OS and CI_OS (2007 case), is a measure of the of the nonlinearity of the atmospheric response to the 2007 ice anomaly.

Table II. Experiments conducted to assess the impact of the 2007 ice anomaly under a variety of observed SST evolution.
ExperimentIce ForcingSST boundary forcing
  1. Each experiment consisted of 100 ensemble members, and was integrated for four months (May–August).

  2. The ensemble was created by sampling the SST evolution from the 20-year period 1987–2006.

  3. Each year is sampled five times by adding random initial perturbations.

I07_VSObserved 2007 icePrescribed Observed SST (daily). Individual years from 1987 to 2006.
CI_VSClimatological ice 

3. Sensitivity of the atmosphere to the 2007/2008 ice anomaly in atmospheric simulations

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Experimental design and analysis methods
  5. 3. Sensitivity of the atmosphere to the 2007/2008 ice anomaly in atmospheric simulations
  6. 4. Coupled versus uncoupled response
  7. 5. Ice forcing versus SST forcing
  8. 6. Implications and conclusions
  9. References

Figure 1 shows the difference in Arctic sea-ice cover between the experiments OI_OS and CI_OS for 2007 and 2008. In both cases the extent of the ice anomalies show large-scale features by July, although in 2008 the July anomalies remained confined to the Alaskan area, only reaching the Eurasian region by August.

thumbnail image

Figure 1. Difference in sea-ice concentration between the experiments with prescribed and climatological ice (OI_OS minus CI_OS) during July–August of 2007 (upper figures) and 2008 (lower figures).

Download figure to PowerPoint

Figure 2 shows the difference in the total heat flux between experiments OI_OS and CI_OS in 2007, which is representative of the surface forcing resulting from the ice anomaly. Results are shown only for 2007, but an equivalent picture emerges for 2008. Over the ice-free areas, the net heat flux going into the ocean exceeds the value of 30 W m−2 due to the increased penetration of solar radiation resulting from the reduced albedo. Kay et al.(2008) estimate that an excess of ocean surface heating of about 30 W m−2, sustained over three months, would increase the ocean surface temperature by 2.4 K, or melt about 0.3 m of ice. The value of the total heat flux for these experiments, where both SST and ice extent are prescribed, the excess of heat into the ocean surface is not relevant for the interpretation of the results. However, the values of total heat flux shown in Figure 2 are very similar to those obtained in coupled experiments, where the SSTs are predicted by the model. (We will return to this point later in this section.) The ocean is not only warming due to the increased short-wave radiation, but it is also releasing heat into the atmosphere in the form of sensible and latent heat flux, at a rate of about 30 W m−2 (right panels of Figure 2). In what follows, results and discussion will be for July and August only.

thumbnail image

Figure 2. Differences in the total (left) and non-solar (right) surface heat flux between the experiments with prescribed and climatological ice (OI_OS minus CI_OS) during June–August 2007. The heat flux into the ocean has positive values.

Download figure to PowerPoint

The impact of the ice anomaly on the July–August atmospheric circulation appears in Figure 3, which shows the difference in Z500 ensemble mean between the experiments with observed and climatological ice for 2007 and 2008, for the case where the observed evolution of SST is prescribed. Although there are differences between the two years, the response in both cases is quite consistent, characterized by a positive anomaly over the Arctic, slightly shifted over the western side, and a negative anomaly over Northwestern Europe and Northeastern America. The ensemble mean anomalies are modest in size (values of about 2 dam), but statistically significant. (The ensemble spread over the areas of largest signal is in the range 3–4 dam, and individual ensemble members exhibit anomalies reaching 10 dam). The patterns and values of the response in Z500 are comparable to those obtained by AL04, although this latter study was for winter conditions using different ice anomalies. The response over the Pacific sector varies between years, and it does not resemble the signal found by BH08 except for the significant anomalous high over the North Pacific in 2008.

thumbnail image

Figure 3. Impact of the summer ice anomalies of (a) 2007 and (b) 2008 on the July–August atmospheric circulation, as measured by the ensemble mean difference in Z500 between two experiments in which the atmosphere model is forced by the analyzed ice coverage and by climatological ice respectively (OI_OS minus CI_OS). The experiments, with 40 ensemble members each, were initialized in May and run for the four months forced by observed SST. Units are dam. The thick black contours denote the 90% significance level.

Download figure to PowerPoint

The response pattern resembles the summertime AO, and would match the observational relation found by Ogi and Wallace (2007; their Figure 4(b)). They identified the summertime anticyclonic circulation as being an important forcing on the reduction of Arctic sea ice, arguing that the anomalous southerly wind stress associated with the strong SLP gradients would affect the transport of sea ice, resulting in a reduced ice extent. Kay et al.(2008) point to additional mechanisms linking the anticyclonic circulation with reduced sea-ice extent: enhanced poleward atmospheric heat transport and reduced cloudiness would produce an increase in the downwelling short-wave radiation. Following these lines of argument, by which the anticyclonic Arctic circulation would be responsible for the decline of the Arctic sea ice, the impact of the Arctic ice anomaly shown in Figure 3 would imply a positive feedback between sea ice and atmospheric circulation.

thumbnail image

Figure 4. Observed Z500 anomalies during July–August of (a) 2007 and (b) 2008. The anomalies are computed with respect to the 1979–2001 ERA-40 climatology. Thick black contours represent the values where the anomaly exceeds the value of the interannual standard deviation.

Download figure to PowerPoint

For guidance, the observed Z500 anomalies during July–August for 2007 and 2008 are shown in Figure 4. The anomalies are relative to the 1979–2001 ERA40 climatology. (Figures 3 and 4 are not directly comparable: Figure 3 shows ensemble mean values while Figure 4 shows results from a single realization, and the amplitude of the ensemble-mean anomalies are smaller than that of an individual realization. Besides, the difference OI_OS minus CI_OS in Figure 3 shows only the impact of the ice anomaly in the atmospheric model, while Figure 4 shows the interannual anomalies, which will be affected by factors other than sea ice.) The observed atmospheric anomalies exhibit a consistent Arctic high during 2007 and 2008. The negative centres of action over Northwestern Europe and Northeastern America are also present in both years. Although the resemblance between the observed anomaly in Figure 4 and the model response to the ice anomaly is encouraging, an attribution statement is beyond the scope of this work.

4. Coupled versus uncoupled response

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Experimental design and analysis methods
  5. 3. Sensitivity of the atmosphere to the 2007/2008 ice anomaly in atmospheric simulations
  6. 4. Coupled versus uncoupled response
  7. 5. Ice forcing versus SST forcing
  8. 6. Implications and conclusions
  9. References

The previous section described the response of the atmosphere mode to the ice anomaly in experiments where the SST evolution was prescribed by observations. In this section, we examine the response to the ice anomaly by comparing experiments where the SST evolution is predicted by the coupled ocean–atmosphere model (OI_CP versus CI_CP).

The time evolution and spatial location of surface heat fluxes associated with the ice anomaly in the coupled model is very similar to the uncoupled case shown in Figure 2. The magnitude is also quite similar, the ensemble mean of the differences not exceeding 2 W m−2. Surprisingly, the response to the sea-ice anomalies of the coupled model (Figure 7 later), although significant, is very different from the response of the forced atmospheric model (OI_OS minus CI_OS in Figure 3), despite the surface fluxes associated with the ice anomaly in coupled and uncoupled mode being very similar. One possible explanation is that the coupled experiments, by predicting the SST, could produce a larger ensemble spread, overshadowing the effect to the sea-ice anomalies. However this is not the case, since the spread of the coupled and uncoupled integrations are comparable (not shown). Another possible explanation for the different response resides in the nonlinear nature of the atmosphere. This idea is explored in the rest of the paper.

As the coupled model is not perfect, the SSTs predicted by the coupled model have errors. The differences between model and observed SST for predictions initialized in May 2008 averaged for months 3–4 (July–August) are in Figure 5(a). The largest differences appear in the regions of the western boundary currents. In the coupled model, the coastal SST along the North American coast is too warm, while the mid North Atlantic is too cold. This might arise if (in the coupled model) the SST gradients associated with the Gulf Stream are too diffused, and/or the Gulf Stream path is not correct, resulting in too much heat transported north along the North American coast, and not enough heat transported towards Europe. These errors in SST are manifest as differences in heat fluxes (Figure 5(b)) as a strong dipole, with too much latent heat flux being released into the atmosphere over the areas of high SST: near the coast in the coupled experiment and towards the middle of the Atlantic in the forced case. (In the case described here, the differences in heat fluxes seem to correspond to differences between model and observed SST. In a general case, it is still possible to have differences in heat fluxes even with no differences in SST, simply because in the uncoupled experiments the heat capacity of the ocean is not defined.)

thumbnail image

Figure 5. Difference in the (a) SST and (b) heat flux forcing between the coupled and forced experiments with climatological ice (CI_CS minus CI_OS) for July–August 2008. (The heat flux into the ocean has positive values). The polygon shows the area used in the partial coupling experiments to prescribed the observed SST.

Download figure to PowerPoint

The difference in Z500 between the coupled and forced integrations (averaged for 2007 and 2008) appear in Figure 6(a). The curvature of the Z500 surface is quite different in coupled and forced mode, with much higher values over the Tropics and a sharper decline at midlatitudes. At the poles, however, Z500 in the coupled model has higher values than the forced model.

thumbnail image

Figure 6. (a) Difference in the July–August atmospheric circulation (Z500 for CI_CS minus CI_OS), computed as the average for 2007 and 2008. (b) Impact on July–August Z500 of correcting the SST over the Northwest Atlantic (coupled experiment, CI_CS, minus the experiment with partial coupling, CI_PS). Units are dam. The thick black lines show the 90% significance level.

Download figure to PowerPoint

The misrepresentation of midlatitude SST is a common error in climate models, which cannot represent adequately the western boundary currents due the relatively coarse resolution in the ocean model (about 1 degree). The large heat flux exchange is likely to affect the atmospheric circulation, as found by Minobe et al.(2008). To find out how the errors in the North Atlantic affect the atmospheric circulation, an additional experiment with partial coupling was conducted, where the ocean model SST was strongly relaxed to the observed SST over the Northwest Atlantic and Gulf Stream area (30–60°N, 80–30°W, shown in the polygon in Figure 5). Everywhere else, the model is fully and freely coupled. The partial coupled integrations were initialized in May 2007 and 2008, and consisted of 40 ensemble members. The effect of the Northwest Atlantic SST on the atmospheric circulation, measured as the differences between the ensemble mean of coupled and partial-coupled experiment, is shown in Figure 6(b). By correcting the SST over the Northwest Atlantic area, it is possible to account for most of the differences between coupled and forced integrations over the Euro-Atlantic sector and Greenland area.

The response of the coupled model to the ice anomaly (OI_CS versus CI_CS) for 2008 is shown in Figure 7(a). The response is very different from that of the forced model (OI_OS minus CI_OS in Figure 3(b)), being almost out of phase over the Arctic and Euro-Atlantic sector. If the response to a given ice anomaly is flow-dependent, the different mean state in the coupled and forced mode will lead to a different response to the anomalous ice forcing. This hypothesis is tested by investigating the effect of the ice anomaly in the partial-coupling experiment. The sensitivity to the 2008 ice anomaly in the partial-coupling experiment (OI_PS versus CI_PS) appears in Figure 7(b). By correcting the values of SST over the Northwest Atlantic, the atmospheric response to the 2008 ice anomaly gets closer to that of the forced model, with high values of Z500 over the Arctic, and a low over Northwestern Europe. However, the results for 2007 are not so striking. The impact of the partial coupling in the mean atmospheric circulation for 2007 and 2008 is similar. However, in 2007, the partial coupling was insufficient to reproduce the response to the 2007 ice anomaly from the experiments with observed SST (OI_OS versus CI_OS, shown in Figure 3(a)). The different response in 2007 between coupled and uncoupled experiments may still be related to errors in the coupled model SST in other regions, such as errors in tropical regions or in the North Pacific, which are not easily cancelled out by the correcting the SST over the Northwest Atlantic.

thumbnail image

Figure 7. Atmospheric response (July–August Z500 for observed ice minus climatological ice) to the 2008 ice anomaly (a) in the coupled model and (b) in the experiment where the Northwest Atlantic SSTs have been corrected. Units are dam. The thick black lines show the 90% significance level.

Download figure to PowerPoint

5. Ice forcing versus SST forcing

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Experimental design and analysis methods
  5. 3. Sensitivity of the atmosphere to the 2007/2008 ice anomaly in atmospheric simulations
  6. 4. Coupled versus uncoupled response
  7. 5. Ice forcing versus SST forcing
  8. 6. Implications and conclusions
  9. References

The response of the atmosphere to a given ice anomaly may well be nonlinear, as discussed in Deser et al.(2004), and suggested by the results from coupled and partial coupled experiments presented above. The nonlinearity could also be one of the reasons for the differences between results from the forced experiments presented here (OI_OS versus CI_OS, in Figure 3) and those found by BH08. An additional pair of experiments (I07_VS and CI_VS) was conducted to explore response of the atmosphere to the 2007 ice anomaly under a variety of SST conditions (Table II). In these experiments, the underlying SST samples the interannual variability from 20 different years for the period 1987–2006.

By comparing the difference between experiments I07_VS and CI_VS, it is possible to assess the impact of the ice anomaly in 2007 under different realizations of SST forcing. If the atmospheric response were linear, the influence of the SST will be cancelled out, and results should be similar to those obtained when the SSTs are prescribed from 2007 data (i.e. differences between OI_OS and CI_OS, in Figure 3(a)). As it turns out, the ensemble mean difference between I07_VS and CI_VS (Figure 8) is almost opposite to that shown in Figure 3(a). The pattern of the ensemble mean difference between I07_VS and CI_VS in Figure 8 resembles the results obtained by BH08, in that that there is a prominent high over the central Pacific and a low over the Arctic. In both cases (OI_OS minus CI_OS in Figure 3(a) and I07_VS minus CI_VS in Figure 8), the ensemble mean differences are highly significant. This apparent contradiction can be explained if the response of the atmosphere model to the ice anomaly is nonlinear and sensitive to the underlying SST forcing. To test this possibility a more sophisticated statistical analysis is needed, as discussed below.

thumbnail image

Figure 8. Atmospheric response (Z500) to the 2007 ice anomaly for ensemble of integrations sampling SST from the years 1987–2006 (I07_AV minus CI_VS). The thick black lines show the 90% significance level.

Download figure to PowerPoint

5.1. Nonlinearity of the atmospheric response to anomalous ice forcing

Consider the null hypothesis that the atmospheric response to the 2007 sea-ice anomaly is linearly superimposed on the response to SST anomalies. In that case, the response to 2007 ice anomaly under a variety of SST forcing (I07_VS minus CI_VS, Figure 8) and the response to 2007 ice anomaly with 2007 SST forcing (OI_OS versus CI_OS, Figure 3(a)) would not be statistically significant. Therefore, asserting that the atmospheric response is nonlinear is equivalent to rejecting the hypothesis that the ensemble means shown in Figures 8 and 3(a) originated from the same statistical distribution. To address this issue, we need to compare the ensemble-mean response to the 2007 ice anomaly with 2007 observed SST with a distribution of different realizations of a 40-member-mean response, where the members are extracted from the I07_VS and CI_VS experiments.

As a first step, in order to reduce the dimensionality of the problem, we computed the first two EOFs of the monthly-mean anomalies of 500 hPa height in July and August from the combined experiments I07_VS and CI_VS. These EOFs appear in Figures 9(a) and (b), and explain respectively 12.1% and 9.1% of the total variance of monthly means. The first EOF resembles the negative phase of the so-called summer Arctic Oscillation (AO): it has an annular structure, with a high over the whole Arctic, and lows over the Central Pacific, Northeast America and Northwest Europe. The second EOF has a nodal line over the Arctic, with anomalies on the Canadian/Greenland side in phase opposition to the anomalies on the Pacific side. The anomalies over the Northeast Pacific and Northwest Atlantic are of opposite sign to each other.

thumbnail image

Figure 9. (a) and (b) show the first two EOFs of the interannual anomalies of the model Z500 used to reduce the dimension of the problem. (c) shows the projection in the PC1/PC2 space of the RMEAN_40_VS distribution (contours) and the mean of RMEAN_S07 (intersection of dotted lines).

Download figure to PowerPoint

We construct 100 realizations of the response to the 2007 ice anomaly in PC space with SST from different years; each realization is the difference between PCs of pairs drawn from I07_VS and CI_VS respectively. Each pair is built by ensemble members with the same date and initial perturbation. We refer to this sample as RES_VS. We can also use the same EOFs to project height fields from the OI_OS and CI_OS experiments with 2007 SST, and compute 40 realizations of the response in PC space with the 2007 SST; this second set is named RES_S07.

The specific question is whether the 40-member ensemble mean of RES_S07 (Figure 3(a)) could have been obtained by subsampling, with only 40 ensemble members, the distribution of RES_VS. To proceed to the statistical test, we performed the following steps:

  • 1000 subsamples of 40-members each (RES_VS_40) were created by selecting, in a quasi-random way, elements from the RES_VS sample. Specifically, we randomly selected two out of five members for each of the 20 years from 1987 to 2006. In this way, the ensemble-mean properties of the subsamples only differ because of internal atmospheric variability, since the mean (and the interannual variability) of SST is the same in each sample.

  • The ensemble mean of each RES_VS_40 subsample is then computed, thus creating a 1000-element sample for the ensemble-mean response in PC space, estimated with 40 ensemble members (RMEAN_VS). Similarly, we computed the ensemble-mean of the 40-member RES_S07 dataset (RMEAN_S07), which originates from the same ice anomaly and the same ensemble size, but using 2007 SST.

  • Finally, we estimated the probability density function (PDF) of the PCs in the RMEAN_VS dataset, and estimated the probability that RMEAN_S07 belongs to the distribution of the RMEAN_VS realizations in PC space.

Figure 9(c) shows two-dimensional PDF of the RMEAN_VS response in the PC1–PC2 plane (i.e. the projection of the 40-member response on EOFs 1 and 2 respectively), computed using a Gaussian kernel estimator (e.g. Silverman, 1986). In the figures, the values of PC1 and PC2 have been normalized by the standard deviation of the monthly-mean anomalies. The numerical values of the mean and standard deviations of PC1 and PC2 appear in Table III. The RMEAN_VS data, i.e. the mean response of the atmosphere to the 2007 ice anomaly with time-varying SST, predominantly project on the negative phase of EOF-1, with the mean value of PC1 = − 0.150 and standard deviation of 0.138. The projection on EOF-2 is weaker (PC2 mean = 0.071; PC2 standard deviation = 0.118). The RMEAN_S07 response (i.e. the response of the atmosphere to the 2007 ice anomaly with underlying values of SST from 2007) is represented in Figure 9(c) by the intersection of the dotted lines. It projects positively onto EOF1, with a value of PC1 = 0.142, and a value of PC2 = 0.052. Finally, we computed what proportion of the 40-member samples with time-varying SST deviates from the ‘grand’ mean more than the 40-member ensemble with 2007 SST: the result is 2.3% along PC1, and 86.8% along PC2.

Table III. Summary of statistics used to test the significance of the influence of SST on the atmospheric response to the 2007 ice anomaly. See text for explanation of naming convention.
 RMEAN_VS_40RMEAN_S07Probability (null hypothesis)
  1. The values of the mean and standard deviation are normalized by the value of the standard deviations of the interannual anomalies.

  2. The null hypothesis is that the mean response with 2007 SST (RMEAN_S07) belongs to the distribution variable-SST responses (RMEAN_VS_40).

 MeanStdv  
PC1− 0.1500.1380.1422.3%
PC20.0710.1180.05286.8%

So, it can be concluded that, while there is no significant difference along PC2, the response along PC1 (corresponding to a summer annular mode) is significantly different with 97.7% confidence; i.e. the annular-mode response to the sea-ice anomaly appears to be SST-dependent. This finding is consistent with the winter-case results of Deser et al.(2004), who found that SST forcing could change the polarity of the ice-induced atmospheric annular mode.

6. Implications and conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Experimental design and analysis methods
  5. 3. Sensitivity of the atmosphere to the 2007/2008 ice anomaly in atmospheric simulations
  6. 4. Coupled versus uncoupled response
  7. 5. Ice forcing versus SST forcing
  8. 6. Implications and conclusions
  9. References

Sensitivity experiments conducted by forcing the ECMWF atmospheric model indicate that the ice anomalies in 2007 and 2008 had a significant impact on the atmospheric circulation over the Euro-Atlantic sector, characterized by a high over the Arctic and low centres over Western Europe and Northwest America. The response projects into the summer Arctic Oscillation, consistent with the observational relationship found by Ogi and Wallace (2007). In their study, they hypothesize that the observed statistical relationship was indicative of one-way coupling, with the anticyclonic circulation reducing the Arctic ice extent by the way of Ekman drift in the marginal seas. Based on the results from Bhatt et al.(2008), they discarded the existence of any positive feedback. However, the results presented here, where the AO response is a consequence of the ice anomaly, would suggest the possibility of a positive feedback between the atmospheric AO and the ice anomaly, the mechanisms of which need further investigation.

The sensitivity of the atmospheric circulation to the ice anomaly is quite different when the ECMWF coupled ocean–atmosphere model is used. Further numerical experimentation indicates that the response of the atmosphere to a given ice anomaly is largely conditioned by the background mean state of the ocean–atmosphere system. Results indicate that the SST in the Northwest Atlantic influences both the mean atmospheric circulation and its sensitivity to the ice anomalies.

The nonlinear nature of the atmospheric response to the ice anomaly has been explored by conducting sensitivity experiments under a variety of SST conditions. Results indicate that, while the atmospheric response to the ice anomaly projects mainly in the AO mode, the polarity of the response is conditioned by the underlying SST. Such a strongly nonlinear response implies that experiments in which the atmospheric sensitivity to ice concentration is estimated using climatological or idealised SST distributions may not be relevant to assess the impact of sea-ice anomalies in specific years. Specifically, conclusions about the existence of a positive or negative feedback from sea ice onto the atmospheric circulation may only be valid if the co-existing SST anomalies are correctly represented in numerical experiments.

The results presented here suggest that the skill of the current ECMWF seasonal forecast system over the Euro-Atlantic sector may be limited by the deficient representation of the ice anomalies and the midlatitude SST. The degree of predictability of the ice anomalies at seasonal time-scales needs to be clearly established, especially in the context of a warming climate. If the large ice anomalies observed in recent years are predictable, seasonal forecasting systems should include accurate initialization and modelling of the sea-ice evolution in order to obtain reliable predictions of extratropical summer conditions. Even if the sea-ice anomalies were not predictable, an ensemble forecasting system should still represent the possible range of ice anomalies, instead of just prescribing climatological or persistent sea-ice conditions.

The results on SST sensitivity discussed here have far-reaching consequences, since they imply that accurate seasonal to decadal predictions and climate projections require an accurate representation of the midlatitude SST gradients associated with western boundary currents. These are difficult to represent in the current generation of coupled models used for climate predictions. If no empirical corrections are to be used, improvements of the SST gradients associated with the western boundary currents are likely to be achieved only by increasing the resolution of the ocean and atmospheric models.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Experimental design and analysis methods
  5. 3. Sensitivity of the atmosphere to the 2007/2008 ice anomaly in atmospheric simulations
  6. 4. Coupled versus uncoupled response
  7. 5. Ice forcing versus SST forcing
  8. 6. Implications and conclusions
  9. References
  • Alexander MA, Bhatt US, Walsh JE, Timlin M, Miller JS. 2004. The atmospheric response to realistic Arctic sea ice anomalies in an AGCM during winter. J. Climate 17: 890905.
  • Anderson DLT, Stockdale T, Balmaseda MA, Ferranti L, Vitart F, Molteni F, Doblas-Reyes F, Mogensen K, Vidard A. 2007. ‘Seasonal Forecast System 3’. ECMWF Newsletter 110.
  • Bhatt US, Alexander MA, Deser C, Walsh JE, Miller JS, Timlin M, Scott JD, Tomas R. 2008. The atmospheric response to realistic reduced summer Arctic sea ice anomalies. In Arctic Sea Ice Decline: Observations, Projections, Mechanisms, and Implications, DeWeaverET, BitzCM, TremblayLB. (eds.) Geophys. Monogr. Series 180: 91110. AGU: Washington DC.
  • Comiso JC, Parkinson CL, Gersten R, Stock L. 2008. Accelerated decline in the Arctic sea ice cover. Geophys. Res. Lett. 35: L01703, DOI: 10.1029/2007GL031972.
  • Deser C, Teng H. 2008. Recent trends in Arctic sea ice and the evolving role of atmospheric circulation forcing, 1979–2007. In Arctic Sea Ice Decline: Observations, Projections, Mechanisms, and Implications, DeWeaverET, BitzCM, TremblayLB. (eds.) Geophys. Monogr. Series 180: AGU: Washington DC.
  • Deser C, Magnusdottir G, Saravanan R, Phillips A. 2004. The effects of North Atlantic SST and sea ice anomalies on the winter circulation in CCM3. Part II: Direct and indirect components of the response. J. Climate 17: 877889.
  • Deser C, Thomas RA, Peng S. 2007. The transient atmospheric circulation response to North Atlantic SST and sea ice anomalies. J. Climate 20: 47514767.
  • Francis JA, Chan W, Leathers DJ, Miller JR, Veron DE. 2009. Winter Northern Hemisphere weather patterns remember summer Arctic sea-ice extent. Geophys. Res. Lett. 36: L07503, DOI: 10.1029/2009GL037274.
  • Kay JE, L'Ecuyer T, Gettelman A, Stephens G, O'Dell C. 2008. The contribution of cloud and radiation anomalies to the 2007 Arctic sea ice extent minimum. Geophys. Res. Lett. 35: L08503, DOI: 10.1029/2008GL033451.
  • L'Heureux ML, Kumar A, Bell GD, Halpert MS, Higgins RW. 2008. Role of the Pacific-North American (PNA) pattern in the 2007 Arctic sea ice decline. Geophys. Res. Lett. 35: L20701, DOI: 10.1029/2008GL035205.
  • Magnusdottir G, Deser C, Saravanan R. 2004. The effects of North Atlantic SST and sea-ice anomalies on the winter circulation in CCM3. Part I: Main features and storm-track characteristics of the response. J. Climate 17: 857876.
  • Minobe S, Kuwano-Yoshida A, Komori N, Xie S-P, Small RJ. 2008. The influence of the Gulf Stream on the troposphere. Nature 452: 206209. DOI: 10.1038/nature06690.
  • Molteni F, Vitart F, Stockdate T, Ferranti L, Balmaseda MA. 2007. ‘Predictions of tropical rainfall with the ECMWF seasonal and monthly forecast systems’. Workshop on Ensemble Prediction, 7–9 November 2007. ECMWF: Reading, UK.
  • Ogi M, Wallace JM. 2007. Summer minimum Arctic sea ice extent and the associated summer atmospheric circulation. Geophys. Res. Lett. 34: L12705, DOI: 10.1029/2007GL029897.
  • Polyakov I, Timokhov L, Hansen E, Piechura J, Walczowski W, Ivanov V, Simmons H, Fahrbach E, Hölemann J, Steele M, Pickart R, Fortier L, Schauer U, Beszczynska-Möller A, Holliday NP, Dmitrenko I, Dickson R, Gascard J-C, Mauritzen C. 2007. Observational program tracks Arctic Ocean transition to a warmer state. EOS Trans. AGU 88(40): 398.
  • Reynolds RW, Rayner NA, Smith TM, Stokes DC, Wang W. 2002. An improved in situ and satellite SST analysis for climate. J. Climate 15.
  • Rigor IG, Wallace JM. 2004. Variations in the age of Arctic sea-ice and summer sea ice extent. Geophys. Res. Lett. 31: L09401, DOI: 10.1029/2004GL019492.
  • Schweiger AJ, Linsay RW, Vavrus S, Francis JA. 2008. Relationships between Arctic sea ice and clouds during autumn. J. Climate 21: 47994810, DOI: 10.1175/2008JCLI2156.1.
  • Serreze MC, Francis JA. 2006. The Arctic amplification debate. Climatic Change 76: 241264.
  • Silverman BW. 1986. Density estimation for statistics and data analysis. Chapman and Hall: New York.
  • Singarayer JS, Bamber JL, Valdes PJ. 2006. Twenty-first century climate impacts from a declining Arctic sea ice cover. J. Climate 19: 11091125.
  • Slingo J, Sutton R. 2007. Sea-ice decline due to more than warming alone. Nature 450(7166): 27.
  • Stroeve J, Serreze M, Drobot S, Gearheard S, Holland M, Maslanik J, Meier W, Scambos T. 2008. Arctic sea ice extent plummets in 2007. EOS Trans. AGU 89(2): 13.
  • Uppala SM, Kållberg PW, Simmons AJ, Andrae U, Da Costa Bechtold V, 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, Dragosazac M, Fisher M, Fuentes M, Hagemann S, Hólm E, Hoskins BJ, Isaksen L, Janssen PAEM, Jenne R, McNally AP, Mahfouf J-F, Morcrette J-J, Rayner NA, Saunders RW, Simon P, Sterl A, Trenberth KA, Untch A, Vasiljevic D, Viterbo P, Woollen J. 2005. The ERA-40 reanalysis. Q. J. R. Meteorol. Soc. 131: 29613012.
  • Vialard J, Vitart F, Balmaseda MA, Stockdale T, Anderson D. 2005. An ensemble generation method for seasonal forecasting with an ocean-atmosphere coupled model. Mon Weather Rev. 133: 441453.
  • Zhang J, Lindsay R, Steele M, Schweiger A. 2008. What drove the dramatic retreat of Arctic sea ice during summer 2007? Geophys. Res. Lett. 35: L11505, DOI: 10.1029/2008GL034005.