1. Introduction
The importance for many societal applications of improved information about near-term climate evolution (from 1 year to a decade in advance) has prompted considerable research in the field of decadal climate prediction (e.g., Meehl et al. 2014). In the framework of phase 5 of the Coupled Model Intercomparison Project (CMIP5; Taylor et al. 2012), decadal forecast experiments initialized with observationally based state information have shown greater skill in predicting the evolution of planetary-averaged temperature compared to traditional noninitialized historical coupled simulations (Kirtman et al. 2013; Bellucci et al. 2015). Beyond global and integrative metrics, robust skill in hindcasts is also found at ocean/continental basin scale for lead times up to 7–8 years and for particular regions such as the North Atlantic sector, which clearly stands out (Doblas-Reyes et al. 2013). In combination with prescribed anthropogenic and natural external forcings (Terray 2012), the so-called Atlantic multidecadal variability (AMV), characterized by basinwide low-frequency variations of the North Atlantic sea surface temperature (SST), has been identified as one of the sources of predictability at decadal time scale (e.g., Kim et al. 2012).
The origin of the AMV is still highly debated due to the shortness, spatial sparsity, and uneven quality of the observational record over the instrumental epoch (e.g., Cassou et al. 2018) and additionally to the probable coexistence and combination of several physical mechanisms yielding low-frequency fluctuations over the North Atlantic, as assessed from modeling results [see Yeager and Robson (2017) and Zhang et al. (2019) for a review]. Although the integration of so-called atmospheric noise by the ocean has been recently proposed as a potential source of decadal variability in the North Atlantic (Clement et al. 2015; Cane et al. 2017), part of the observed AMV is commonly considered as the surface fingerprint of ocean heat content anomalies driven by internal climate dynamics (O’Reilly et al. 2016). This involves large-scale changes in both air–sea fluxes and ocean heat transport through the variability of the North Atlantic subpolar and subtropical horizontal gyres and the Atlantic meridional overturning circulation (AMOC) (e.g., Zhang and Wang 2013) in the presence of complex feedbacks (Ruprich-Robert and Cassou 2015; Peings et al. 2016). Consistently, Msadek et al. (2014), Robson et al. (2012), and Yeager et al. (2015), among others, show that the prediction of the North Atlantic SST and ocean heat content as well as sea ice extent in subarctic basins clearly benefits from the initialization of the three-dimensional thermodynamical ocean.
However, the added value of the initialization is considerably reduced over the North Atlantic adjacent continents, as found in most CMIP5 decadal prediction systems (Goddard et al. 2013; Doblas-Reyes et al. 2013). Such a loss of predictability over land is somewhat paradoxical given the tight links that exist in observations between AMV and the decadal variations in summertime temperature and precipitation over the North American continent (Sutton and Hodson 2005; Ruprich-Robert et al. 2017), over Europe (Sutton and Dong 2012; O’Reilly et al. 2017), and over Africa for Sahel rainfall (Zhang and Delworth 2006). Note that greater predictive model performance is found for specific decadal shifts (e.g., the mid-1990 warming of the subpolar gyre; Robson et al. 2013), but despite the existence of such a conditional skill, the use of decadal prediction systems still remains limited for operational purposes (Towler et al. 2018).
More optimistic views and opportunities for progress in the science of decadal forecasting have been recently presented in Yeager et al. (2018). Based on an updated version of the CESM model, they insist on the importance of minimizing intrinsic model biases and inhomogeneities in initialized fields in key regions, such as the North Atlantic Ocean, in order to limit spurious drifts and shocks (Sanchez-Gomez et al. 2016), which deteriorate the levels of skill. They also clearly illustrate the crucial need for large ensembles to robustly extract the predictive signals over land that can be attributable to the initialized decadal ocean fluctuations. Despite these new promising results, a key outstanding challenge for the climate research community is to better understand how decadal changes in the ocean affect surface climate over land and ultimately translate into useful prediction. Over the North Atlantic, obstacles stand in the diversity of the statistical (amplitude, intrinsic frequency, etc.) and physical (spatial patterns, role of AMOC, etc.) properties of the AMV simulated by the current generation of models, leading to large uncertainties in teleconnectivity over Europe as shown in Qasmi et al. (2017). In their study based on long control model simulations and historical ensembles, they further insist on the nonstationarity of the level, even sign, of the AMV–Europe teleconnection as a function of the considered period. Using proxy-based reconstructions of atmospheric modes over the Atlantic, Raible et al. (2014) also found temporal diversity in teleconnection patterns involving ocean–atmosphere coupling over the last millennium. This clearly adds a degree of complexity to extract the AMV-forced fingerprint and associated physical processes at the origin of the observed ocean–land relationship.
As discussed above, most of the robust AMV impacts over Europe have been reported for summertime but the seasonality of the teleconnection remains an open question. Based on observations, contradictory results may be found in the literature for winter. For instance, Sutton and Dong (2012) could not find any significant anomalous atmospheric circulation over the North Atlantic concomitant with AMV phases and they claim that no significant signal in temperature and precipitation could be detected over Europe. O’Reilly et al. (2017) confirm the missing continental AMV fingerprint and attribute the lack of teleconnectivity to the dominance of atmospheric noise whose intensity/weight is maximum in winter and may thus overcome any potential oceanic influences. Based on a Lagrangian approach, Yamamoto and Palter (2016) alternatively interpret the “seasonal teleconnectivity hole” as the result of compensation between AMV-driven anomalies in atmospheric dynamics on the one hand and direct thermodynamic influence through air–sea fluxes on the other hand. Introducing temporal lags between atmospheric and oceanic fields in the observations, Gastineau and Frankignoul (2015) suggest that the large-scale wintertime atmosphere response to positive AMV (AMV+) projects onto the negative phase of the North Atlantic Oscillation (NAO−). Peings et al. (2016) find a similar response but only for two models out of the full CMIP5 archive. They attribute the weak feedback of the AMV onto the wintertime atmosphere to the coupled model deficiencies in generating sufficiently strong and persistent multidecadal variability over the North Atlantic in line with Kavvada et al. (2013) and Qasmi et al. (2017), among others.
Results appear more robust in dedicated sensitivity model experiments with prescribed or restored SSTs. Peings and Magnusdottir (2014) provide evidence for favored NAO− (NAO+) circulation regimes during AMV+ (AMV−) and Davini et al. (2015), consistently with earlier studies (e.g., Cassou et al. 2004; Hodson et al. 2010), interpret this relationship as a by-product of forced atmospheric Rossby waves generated in the Caribbean basin by altered convection in response to AMV-related SSTs anomalies. Peings et al. (2016), similarly to Drévillon et al. (2003), confirm the importance of ocean–atmosphere feedbacks at midlatitudes to allow a full northward extension of the tropical-initiated Rossby wave in order to generate significant impacts over Europe located at the tail end of the teleconnection. The relative weight of tropical versus extratropical AMV-related SST anomalies is analyzed as well in Ruprich-Robert et al. (2017). But, beyond this issue per se, they also insist based on modeling experiments on the overall weak signal-to-noise (SNR) ratio in terms of wintertime atmospheric response over Europe, which challenges the actual existence of any AMV teleconnection. Whether this absence of teleconnection in winter is real or due to deficiencies of the current generation of models remains a key scientific question, which definitely conditions the level of potential predictability in decadal forecast systems as raised in Yeager et al. (2018).
Within this context and built on lessons drawn from CMIP5, the Decadal Climate Prediction Project (DCPP) has proposed for CMIP6 (Eyring et al. 2016) a new targeted multimodel framework (named Component C; Boer et al. 2016) aiming at increasing knowledge and physical understanding of the worldwide impacts of the AMV through teleconnectivity (Cassou et al. 2018). The CMIP6-endorsed coordinated experiments are inspired by Ruprich-Robert et al. (2017) and rely on so-called pacemaker simulations where the North Atlantic SSTs are restored toward a specific anomalous pattern that is representative of AMV phases, whereas the rest of the coupled model remains free to evolve. We here conducted those specific DCPP Component C experiments using the CNRM-CM5 global circulation model (Voldoire et al. 2013). In the following paper, we concentrate our analyses and physical interpretations of the model AMV-forced response in winter over Europe. We employ a so-called circulation analog technique inspired from Boé et al. (2009), Deser et al. (2016), and O’Reilly et al. (2017) to decompose the impact of the AMV on surface air temperature and precipitation over Europe into dynamical versus so-called thermodynamical relative contributions. Considering the weak SNR properties documented in many studies, we have performed additional experiments in which the AMV-related SST anomalies are artificially boosted to potentially increase the forced response. Concurrently, we have produced larger ensembles than current protocols recommend for CMIP6 in order to get a better estimation of the AMV-forced response in the presence of prevalent climate noise, following the advice of Deser et al. (2012) and Yeager et al. (2018).
The structure of the paper is organized around three main objectives: 1) to isolate the dynamical and thermodynamical fingerprints of the AMV in the North Atlantic/European climate assessed from our large ensembles and revisit the results presented in Ruprich-Robert et al. (2017), Yamamoto and Palter (2016), and O’Reilly et al. (2017); 2) to identify the physical processes explaining the modeled AMV teleconnectivity over Europe in winter; and 3) to evaluate the sensitivity of the model atmospheric response and related mechanisms to the amplitude of the AMV. After a description of the modeling protocols in section 2, the mean wintertime response to the AMV, as well as the dynamical and thermodynamical mechanisms, is detailed in section 3. The sensitivity of the AMV-forced atmospheric response to the AMV amplitude is discussed, followed by conclusions and perspectives in section 4.
2. Methods
a. Model pacemaker sensitivity experiments
As mentioned in the introduction, the pacemaker or partial coupling simulations analyzed in the paper follow the protocol endorsed by CMIP6 and commonly labeled as DCPP-C AMV experiments. The reader is invited to refer to Boer et al. (2016, Components C1.2 and C1.3 in their Table C1) for a thorough presentation of the coordinated experimental framework and related input datasets shared through input4MIPs (Durack et al. 2018). The AMV anomalous pattern toward which the model is restored corresponds to an estimation of the internal component of the observed SST decadal variability over the North Atlantic after subtraction of the forced component. The latter is obtained following the procedure detailed by Ting et al. (2009). It is based on a signal-to-noise maximizing empirical orthogonal function analysis, which was applied to global annual mean SST given by the CMIP5 multimodel historical simulations. The observed AMV index is thus defined as the residual of the forced component and the AMV anomalous SST pattern used for restoring in DCPP-C is then obtained from observation by regression over the 1900–2013 period (Fig. 1a). In CNRM-CM5, this is achieved through the addition of a feedback term to the nonsolar total heat flux in the surface temperature derivative equation following Haney (1971). This flux formulation affects the entire ocean mixed layer depth. In compliance with DCPP recommendations (see the technical notes at https://www.wcrp-climate.org/experimental-protocol), we set the restoring coefficient to a spatially and temporally constant value equal to −40 W m−2 K−1. For the SST, it is equivalent to a damping time scale of ~2 months for a 50-m-deep mixed layer.
(a) Anomalous SST pattern used for restoring and taken from the input4MIPs archive [https://esgf-node.llnl.gov/projects/input4mips/; units are °C per σ(AMV), shading interval is every 0.03°C]. (b) Simulated raw annual SST averaged over the North Atlantic restored sector for AMV+ (black) and AMV− (gray) experiments. Each boxplot stands for the distribution obtained from 360 years for each ensemble (40 members × 9 years, the first year being discarded to account for the model SST initial adjustment to restoring). The top (bottom) of the box represents the first (last) tercile of the distribution and the upper (lower) whisker represents the first (ninth) decile. Dots and inside line stand for the mean and the median of the distribution, respectively. The green, orange, and magenta horizontal lines show the SST targets for the 1xAMV, 2xAMV, and 3xAMV ensembles corresponding to 1, 2, and 3 standard deviations of the observed AMV index, respectively. Solid and dashed stand respectively for AMV+ and AMV− experiments.
Citation: Journal of Climate 33, 7; 10.1175/JCLI-D-19-0428.1
(a) Anomalous SST pattern used for restoring and taken from the input4MIPs archive [https://esgf-node.llnl.gov/projects/input4mips/; units are °C per σ(AMV), shading interval is every 0.03°C]. (b) Simulated raw annual SST averaged over the North Atlantic restored sector for AMV+ (black) and AMV− (gray) experiments. Each boxplot stands for the distribution obtained from 360 years for each ensemble (40 members × 9 years, the first year being discarded to account for the model SST initial adjustment to restoring). The top (bottom) of the box represents the first (last) tercile of the distribution and the upper (lower) whisker represents the first (ninth) decile. Dots and inside line stand for the mean and the median of the distribution, respectively. The green, orange, and magenta horizontal lines show the SST targets for the 1xAMV, 2xAMV, and 3xAMV ensembles corresponding to 1, 2, and 3 standard deviations of the observed AMV index, respectively. Solid and dashed stand respectively for AMV+ and AMV− experiments.
Citation: Journal of Climate 33, 7; 10.1175/JCLI-D-19-0428.1
(a) Anomalous SST pattern used for restoring and taken from the input4MIPs archive [https://esgf-node.llnl.gov/projects/input4mips/; units are °C per σ(AMV), shading interval is every 0.03°C]. (b) Simulated raw annual SST averaged over the North Atlantic restored sector for AMV+ (black) and AMV− (gray) experiments. Each boxplot stands for the distribution obtained from 360 years for each ensemble (40 members × 9 years, the first year being discarded to account for the model SST initial adjustment to restoring). The top (bottom) of the box represents the first (last) tercile of the distribution and the upper (lower) whisker represents the first (ninth) decile. Dots and inside line stand for the mean and the median of the distribution, respectively. The green, orange, and magenta horizontal lines show the SST targets for the 1xAMV, 2xAMV, and 3xAMV ensembles corresponding to 1, 2, and 3 standard deviations of the observed AMV index, respectively. Solid and dashed stand respectively for AMV+ and AMV− experiments.
Citation: Journal of Climate 33, 7; 10.1175/JCLI-D-19-0428.1
A common issue associated with this type of protocol is that the AMV pattern used for restoring is derived from observations and may not be necessarily representative and consistent with the AMV simulated intrinsically by the model internal variability. This might generate a response that may be not physically coherent with the model’s dynamics. To test this potential caveat, we conducted additional ensemble experiments replicating the 1xAMV ensemble, but using the intrinsic AMV pattern of CNRM-CM5 extracted from the 850-yr-long preindustrial control (piControl) experiments of the model (see Fig. 14a in Ruprich-Robert and Cassou 2015). Note that AMV-observed and AMV-intrinsic patterns are rather similar except for the tropical branch with underestimated anomalies. We found that the forced AMV response is virtually indistinguishable over Europe between the 1xAMV reference ensemble and the sensitivity one using the model AMV SST anomalies (not shown). This suggests that the model shows a qualitatively consistent response to 1xAMV forcing, whether the model-intrinsic or the observed 1xAMV pattern is used in the pacemaker simulations, increasing our confidence in the usefulness of the experimental setup.
Two large ensembles of 40 members of 10-yr-long simulations are conducted. They differ by the sign of the targeted anomalous SST pattern, corresponding to either positive (i.e., warm; the AMV+ ensemble) or negative (i.e., cold; the AMV− ensemble, with a sign reversal of Fig. 1a) phases of the AMV. The initial conditions for the 40 members are ocean + atmosphere + land + sea ice states [so-called macroperturbation, the nomenclature of Hawkins et al. (2016)] taken arbitrarily every 20 years from the piControl experiment of CNRM-CM5. The same set of initial conditions is used for the AMV+ and AMV− ensembles. The ensemble size has been increased here to 40 instead of 25 (the minimum number recommended in DCPP-C; Boer et al. 2016) to ensure a better estimation of AMV-forced signals. Additional twin experiments are conducted by multiplying by 2 and 3 the anomalous SST pattern toward which the model is restored over the North Atlantic (Fig. 1a). Those additional ensembles are hereafter termed 2xAMV and 3xAMV, respectively, and the reference DCPP-compliant ensemble is referred to as 1xAMV.
Figure 1b provides a crude evaluation of the pacemaker protocol and, importantly, an indication of the actual SST forcing in each ensemble. Independently of the sign of the AMV experiments, a spread exists in simulated annual SSTs for all ensembles. The corresponding ensemble means are always lower than the targeted SSTs toward which the coupled model is restored. Both features are attributable to the weak restoring coefficient used here. We tested stronger values, which do allow the actual SST to be closer to the targeted SST (not shown). However, those lead to spurious energy imbalance and perturb the modeled high-frequency air–sea interactions, the ocean heat content, and meridional transports through AMOC, etc., which ultimately affects the interpretation of the atmospheric response to the AMV forcing (see also the DCPP technical note at https://www.wcrp-climate.org/wgsip/documents/Tech-Note-2.pdf). Despite a weak restoring coefficient, the interannual variance of the modeled SST averaged over the North Atlantic is reduced by a factor of 10 compared to the free piControl CNRM-CM5 experiment. Because the restoring coefficient is fixed, its efficiency is strongly dependent on the ocean mixed layer depth. The reduction of the basinwide variance thus masks considerable regional heterogeneities, such as the subpolar gyre characterized by seasonal deep ocean mixing and the more stratified tropical Atlantic regions (see also Ruprich-Robert et al. 2017; Ortega et al. 2017).
Since the restoring is not perfect, the multiplication by 2 or 3 of the anomalous SST-forcing pattern in our additional ensembles is not as artificial or unrealistic as it may appear at first glimpse. Actual North Atlantic SSTs obtained in 3xAMV are in fact close to the targeted SSTs of 2xAMV, which correspond to ± two standard deviations of the observed AMV index over the instrumental record. Actual SSTs in 2xAMV are close to the targeted SSTs of 1xAMV (Fig. 1b). These additional experiments will be useful to assess the sensitivity of the teleconnection to the intensity of the AMV-forced SST anomalies and, in particular, its degree of linearity.
b. Flow analog technique
Boé et al. (2009), Cattiaux and Yiou (2013), and Deser et al. (2016), among others, employed so-called dynamical flow analog techniques to quantify the relative roles of the dynamical versus nondynamical processes in either observed or projected climate change signals. More recently, O’Reilly et al. (2017) applied the same technique to study the AMV teleconnection over North America, Europe, and Africa based on observational datasets (reanalyses) over the historical period. We here adapt the methodology to our ensemble approach aiming at better understand the involved mechanisms of the AMV-forced surface temperature and precipitation anomalies simulated in CNRM-CM5 over Europe in winter.
Technically, for each winter day K of the AMV+ experiment, we seek for the N best analogs of the atmospheric circulation in the population of winter days from the twin AMV− experiment. We use sea level pressure (SLP) maps centered over Europe (EU; 35°–70°N, 15°W–35°E) and the similarity criterion to define the circulation analogs as the Euclidean distance. The N best analogs for K are the N days in AMV− for which the Euclidean distances to K are minimum. We then reconstruct the temperature/precipitation map of day K of AMV+ by averaging the N temperature/precipitation maps of the best N SLP analogs found in AMV−. Assuming the absence of feedback processes between the surface and the circulation, the latter reconstructed temperature/precipitation is interpreted as the surface fingerprint of the atmospheric circulation (hereafter named the dynamical part of the field) and the residual with respect to the actual raw AMV+ temperature/precipitation is treated as the thermodynamical part for the sake of simplicity. Note that to account for the seasonal cycle of the reconstructed surface fields, which could be particularly pronounced (e.g., for temperature), the analog search for AMV+ day K is constrained to be in a 15-day window around day K in the AMV− pool of days as done in Dayon et al. (2015), for instance. To sum up, let us take a concrete example. Let day K be 1 February of year 4 of member 18 (01 February-y4-m18) of AMV+. Let N = 2. Research of analogs is done in the pool of days formed by the 40 members and 10 years of AMV− between 24 January and 7 February. We find the two best SLP analogs be 29 January-y2-m38 and 06 February-y10-m3 and average the corresponding raw temperature/precipitation of those 2 days of AMV− to form the reconstructed dynamical temperature/precipitation of day K of AMV+. The computation is repeated for all the winter (1 December to 28 February) days of AMV+.
Sensitivity tests to 1) the spatial domain used for analog seek and 2) the number N of retained analogs used for reconstruction have been conducted a priori. To do so, the above-described stepwise process is applied within the AMV+ ensemble itself; this additionally provides a quantitative evaluation/validation of the proposed methodology. Technically speaking, a given day K-yY-mM of AMV+ is here reconstructed from the N best SLP analogs found in the pool of AMV+ days excluding in that case the year Y to which day K belongs to account for the day-to-day persistence of the circulation. Regarding the geographical domains, results from SLP analog extracted from a larger region corresponding to the North Atlantic–Europe region (NAE; 20°–80°N, 80°W–30°E) used traditionally for large-scale dynamical purposes (see, e.g., Cassou et al. 2011; Michel and Rivière 2014, etc.) have been contrasted to the above-mentioned EU sector used as reference.
A two-step evaluation of the performance of the methodology is carried out based on spatial root-mean-square error (RMSE) and spatial correlation metrics between 1) the reconstructed SLP with the analog method (the predictor) and the actual SLP in AMV+ (Table 1) and 2) the reconstructed T2m (the predictand) with the actual one in AMV+ (Table 2). For SLP reconstruction, we show that the EU domain clearly outperforms the NAE one with lower RMSE and higher correlation whatever the number of selected analogs (Table 1). The optimal number of analogs N lies between 10 and 15 since 1) the highest correlation value is found for N = 10 analogs and becomes insensitive to the inclusion of additional ones and 2) RMSE is concomitantly the smallest for N = 10, being slightly degraded with increased number. Results appear to be much less sensitive to the geographical domain for T2m but the overall above-listed conclusions for the choice in N still hold (Table 2). The temporal variance of monthly T2m in DJF explained by SLP through dynamical reconstruction is also quantified for different number of analogs. Five to ten analogs are required to reach a maximum of explained variance equal to ~70% in average over Europe with higher values along the Atlantic flank of Europe (up to ~80%) whose variability is overly dominated by synoptic storm influence (not shown). The explained variance is weaker for precipitation (~60% on average). Note that the explained temporal variance of the dynamical contribution is not very dependent on the amplitude of the AMV forcing. To further verify the robustness of the method, all these validation steps are also repeated with AMV− instead of AMV+ experiments and additionally with the 2xAMV and 3xAMV ensembles. We have verified that the dynamical reconstruction discussed here is not sensitive to the ensemble of experiments in which analogs are sought. In other words, seeking analogs of 2xAMV in 1xAMV ensembles does not change the dynamical-reconstructed spatial pattern and amplitude. Notably, the same results are obtained when analogs are taken from piControl. Altogether this suggests that the feedback between AMV-related surface conditions (including the dominant thermodynamical component) and atmospheric circulation is weak (not shown). Results remain valid whatever the case (not shown) and the combination EU domain/N = 10 is therefore retained for our study.
Mean RMSE and spatial correlation between the analog daily SLP and the actual daily SLP estimated from all winter (1 Dec–28 Feb) days over the 40 members and 10 years in 1xAMV+, as a function of the number of analogs and domains.
In the rest of the paper, the AMV-forced anomalies for any fields (also called response for short) are defined as the ensemble mean differences between the AMV+ and AMV− experiments. The dynamical component of the AMV-forced anomalies is defined as the ensemble mean difference between the reconstructed field of AMV+ based on the SLP analog seek in the counterpart AMV− experiment and the reconstructed field of AMV− based on the SLP analog seek in AMV− itself. This accounts for the methodology error linked intrinsically to the analog technique. The thermodynamical component of the AMV-forced anomalies is defined as the residual anomaly calculated by subtracting the dynamical AMV-forced anomaly from the full response.
3. Results
a. Mean wintertime response to AMV over Europe
Figure 2 summarizes the AMV-forced winter anomalies for some surface and dynamical atmospheric fields.
AMV-forced anomalies for December–February seasonal mean for (a)–(c) T2m (shading interval is 0.075°C), (d)–(f) precipitation (in relative percentage; shading interval is 0.8%), and (g)–(i) land snow cover and sea ice (shading intervals are 0.4% and 4%, respectively) and Z500* (contour interval is 4 m and the thicker black contour stands for the zero line) superimposed on MSLP (shading interval is 0.1 hPa) for (left) 1xAMV, (middle) 2xAMV, and (right) 3xAMV. Stippling indicates regions that are above the 95% confidence level of statistical significance based on a two-sided Student’s t test.
Citation: Journal of Climate 33, 7; 10.1175/JCLI-D-19-0428.1
AMV-forced anomalies for December–February seasonal mean for (a)–(c) T2m (shading interval is 0.075°C), (d)–(f) precipitation (in relative percentage; shading interval is 0.8%), and (g)–(i) land snow cover and sea ice (shading intervals are 0.4% and 4%, respectively) and Z500* (contour interval is 4 m and the thicker black contour stands for the zero line) superimposed on MSLP (shading interval is 0.1 hPa) for (left) 1xAMV, (middle) 2xAMV, and (right) 3xAMV. Stippling indicates regions that are above the 95% confidence level of statistical significance based on a two-sided Student’s t test.
Citation: Journal of Climate 33, 7; 10.1175/JCLI-D-19-0428.1
AMV-forced anomalies for December–February seasonal mean for (a)–(c) T2m (shading interval is 0.075°C), (d)–(f) precipitation (in relative percentage; shading interval is 0.8%), and (g)–(i) land snow cover and sea ice (shading intervals are 0.4% and 4%, respectively) and Z500* (contour interval is 4 m and the thicker black contour stands for the zero line) superimposed on MSLP (shading interval is 0.1 hPa) for (left) 1xAMV, (middle) 2xAMV, and (right) 3xAMV. Stippling indicates regions that are above the 95% confidence level of statistical significance based on a two-sided Student’s t test.
Citation: Journal of Climate 33, 7; 10.1175/JCLI-D-19-0428.1
For surface air or 2-m temperature (T2m), an overall weak response is obtained in CNRM-CM5 in 1xAMV (Fig. 2a) consistent with previous studies, which highlighted the absence of detectable impact of the AMV on wintertime European climate as assessed both from models (Ruprich-Robert et al. 2017) and observations (Yamamoto and Palter 2016; O’Reilly et al. 2017). Despite very marginal pointwise significance, which is limited to a weak warming over the Atlantic side of the Iberian Peninsula, note that a robust spatial pattern interestingly emerges at continental scale. Cooling dominates in central Europe, from western Russia to the North Sea shoreline including the Alpine region and southern Sweden, whereas warming occurs in the northernmost part of Scandinavia and along the Mediterranean Sea, to a lesser extent. Similar qualitative conclusions can be drawn in terms of precipitation (Fig. 2d). In 1xAMV, the AMV-forced response is overall weak but characterized by large-scale coherence. Drier conditions extend from the United Kingdom and northern France to Sweden and the Baltic shore of Finland where the AMV-forced response is the most pronounced. This contrasts to wetter winters along the western windward coast of Scandinavia and around the entire Mediterranean Sea with regional features that are indicative of orographic effects.
Consistently with colder and wetter conditions, albeit weak and insignificant, increased snow cover is found on the south side of the Baltic Sea along a narrow latitudinal band from the Netherlands to Belarus (Fig. 2g). Over the ocean, sea ice extent is diminished in all the Nordic seas with maximum amplitude of the AMV-response along the seasonal ice edge. In terms of atmospheric circulation (Fig. 2j), higher geopotential height dominates the northern part of the Atlantic basin at 500 hPa (Z500*) with maximum loading between Iceland and the United Kingdom. Note that Z500 zonal means have been subtracted to account for the mean dilatation of the atmosphere due to the artificial heat source introduced in the model in pacemaker experiments via the flux restoring term. The signal is barotropic with a nominal eastward shift at the surface [significantly higher mean sea level pressure (MSLP) centered in the North Sea] but baroclinic over the retracted sea ice regions (Labrador and Greenland Seas). Negative MSLP and Z500* anomalies cover most of the European continent from the Iberian Peninsula to western Russia south of 50°N.
In 2xAMV, the AMV-forced temperature anomalies are positive over the entire continent with maximum loading in Scandinavia over Sweden/Finland and along the Atlantic flank of Europe (including the entire Iberian Peninsula; Fig. 2b). Continental-scale warming is further intensified in 3xAMV (Fig. 2c) and penetrates deeper inland with significant and amplified response along an axis going from the Baltic Sea up to southern France/northern Spain. For precipitation, despite limited pointwise significance, wetter conditions tend to prevail over the entire continent (except Scandinavia) with maximum anomalies found over the Balkans in 2xAMV (Fig. 2e). Interestingly, although precipitation further increases on average over Europe in 3xAMV, the regional structure of the response greatly differs from the other two ensembles (Fig. 2f). Maximum excess is not found any more in the Balkans like in 2xAMV but over eastern Spain and a large part of central Europe along the stretch of maximum warming (Fig. 2c).
Strong reduction of snow cover (Fig. 2i) is also collocated with the greatest positive temperature anomalies, which is indicative for enhanced rainfall at the expense of snowfall in 3xAMV. This is less valid for 2xAMV where the reinforced precipitation over the Balkans is accompanied by locally increased snow cover, albeit marginally (Fig. 2h). Loss of sea ice is considerably reinforced with the amplitude of the AMV (Figs. 2h,i) on both sides of the Atlantic basin, with maximum ice decline in the Odden region at the eastern edge of the Greenland Sea and along the Eastern Labrador Current. In terms of atmospheric dynamics, lower Z500* and negative SLP anomalies are considerably reinforced south of 55°N and become significant from Newfoundland to the Mediterranean Sea in 2xAMV (Fig. 2k). Note though that both positive MSLP and Z500* anomalies are northwestward shifted compared to 1xAMV with a degree of intensification and significance at polar latitudes that is considerably less than their negative counterpart to the south. In 3xAMV, the AMV-forced signal in Z500* is wavier with two cyclonic cores (one between Newfoundland and the Azores and a second one over western Russia), which sandwiches positive Z500* anomalies from Greenland to France (Fig. 2l). At the surface, stronger negative MSLP anomalies cover most of the Atlantic Ocean except in the Norwegian Sea and over Greenland where positive MSLP signals, although slackened, remain.
To deepen our understanding of the full response in temperature and precipitation over Europe, we use in the following section the flow analog method described in section 2b to assess the respective weight of the dynamical versus thermodynamical related processes.
b. Decomposition in dynamical and thermodynamical components of the AMV-forced anomalies over Europe
In 1xAMV, the atmospheric dynamical response is characterized by an anticyclonic circulation centered over Scotland, which leads to prevalent northeasterly wind anomalies over most of Europe, except Scandinavia (Fig. 3a). The latter flow is responsible for dominant, yet marginally significant, negative temperature anomalies over the entire continent. The presence of an anomalous high just off Europe tends to favor the advection of colder and drier air from the east and/or to block storms from penetrating inland, which explains the large-scale deficit in rainfall along an axis going from northern France to the Baltic shore of Sweden/Finland (Fig. 3g). Dominant northeasterlies over the Mediterranean basin lead to onshore anomalous flow over Spain leading locally to wetter significant conditions. Elsewhere, slight enhanced rainfall dominates with some orographic effect over the Carpathians and the Balkans. The thermodynamical component of the AMV-forced T2m signal is characterized by large-scale warming with maximum loading along the Atlantic shore and in the northernmost part of Scandinavia (Fig. 3d). It counteracts the dynamical component dominated by chillier conditions (Fig. 3a), leading to a weak total response in temperature (Fig. 2a), in line with Yamamoto and Palter’s (2016) findings. For precipitation, the thermodynamical part (Fig. 3j) reinforces the dynamically induced wetter conditions along the Mediterranean shore (Fig. 3g). It is also responsible for increased snowfall along the south shore of the Baltic Sea (Fig. 2g), while the dynamical part clearly explains the largest portion of the drier conditions over most of Scandinavia (cf. Figs. 3g,j with Fig. 2d).
Decomposition of the AMV-forced anomalies into dynamical and thermodynamical components estimated from analog reconstruction for December–February seasonal mean for (top two rows) T2m (shading interval is 0.05°C) and (bottom two rows) for precipitation (in relative percentage; shading interval is 0.8%). See section 2b for the description of the method. Stippling indicates regions that are above the 95% confidence level of statistical significance based on a two-sided Student’s t test.
Citation: Journal of Climate 33, 7; 10.1175/JCLI-D-19-0428.1
Decomposition of the AMV-forced anomalies into dynamical and thermodynamical components estimated from analog reconstruction for December–February seasonal mean for (top two rows) T2m (shading interval is 0.05°C) and (bottom two rows) for precipitation (in relative percentage; shading interval is 0.8%). See section 2b for the description of the method. Stippling indicates regions that are above the 95% confidence level of statistical significance based on a two-sided Student’s t test.
Citation: Journal of Climate 33, 7; 10.1175/JCLI-D-19-0428.1
Decomposition of the AMV-forced anomalies into dynamical and thermodynamical components estimated from analog reconstruction for December–February seasonal mean for (top two rows) T2m (shading interval is 0.05°C) and (bottom two rows) for precipitation (in relative percentage; shading interval is 0.8%). See section 2b for the description of the method. Stippling indicates regions that are above the 95% confidence level of statistical significance based on a two-sided Student’s t test.
Citation: Journal of Climate 33, 7; 10.1175/JCLI-D-19-0428.1
The storyline is very different for 2xAMV and 3xAMV temperature since the thermodynamical component controls most of the large-scale AMV-forced total warming found over all of Europe (cf. Figs. 3e,f with Figs. 2b,c) and thus clearly outpaces the dynamical one. The inland penetration of the thermodynamical signal is clearly a function of the amplitude of the AMV SST forcing with some amplification over central Europe, as noted earlier from Fig. 2c. Conversely, the strength of the dynamical cooling is similar in all ensembles (Figs. 3a–c) and does not increase with the AMV forcing; in 3xAMV, it is even barely significant. The dynamical cooling is very much sensitive to subtle changes that occur in the position of the AMV-forced MSLP anomalies. The positive core around 60°N progressively shifts northwestward with the amplitude of the AMV forcing, while anomalous cyclonic circulations farther south move to the west from the Black Sea in 1xAMV to the Adriatic Sea in 2xAMV and off Portugal in 3xAMV. These displacements, without significant simultaneous amplification, control the strength and direction of the dominant easterly anomalies over Europe; they explain a large portion of the regional changes in the dynamical component as a function of AMV forcing.
The above conclusions for temperature are also valid for precipitation (Figs. 3h,i). At large scale, the anomalous AMV-forced circulation is responsible for wetter conditions over the European Mediterranean coast with maximum response in 2xAMV where minimum MSLP and associated cyclonic flow are the most pronounced. Concurrent drier conditions prevail north of 50°N and are related to the anomalous advection of dry and cold air from the east or to the reduced penetration of warm and humid air masses from the west. These dynamical features are found in all ensembles, along with a southward displacement of the storm track over Europe (not shown). The thermodynamical response tends to increase with the amplitude of the AMV SST forcing and leads to wetter conditions at continental scale, except over the Mediterranean domain where the signals are very weak (Figs. 3k,l). At first glimpse, dynamical and thermodynamical contributions oppose each other in 2xAMV (Figs. 3h,k) whereas clear rainfall excess (Fig. 3l) dominates in 3xAMV in central Europe, in collocation with the area of maximum warming and snow cover reduction extending from Catalonia to the Baltic countries, as documented above (Figs. 2c,f).
In the following sections, we concentrate on the physical mechanisms and phenomena at the origin of the dynamical and thermodynamical responses as a function of the amplitude of the AMV forcing.
c. Mechanisms for the dynamical component of the AMV-forced response
Zonally averaged anomalies over the North Atlantic basin are presented in Fig. 4 as a function of height and as a function of the intensity of the AMV forcing. In 1xAMV, the warming imposed at the surface ocean is exported throughout the entire atmospheric column with maximum signals in the subtropics between 20° and 30°N and more importantly at high latitudes from 50°N northward (Fig. 4a). In the polar regions, there is a clear amplification of the atmospheric temperature response to the restored SST anomalies, which is caused by the pronounced AMV-forced reduction of sea ice acting as an additional source of heat at the surface in all the subarctic basins (Fig. 2g). In response to warmer SST, humidification occurs in the lower atmosphere and is exported upward to the upper troposphere between the equator and 15°N (Fig. 4b); this is collocated with a reduction of the mean upper-level westerlies in the deep tropics (Fig. 4c). At higher latitudes, albeit barely significant, the AMV-forced response is characterized by a weakening on the northern flank of the climatological maximum westerly jet around 45°N while no signal is found elsewhere.
Zonal average over the North Atlantic (15°S–80°N, 80°–15°W) of the AMV-forced anomalies for December–February seasonal mean for (from left to right) temperature (shading interval is 0.1°C), specific humidity (in 10−4 kg kg−1; shading interval is 0.5. 10−4 kg kg−1), and zonal wind (shading interval is 0.2 m s−1) with the climatological value superimposed from AMV− (contour interval is 4 m s−1 and the thicker black contour stands for the zero line) for (a)–(c) 1xAMV, (d)–(f) 2xAMV, and (g)–(i) 3xAMV. Stippling indicates regions that are above the 95% confidence level of statistical significance based on a two-sided Student’s t test.
Citation: Journal of Climate 33, 7; 10.1175/JCLI-D-19-0428.1
Zonal average over the North Atlantic (15°S–80°N, 80°–15°W) of the AMV-forced anomalies for December–February seasonal mean for (from left to right) temperature (shading interval is 0.1°C), specific humidity (in 10−4 kg kg−1; shading interval is 0.5. 10−4 kg kg−1), and zonal wind (shading interval is 0.2 m s−1) with the climatological value superimposed from AMV− (contour interval is 4 m s−1 and the thicker black contour stands for the zero line) for (a)–(c) 1xAMV, (d)–(f) 2xAMV, and (g)–(i) 3xAMV. Stippling indicates regions that are above the 95% confidence level of statistical significance based on a two-sided Student’s t test.
Citation: Journal of Climate 33, 7; 10.1175/JCLI-D-19-0428.1
Zonal average over the North Atlantic (15°S–80°N, 80°–15°W) of the AMV-forced anomalies for December–February seasonal mean for (from left to right) temperature (shading interval is 0.1°C), specific humidity (in 10−4 kg kg−1; shading interval is 0.5. 10−4 kg kg−1), and zonal wind (shading interval is 0.2 m s−1) with the climatological value superimposed from AMV− (contour interval is 4 m s−1 and the thicker black contour stands for the zero line) for (a)–(c) 1xAMV, (d)–(f) 2xAMV, and (g)–(i) 3xAMV. Stippling indicates regions that are above the 95% confidence level of statistical significance based on a two-sided Student’s t test.
Citation: Journal of Climate 33, 7; 10.1175/JCLI-D-19-0428.1
Amplification of the AMV-forced response is found throughout the depth of the troposphere in 2xAMV and 3xAMV for temperature (Figs. 4d,g) and specific humidity (Figs. 4e,h). The intensification of the signals is rather linear with respect to the amplitude of the AMV forcing. Such a linear behavior is also valid for the reduction of the westerly wind in the subtropics, whereas a different picture emerges in the extratropics (Figs. 4f,i). A clear meridional dipole in zonal wind is found straddling the climatological jet core in both 2xAMV and 3xAMV, with a significant strengthening on its equatorward side and a slackening on its poleward side, implying an equatorward shift of the midlatitude mean westerly flow. Notably, the extratropical response in 2xAMV and 3xAMV is in quadrature compared to 1xAMV; it is extremely similar in both ensembles in terms of not only spatial structure but also intensity. The alteration of the midlatitude North Atlantic dynamics in response to AMV can be interpreted as a combination of 1) local forcing associated with the subpolar gyre SST anomalies and 2) tropical–extratropical teleconnectivity (Davini et al. 2015; Ruprich-Robert et al. 2017). The respective weight between the two mechanisms is expected to control the total model response and to explain part of its sensitivity to the amplitude of the AMV forcing as assessed here. Note that the latitudinal displacement of jet found here is consistent with Simpson et al.’s (2018, 2019) results, which further suggests a reinforced response in March compared to DJF. The latter conclusion is also valid in CNRM-CM5 in 1xAMV and 2xAMV ensembles but not in 3xAMV (not shown). However, it is beyond the scope of this paper to investigate more deeply why the 3xAMV ensemble behaves uniquely in such a way.
Regarding the tropical pathway of influence, evidence is provided in the literature based on both theory and global circulation models that warmer SST in the subtropics is associated with increased precipitation on the northern flank of the climatological intertropical convergence zone (ITCZ), yielding a Gill–Matsuno type of atmospheric response in the tropical Atlantic. Such a feature is consistently found in CNRM-CM5, which simulates, in all AMV ensembles, enhanced rainfall between the equator and 15°N and a concomitant dipole in upper-tropospheric streamfunction straggling the equator, as depicted in Fig. 5. The anomalous anticyclonic circulations are located at 20°–30°N and 10°S and are maximum on the northwestern and southwestern side, respectively, of the main source of latent heat released throughout the troposphere by enhanced convection and ascendant motion (not shown), as featured in Figs. 4b, 4e, and 4h showing the vertical zonally averaged profile of specific humidity. The overall response is largely linear in the tropics with respect to the amplitude of the AMV forcing and spatially matches with the linear framework presented in Gill (1980, their Fig. 3) in the presence of off-equatorial diabatic heating.
AMV-forced anomalies for December–February seasonal mean for precipitation (shading interval is 0.07 mm day−1), and streamfunction at 200 hPa (black contour interval is 0.2 × 106 m2 s−1; the thicker black contour stands for the zero line) for (a) 1xAMV, (b) 2xAMV, and (c) 3xAMV. Climatological zonal wind speed at 300 hPa from AMV- is superimposed (2 magenta contours at 20 and 25 m s−1).
Citation: Journal of Climate 33, 7; 10.1175/JCLI-D-19-0428.1
AMV-forced anomalies for December–February seasonal mean for precipitation (shading interval is 0.07 mm day−1), and streamfunction at 200 hPa (black contour interval is 0.2 × 106 m2 s−1; the thicker black contour stands for the zero line) for (a) 1xAMV, (b) 2xAMV, and (c) 3xAMV. Climatological zonal wind speed at 300 hPa from AMV- is superimposed (2 magenta contours at 20 and 25 m s−1).
Citation: Journal of Climate 33, 7; 10.1175/JCLI-D-19-0428.1
AMV-forced anomalies for December–February seasonal mean for precipitation (shading interval is 0.07 mm day−1), and streamfunction at 200 hPa (black contour interval is 0.2 × 106 m2 s−1; the thicker black contour stands for the zero line) for (a) 1xAMV, (b) 2xAMV, and (c) 3xAMV. Climatological zonal wind speed at 300 hPa from AMV- is superimposed (2 magenta contours at 20 and 25 m s−1).
Citation: Journal of Climate 33, 7; 10.1175/JCLI-D-19-0428.1
At the midlatitudes, as a direct consequence of the anomalous anticyclonic circulation/enhanced momentum convergence due to the tropical wind divergence related to increased precipitation, there is an acceleration of the zonal wind at the entrance of the climatological upper-level subtropical jet around 30°N and 70°W. This acceleration extends downstream to the center of the North Atlantic toward the Azores in 2xAMV and 3xAMV (as seen from streamfunction anomalies in Figs. 5b and 5c), whereas it is weak and rather confined to the western part of the basin in 1xAMV (Fig. 5a). This leads altogether in 2xAMV and 3xAMV to a strengthening of the westerly wind along the southern edge of the subtropical jet (as already described in Figs. 4f and 4i from zonal averages), a feature that is not present in 1xAMV. At higher latitudes, circulation anomalies display a wave pattern along a southwest–northeast great circle from the Caribbean to Scandinavia in agreement with the classical stationary Rossby wave theory (Hoskins and Karoly 1981). Again, signals are rather weak in 1xAMV (Fig. 5a) but are considerably reinforced in 2xAMV and 3xAMV with pronounced cyclonic circulation off Newfoundland and dominant anomalous anticyclonic flow from 55°N northward (Figs. 5b,c and 2k,l); this is associated with a weakening of the westerly wind on the northern side of the climatological upper-level jet in the latter two ensembles (Figs. 4f,i). Consistently, the AMV-forced response is then characterized by large-scale enhanced (reduced) baroclinicity in the southern (northern) side of the jet and equatorward shift in storm track driven by planetary wave changes that are reinforced and/or maintained through eddy–mean flow interaction and favor cyclonic Rossby wave breakings at short time scales (synoptic eddies; Rivière 2009; Davini and Cognazzo 2014), when the Atlantic is warmer (not shown).
Despite differences in amplitude, these mechanisms share many features previously identified in Hodson et al. (2010), Peings et al. (2016), and Davini et al. (2015), among others, in particular for the attribution of the midlatitude anomalous cyclonic circulation to a tropical forcing, which originates from the Caribbean basin in response to warmer tropical Atlantic SST. For instance, there is a remarkable agreement between Fig. 5 and Figs. 2j–l in the present paper with Fig. 10 in Terray and Cassou (2002), based on an earlier version of the ARPEGE model used in sensitivity experiments to isolate the respective role of tropical versus extratropical North Atlantic SST anomalies, and with Fig. 9 in Ruprich-Robert et al. (2017). Results from decomposition of the daily circulation into weather regimes reveal a significant predominance of NAO− events at the expense of NAO+ in 2xAMV and 3xAMV (not shown), consistently with the mean circulation changes displayed in Figs. 2k and 2l; this is indicative of a large contribution of the AMV tropical component in line with above-cited papers and Cassou et al. (2004). In 1xAMV, there is no change in NAO-related regimes but a slight, albeit nonsignificant, prevalence of blocking circulations (not shown). We speculate here that the tropical influence is less dominant and that the extratropical SST component of the AMV is a key factor to explain the total response. Warmer SST in the subpolar gyre induces a reduction of the North Atlantic meridional temperature gradient exported upward throughout the troposphere with maximum loading at the intergyre around 45°N; it is responsible for a collocated weakening of the westerly circulation as shown in Fig. 4c (Peings and Magnusdottir 2014). Amplification due to sea ice loss is also expected to play a role in the slowdown of the jet [Deser et al. 2015; Oudar et al. 2017; see the review of Screen et al. (2018) including CNRM-CM5]. Note also that the anticyclonic circulation (Figs. 2j and 5a) is located between Iceland and the United Kingdom in 1xAMV, namely downstream of the maximum SST anomalies over the subpolar gyre (Fig. 1a) and related diabatic heating (precipitation anomalies in Fig. 5a), as opposed to 2xAMV and 3xAMV where the anomalous core in geopotential (Figs. 2k,l) and upper-level streamfunction (Figs. 5b,c) is centered over eastern Greenland. The 1xAMV pattern is consistent with an equivalent barotropic atmospheric response to extratropical SST anomalies resulting from changes in the position or strength of the storm tracks in the presence of an anomalous meridional SST gradient and related altered baroclinicity, as described in Kushnir et al. (2002).
Note finally that the North Pacific–North Atlantic connection due to the remote effects of AMV on the Pacific basin-scale climate may also play a role. Consistently with Ruprich-Robert et al. (2017), results with CNRM-CM5 show a forced response in the Pacific that is reminiscent of the negative phase of the Pacific decadal variability (PDV) in SST. In terms of atmospheric circulation, a slackened Aleutian low as part of a larger-scale Rossby wave pattern cascading into the North Atlantic is also detectable in the model (not shown), but the amplitude of the Pacific–Atlantic teleconnection is lower in CNRM-CM5 compared to other models taken from the DCPP-C database (Ruprich-Robert et al. 2019). We thus interpret the dynamical response over Europe as primarily driven by the local Atlantic influence with some modulation from the Pacific-initiated wave train in line with Ding et al.’s (2017) findings from observations. The role of the PDV, which is likely a function of AMV-forcing amplitude, remains to be better quantified but a deeper analysis goes beyond the scope of the present paper.
d. Mechanisms of the thermodynamical component of the AMV-forced response
As above detailed, the thermodynamical component is computed as the difference between the total modeled response and the estimated contribution of the dynamical changes: it is a residual term encompassing multiple processes. Advection of heat at low-level atmosphere is one of them and has been shown to be a key factor to understand temperature anomalies over a given sector (see, e.g., de Vries et al. 2012). It is assessed from the advection term V ⋅ ∇T, where V stands for wind speed and ∇T for temperature gradient usually taken at 850-hPa level to exclude turbulent and direct surface radiative influence in the boundary layer. In winter, the climatological westerly flow tends to advect relatively warm and humid oceanic air inland toward Europe. During positive AMV, an increase of the thermal advection by the climatological westerly flow is therefore expected because of warmer ocean but we showed that the latter thermodynamical term is counteracted by anomalous easterlies associated with the forced anticyclonic dynamical anomalies located off Europe (Figs. 3a–c). As a final result, individual terms tend to cancel each other in 1xAMV, yet with a slight weakening of the total advection along the Atlantic flank (except Scandinavia), thus contributing to cooling (Fig. 6a), and a slight intensification around 55°N in central Europe leading to warming. Figures 6b and 6c shows a progressive reinforcement of the thermal advection with the amplitude of the AMV along the Atlantic flank. The larger changes are found in 3xAMV with an increase of the advection up to 30% over Germany contributing to large-scale warming from the United Kingdom to Poland. As the easterly wind anomalies are very similar in all AMV ensembles (Fig. 3), this suggests that the weight of the thermodynamical term in the total advection, that is, the transport of temperature anomalies by the climatological westerly flow, becomes dominant and contributes to explain a significant fraction of the full T2m positive anomalies found in 2xAMV and 3xAMV (Figs. 2a–c) consistent with the outcomes of the thermodynamical–dynamical decomposition (Figs. 3a–f).
AMV-forced anomalies for December–February seasonal mean for temperature advection anomalies at 850 hPa (shading interval is 0.2 10−6 K s−1) for (a) 1xAMV, (b) 2xAMV, and (c) 3xAMV. Climatological advection from AMV− is superimposed (black contour interval is 10−6 K s−1; the thicker black contour stands for the zero line). Stippling indicates regions that are above the 95% confidence level of statistical significance based on a two-sided Student’s t test.
Citation: Journal of Climate 33, 7; 10.1175/JCLI-D-19-0428.1
AMV-forced anomalies for December–February seasonal mean for temperature advection anomalies at 850 hPa (shading interval is 0.2 10−6 K s−1) for (a) 1xAMV, (b) 2xAMV, and (c) 3xAMV. Climatological advection from AMV− is superimposed (black contour interval is 10−6 K s−1; the thicker black contour stands for the zero line). Stippling indicates regions that are above the 95% confidence level of statistical significance based on a two-sided Student’s t test.
Citation: Journal of Climate 33, 7; 10.1175/JCLI-D-19-0428.1
AMV-forced anomalies for December–February seasonal mean for temperature advection anomalies at 850 hPa (shading interval is 0.2 10−6 K s−1) for (a) 1xAMV, (b) 2xAMV, and (c) 3xAMV. Climatological advection from AMV− is superimposed (black contour interval is 10−6 K s−1; the thicker black contour stands for the zero line). Stippling indicates regions that are above the 95% confidence level of statistical significance based on a two-sided Student’s t test.
Citation: Journal of Climate 33, 7; 10.1175/JCLI-D-19-0428.1
Changes in surface energy balance are also included in the temperature anomalies driven by thermodynamical processes. No significant anomalies of latent and sensible heat fluxes are observed over Europe (not shown). Significant changes in cloud cover (Figs. 7a–c) are noted, with potential impacts on net longwave (Figs. 7d–f) and net shortwave (Figs. 7g–i) radiation at the surface. In 1xAMV, a significant increase in longwave radiation and decrease in shortwave radiation are noted over the north of Germany, Poland, etc. (Figs. 7d,g), associated with an increase, although nonsignificant, in cloud cover there (Fig. 7a). These anomalies are consistent with the radiative impact of clouds, with a greenhouse effect that tends to increase longwave radiation at surface and a parasol effect that tends to reduce shortwave radiation at the surface. In 2xAMV, the cloud cover decreases almost everywhere over Europe, with significant values over the north of Poland again and over Greece and Turkey (Fig. 7b). These negative cloud anomalies are also associated with a significant increase in longwave radiation and a decrease in shortwave radiation at the surface (Figs. 7e,h). In 3xAMV, the cloud cover increase is particularly pronounced over the northeast of France, the Benelux region, and the north of Germany (Fig. 7c), with a large and significant increase in longwave radiation there (Fig. 7f). The impact of cloud cover on net shortwave anomalies at surface is less clear (Fig. 7i). This might be related to the large decrease in snow cover over large parts of Europe seen in 3xAMV (Fig. 2i), consistent with Lehner et al. (2017), who also found a contribution of snow cover in the temperature residual signal after a similar decomposition based on circulation analogs.
AMV-forced anomalies for December–February seasonal mean for (a)–(c) total cloud cover (shading interval is 0.2%), (d)–(f) net longwave radiation at surface (shading interval is 0.2 W m−2), (g)–(i) net shortwave radiation at surface, (j)–(l) radiative effect due to surface albedo changes in shortwave, and (m)–(o) net shortwave and longwave radiation at surface for (left) 1xAMV, (middle) 2xAMV, and (right) 3xAMV. Positive values represent energy moving toward the surface. Stippling indicates regions that are above the 95% confidence level of statistical significance based on a two-sided Student’s t test.
Citation: Journal of Climate 33, 7; 10.1175/JCLI-D-19-0428.1
AMV-forced anomalies for December–February seasonal mean for (a)–(c) total cloud cover (shading interval is 0.2%), (d)–(f) net longwave radiation at surface (shading interval is 0.2 W m−2), (g)–(i) net shortwave radiation at surface, (j)–(l) radiative effect due to surface albedo changes in shortwave, and (m)–(o) net shortwave and longwave radiation at surface for (left) 1xAMV, (middle) 2xAMV, and (right) 3xAMV. Positive values represent energy moving toward the surface. Stippling indicates regions that are above the 95% confidence level of statistical significance based on a two-sided Student’s t test.
Citation: Journal of Climate 33, 7; 10.1175/JCLI-D-19-0428.1
AMV-forced anomalies for December–February seasonal mean for (a)–(c) total cloud cover (shading interval is 0.2%), (d)–(f) net longwave radiation at surface (shading interval is 0.2 W m−2), (g)–(i) net shortwave radiation at surface, (j)–(l) radiative effect due to surface albedo changes in shortwave, and (m)–(o) net shortwave and longwave radiation at surface for (left) 1xAMV, (middle) 2xAMV, and (right) 3xAMV. Positive values represent energy moving toward the surface. Stippling indicates regions that are above the 95% confidence level of statistical significance based on a two-sided Student’s t test.
Citation: Journal of Climate 33, 7; 10.1175/JCLI-D-19-0428.1
As the increase in net longwave radiation at surface induced by an increase in cloud cover tends to be greater than the associated decrease in net shortwave radiation in winter over Europe, and additionally given the decrease in upward solar radiation due to snow cover reduction, the net total radiation at surface tends to increase in 1xAMV, 2xAMV, and 3xAMV over most of Europe (Figs. 7m–o) although these differences are mostly significant in 3xAMV (Fig. 7o).
The thermodynamical temperature anomalies seem to be mainly explained by three mechanisms: 1) the strengthening of the advection of warmer and moister oceanic air by the climatological westerly flow during positive AMV phases, 2) the decrease of snow cover, and 3) the increase of clouds, both driving changes in shortwave and longwave radiation at surface. The increase in cloud cover may be related to the eastward advection of cloud and/or of warm and moist air by the mean flow from the Atlantic to the European cold continental areas where cloud formation would be enhanced.
4. Conclusions and discussion
In this study, the teleconnection between the AMV and the wintertime climate over Europe is assessed with the CNRM-CM5 coupled model via DCPP-C pacemaker experiments (Boer et al. 2016), in which the modeled North Atlantic SSTs are restored toward anomalies that are characteristic of the observed AMV. The sensitivity of the teleconnection to the AMV amplitude is evaluated thanks to three ensembles of simulations with different amplitudes of targeted SST anomalies. In the first ensemble (1xAMV), which strictly follows the DCPP-C coordinated protocol, the targeted SST anomalies correspond to one standard deviation on the observed AMV. They are respectively doubled and tripled for the 2xAMV and 3xAMV ensembles.
Figure 8 wraps up our findings for surface temperature averaged over Europe. Positive AMV tends to be associated in winter with positive temperature anomalies especially in the 2xAMV and 3xAMV experiments as assessed from ensemble means; in 1xAMV, signals are very weak and barely significant (Fig. 8a). Spatial averages mask some regional features in 1xAMV with a slight cooling over a broad central Europe compensated by warming in Scandinavia and along the Mediterranean Sea to a lower extent. Precipitation anomalies tends to be positive over Europe, except Scandinavia, but their significance is marginal and does not evolve consistently with the amplitude of the AMV.
Spatial average of AMV-forced anomalies for December–February seasonal mean of T2m anomalies over Europe (same domain as in Fig. 2) vs North Atlantic SST (0°–60°N) for the (a) dynamical part, (b) thermodynamical (or residual) part, and (c) total anomalies for 1xAMV (green), 2xAMV (orange), and 3xAMV (magenta). The small dots represent the 10-yr mean response of each member and the big dot stands for the ensemble mean. The slope β obtained from the linear regression between the T2m and the SST anomalies distributions from all the experiments is given in the upper right title of each panel.
Citation: Journal of Climate 33, 7; 10.1175/JCLI-D-19-0428.1
Spatial average of AMV-forced anomalies for December–February seasonal mean of T2m anomalies over Europe (same domain as in Fig. 2) vs North Atlantic SST (0°–60°N) for the (a) dynamical part, (b) thermodynamical (or residual) part, and (c) total anomalies for 1xAMV (green), 2xAMV (orange), and 3xAMV (magenta). The small dots represent the 10-yr mean response of each member and the big dot stands for the ensemble mean. The slope β obtained from the linear regression between the T2m and the SST anomalies distributions from all the experiments is given in the upper right title of each panel.
Citation: Journal of Climate 33, 7; 10.1175/JCLI-D-19-0428.1
Spatial average of AMV-forced anomalies for December–February seasonal mean of T2m anomalies over Europe (same domain as in Fig. 2) vs North Atlantic SST (0°–60°N) for the (a) dynamical part, (b) thermodynamical (or residual) part, and (c) total anomalies for 1xAMV (green), 2xAMV (orange), and 3xAMV (magenta). The small dots represent the 10-yr mean response of each member and the big dot stands for the ensemble mean. The slope β obtained from the linear regression between the T2m and the SST anomalies distributions from all the experiments is given in the upper right title of each panel.
Citation: Journal of Climate 33, 7; 10.1175/JCLI-D-19-0428.1
We apply a flow analog method in the three ensembles to decompose the total temperature and precipitation response in a dynamical part and a residual signal mostly including thermodynamical processes. During a positive phase of the AMV, in all the ensembles, the thermodynamical response is characterized by large-scale and positive T2m (Fig. 8b) and precipitation anomalies over most of Europe. Different mechanisms govern this net response: 1) the advection of positive temperature anomalies by the climatological westerly flow, 2) the radiative effect of increase of cloud cover at the surface, and 3) the decrease of snow cover over central Europe. The intensity of the thermodynamical warming migrates deeper and deeper inland from the Atlantic coast with respect to the amplitude of the AMV forcing with some positive feedback associated with the progressive snow cover disappearance (eastward retraction).
By contrast, the dynamical response is characterized by negative temperature (Fig. 8c) and precipitation anomalies mostly over the northern half of Europe, because of the presence of large-scale AMV-forced northeasterly wind anomalies that counter the climatological advection of relatively warm and moist air from the ocean. We speculate that the net response to AMV forcing in terms of atmospheric dynamics can be understood as a combined effect of extratropical and tropical influences: 1) the tropical branch of the AMV SST anomalies enhances local diabatic heating at the northern edge of the climatological ITCZ acting as a Rossby wave source via a Gill–Matsuno response, which cascades over northern Europe, and 2) positive SST anomalies over the subpolar gyre and associated sea ice melting in all the Nordic seas, responsible for polar amplification, lead to the development of an anomalous high at mid- to high latitudes. Preliminary results from additional twin ensemble experiments (also proposed in DCPP-C), where the 1xAMV full pattern is split into tropical and extratropical anomalies, tend to confirm our interpretation. A southward shift/contraction of the jet-stream is obtained in the tropical-only experiment and is very much similar to 2xAMV and 3xAMV responses. The extratropical-only response is in quadrature and dominated by a decrease of the jet at its northern flank. The latter is consistent with 1xAMV (not shown). We conclude that the tropical component dominates the total response in 2xAMV and 3xAMV with some modulation by the extratropical forcing whereas contributions are comparable in 1xAMV. This has some strong implication for impacts over Europe located at the tail end of the chain of influences. As shown in Fig. 3, despite large-scale easterly anomalies over most of the continent in all AMV ensembles, the precise position of the anomalous pressure centers of action in response to AMV matters a lot (especially for precipitation). The T2m and precipitation responses over Europe are found to be shifted northward as the AMV amplitude increases. Based on our interpretation, the position is very likely controlled by the respective weight between the strength, curvature, waviness, and northeastward extension of the tropically forced Rossby wave on the one hand and the extratropical forcing included sea ice on the other hand, with some possibility for partial nonadditivity of the dynamical signals because of nonlinear processes (polar amplification, tropical convective responses, etc.). We believe that the methodological framework proposed here, namely the decomposition of the AMV impacts over Europe into dynamical and thermodynamical components, could be a useful process-based approach to characterize and understand the intermodel differences regarding the AMV-forced teleconnection obtained from all the CMIP6 models involved in DCPP-C.
As a summary, the thermodynamical and dynamical impacts of the AMV on European temperatures tend to be opposed in CNRM-CM5 and confirm previously results shown by Yamamoto and Palter (2016) and O’Reilly et al. (2017) based on observations. Although it is impossible to strictly isolate the AMV-forced component in the latter, the response in CNRM-CM5 bears some similarities with composite differences between AMV phases for observed T2m and SLP (not shown). However, the dynamical component seems to be more pronounced in the observations than in the model and it is important to highlight here that a large ensemble is needed to get a robust estimation of the model responses. This is consistent with the so-called signal-to-noise paradox raised by Scaife and Smith (2018), which stipulates that climate models tend to underestimate the atmospheric dynamical response to a given forcing and in particular SST. Hence, the real-world weights between the dynamical and thermodynamical response could actually be different from the one estimated here with CNRM-CM5. For weak AMV forcing in CNRM-CM5, both dynamical and residual terms compensate each other and no significant impact of the AMV is obtained over Europe, while a significant warming is found during positive AMV phases in 2xAMV and 3xAMV experiments due to the thermodynamical response, which becomes dominant. Notably, the net temperature anomaly averaged over Europe scales linearly with the amplitudes of the AMV − SST anomalies mostly because of the thermodynamical component as evidenced from Fig. 8. But recall that is not the case regionally, therefore challenging the validity of the so-called pattern-scaling technique to evaluate teleconnectivity and related impacts associated with AMV-type variability. Dividing respectively by 2 and 3 the 2xAMV and 3xAMV forced response in order to get a proxy for the 1xAMV spatial fingerprint does not reproduce the actual map of the 1xAMV response obtained from the model. The pattern correlation in T2m between the patterns from the 2xAMV and 3xAMV ensembles and the real 1xAMV outcome is equal to 0.83 and 0.67, respectively. The same conclusions are found for precipitation with values equal to 0.47 and 0.33. The limit for pattern scaling is speculated here to be related to the relative changes between dynamical and thermodynamical influences partly governed by nonlinear processes (polar amplification, eddy–mean flow interaction, which is crucial at the tail end of the cascading Rossby waves over Europe, snow cover effects, etc.).
Assessing the true degree of linearity of the response is further complicated by some intrinsic limitations related to the experimental protocol based on pacemaker techniques. Since the SST restoring coefficient is fixed, it is more efficient in the tropics than in the extratropics because its strength is function of the ocean mixed layer depth. The actual SSTs are therefore closer to the targeted SSTs in the tropical band dominated by pronounced stratification (shallow mixed layer) than at midlatitudes characterized by deep mixing (thick mixed layer) (not shown). More weight might be therefore artificially given to the AMV-forcing originating from the tropics relative to the one induced by extratropical SST anomalies. To partly overcome this issue, we performed additional test experiments with a restoring coefficient that varies with the mixed layer depth as in Ortega et al. (2017) but only for 1xAMV. Preliminary results suggest that the model response presented here for 1xAMV is robust (not shown) but further analyses would be needed to firmly conclude. Note finally that the relative weight between tropical and extratropical influences is not conserved when the AMV SST anomalies are multiplied by 2 and 3 as done in our study; this hampers a clean investigation of the linearity.
Figure 8 also highlights the weak signal-to-noise ratio of the AMV-forced temperature response over Europe. Even if the spread in North Atlantic SST anomalies for a given AMV forcing is very constrained by the restoring framework, a very large intermember spread is noted in all ensembles for the mean temperature response; it is maximum for the thermodynamical term. The spread is so large that even in the 3xAMV (2xAMV) experiment, 7 (16) members out of 40 have a 10-yr averaged temperature response over Europe of negative sign, despite overall warming effect of the AMV. Note that the AMV experiments are named according to the targeted SSTs and not to the actual SST anomalies obtained in the pacemaker ensembles. As shown in Fig. 1, the actual SST values in 2xAMV and 3xAMV are closer to the targeted SST of 1xAMV and 2xAMV, respectively. As a result, the actual SST anomalies in 2xAMV are far from rare, while those in 3xAMV become extreme but may still be observed since they correspond to two standard deviations of the observed AMV over the historical period.
A last perspective but with potentially strong implications would be to evaluate the sensitivity of the AMV-forced teleconnectivity to the model mean background state. In this work, we investigated the impact of the AMV in so-called preindustrial climate. Nothing guarantees that the AMV teleconnection over Europe is independent of the climate mean state and could then change (or has already changed) as climate is warming due to anthropogenic factors. New twin experiments in which the North Atlantic SST would be restored to the same anomalies as in this study but with a mean state characteristic of current climate (about +1°C) or future climate warming targets depending on future CO2 emission scenarios could be of particular interest. Beyond climate change, linking AMV-forced response to model mean states (and consequently intrinsic biases) could be also a pertinent framework to understand the CMIP6 model diversity within the coordinated DCPP initiative.
Acknowledgments
This work was supported by a grant from Electricité de France (EDF) and by the French National Research Agency (ANR) in the framework of the MORDICUS project (ANR-13-SENV-0002). The authors are grateful to Marie-Pierre Moine, Laure Coquart, and Isabelle d’Ast for technical help to run the model. Computer resources have been provided by Cerfacs. The figures were produced with the NCAR Command Language Software (http://dx.doi.org/10.5065/D6WD3XH5).
REFERENCES
Bellucci, A., and Coauthors, 2015: Advancements in decadal climate predictability: The role of nonoceanic drivers. Rev. Geophys., 53, 165–202, https://doi.org/10.1002/2014RG000473.
Boé, J., L. Terray, C. Cassou, and J. Najac, 2009: Uncertainties in European summer precipitation changes: Role of large scale circulation. Climate Dyn., 33, 265–276, https://doi.org/10.1007/s00382-008-0474-7.
Boer, G. J., and Coauthors, 2016: The Decadal Climate Prediction Project (DCPP) contribution to CMIP6. Geosci. Model Dev., 9, 3751–3777, https://doi.org/10.5194/gmd-9-3751-2016.
Cane, M. A., A. C. Clement, L. N. Murphy, and K. Bellomo, 2017: Low-pass filtering, heat flux, and Atlantic multidecadal variability. J. Climate, 30, 7529–7553, https://doi.org/10.1175/JCLI-D-16-0810.1.
Cassou, C., L. Terray, J. W. Hurrell, and C. Deser, 2004: North Atlantic winter climate regimes: Spatial asymmetry, stationarity with time, and oceanic forcing. J. Climate, 17, 1055–1068, https://doi.org/10.1175/1520-0442(2004)017<1055:NAWCRS>2.0.CO;2.
Cassou, C., M. Minvielle, L. Terray, and C. Périgaud, 2011: A statistical–dynamical scheme for reconstructing ocean forcing in the Atlantic. Part I: Weather regimes as predictors for ocean surface variables. Climate Dyn., 36, 19–39, https://doi.org/10.1007/s00382-010-0781-7.
Cassou, C., Y. Kushnir, E. Hawkins, A. Pirani, F. Kucharski, I.-S. Kang, and N. Caltabiano, 2018: Decadal climate variability and predictability: Challenges and opportunities. Bull. Amer. Meteor. Soc., 99, 479–490, https://doi.org/10.1175/BAMS-D-16-0286.1.
Cattiaux, J., and P. Yiou, 2013: U.S. heat waves of spring and summer 2012 from the flow-analogue perspective [in “Explaining Extreme Events of 2012 from a Climate Perspective”]. Bull. Amer. Meteor. Soc., 94 (9), S10–S13, https://doi.org/10.1175/BAMS-D-13-00085.1.
Clement, A., K. Bellomo, L. N. Murphy, M. A. Cane, T. Mauritsen, G. Rädel, and B. Stevens, 2015: The Atlantic Multidecadal Oscillation without a role for ocean circulation. Science, 350, 320–324, https://doi.org/10.1126/science.aab3980.
Davini, P., and C. Cagnazzo, 2014: On the misinterpretation of the North Atlantic Oscillation in CMIP5 models. Climate Dyn., 43, 1497–1511, https://doi.org/10.1007/s00382-013-1970-y.
Davini, P., J. von Hardenberg, and S. Corti, 2015: Tropical origin for the impacts of the Atlantic multidecadal variability on the Euro-Atlantic climate. Environ. Res. Lett., 10, 094010, https://doi.org/10.1088/1748-9326/10/9/094010.
Dayon, G., J. Boé, and E. Martin, 2015: Transferability in the future climate of a statistical downscaling method for precipitation in France. J. Geophys. Res. Atmos., 120, 1023–1043, https://doi.org/10.1002/2014JD022236.
Deser, C., A. Phillips, V. Bourdette, and H. Teng, 2012: Uncertainty in climate change projections: The role of internal variability. Climate Dyn., 38, 527–546, https://doi.org/10.1007/s00382-010-0977-x.
Deser, C., R. A. Tomas, and L. Sun, 2015: The role of ocean–atmosphere coupling in the zonal-mean atmospheric response to Arctic Sea ice loss. J. Climate, 28, 2168–2186, https://doi.org/10.1175/JCLI-D-14-00325.1.
Deser, C., L. Terray, and A. S. Phillips, 2016: Forced and internal components of winter air temperature trends over North America during the past 50 years: Mechanisms and implications. J. Climate, 29, 2237–2258, https://doi.org/10.1175/JCLI-D-15-0304.1.
de Vries, H., R. J. Haarsma, and W. Hazeleger, 2012: Western European cold spells in current and future climate. Geophys. Res. Lett., 39, L04706, https://doi.org/10.1029/2011GL050665.
Ding, S., W. Chen, J. Feng, and H.-F. Graf, 2017: Combined impacts of PDO and two types of La Niña on climate anomalies in Europe. J. Climate, 30, 3253–3278, https://doi.org/10.1175/JCLI-D-16-0376.1.
Doblas-Reyes, F. J., and Coauthors, 2013: Initialized near-term regional climate change prediction. Nat. Commun., 4, 1715, https://doi.org/10.1038/ncomms2704.
Drévillon, M., C. Cassou, and L. Terray, 2003: Model study of the North Atlantic region atmospheric response to autumn tropical Atlantic sea-surface-temperature anomalies. Quart. J. Roy. Meteor. Soc., 129, 2591–2611, https://doi.org/10.1256/qj.02.17.
Durack, P., and Coauthors, 2018: Toward standardized data sets for climate model experimentation. Eos, Trans. Amer. Geophys. Union, 99, https://doi.org/10.1029/2018EO101751.
Eyring, V., S. Bony, G. A. Meehl, C. A. Senior, B. Stevens, R. J. Stouffer, and K. E. Taylor, 2016: Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization. Geosci. Model Dev., 9, 1937–1958, https://doi.org/10.5194/gmd-9-1937-2016.
Gastineau, G., and C. Frankignoul, 2015: Influence of the North Atlantic SST variability on the atmospheric circulation during the twentieth century. J. Climate, 28, 1396–1416, https://doi.org/10.1175/JCLI-D-14-00424.1.
Gill, A. E., 1980: Some simple solutions for heat-induced tropical circulation. Quart. J. Roy. Meteor. Soc., 106, 447–462, https://doi.org/10.1002/qj.49710644905.
Goddard, L., and Coauthors, 2013: A verification framework for interannual-to-decadal predictions experiments. Climate Dyn., 40, 245–272, https://doi.org/10.1007/s00382-012-1481-2.
Haney, R. L., 1971: Surface thermal boundary condition for ocean circulation models. J. Phys. Oceanogr., 1, 241–248, https://doi.org/10.1175/1520-0485(1971)001<0241:STBCFO>2.0.CO;2.
Hawkins, E., R. S. Smith, J. M. Gregory, and D. A. Stainforth, 2016: Irreducible uncertainty in near-term climate projections. Climate Dyn., 46, 3807–3819, https://doi.org/10.1007/s00382-015-2806-8.
Hodson, D. L. R., R. T. Sutton, C. Cassou, N. Keenlyside, Y. Okumura, and T. Zhou, 2010: Climate impacts of recent multidecadal changes in Atlantic Ocean sea surface temperature: A multimodel comparison. Climate Dyn., 34, 1041–1058, https://doi.org/10.1007/s00382-009-0571-2.
Hoskins, B. J., and D. J. Karoly, 1981: The steady linear response of a spherical atmosphere to thermal and orographic forcing. J. Atmos. Sci., 38, 1179–1196, https://doi.org/10.1175/1520-0469(1981)038<1179:TSLROA>2.0.CO;2.
Kavvada, A., A. Ruiz-Barradas, and S. Nigam, 2013: AMO’s structure and climate footprint in observations and IPCC AR5 climate simulations. Climate Dyn., 41, 1345–1364, https://doi.org/10.1007/s00382-013-1712-1.
Kim, H.-M., P. J. Webster, and J. A. Curry, 2012: Evaluation of short-term climate change prediction in multi-model CMIP5 decadal hindcasts. Geophys. Res. Lett., 39, L10701, https://doi.org/10.1029/2012GL051644.
Kirtman, B., and Coauthors, 2013: Near-term climate change: Projections and predictability. Climate Change 2013: The Physical Science Basis, T. F. Stocker et al., Eds., Cambridge University Press, 953–1028.
Kushnir, Y., W. A. Robinson, I. Bladé, N. M. J. Hall, S. Peng, and R. Sutton, 2002: Atmospheric GCM response to extratropical SST anomalies: Synthesis and evaluation. J. Climate, 15, 2233–2256, https://doi.org/10.1175/1520-0442(2002)015<2233:AGRTES>2.0.CO;2.
Lehner, F., C. Deser, and L. Terray, 2017: Toward a new estimate of “time of emergence” of anthropogenic warming: Insights from dynamical adjustment and a large initial-condition model ensemble. J. Climate, 30, 7739–7756, https://doi.org/10.1175/JCLI-D-16-0792.1.
Meehl, G. A., and Coauthors, 2014: Decadal climate prediction: An update from the trenches. Bull. Amer. Meteor. Soc., 95, 243–267, https://doi.org/10.1175/BAMS-D-12-00241.1.
Michel, C., and G. Rivière, 2014: Sensitivity of the position and variability of the eddy-driven jet to different SST profiles in an aquaplanet general circulation model. J. Atmos. Sci., 71, 349–371, https://doi.org/10.1175/JAS-D-13-074.1.
Msadek, R., and Coauthors, 2014: Predicting a decadal shift in North Atlantic climate variability using the GFDL forecast system. J. Climate, 27, 6472–6496, https://doi.org/10.1175/JCLI-D-13-00476.1.
O’Reilly, C. H., M. Huber, T. Woollings, and L. Zanna, 2016: The signature of low-frequency oceanic forcing in the Atlantic Multidecadal Oscillation. Geophys. Res. Lett., 43, 2810–2818, https://doi.org/10.1002/2016GL067925.
O’Reilly, C. H., T. Woollings, and L. Zanna, 2017: The dynamical influence of the Atlantic multidecadal oscillation on continental climate. J. Climate, 30, 7213–7230, https://doi.org/10.1175/JCLI-D-16-0345.1.
Ortega, P., E. Guilyardi, D. Swingedouw, J. Mignot, and S. Nguyen, 2017: Reconstructing extreme AMOC events through nudging of the ocean surface: A perfect model approach. Climate Dyn., 49, 3425–3441, https://doi.org/10.1007/s00382-017-3521-4.
Oudar, T., E. Sanchez-Gomez, F. Chauvin, J. Cattiaux, L. Terray, and C. Cassou, 2017: Respective roles of direct GHG radiative forcing and induced Arctic sea ice loss on the Northern Hemisphere atmospheric circulation. Climate Dyn., 49, 3693–3713, https://doi.org/10.1007/s00382-017-3541-0.
Peings, Y., and G. Magnusdottir, 2014: Forcing of the wintertime atmospheric circulation by the multidecadal fluctuations of the North Atlantic Ocean. Environ. Res. Lett., 9, 034018, https://doi.org/10.1088/1748-9326/9/3/034018.
Peings, Y., G. Simpkins, and G. Magnusdottir, 2016: Multidecadal fluctuations of the North Atlantic Ocean and feedback on the winter climate in CMIP5 control simulations. J. Geophys. Res. Atmos., 121, 2571–2592, https://doi.org/10.1002/2015JD024107.
Qasmi, S., C. Cassou, and J. Boé, 2017: Teleconnection between Atlantic multidecadal variability and European temperature: Diversity and evaluation of the Coupled Model Intercomparison Project phase 5 models. Geophys. Res. Lett., 44, 11 140–11 149, https://doi.org/10.1002/2017GL074886.
Raible, C. C., F. Lehner, J. F. González-Rouco, and L. Fernández-Donado, 2014: Changing correlation structures of the Northern Hemisphere atmospheric circulation from 1000 to 2100 AD. Climate Past, 10, 537–550, https://doi.org/10.5194/cp-10-537-2014.
Rivière, G., 2009: Effect of latitudinal variations in low-level baroclinicity on eddy life cycles and upper-tropospheric wave-breaking processes. J. Atmos. Sci., 66, 1569–1592, https://doi.org/10.1175/2008JAS2919.1.
Robson, J., R. Sutton, and D. Smith, 2012: Initialized decadal predictions of the rapid warming of the North Atlantic Ocean in the mid 1990s. Geophys. Res. Lett., 39, L19713, https://doi.org/10.1029/2012GL053370.
Robson, J., R. Sutton, and D. Smith, 2013: Predictable climate impacts of the decadal changes in the ocean in the 1990s. J. Climate, 26, 6329–6339, https://doi.org/10.1175/JCLI-D-12-00827.1.
Ruprich-Robert, Y., and C. Cassou, 2015: Combined influences of seasonal East Atlantic pattern and North Atlantic Oscillation to excite Atlantic Multidecadal Variability in a climate model. Climate Dyn., 44, 229–253, https://doi.org/10.1007/s00382-014-2176-7.
Ruprich-Robert, Y., R. Msadek, F. Castruccio, S. Yeager, T. Delworth, and G. Danabasoglu, 2017: Assessing the climate impacts of the observed Atlantic multidecadal variability using the GFDL CM2.1 and NCAR CESM1 global coupled models. J. Climate, 30, 2785–2810, https://doi.org/10.1175/JCLI-D-16-0127.1.
Ruprich-Robert, Y., and Coauthors, 2019: The impacts of the Atlantic multidecadal variability on tropical Pacific as assessed from CMIP6/DCPP-C idealized simulations. AGU Fall Meeting 2019, San Francisco, CA, Amer. Geophys. Union, Abstract OS23A-04, https://agu.confex.com/agu/fm19/meetingapp.cgi/Paper/591495.
Sanchez-Gomez, E., C. Cassou, Y. Ruprich-Robert, E. Fernandez, and L. Terray, 2016: Drift dynamics in a coupled model initialized for decadal forecasts. Climate Dyn., 46, 1819–1840, https://doi.org/10.1007/s00382-015-2678-y.
Scaife, A. A., and D. Smith, 2018: A signal-to-noise paradox in climate science. npj Climate Atmos. Sci., 1, 1–8, https://doi.org/10.1038/s41612-018-0038-4.
Screen, J. A., T. J. Bracegirdle, and I. Simmonds, 2018: Polar climate change as manifest in atmospheric circulation. Curr. Climate Change Rep., 4, 383–395, https://doi.org/10.1007/s40641-018-0111-4.
Simpson, I. R., C. Deser, K. A. McKinnon, and E. A. Barnes, 2018: Modeled and observed multidecadal variability in the North Atlantic jet stream and its connection to sea surface temperatures. J. Climate, 31, 8313–8338, https://doi.org/10.1175/JCLI-D-18-0168.1.
Simpson, I. R., S. G. Yeager, K. A. McKinnon, and C. Deser, 2019: Decadal predictability of late winter precipitation in western Europe through an ocean–jet stream connection. Nat. Geosci., 12, 613–619, https://doi.org/10.1038/s41561-019-0391-x.
Sutton, R. T., and D. L. R. Hodson, 2005: Atlantic Ocean forcing of North American and European summer climate. Science, 309, 115–118, https://doi.org/10.1126/science.1109496.
Sutton, R. T., and B. Dong, 2012: Atlantic Ocean influence on a shift in European climate in the 1990s. Nat. Geosci., 5, 788–792, https://doi.org/10.1038/ngeo1595.
Taylor, K. E., R. J. Stouffer, and G. A. Meehl, 2012: An overview of CMIP5 and the experiment design. Bull. Amer. Meteor. Soc., 93, 485–498, https://doi.org/10.1175/BAMS-D-11-00094.1.
Terray, L., 2012: Evidence for multiple drivers of North Atlantic multi-decadal climate variability. Geophys. Res. Lett., 39, L19712, https://doi.org/10.1029/2012GL053046.
Terray, L., and C. Cassou, 2002: Tropical Atlantic sea surface temperature forcing of quasi-decadal climate variability over the North Atlantic–European region. J. Climate, 15, 3170–3187, https://doi.org/10.1175/1520-0442(2002)015<3170:TASSTF>2.0.CO;2.
Ting, M., Y. Kushnir, R. Seager, and C. Li, 2009: Forced and internal twentieth-century SST trends in the North Atlantic. J. Climate, 22, 1469–1481, https://doi.org/10.1175/2008JCLI2561.1.
Towler, E., D. PaiMazumder, and J. Done, 2018: Toward the application of decadal climate predictions. J. Appl. Meteor. Climatol., 57, 555–568, https://doi.org/10.1175/JAMC-D-17-0113.1.
Voldoire, A., and Coauthors, 2013: The CNRM-CM5.1 global climate model: Description and basic evaluation. Climate Dyn., 40, 2091–2121, https://doi.org/10.1007/s00382-011-1259-y.
Yamamoto, A., and J. B. Palter, 2016: The absence of an Atlantic imprint on the multidecadal variability of wintertime European temperature. Nat. Commun., 7, 10930, https://doi.org/10.1038/ncomms10930.
Yeager, S. G., and J. I. Robson, 2017: Recent progress in understanding and predicting Atlantic decadal climate variability. Curr. Climate Change Rep., 3, 112–127, https://doi.org/10.1007/s40641-017-0064-z.
Yeager, S. G., A. R. Karspeck, and G. Danabasoglu, 2015: Predicted slowdown in the rate of Atlantic sea ice loss. Geophys. Res. Lett., 42, 10 704–10 713, https://doi.org/10.1002/2015GL065364.
Yeager, S. G., and Coauthors, 2018: Predicting near-term changes in the Earth system: A large ensemble of initialized decadal prediction simulations using the Community Earth System Model. Bull. Amer. Meteor. Soc., 99, 1867–1886, https://doi.org/10.1175/BAMS-D-17-0098.1.
Zhang, L., and C. Wang, 2013: Multidecadal North Atlantic sea surface temperature and Atlantic meridional overturning circulation variability in CMIP5 historical simulations. J. Geophys. Res. Oceans, 118, 5772–5791, https://doi.org/10.1002/jgrc.20390.
Zhang, R., and T. L. Delworth, 2006: Impact of Atlantic multidecadal oscillations on India/Sahel rainfall and Atlantic hurricanes. Geophys. Res. Lett., 33, L17712, https://doi.org/10.1029/2006GL026267.
Zhang, R., R. Sutton, G. Danabasoglu, Y.-O. Kwon, R. Marsh, S. G. Yeager, D. E. Amrhein, and C. M. Little, 2019: A review of the role of the Atlantic meridional overturning circulation in Atlantic multidecadal variability and associated climate impacts. Rev. Geophys., 57, 316–375, https://doi.org/10.1029/2019RG000644.
