Jinghu Cho , Gungzhou Fn , Ruiqing Ding

a School of Atmospheric Sciences, Chengdu University of Information Technology, Chengdu, China

b State Key Laboratory of Earth Surface Processes and Resource Ecology, Beijing Normal University, Beijing, China

Keywords:North American Dipole ENSO Coupled ocean—atmosphere model

ABSTRACT The North American Dipole (NAD) is a north—south seesaw pattern of sea level pressure anomalies over the western tropical North Atlantic and northeastern North America. Previous observational studies have demonstrated that the NAD can affect the outbreak of El Niño—Southern Oscillation (ENSO) events. The present study analyzed the NAD—ENSO relationship as simulated by a coupled ocean—atmosphere model —namely, the Flexible Global Ocean—Atmosphere—Land System model, gridpoint version 2 (FGOALS-g2). Results indicated that the model can replicate a distinct dipole comprised of a low over northeastern North America and a high over the western tropical North Atlantic, which is the signature feature of the NAD. Further analysis verified that the winter NAD can initiate the central equatorial Pacific warming in the subsequent winter by effectively forcing an anticyclonic flow and sea surface temperature (SST) warming over the northeastern subtropical Pacific (NESP) during late winter or early spring. In addition, the probability of an El Niño event was increased by a factor of 1.8 in the assimilation experiment with the NAD. By comparison, the winter Northern Atlantic Oscillation had no significant impact on the occurrence of ENSO a year later owing to its failure to induce the SST and surface wind anomalies over the NESP.

1. Introduction

El Niño—Southern Oscillation (ENSO) is the strongest signal in the interannual variation of Earth’s ocean—atmosphere system and has received extensive attention owing to its impacts on global climate( Bjerknes, 1969 ; Van Loon and Madden, 1981 ; Rasmusson and Wallace, 1983 ; Schopf and Suarez, 1988 ; Jin, 1997 ). Given its far-reaching influence, it is pertinent to examine the factors contributing to its occurrence.

Previous studies have suggested that extratropical sea level pressure (SLP) variations may influence the onset of ENSO ( Yu and Rienecker, 1998 ; Yu and Kim, 2011 ). The North Pacific Oscillation (NPO;Rogers, 1981 ) has been proven to influence the onset of ENSO through the seasonal foot-printing mechanism ( Vimont et al., 2001 , 2003 , 2009 ).Also, while numerous studies have focused on the possible link between the North Atlantic Oscillation (NAO; Rogers and Van Loon, 1979 ) and ENSO, many other studies have demonstrated a great deal of uncertainty in the relationship between them ( Li and Lau, 2012 ; Zhang et al., 2015 ,2019 ).

Recently, Ding et al. (2017) proposed that there is a meridional seesaw pattern of SLP anomalies between the western tropical North Atlantic and northeastern North America, named the North American Dipole (NAD). Although both the winter NAD and the NAO present seesaw characteristics of SLP anomalies in the North Atlantic, the center of the SLP anomalies associated with the NAD is further west and south compared to the NAO (Fig. S1). In particular, the winter NAD is an extratropical forcing for the onset of ENSO in the subsequent winter( Ding et al., 2017 , 2019 ). Also, it may act as a distinctive precursor for the central Pacific (CP)-type El Niño as compared to other ENSO precursors such as the NPO and the Pacific—North America pattern ( Horel and Wallace, 1981 ). Using observational data, Ding et al. (2017) showed that the mechanism through which the NAD affects ENSO events bears a resemblance to the subtropical teleconnection mechanism suggested by Ham et al. (2013) , which is used to account for the influence of SST anomalies over the northern tropical Atlantic on subsequent ENSO events.

Fig. 1. Correlation map of the three-month-averaged SST (shaded; units:°C), surface wind (vectors; units: m s —1 ) and precipitation (stippled; units: mm d —1 ) anomalies with the DJFM-averaged (a—d) NADI and (e—h) NAOI for DJF, MAM, JJA, and SON. Positive (red) and negative (blue) SST anomalies with correlations significant at or above the 90% confidence level are shaded. Only wind vectors significant at the 90% confidence level are shown. Positive (green) and negative (red) precipitation anomalies with correlations significant at or above the 90% confidence level are stippled.

Based on observational analysis, the NAD may be closely linked to the occurrence of ENSO, which implies that the NAD variability may provide a good reference for the prediction of ENSO. However, there is still a degree of uncertainty regarding how coupled ocean—atmosphere models perform when it comes to simulating the NAD—ENSO relationship. Thus, the purpose of this study was to examine if the influence of the winter NAD on the subsequent ENSO can be reproduced in the coupled ocean—atmosphere model FGOALS-g2 (Flexible Global Ocean—Atmosphere—Land System Model, gridpoint version 2). By analyzing the NAD—ENSO connection and the associated physical processes between them in FGOALS-g2, this study contributes to improving the predictive precision of ENSO. The remainder of the paper is organized as follows:Section 2 introduces the data and methods. Section 3 reports on the NAD—ENSO relationship and the associated physical processes between them in FGOALS-g2. Finally, a summary of the study’s key findings is presented in Section 4 .

Fig. 2. (a) Composite differences in the DJFM-averaged SLP anomalies between the ASSIM-NAD and CTL experiments. (b) As in (a) but between the ASSIM-NAO and CTL experiments. Positive (red) and negative (blue) SLP anomalies with correlations significant at or above the 90% confidence level are shaded.

2. Data and methods

2.1. Data

The monthly SST data used in this study were from the Hadley Centre Sea Ice and SST dataset ( Rayner et al., 2003 ). The monthly precipitation data were extracted from NOAA’s Climate Prediction Center Merged Analysis of Precipitation dataset ( Xie and Arkin, 1997 ). Monthly atmospheric circulation data, including winds, SLP, specific humidity, velocity potential, and air temperature measurements, were procured from the Kalnay et al. (1996) and Kistler et al. (2001) NCEP—NCAR reanalyses. The period of coverage was from 1979 to 2019. The long-term linear trend and seasonal cycles were removed from all the variables prior to analysis in this study.

2.2. Index

Similar to Ding et al. (2017), the NAD index (NADI) used in this study was defined as the difference in normalized SLP anomalies between the southern (10°—30°N, 90°—60°W) and northern (55°—70°N, 75°—45°W)poles (southern minus northern pole) after linearly excluding the effect of ENSO. As a comparison, we also calculated the NAO index (NAOI),which was defined as the difference in the normalized SLP zonally averaged over the North Atlantic sector from 80°W to 30°E between 35°N and 65°N ( Li and Wang, 2003) . According to Zhou et al. (2019), the maximum seasonal variance between the NAD and NAO ranges from December to March (DJFM). Therefore, our analysis was based on the four-month-average DJFM NADI and NAOI. Fig. S1(c) depicts the interannual time series of the winter NADI (blue line) and NAOI (red line)for 1979—2018, with a correlation of 0.49 (significant at the 99% confidence level), which is an indication that the variability in the NAD is independent of the NAO.

2.3. Numerical experiment

The coupled model used in this study, FGOALS-g2, was developed at the State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics (LASG), Institute of Atmospheric Physics (IAP), at the Chinese Academy of Sciences ( Li et al., 2013 ). The model is composed of four separate component models —namely, an atmospheric model (GAMIL2: Grid-point Atmospheric Model of IAP LASG,version 2); an oceanic model (LICOM2: LASG IAP Climate System Ocean Model, version 2); a sea-ice model (CICE4-LASG), which is an improved version of CICE4.0 (the Los Alamos Sea Ice Model or Community Ice Code, version 4); and a land model (CLM3: Community Land Model,version 3).

Although the effect of the NAD on ENSO has been demonstrated observationally, we need to further verify this relationship by using a coupled circulation model. Thus, three experiments were designed as follows: a control (CTL), and two assimilation experiments. The CTL experiments was a historical run driven by observed external forces from 1979 to 2018.

The assimilation method employed in this study used the DRP-4DVar(dimension-reduced projection four-dimensional variational data assimilation) scheme of Wang et al. (2010) . The data used in the assimilation were the monthly air temperature (T), zonal wind (U), and meridional wind (V), with a horizontal grid resolution of 3° × 3°, from the ERAInterim dataset at 1000 hPa, 925 hPa, and 850 hPa from 1979 to 2018( Dee et al., 2011 ). The first assimilation experiment (ASSIM-NAD) added the NAD-relatedU,V, andTanomalies over northeastern North America and the western tropical North Atlantic (0°—80°N, 120°—30°W), which were obtained from regressions of the DJFMU,V, andTanomalies onto the concurrent NADI (left-hand column of Fig. S2). The second assimilation experiment (ASSIM-NAO) was the same as ASSIM-NAD but with NAO-relatedU,V, andTanomalies over the North Atlantic (0°—80°N,120°W—30°E; right-hand column of Fig. S2). The initialization was cycled for 40 years from 1978 to 2018 with one-month assimilation windows and the output from the past 30 years was applied in the composite analysis.

2.4. Methods

This study adopted the use of the following statistical techniques: composite analysis, correlation analysis, and linear regression.The statistical significance is based on a two-tailed Student’st-test,

Fig. 3. (a—d) Composite differences in the three-month-averaged SST (shaded; units:°C), surface wind (vectors; units: m s —1 ), and precipitation (stippled; units: mm d —1 ) anomalies between the ASSIM-NAD and CTL experiments for DJF, MAM, JJA, and SON. (e—g) As in (a—d) but between the ASSIM-NAO and CTL experiments.Positive (red) and negative (blue) SST anomalies with correlations significant at or above the 90% confidence level are shaded. Only wind vectors significant at the 90% confidence level are shown. Positive (green) and negative (red) precipitation anomalies with correlations significant at or above the 90% confidence level are stippled.

where the effective number of degrees of freedom (N∗) based on Bretherton et al. (1999) is calculated as follows:

whereNis the sample size, andRxandRyrepresent the lag-one autocorrelations of two time seriesxandy, respectively.

3. Results

Analyzing the main evolutionary characteristics of the observed data during the period when an ENSO event is influenced by the NAD is the key to establishing the feasibility of simulating the NAD—ENSO relationship in the coupled ocean—atmosphere model. Fig. 1 illustrates the concurrent and lagged correlation maps of the DJFM-averaged NADI and NAOI with the three-month-averaged surface wind and SST anomalies for December—February (DJF), March—May (MAM), June—August (JJA),and September—November (SON).

Fig. 4. (a) Composite differences in the SST over the western Pacific (Niño3.4 region; 5°S—5°N, 170°—120°W) between the ASSIM-NAD and CTL experiments from January to November. Error bars indicate the 95% confidence intervals. (b) Probability distribution of November Niño3.4 SST in the ASSIM-NAD (orange bars) and CTL (blue bars) experiments. Note that an El Niño (La Niña) event is defined as the magnitude of the Niño3.4 SST index being greater (less) than 1.0 with positive(negative) standard deviation.

It can be concluded that there are two processes by which the NAD affects the occurrence of ENSO. The first is that the winter NAD influences the SST and surface wind over the northeastern subtropical Pacific(NESP) via simultaneous anticyclonic flow related to the NAD, which is conducive to stimulating an initial warming over the NESP in late winter or early spring. On the other hand, the spring northern tropical Atlantic SST cooling forced by the winter NAD enhances this anticyclonic flow over the NESP with a Gill-type Rossby wave response ( Gill, 1980 ), and ultimately the NAD-induced NESP warming can favor the development of a Pacific meridional mode, thereby causing the onset of CP-type El Niño in the subsequent winter ( Ding et al., 2019 ). Conversely, the winter NAO shows no significant effect on the onset of ENSO a year later owing to its inability to stimulate the NESP SST warming. Note that, while the observational results above demonstrate the existence of a NAD—ENSO relationship, it is unclear how well the climate model, FGOALS-g2, performs in simulating the link between the NAD and ENSO. Next, therefore, further analysis of the coupled model results is presented.

First, we calculated the composite differences in winter SLP anomalies between the ASSIM-NAD and CTL experiments ( Fig. 2 (a)). As we can see, the model can replicate a distinct dipole comprised of a high over the western tropical North Atlantic and a low over northeastern North America, which is the signature feature of the NAD. In terms of the location of the NAD’s center, the model-simulated magnitudes of the NAD positive center show a northward displacement when compared to the observation (Fig. S1(a)). Also, the NAD negative center is more extensive than observed, covering almost all of North America.Likewise, the composite differences in winter SLP anomalies between the ASSIM-NAO and CTL experiments were also calculated ( Fig. 2 (b)).From the results, we can see that the model is capable of reproducing the NAO-like dipole pattern over the North Atlantic. In terms of the location of the NAO center, the model-simulated NAO negative center is located north of 60°N over the Western Hemisphere, which is clearly more westward than the observed NAO, albeit with weaker intensity.In short, the available evidence from the analysis gives reason to infer that the model overestimates the northern pole of both the NAD and the NAO.

To investigate the NAD—ENSO relationship in FGOALS-g2, the composite differences in the seasonal evolutions of SST, surface wind, and precipitation anomalies between the ASSIM-NAD and CTL experiments are illustrated in the left-hand column of Fig. 3 . The model predicts a slightly weaker tripole-like SST anomaly pattern over the North Atlantic and the center of the SST anomalies is situated more to the east than observed during the winter and spring. Notably, there are much warmer SSTs over the NESP from winter to summer, which we believe to be a consequence of the concurrent anticyclonic flow related with the winter NAD. Meanwhile, the spring northern tropical Atlantic SST cooling induced by the winter NAD also enhances the anticyclonic flow and the SST warming over the NESP, which promotes the development of the Pacific meridional mode in several months. Ultimately, a distinct CP-like El Niño pattern can be seen in autumn. In addition, similar composite differences of the seasonal evolutions of the SST, surface wind and precipitation anomalies between the ASSIM-NAO and CTL experiments are displayed in the right-hand column of Fig. 3 . Comparatively, the winter NAO has no significant impact on the occurrence of ENSO a year later owing to its failure to induce the SST and surface wind anomalies over the NESP. The results above are broadly consistent with those described by the observational data ( Fig. 1 ).

To further estimate the relationship between the NAD and ENSO quantitatively, we computed the composite differences in the SST over the western Pacific (Niño3.4 region: 5°S—5°N, 170°—120°W) between the ASSIM-NAD and CTL experiments from January to November( Fig. 4 (a)). It can be seen that the positive SST over the equatorial Pacific is emerging from late spring or boreal summer, with further intensification in the following seasons, and this is consistent with the depiction in Fig. 3 . In addition, the model appears to indicate that El Niño events are more likely to occur in the ASSIM-NAD experiment ( Fig. 4 (b)). Specifically, the probability of an El Niño event is increased by a factor of 1.8 in the assimilation experiment with the NAD. Conversely, the probability of La Niña is reduced to 55% in the ASSIM-NAD experiment, which clearly implies that the winter NAD favors the onset of El Niño in the following winter.

4. Summary and discussion

This study investigated the NAD—ENSO relationship as simulated by a coupled ocean—atmosphere model, FGOALS-g2. The results indicated that the spatial pattern of the winter NAD and the NAO can be reproduced by the model. There was also evidence that the NAD is situated more to the west than the NAO, albeit with weaker intensity. In addition, FGOALS-g2 overestimates the northern pole of both the NAD and the NAO.

Although only the atmospheric component of the coupled model is assimilated, the model can transfer observed information from the atmosphere to the ocean through processes of air—sea coupling. The numerical experiments indicated that the winter NAD can excite a simultaneous anticyclonic flow and initial warming over the NESP, which is conducive to developing the NESP SST warming in late winter or early spring. Also, the springtime northern tropical Atlantic SST associated with the NAD strengthens the SST warming over the NESP, extending to the tropics in favor of the development of the Pacific meridional mode.Finally, more warming in the central equatorial Pacific than in the eastern Pacific occurs, which is an indication of the onset of a CP-like El Niño event in the subsequent winter.

Furthermore, the model appears to predict that El Niño events are more likely to occur in the ASSIM-NAD experiment. Specifically, the probability of an El Niño is increased by a factor of 1.8 in the assimilation experiment with the NAD. Conversely, the probability of a La Niña is reduced to 55% with the ASSIM-NAD experiment, and this clearly shows that the winter NAD favors the development of an El Niño during the subsequent winter season. By comparison, the winter NAO has no significant impact on the occurrence of ENSO a year later owing to its failure to induce the SST and surface wind anomalies over the NESP.With the frequent occurrence of CP-type El Niño in recent decades, a better understanding of the NAD can provide a clear predictor for the occurrence of different El Niño events.

Declaration of Competing Interest

No potential conflict of interest was reported by the authors.

Funding

This work was jointly supported by the National Natural Science Foundation of China [grant number 41975070 ], and the State Key Laboratory of Tropical Oceanography, South China Sea Institute of Oceanology, Chinese Academy of Sciences [project number LTO1901 ].

Supplementary materialsSupplementary material associated with this article can be found, in the online version, at doi: 10.1016/j.aosl.2021.100058 .