Asymmetrical Modulation of the Relationship Between the Western Pacific Pattern and El Niño–Southern Oscillation by the Atlantic Multidecadal Oscillation in the Boreal Winter

Tropical atmospheric convection generated by the El Niño–Southern Oscillation (ENSO) plays a crucial role in affecting the western Pacific pattern (WP) in the boreal winter by triggering an atmospheric teleconnection. Here we show from analysis of observations and model simulations that the Atlantic Multidecadal Oscillation (AMO) asymmetrically modulates the relationship between ENSO and WP. We find a significant modulatory effect of AMO on the relationship between wintertime El Niño and WP. A robust El Niño−WP relation can be attributed to the negative AMO phase (−AMO), yet a weak relationship during the positive AMO phase (+AMO). In contrast, the relationship between La Niña and WP is independent of AMO modulation. Furthermore, during the −AMO period, stronger El Niño amplitudes lead to stronger atmospheric convection anomalies over the tropical western North Pacific, which excites stronger atmospheric teleconnection and thus has a more significant effect on WP than during the +AMO period.

• The connection between El Niño-Southern Oscillation (ENSO) and the western Pacific pattern (WP) varies markedly in different Atlantic Multidecadal Oscillation (AMO) phases • A robust connection between ENSO and the WP in the boreal winter can only be observed during the negative AMO phase • The AMO influences the ENSO-WP relationship via modulating the ENSO amplitude and the associated change in atmospheric convection

Supporting Information:
Supporting Information may be found in the online version of this article.
There are two lines of thought concerning the formation and maintenance of the WP. On the one hand, it is generally well-accepted that the WP is an intrinsic mode of atmospheric circulation. Lau (1988) and Lau and Nath (1991) revealed the critical role of the interaction between synoptic-scale eddies and mean flows in the formation and maintenance of the WP. Tanaka et al. (2016) proposed that the WP is a dynamical mode that can maintain itself through efficient baroclinic energy conversion from the climatological mean flow. More recently, Zhuge and Tan (2021) suggested that baroclinic energy conversion is the primary source of energy that drives the WP, and in particular, the feedback forcing of transient eddies acts as a key source of kinetic energy for the WP its growing stage of WP, while a strong kinetic energy sink during the decay stage. On the other hand, several studies have indicated that the WP can be excited by external forcings, such as sea surface temperature (SST) anomalies in the tropical Pacific and Arctic sea-ice anomalies (Furtado et al., 2012;Horel & Wallace, 1981;Kodera, 1998;Linkin & Nigam, 2008;Nakamura et al., 2015). Horel and Wallace (1981) first proposed that SST anomalies related to the El Niño-Southern Oscillation (ENSO) in the tropical Pacific could exert an impact on the WP by triggering an atmospheric teleconnection. Dai and Tan (2016) reported a significant modulation effect of ENSO on WP events, which may have additional implications for triggering the stratospheric sudden warming events. However, the connections between the WP and its related external forcings are unstable. The evidence for the various ENSO and WP connections can be seen in the distinct extratropical atmospheric responses arising from the varying tropical Pacific SST anomalies (Furtado et al., 2012;Yeh et al., 2015). In this sense, Aru et al. (2022) initially determined the pronounced interdecadal variability of the WP, highlighting that the distinct ENSO-WP relationship can primarily be attributed to changes in the WP variability over the past few decades. Nevertheless, the reason for the unstable ENSO-WP relationship remains to be explored.
ENSO typically maintains a sensational impact on the extratropical North Pacific through the "atmospheric bridge," which generally refers to atmospheric teleconnection patterns over the North Pacific (e.g., Alexander et al., 2002;Bjerknes, 1969;Cai et al., 2019;W. Chen et al., 2019;Trenberth et al., 1998). The original idea describes physical links between the ENSO-related tropical Pacific SST amplitudes and ENSO-induced climate anomalies (Chung et al., 2014;Hoerling et al., 2001). For instance, in Asia (W. Chen et al., 2013;Wang et al., 2017), the United States (Hoell et al., 2016), and Maritime Continent (Jia et al., 2016), the precipitation response is sensitive to the ENSO intensity. Subsequently, studies have emphasized that the ENSO amplitude directly contributes to the extratropical atmospheric circulation response, which can further be attributed to striking climatic impacts (Z. Chen et al., 2022;Jia & Ge, 2017;Krishnamurthy et al., 2016). For instance, Z. Chen et al. (2022) proposed that strong El Niño events may play a crucial role in central Asian precipitation anomalies by significantly modulating the anomalous upper-level moisture divergence in the tropical central-eastern Pacific. Although the ENSO variability and ENSO-related climate impacts have been extensively investigated, questions remain as to how ENSO variability will affect the extratropical circulation and how the ENSO-WP relationship will respond to the ENSO variability.
Existing research recognizes the essential role played by the Atlantic Multidecadal Oscillation (AMO) in modulating the ENSO amplitude (Cai et al., 2019;Dong et al., 2006;Kang et al., 2014;J. Y. Yu et al., 2015). The negative (positive) phase of the AMO corresponds to an increase (decrease) in the ENSO amplitude (Dong & Sutton, 2007;Timmermann et al., 2007). Because the AMO causes significant changes in the ENSO amplitude, which in turn drives the extratropical circulation and climate effects (Geng et al., 2017(Geng et al., , 2020Gong et al., 2020;Zhao et al., 2022), a question naturally arises: is the relationship between ENSO and WP modulated by the AMO?
If so, what are the underlying mechanisms of this modulation? In the present study, we provide observational evidence for distinct differences in the modulation of the AMO phases on the ENSO-WP relationship. The mechanisms behind the modulating effect of the AMO are further examined.

Data and Methods
The present study utilizes monthly data, including the geopotential height, 10-m wind, zonal and meridional winds, precipitation, and relative and specific humidity, from the European Centre for Medium-Range Weather Forecaste twentieth-century reanalysis (ERA-20C; Poli et al., 2016), with a 1° × 1° horizontal resolution and 37 pressure levels from 1900 to 2010. The monthly mean SST data are derived from the NOAA Extended Re-construction SST, version 5 (Huang et al., 2017). The data set maintains a resolution of 2° × 2° in a latitudelongitude grid and is available from 1854 to the present.
Monthly anomalies for variables are calculated with respect to the period 1900-2010 climatology. Then, we calculated the winter (DJF) average from the monthly anomaly data. All data were quadratically detrended prior to the analysis. A 2-9-a bandpass Lancozs filter is applied to extract the interannual variability (Duchon, 1979).

of 10
The Niño3.4 index, which is defined as the area-averaged SST anomalies over 5°S-5°N, 120°-170°W, is used to delineate the ENSO characteristics. The rest of the definitions of indices applied in this study are specified in Text S1 of Supporting Information S1.
A composite analysis is utilized in this study, and the statistical significance is evaluated by a two-tailed Student's t-test. Hereafter, unless otherwise stated, this study focused on the 1900-2010 DJF variations. A linear baroclinic model (LBM) is adopted to verify the observational hypotheses; it is based on linearized primitive equations, with a T42L20 resolution in the dry version (Watanabe & Kimoto, 2000). Figure 1a shows the spatial pattern of the winter WP, represented by a regression of the quadratically detrended Z500 anomalies onto the normalized WP index (definition of the WP index presented in the caption of Figure 1). The winter WP pattern is characterized by a meridional dipole mode over the WNP. The winter WP-related winter SST anomalies present a distinctive El Niño warming pattern in the tropical central-eastern Pacific, indicative of the close connection between the WP and ENSO in the winter (Figure 1b), which is consistent with the findings of previous studies (Aru et al., 2022;Linkin & Nigam, 2008). Figures 1c and 1d reveal that the circulation and SST associated with the Niño3.4 index exhibit a typical pattern of El Niño SST warming; they also indicate that the extratropical circulation is WP-like in the WNP and accompanied by a Pacific North America (PNA)-like circulation pattern downstream, indicating a significant correlation between ENSO and the WP. In addition, the time series of the WP and Niño3.4 index is highly correlated, with a correlation coefficient of 0.45 (Figure 1e), significant at the 99% confidence level.

Modulation of the Relationship Between ENSO and the WP by the AMO
To inspect the decadal variation in the ENSO and WP relationship, we further calculated the 21-a sliding correlation of the winter WP index and Niño3.4 index from 1900 to 2010 (blue line in Figure 1f). A prominent ENSO-WP relationship occurs from the early 1910s to the late 1920s and over the 1950s to the late 1980s, principally emerges in the negative AMO phase, but with AMO in the positive phase, there seems to be a delay in late 1920s-1930s and between the 1950s and early 1960s ( Figure 1f). However, there are no clear connections between the WP and ENSO from the early 1930s to the mid-1950s-a time period generally associated with a positive AMO phase (Figure 1f). This breakdown of the interdecadal ENSO-WP relationship and the AMO phase transition implies that the AMO may have played a significant role in the varying ENSO-WP relations.
In terms of the correspondence between the AMO phase-switching and the interdecadal variations of the ENSO-WP relationship, while also considering the asymmetric characteristics of ENSO, we investigated how the modulation varied under different combinations of AMO and ENSO polarities. We first classified the ENSO events based on the winter ENSO index in Figure 1e. A typical El Niño (La Niña) event is identified when the DJF-averaged ENSO index exceeds a +0.5 standard deviation (is below a −0.5 standard deviation). These ENSO events are then grouped into positive and negative AMO phases, depending on the AMO state at the time they occur, thereby resulting in four categories. Namely, El Niño events within a positive AMO phase (+AMO/ El Niño), El Niño events within a negative AMO phase (−AMO/El Niño), La Niña events within a positive AMO phase (+AMO/La Niña), and La Niña events within a negative AMO phase (−AMO/La Niña) ( Table  S1 in Supporting Information S1). During the +AMO periods, 15 (14) El Niño (La Niña) years were identified. In comparison, 19 (20) El Niño (La Niña) years were identified during the −AMO periods (Table S1 in Supporting Information S1). Figures 2a-2d present the composite Z500 anomalies for the four categories in the boreal winter. The −AMO/El Niño shows a distinct atmospheric response in the WNP-a pattern resembling the WP (Figure 2b). Concurrently, a marked PNA-like pattern is also evident in the −AMO/El Niño composi tion ( Figure 2b). By contrast, for the +AMO/El Niño, the composition of the Z500 anomalies shows a cyclonic anomaly over the region in which the PNA is generally located (Figure 2a). We note that spatial patterns of the Z500 anomalies are highly similar between the −AMO/La Niña and +AMO/La Niña, both showing an anticyclonic anomaly extending from the Far East of Russia to the west coast of Alaska, as well as a cyclonic anomaly in the mid-latitudes of the western Pacific (Figures 2c and 2d). This suggests that the impacts of the La Niña events on the atmospheric circulation and climate anomalies are unlikely to be modulated by the AMO. In sharp contrast, the AMO has a notable modulation effect on the El Niño-WP relationship. Thus, we conclude that the AMO has an asymmetric modulation effect on the relationship between the winter ENSO and WP. Figures 2e and 2f further shows scatterplots between the Niño3.4 and WP or PNA indices. In the +AMO/El Niño case, the ENSO-related extratropical circulation has a closer relationship with the PNA (correlation coefficient of 0.56) than with the WP (correlation coefficient of 0.35). In contrast, the correlation coefficient with the WP is significantly enhanced in the −AMO/El Niño event, reaching 0.61, while also exhibiting a link with the PNA. This suggests that ENSO-related extratropical atmospheric anomalies correspond to a combination of the PNA and WP during the −AMO. Statistically, 13 of the selected years (68%) exhibit strong WP events during the −AMO/ El Niño period (bolded years in Table S1 of Supporting Information S1). In addition, only 5 years (33%) were identified as having strong WP indices during the +AMO/El Niño period (bolded years in Table S1 of Supporting Information S1). Nonetheless, there is no clear distinction in the years with a strong WP between −AMO/La Niña and +AMO/La Niña periods (∼40% vs. 50%). Collectively, the aforementioned evidence demonstrates that the El Niño teleconnections are significantly modulated by the AMO, with a close relationship occurring between ENSO and WP during the −AMO period and no such relationship occurring during the +AMO period. Thus, we will now examine the mechanisms underpinning the asymmetric modulation effect of the different AMO phases on the ENSO-WP relationship.

The Asymmetric Responses of SSTs and Precipitation to El Niño
Given the distinct El Niño-WP relation modulated by the −AMO, we now turn to investigate the plausible mechanism of the AMO's modulation of the El Niño-WP relationship. As previously demonstrated, −AMO contributes to ENSO variability by modulating the mean state of the eastern tropical Pacific (Dong et al., 2006;Timmermann et al., 2007). We define the ENSO amplitude as the winter area-mean SST anomalies (obtained by regression on the winter Niño3.4 SST index) over the Niño3.4 region. It is worth noting that the amplitude of El Niño and La Niña events exhibit distinct differences ( Figure S1a in Supporting Information S1). The SST anomalies for the El Niño events are stronger during −AMO than +AMO (Figures S1b and S1c in Supporting Information S1); whereas, during −AMO periods, the SST anomalies for La Niña events are comparable to those of the +AMO periods (Figures S1d and S1e in Supporting Information S1). This asymmetric intensification of El Niño and La Niña events is consistent with the findings of Sung et al. (2015). Therefore, the switching of the AMO phase plays a more crucial role in modulating the amplitude of El Niño events than in modulating La Niña events.
Since ENSO exerts its influence on the WP via triggering extratropical teleconnections through tropical convection, a question naturally arises as to whether the El Niño-related tropical SST and precipitation (also referred to as atmospheric convection) anomalies differ across AMO phases. To address this issue, we examine the composite SST and precipitation anomalies for different combinations of AMO and ENSO phases. During the La Niña years, there were no significant differences between the SST and precipitation anomalies for the −AMO and +AMO ( Figure S2 in Supporting Information S1). This is also consistent with the aforementioned insignificant role of the AMO in modulating La Niña-related anomalies. Figures 3a and 3b shows the composite SST anomalies over the tropical Pacific for +AMO/El Niño and −AMO/El Niño cases. The enhanced SST anomaly during the −AMO phase compared to the +AMO phase is consistent with the idea that the −AMO leads to enhanced ENSO amplitudes (Figure 3b vs. 3a), which is consistent with the results of previous studies (Dong et al., 2006;Park & Li, 2019;J. Y. Yu et al., 2015). In addition to the SST anomalies, Figures 3c and 3d illustrates the composite of the 10-m wind and precipitation anomalies for the above two cases. El Niño events coinciding with +AMO and −AMO phases present marked dipole precipitation anomalies in the tropical Pacific, superimposed with significant meridional wind shear of the anomalous zonal wind in the tropical western North Pacific (TWNP) and the equatorial central Pacific (ECP) (Figures 3c and 3d). However, composite precipitation anomaly differences display more pronounced precipitation in the −AMO ( Figure S3 in Supporting Information S1). This inconsistency might be associated with changes in the atmospheric mean state in the Pacific due to the AMO forcing. Meanwhile, the presence of positive precipitation and westerly wind anomalies in the ECP is attributable to a weakening of the Walker circulation arising from the positive phase of the AMO (Kucharski et al., 2011(Kucharski et al., , 2016Levine et al., 2017;Zanchettin et al., 2016;Zhao et al., 2022). Taken together, these findings suggest the possible role of anomalous atmospheric convection in bridging the effects of El Niño and the WP.

Role of the Tropical Atmospheric Convection
Given the above information, we conducted three LBM experiments based on climatological mean states during the +AMO and −AMO periods to quantify the specific role of regional convection: the first experiment (EXPA) shared the same horizontal heat source in the TWNP but with different amplitudes; the second experiment (EXPB) incorporated horizontal heat sources of different intensities but located in the ECP region; and the third experiments (EXPC) was equipped with a combined heat source of the TWNP and ECP, with all three sets of experiments having the same vertical profile of the heat source (see Table S2, Text S2, and Figure S4 in Supporting Information S1 for details). Figure 4 illustrates the response of the 500-hPa atmosphere to the TWNP and ECP heating sources at different basic mean states. In EXPA, the atmospheric anomalies in the extratropical North Pacific closely resemble the WP dipole pattern (cf., Figure 1a), yet display a stronger correspondence in the mean flow of the −AMO phase (Figures 4a and 4b). This further suggests the role of the negative AMO phase in facilitating the connection between TWNP heating and the WP (Figures 4a and 4b). In contrast, in the EXPB, the WNP presents no significant signal in either the +AMO or −AMO (Figures 4c and 4d). Specifically, ECP heating along with the −AMO mean flow contributes slightly to the southern center of the WP, with a positive geopotential height response at mid-latitudes in the North Pacific (Figure 4d). It is worth noting that the dipole-like atmospheric response of the −AMO mean flow in the WNP is enhanced in EXPC relative to EXPA (Figure 4f vs. Figure 4b). However, few discrepancies are observed in the +AMO mean flow atmospheric circulation response in Figure 4e versus Figure 4a, in contrast to the aforementioned −AMO integrated-heat-source response. To disentangle the role of the AMO mean states and regional heating sources, we implement a set of symmetric ideal heat source experiment ( Figure S5 in Supporting Information S1). Considering the same amplitude of the TWNP and ECP anomalous heating, there are no distinctions in the atmospheric circulation response between the +AMO and −AMO climatological mean states. This further confirms the key role of the heat sources with different intensities in the context of different AMO phases ( Figure S5d-S5j in Supporting Information S1). Thus, our story attributes the primary factor of the various ENSO-WP relationships to the predominant influence of the diabatic forcing amplitude, which is caused by the distinct ENSO amplitudes during the different AMO phases.

Summary and Discussion
Asymmetric modulation of the winter ENSO and WP relationship, as well as its associated physical mechanisms, are outlined on the basis of observational analyses and LBM experiments. The distinct 21-a sliding-window ENSO-WP relationships vary consistently with the transition of the AMO phases. During the −AMO period, the El Niño-WP relationship strengthens significantly, while an insignificant El Niño-WP relationship is observed when El Niño events occur during the +AMO period. Typically, a significant projection upon the WP is captured when the El Niño events occur during a −AMO phase, with a concomitant PNA pattern. However, the relationship between La Niña and WP is independent of AMO modulation. Further studies suggest that during the −AMO period, the strengthening of the El Niño amplitude and the intensification of the tropical atmospheric convection in the TWNP and ECP largely favor a positive WP event. In contrast, the amplitude of La Niña is similar during positive and negative AMO periods, namely, it is not modulated by AMO. Therefore, the effect of La Niña events on WP is also not subject to AMO modulation. LBM experiments further demonstrate the role of tropical atmospheric convection in the TWNP in linking anomalous tropical atmospheric convection and the WP. However, as the LBM is only a dry linear model that does not incorporate moisture processes or nonlinear feedback, further analyses should be performed with more complicated models to investigate this issue. It is noted that Pacific Decadal Oscillation (PDO) also contributes significantly to ENSO and its related North Pacific variability. A parallel examination of the link between PDO and ENSO-WP is also made ( Figure  S6 in Supporting Information S1). It reveals no clear correspondence between the variations of the PDO and interdecadal changes in the ENSO-WP relationship. This further suggests that the PDO is unlikely to have a modulating effect on the ENSO-WP relationship. Despite the near-synchronous matching relationship between AMO and ENSO-WP, a closer inspection reveals that their changes are not perfectly synchronous. To explicitly address whether there is a leading or lagging correlation, a 20-year lead-lag correlation was calculated ( Figure  S7 in Supporting Information S1). It reveals that the simultaneous correlation reaches a maximum. We are there- fore confident in concluding that AMO has a synchronous modulatory effect on the changes in the ENSO-WP relationship. Further attention should be paid to why the tropical atmospheric convection related to El Niño is stronger during the −AMO period than during the +AMO period. One plausible explanation is the change in the mean state (Martín-Rey et al., 2018;Sung et al., 2015). For example, Sung et al. (2015) highlighted that the warmer and wetter basic state over the central to eastern tropical Pacific during the -AMO period accompanied by enhanced air-sea coupling is more favorable for enhanced El Niño events. During the −AMO period, anomalous warming of the SSTs over the central-eastern equatorial Pacific occurs as anomalous cooling of the North Atlantic SSTs occurs (figure not shown). In this context, the increase in SSTs over the equator further enhances the equatorial SST variability, thereby favoring the enhanced atmospheric convection associated with El Niño. Nevertheless, a detailed analysis of the tropical Pacific air-sea interaction that contributes to enhanced atmospheric convection requires further discussion.

Data Availability Statement
The ERA-20C data (Poli et al., 2016) are provided by the ECMWF; The ERSSTv5 SST data (Huang et al., 2017) are provided by the NOAA. All figures were made with the NCAR Command Language version 6.6.2 (UCAR/ NCAR/CISL/TDD, 2019).