**3. Results**

*3.1. Comparison of Different Types of ENSO and Related NAO*

3.1.1. SSTA Patterns and Their Associated Tropic Atmospheric Responses

First, since the mega-ENSO index can well represent the strong and weak gradient ENSOs, combined with the Niño3.4 index, the ENSO events with different intensities and gradients might be distinguished. According to the method displayed in Section 2, three types of El Nino and three types of La Nina are classified (Table 1). Interestingly, the strong ENSOs commonly contain a sizeable zonal gradient; only the weak ENSOs can be separated based on the east-west gradient to the weak intensity-strong gradient (WSG) and weak intensity-weak gradient events. Since the weak intensity-weak gradient events are confined in the equatorial region, they are labeled as equatorial (Eq) ENSO. Previous studies subdivided ENSO into the CP and EP groups according to the shift of the maximum SSTA [24,25,61,62]. Therefore, the CP ENSOs are bolded in Table 1, showing that the CP events occur randomly in the three flavors of El Niño; however, they coincide with the strong gradient (SG) La Niña, which consists of the strong and WSG La Niña.

Figure 2 compares the large-scale SST anomalies for the El Niño and La Niña types. Generally speaking, the spatial patterns are well classified. In the strong El Niño (SEN) winters, the obvious warm SSTAs dominate the eastern Pacific and Indian Ocean (IO), while the cold SSTAs appear in the western Pacific (Figure 2a). The strong La Niña (SLN) displays an SSTA distribution opposite to the SEN (Figure 2d). The WSG event also displays a Pacific "seesaw" pattern but a neutral IO SSTA (Figure 2b,e). In contrast, a salient narrow positive (negative) SSTA controls the equatorial central and eastern Pacific for the equatorial EN (equatorial LN) with significant warm (cold) SSTA in the IO (western IO) (Figure 2c,f). In addition, as shown in Table 1, several equatorial (SG) La Niña events coincide with the years of EP (CP) La Niña. Their spatial patterns show the maximum negative SSTA center in the tropical CP for SGLN (Figure 2d,e) but in the tropical EP for equatorial LN (Figure 2f). However, the centers of maximum SST anomalies shifted very little for the three types of El Niño (Figure 2a–c).

**Figure 2.** DJF sea surface temperature (SST; K, interval: 0.3 K) composite differences of (**a**) strong El Niño, (**b**) weak intensity-strong gradient (WSG) El Niño, and (**c**) equatorial (Eq) El Niño. (**d**–**f**) same as (**a**–**c**), but for La Niña events. The shadings in each panel represent the region with anomalies significant at the 95% confidence level (Student's *t*-test). The green contour represents 28 ◦C isotherm.

Due to the diversity in ENSO intensity and spatial patterns, the precipitation anomalies associated with the different ENSOs exhibit distinct features (Figure 3). The tropical precipitation anomalies are stronger in response to intense ENSOs than weak events. For El Niños, the salient positive precipitation anomalies straddle the dateline with the negative anomalies controlling the western Pacific (WP) (Figure 3a–c). Significant discrepancies are found over the IO region with rich tropical rainfall generated by the warm local SSTA during the SEN and equatorial EN winters. However, no salient precipitation anomalies are seen in the WSGEN winters. For the La Niña events (Figure 3d–f), the neutral precipitation anomalies over the IO indicate a relatively weak local air-sea interaction compared to those of the SEN and WSGEN. It also shows the poor precipitation centers in the tropical central Pacific for the two SGLN cases but extends eastward during equatorial LN, which is attributed to the shift of the maximum center of SSTA.

**Figure 3.** The same as Figure 2, except for tropical precipitation (mm/mon, interval: 30 mm/mon).

3.1.2. Unsteady Relationship of the ENSO with NAO

Next, we display the scatterplot of NAO and Niño3.4 indices in winter (Figure 4), and a complex relationship is detected between ENSO and NAO events. It shows that six WSGENs emerge with a negative NAO; however, the negative or positive NAO occurrence is evenly balanced during the SEN and equatorial EN winters, indicating a steady relationship between the NAO signal and the WSGEN. For the La Niña events, most of the SLN (6 of 8) and WSGLN (4 of 5) events, in other words, 10 of 13 SGLNs, are accompanied by a positive phase of NAO, but such relationships cannot be seen in the equatorial LN cases.

**Figure 4.** Scatter map of the DJF Niño3.4 and NAO indices for WSG-ENSO (red dots), equatorial ENSO (green crosses), and strong ENSO (purple snowflakes). The blue dashed line denotes ±1.1.

Figure 5 displays the anomalous 850-hPa geopotential height (Z850) and zonal wind at 200-hPa (U200) associated with the different types of ENSO to further inspect the extratropical atmospheric responses. During the EN winters (Figure 5a–c), anomalous positive SST in the tropical central and eastern Pacific induces salient diabatic heating, generating a large-scale Rossby wave train that resembles the PNA teleconnection pattern over the North Pacific (NP) and North America regions. Significant discrepancies are observed

over the NP as well as the North Atlantic. Over the NP region, although the apparent negative anomalies, which imply the enhanced Aleutian Low (AL), are seen in the three EN cases, they exhibit different intensities and spatial locations. Generally, the negative Z850 anomalies over NP for SEN (Figure 5a) are much stronger than the weak events (Figure 5b,c). Compared with the SEN and equatorial EN composites (Figure 5a,c), the enhanced AL associated with WSGEN moves eastward to some extent (Figure 5b). Over the North Atlantic, the Z850 anomalies for WSGEN (Figure 5b) exhibit a negative NAO-like anomaly with negative and positive centers over the Azores Island and Iceland, respectively. However, such an anomalous pattern over the North Atlantic cannot be seen in the SEN and equatorial EN cases (Figure 5a,c). For the LN cases, the Z850 anomalies feature a decreased AL over NP in the SGLN winters (Figure 5d). Simultaneously, negative and positive Z850 anomalies are evident in Iceland and Azores Island, respectively, reflecting a positive phase of NAO-like anomalies pattern. In contrast, the equatorial LN-related anomalous Z850 field (Figure 5e) displays the salient negative anomalies over the subtropical North Pacific, representing the southward movement of the AL. Meanwhile, the positive Z850 anomalies, instead of the NAO-like pattern, dominate the mid- to high-latitude North Atlantic.

**Figure 5.** DJF 850-hPa geopotential height (Z850; m) composite differences of (**a**) strong El Niño—neutral, (**b**) WSG-El Niño—neutral, (**c**) Eq-El Niño—neutral, (**d**) SG-La Niña—neutral, and (**e**) Eq-La Niña—neutral. (**f**–**j**) same as (**a**–**e**) but for zonal wind at 200-hPa (U200; m/s). The dots in each panel represent the region with anomalies significant at the 90% confidence level (Student's *t*-test).

We observed that only WSGENs, rather than the SEN and equatorial EN, are concurrent with a negative phase of the NAO event, indicating that both the change of intensity and east-west gradient in SST determine the linking of El Niño and NAO. For the La Niña cases, although the SLN and WSGLN display different amplitudes, they both show an obvious east-west gradient in SST. They are accompanied by a positive phase of NAO, which cannot be seen during the equatorial LN winter. Therefore, the influence of LN events on NAO mainly depends on the spatial distribution rather than the intensity of ENSO. We also noticed that compared to the equatorial LN events, the SSTA of SLN and WSGLN show strong east-west gradients, and the maximum negative values center in the equatorial CP. We, therefore, combine the SLN and WSGLN into one category, namely, the SGLN, and discuss the potential mechanisms for the impact of the WSGEN and SGLN on NAO in the next section.

#### 3.1.3. Possible Mechanisms for the Impact of WSGEN and SGLN on NAO

The previous study [25] argued that the jet streams commonly act as an atmospheric bridge, transporting ENSO-induced planetary wave energy to the downstream region and resulting in an oscillation of the remote atmosphere, for instance, the air mass over the North Atlantic. The lower panel of Figure 5 displays the composite U200 anomalies in each ENSO winter.

In the El Niño cases, anomalous U200 displays the tripolar pattern in the NP and extends eastward of a different degree to the North Atlantic (Figure 5f–h). The intensified subtropical jet stream and the weakened mid-latitude westerly jet over North Atlantic tend to elongate farther eastward during the WSGEN winters (Figure 5g) than during the SEN and the equatorial El Niño winters (Figure 5f,h). Such a phenomenon implies that the two branches of westerly jets may serve as the "bridge" connecting the WSGEN and the negative phase of the NAO signal. However, this "bridge" effect is invalid for SEN and equatorial El Niño. For the La Niña events, an opposite situation is observed for the SGLN cases over the North Pacific-North Atlantic sector with the negative and positive U200 anomalies over the subtropical and mid-latitude North Atlantic, indicating the weakened subtropical and the enhanced mid-latitude jets, respectively (Figure 5i). In contrast, the equatorial La Niña-related U200 anomalies show a weakened and northeastward tilted westerly jet from the subtropical North Pacific extent far east to the Barents Sea, but without the salient tripolar structure over the North Pacific region and the dipolar anomalies over the North Atlantic (Figure 5j).

Numerical experiments were performed to identify the potential effects of westerly jets on connecting the tropical SSTAs corresponding to diverse ENSOs and NAO signals. The detailed experiment designs are listed in Table 2. Figure 6 displays the abnormal tropical precipitation, Z850, and U200 response to the SEN\_IO, WSGEN, EqEN\_IO, SGLN, and EqLN SSTAs forcing versus the CTRL run. In general, the simulated results can well reproduce the observational analysis counterparts. In the EN-type simulations, the prescribed warm SSTA in EP and IO with the cold SSTA in WP for the SEN\_IO simulation leads to stronger tropical precipitation responses (Figure 6a) than those of the WSGEN (Figure 6b) and EqEN\_IO (Figure 6c) experiments. Additionally, significant positive precipitation anomalies dominate tropical IO in the SEN\_IO (Figure 6a) and EqEN\_IO (Figure 6c) experiments but disappear in the WSGEN (Figure 6b) experiments. Further inspecting the Z850 responses to each EN-type SSTA forcing can reveal an enhanced AL over the NP in each anomalous Z850 field (Figure 6f–h). Over the North Atlantic region, the WSGEN SSTA forcing (Figure 6g) tends to induce a negative NAO anomaly, which cannot be seen in the SEN\_IO (Figure 6f) and EqEN\_IO (Figure 6h) simulations. In addition, the anomalous U200 exhibits a tripolar structure over the NP in each simulation (Figure 6k–m), and the salient discrepancies also appear over the North Atlantic region. The WSGEN SSTA forcing triggers the eastward spread of the subtropical and mid-latitude westerly jets (Figure 6l), confirming the observational result. In comparison, the two westerly jets cannot extend that far eastward, indicating the SEN\_IO and EqEN\_IO-types SSTAs are hard to trigger the NAO-like atmospheric anomaly through the "atmospheric bridge"—the westerly jets (Figure 6k,m).

In the two La Niña-type simulations, considering the weak air-sea interaction over the tropical IO region (Figure 3d–f), we imposed the SSTA associated with the SGLN and the equatorial LN mainly within the Pacific basin. The significant negative precipitation anomalies over the equatorial-CP and EP exhibit a much stronger intensity and extent more westward in response to the SGLN (Figure 6d) than to the EqLN forcing (Figure 6e). This may be due to the maximum SSTA centers straddling the equatorial CP for SGLN but in the equatorial-EP for equatorial LN. The resultant circulation fields for the SGLN simulation are contrary to that of the WSGEN simulation, showing a salient positive phase of NAOlike anomaly (Figure 6i) with the weakened subtropical westerly jet and the intensified mid-latitude jet (Figure 6n) over the North Atlantic. However, the Z850 and U200 responses over the North Atlantic exhibit a neutral pattern in the EqLN simulation (Figure 6j,o).

The above numerical experiments proved that the two branches of westerly jets serve as atmospheric bridges to extend the WSGEN/SGLN signals to the downstream regions, inducing NAO-like anomalies. However, why can the other types of ENSO SSTAs not expand their impact to the downstream North Atlantic?

**Figure 6.** Tropical precipitation responses in the ECHAM5 regarding a difference between (**a**) SEN\_IO, (**b**) WSGEN, (**c**) EqEN\_IO, (**d**) SGLN, and (**e**) EqLN forcings and the control run (interval: 2 mm/mon). (**f**–**j**) same as (**a**–**e**) but for geopotential at 850-hPa (Z850; 10 m). (**k**–**o**) same as (**a**–**e**) but for zonal wind at 200-hPa (U200; m/s). The dots represent the anomalies significant at the 90% confidence level (Student's *t*-test).

#### 3.1.4. The Causes for the Abruption of ENSO and NAO Relations

The observational and numerical analysis demonstrated that different El Niño/La Niña flavors exhibit intricate relations with the NAO phase. For the EN cases, we observed the salient positive SSTAs in the tropical IO for the SEN and equatorial EN cases (Figure 2a,c), reflecting the inter-basin coupling of the tropical Pacific and IO, which is the key to comprehending the teleconnection pathway on the globe that is rooted in the tropics [63–66]. Furthermore, the IO is also identified as a principal contributor to the North Atlantic extratropical atmospheric anomalies in winter [10,67,68]. Therefore, it is natural to speculate that the anomalous warm SST in the IO may modulate the linking of El Niño and NAO.

To verify our speculation, an IO warming (IOW) experiment was carried out with the positive SSTAs associated with SEN imposed onto the tropical IO, which induces rich local precipitation (Figure 7a) and a positive phase of NAO anomaly (Figure 7b). This atmosphere experiment confirms the previous studies that the slow variations of IO force the NAO on the seasonal time scale [67,69,70]. Next, the SEN and EqEN experiments were conducted without warm IO SSTA. The dipolar precipitation responses in the Pacific are much weaker in the EqEN simulation (Figure 7b) than in the SEN simulation (Figure 7c); both of them can trigger a negative NAO-like Z850 anomaly (Figure 7e,f). However, the NAO-like Z850 anomalies disappeared when the IO and SEN/EqEN forcings were imposed simultaneously (Figure 6f,h), indicating that the IO SSTA modulates the forcing of El Niño onto the NAO signals.

For the SGLN cases, the SSTA shows a sizeable east-west gradient with the maximum center in the equatorial CP (Figure 2d,e), generating a positive NAO-like anomaly over the North Atlantic (Figures 5d and 6i). By contrast, the equatorial LN features a weak east-west gradient with the maximum SSTA in the equatorial EP (Figure 2f), and the resultant atmospheric anomalies over the North Atlantic exhibit the NAO-neutral pattern (Figures 5e and 6j). However, although the amplitude of WSGLN (Figure 2e) is comparable with equatorial LN (Figure 2f) (their average intensities are −0.72 and −0.82, respectively), their resultant precipitation anomalies exhibit significant discrepancies in

intensity (Figure 3e,f). Therefore, the differences in atmospheric anomalies associated with SGLN (Figure 5d) and equatorial LN (Figure 5e) may not depend on the amplitude of SSTA in the Niño3.4 region.

**Figure 7.** Tropical precipitation responses in the ECHAM5 regarding a difference between (**a**) IOW, (**b**) SEN, (**c**) EqEN forcings, and the control run (interval: 2 mm/mon). (**d**–**f**) same as (**a**–**e**) but for geopotential at 850-hPa (Z850; 10 m). The dots represent the anomalies significant at the 90% confidence level (Student's *t*-test).

Regarding the SSTA distributions, the SGLN and equatorial LN correspond to the CP and EP La Niñas, respectively. We know that the climatology SST, to a great extent, determines the occurrence of deep tropical convection; the threshold commonly is about 28 ◦C [71,72], and the variability of convection and atmospheric responses are primarily dependent on the anomalous tropical SST. Owing to the proximity of the central Pacific to the western Pacific warm pool, the climatology SSTs above 28 ◦C (green contours in Figure 2), abnormal convective motions, and extra-tropical atmospheric circulation usually present a quasi-linear response to SST anomalies [25]. During the SGLN winter, the negative SSTA in the CP superimposes on the climatology SST, significantly dropping local SST below 28 ◦C, reducing the deep convection and precipitation over the CP (Figure 3d,e). As a result, a steady forcing of SGLN on the atmospheric anomalies over the North Atlantic is established. In comparison, the low climatological SSTs normally do not reach the threshold to generate convection in the eastern Pacific. Consequently, the negative SSTAs in EP exert a weaker influence on the tropical convective motions and precipitation in the equatorial LN cases (Figure 3f). Therefore, changes in the maximum SSTA center in the tropical Pacific profoundly alter global teleconnections [38,73]. We further executed the WSGLN simulations with the associated SSTA imposed onto the Pacific region. Figure 8a displays the tropical precipitation anomalies in response to the WSGLN forcing versus CTRL run. The negative precipitation anomalies induced by WSGLN are concentrated in the tropical CP with rich rainfall over the WP. The large precipitation differences between WSGLN and EqLN simulations are seen over the tropical CP and EP (Figure 8b). The anomalous dipole mode over the Pacific indicates that WSGLN-type SSTA tends to reduce the tropical convection, weakening the convective motion more pronouncedly over the CP than the EP. The resultant Z850 anomalies for WSGLN forcing also highly resemble a positive phase of the NAO pattern (Figure 8c).

**Figure 8.** Tropical precipitation responses in the ECHAM5 regarding a difference between (**a**) WSGLN and the control run, (**b**) WSGLN and EqLN (interval: 2 mm/mon), (**c**) same as (**a**) but for geopotential at 850-hPa (Z850; 10 m). The dots represent the anomalies significant at the 90% confidence level (Student's *t*-test).
