a. Datasets

The monthly SST data are from the National Oceanic and Atmospheric Administration (NOAA) Optimum Interpolation SST, version 2, with a horizontal grid resolution of 1° × 1°, which is provided by the NOAA Earth Research Laboratory Physical Science Division (http://www.esrl.noaa.gov/psd/data). The monthly atmospheric data are from the fifth major global reanalysis developed by the European Centre for Medium-Range Weather Forecasts (ERA5; https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-single-levels-monthly-means?tab=form), with a horizontal resolution of 0.25° × 0.25°, including the surface latent heat flux, sensible heat flux, net shortwave radiation, net longwave radiation, precipitation, boundary layer height, surface zonal and meridional winds, surface wind speed, air temperature, and three-dimensional relative humidity. Besides, the monthly SST from ERA5 is chosen only for computing the regressions between SSTAs and relative humidity anomalies, and between SSTAs and boundary layer height anomalies. The monthly oceanic three-dimensional data are from the National Centers for Environmental Prediction (NCEP) Global Ocean Data Assimilation System (GODAS; https://www.esrl.noaa.gov/psd/data/gridded/data.godas.html), with a horizontal resolution of 1/3° longitude × 1° latitude. In addition, we also use surface net heat fluxes from GODAS and the NCEP–National Center for Atmospheric Research (NCAR) reanalysis (https://psl.noaa.gov/data/gridded/data.ncep.reanalysis.html) to confirm the results derived from ERA5. All the datasets are chosen for the period 1982–2018 during which all variables are available. The monthly anomalies are obtained by removing the long-term trend as well as the climatological annual cycle of the chosen time period, and then a 3-month running mean is applied to reduce the intraseasonal variability.

b. Fuzzy clustering method

The members applied to the FCM here are a subset of the monthly SSTAs in the tropical Pacific (20°S–20°N, 150°E−90°W) during El Niño events. We first use a 40° × 10° window zonally sliding by 2.5° along the equator (5°S–5°N), starting from 150°E to 90°W, in order to obtain a set of regional mean SSTAs and the corresponding standard deviations (STDs). The month in which any regional-mean SSTA is greater than the corresponding positive STD and 0.5°C is then regarded as a warm record. When all the warm records are extracted, those segments with less than five successive months in the set of warm records are deleted. Moreover, as the peak time of El Niño tends to be phase locked in boreal winter (Tziperman et al. 1998), the warm segments that do not contain boreal wintertime (November–January) are also discarded. The remaining warm months are then used for our classification of different El Niño types. In addition, the type of a specific El Niño event is based on the type into which its DOM in boreal winter falls. Details regarding the application of the FCM technique in El Niño classification can also be found in Chen et al. (2015).

c. Ocean mixed layer heat budget analysis

d. Decomposition of the surface latent heat flux anomaly

a. Classification of El Niños based on the FCM

The FCM is applied to classify El Niño events during 1982–2018 into two types. As shown in Fig. 1, the first warm pattern displays robust positive SSTAs in the central and eastern Pacific and has its largest warming in the eastern equatorial Pacific near the South American coast (Fig. 1a), which is a typical feature of extreme El Niños (Takahashi et al. 2011; Chen et al. 2015; Xie et al. 2018). Three historical El Niños, commonly known as the extreme El Niño events of 1982/83, 1997/98, and 2015/16 (Cai et al. 2017; Lian et al. 2017), fall into the first pattern classification (Fig. 1c, red curve). The second warm pattern exhibits moderately positive SSTAs centered in the central equatorial Pacific east of the date line around 170°W (Fig. 1b). Nine historical El Niños other than the three aforementioned extreme ones—in 1986/87, 1987/88, 1991/92, 1994/95, 2002/03, 2004/05, 2006/07, 2009/10, and 2014/15—are all classified as the second warm pattern (Fig. 1c, blue curve). Thus, the FCM naturally separates the extreme El Niños from other moderate El Niños when two clusters are set. Moreover, the classified result by the FCM indicates that the pattern differences between extreme and moderate El Niños appear to be the most robust among different El Niño types.

View Full Size

The two El Niño clusters identified by the FCM and the associated DOMs: (a) the extreme El Niño cluster, which involves three historical extreme El Niño events; (b) the moderate extreme El Niño cluster, which includes nine historical moderate El Niño events; and (c) the DOM for extreme El Niño (red curve), moderate El Niño (blue curve), and neither (black curve). Stippling in (a) and (b) indicates that the compositions are significant at the 95% confidence level based on the Student’s t test.

Citation: Journal of Climate 34, 13; 10.1175/JCLI-D-20-0757.1

• Download Figure
• Download figure as PowerPoint slide

b. Discrepant roles of $Q_{\text{net}}^{\prime }$ for the development of SSTA patterns between extreme and moderate El Niños

Figure 2 presents the spatial patterns of SSTAs, SSTA tendencies, and $Q_{\text{net}}^{\prime }$ during developing phase (from May to December of the developing year) of the extreme and moderate El Niños. It is shown that the largest SSTAs during the developing phase appear to be anchored basically in the eastern equatorial Pacific east of 150°W in extreme El Niños (Fig. 2a), while those in moderate El Niños are mostly confined to the central Pacific around 150°–170°W (Fig. 2c). Such a difference is consistent with the different warm patterns classified by the FCM (Figs. 1a,b). Moreover, the SSTA tendencies during the developing phase display similar zonal patterns to the corresponding SSTAs, with more positive values in the eastern (central) than in the central (eastern) equatorial Pacific in extreme (moderate) El Niños (Figs. 2b,d, contours). On the other hand, the damping effects of $Q_{\text{net}}^{\prime }$ in extreme and moderate El Niños are both larger in the eastern equatorial Pacific east of 140°W, albeit with a larger amplitude for the extreme ones (Figs. 2b,d). The former matches well with the corresponding gradual increases of positive SSTAs from the central to the eastern equatorial Pacific (Fig. 2e, solid curves), thus generally displaying a “larger warming gets more damping” zonal paradigm, while the latter zonally deviates from the corresponding larger positive SSTAs in the central equatorial Pacific west of 140°W (Fig. 2e, dashed curves). Accordingly, in moderate El Niños, the more damping effects of $Q_{\text{net}}^{\prime }$ and the weaker positive SSTA tendencies, both in the eastern equatorial Pacific, imply that the damping effect of $Q_{\text{net}}^{\prime }$ may help contribute to the local weaker SSTA tendencies, favoring larger SSTA tendencies as well as larger SSTAs being located in the central equatorial Pacific (Fig. 2d). Similar results can be found based on the $Q_{\text{net}}^{\prime }$ from the GODAS and NCEP–NCAR datasets (Fig. 3). In these two datasets, the larger damping effects of $Q_{\text{net}}^{\prime }$ in extreme El Niños generally match well with the larger SSTA tendencies (Figs. 3a,b), while those in moderate El Niños zonally deviate from the larger SSTA tendencies (Figs. 3c,d).

View Full Size

Spatial patterns of (a) SSTAs and (b) $Q_{\text{net}}^{\prime }$ during the developing phase (May–December of the developing year) for extreme El Niños. Contours in (b) are the SSTA tendencies during the developing phase (units: °C month⁻¹, with an interval of 0.025°C month⁻¹; zero contour thickened and negative dashed). (c),(d) As in (a) and (b) but for moderate El Niños. Stippling indicates that the compositions of shaded values are significant at the 95% confidence level based on the Student’s t test. (e) Zonal distributions of equatorial (2.5°S–2.5°N) $Q_{\text{net}}^{\prime }$ (blue curves) and SSTA (red curves) for extreme (solid curves) and moderate (dashed curves) El Niños.

Citation: Journal of Climate 34, 13; 10.1175/JCLI-D-20-0757.1

• Download Figure
• Download figure as PowerPoint slide

View Full Size

As in Figs. 2b and 2d, but for $Q_{\text{net}}^{\prime }$ data from (a),(c) GODAS and (b),(d) NCEP–NCAR.

Citation: Journal of Climate 34, 13; 10.1175/JCLI-D-20-0757.1

• Download Figure
• Download figure as PowerPoint slide

With regard to each individual El Niño event, it is shown that all the three extreme El Niños exhibit larger positive (negative) SSTAs ($Q_{\text{net}}^{\prime }$) in the eastern than central equatorial Pacific, and most of the moderate El Niños display larger positive SSTAs (negative $Q_{\text{net}}^{\prime }$) in the central (eastern) than eastern (central) equatorial Pacific, leading to the average of positive SSTAs (negative $Q_{\text{net}}^{\prime }$) in moderate El Niños being larger in the central (eastern) equatorial Pacific (Fig. 4). Note that the 1994/95 El Niño event is an outlier of moderate El Niño with larger negative $Q_{\text{net}}^{\prime }$ in the central equatorial Pacific. In addition, there are slightly larger positive SSTAs but much larger negative $Q_{\text{net}}^{\prime }$ in the eastern equatorial Pacific for the 1987/88, 2009/10, and 2014/15 El Niño events. These outliers imply that there could be an intermediate state of SSTA pattern with no explicit difference between central and eastern Pacific warm anomalies (Chen et al. 2015). Nevertheless, they are classified into moderate El Niños as the zonal SSTA patterns of these three El Niños are closer to the second type based on the FCM (Fig. 1b). In the following section, we will reveal that the effects of $Q_{\text{net}}^{\prime }$ on the zonal SSTA pattern formations for these outliers are physically consistent with the common moderate El Niños.

View Full Size

Scatterplot of difference of SSTAs vs that of $Q_{\text{net}}^{\prime }$ between eastern Pacific (2.5°S–2.5°N, 150°–90°W) and central Pacific (2.5°S–2.5°N, 180°–150°W) for each individual El Niño event during the developing phase. The horizontal (vertical) red bar and the square box in the red bar denote the standard deviations and mean of SSTAs ($Q_{\text{net}}^{\prime }$) only for moderate El Niños, respectively. The standard deviations of SSTAs and $Q_{\text{net}}^{\prime }$ for moderate El Niños are indicated by red horizontal and vertical bars, respectively. The red square box in the horizontal (vertical) red bar denotes the mean of SSTAs ($Q_{\text{net}}^{\prime }$) for moderate El Niños.

Citation: Journal of Climate 34, 13; 10.1175/JCLI-D-20-0757.1

• Download Figure
• Download figure as PowerPoint slide

Figure 5 displays a Hovmöller diagram (averaged over 2.5°S–2.5°N) that compares the temporal evolutions of equatorial SSTAs as well as $Q_{\text{net}}^{\prime }$ during the developing year between extreme and moderate El Niño events. Extreme and moderate El Niños both present warm SSTAs first in the central equatorial Pacific in early spring of the developing year, and follow discrepant developing trajectories of the zonal SSTA pattern afterward. In extreme El Niños, the largest SSTAs appear to be anchored basically in the eastern equatorial Pacific east of 150°W after May of the developing year (Fig. 5a), while those in moderate El Niños are mostly confined to the central Pacific around 150°–170°W (Fig. 5c). Such a difference is consistent with the different SSTA patterns averaged over the developing phase (Figs. 2a,c). Meanwhile, the SSTA tendencies during the developing phase show overall similar zonal distributions to the corresponding SSTAs, with more positive values in the eastern (central) than central (eastern) equatorial Pacific in extreme (moderate) El Niños (Figs. 5a,c, contours). On the other hand, the more damping effects of $Q_{\text{net}}^{\prime }$ in extreme and moderate El Niños are both located in the eastern equatorial Pacific during the developing phase (Figs. 5b,d). The former matches well with the corresponding zonal pattern of SSTAs, while the latter is anchored in the eastern equatorial Pacific and zonally deviates from the corresponding more positive SSTAs in the central equatorial Pacific.

View Full Size

Hovmöller diagram for equatorial (2.5°S–2.5°N) (a),(c) SSTAs and (b),(d) surface net heat flux anomalies during the developing year in (a),(b) extreme and (c),(d) moderate El Niños. Contours in (a) and (c) denote the tendency of SSTAs (units: °C month⁻¹, with an interval of 0.05°C month⁻¹; zero contour thickened and negative dashed), and in (b) and (d) denote the SSTAs (units: °C, with an interval of 0.25°C; zero contour thickened and negative dashed). Stippling indicates that the compositions of shaded values are significant at the 95% confidence level based on the Student’s t test.

Citation: Journal of Climate 34, 13; 10.1175/JCLI-D-20-0757.1

• Download Figure
• Download figure as PowerPoint slide

To quantify the discrepant effects of $Q_{\text{net}}^{\prime }$ on the development of zonal SSTA patterns between extreme and moderate El Niño, an ocean mixed layer heat budget analysis is further conducted based on the GODAS dataset (Fig. 6) during the developing phase of the two El Niño clusters both in the eastern (2.5°S–2.5°N, 140°–90°W; red bars) and central equatorial Pacific (2.5°S–2.5°N, 180°–140°W; blue bars), together with their differences (black bars). Note that the spatial patterns of the chosen mixed layer temperature anomalies are quite similar to those of the SSTAs both in extreme and moderate El Niños (not shown). Concurrent with the SSTA tendencies (Figs. 2b,d, contours), the ocean mixed layer temperature tendencies ($C\partial T_O^{\prime }/\partial t$) for extreme El Niños are larger in the eastern than central equatorial Pacific (Fig. 6a). Such zonal distribution is contributed by the ocean three-dimensional heat transport anomalies, among which the $Q_{\mathit{w}}^{\prime }$ contributes the most, consistent with previous studies (Kug et al. 2009; Chen et al. 2015). However, $Q_{\text{net}}^{\prime }$, acting as the major damping term, displays a much more damping effect in the eastern than in the central equatorial Pacific. This indicates that the damping effect of $Q_{\text{net}}^{\prime }$ could not essentially alter the zonal distribution of $C\partial T_O^{\prime }/\partial t$ owing to the overwhelming positive effect from ocean heat transport anomalies; rather, it is merely a response to positive SSTAs.

View Full Size

The ocean mixed layer heat budget during the developing phase of (a) extreme and (b) moderate El Niños based on GODAS. The red, blue, and black bars denote the regional-mean values in the eastern equatorial Pacific (EEP; 2.5°S–2.5°N, 140°–90°W), the central equatorial Pacific (CEP; 2.5°S–2.5°N, 180°–140°W), and their differences (EEP minus CEP). The $C\partial T_{\mathit{O}}^{\prime }/\partial t$ $Q_{\mathit{u}}^{\prime }$, $Q_{\upsilon }^{\prime }$, $Q_{\mathit{w}}^{\prime }$, $Q_{\text{net}}^{\prime }$, and $Q_{\text{res}}^{\prime }$ terms represent the tendency of mixed layer temperature anomalies, the mixed layer zonal, meridional and vertical heat transport anomalies, surface net heat flux anomalies, and residual term, respectively. Note that the values on the y axis are different between (a) and (b).

Citation: Journal of Climate 34, 13; 10.1175/JCLI-D-20-0757.1

• Download Figure
• Download figure as PowerPoint slide

By contrast, the $C\partial T_O^{\prime }/\partial t$ values in the central equatorial Pacific are a little bit larger than those in the eastern equatorial Pacific for moderate El Niños (Fig. 6b). Similar to the extreme El Niños, the contribution of ocean three-dimensional heat transport anomalies in moderate El Niños, albeit with a much smaller magnitude, also favors more positive SSTAs in the eastern than in the central Pacific, while the $Q_{\text{net}}^{\prime }$ acts to suppress such effect. This indicates that the more damping effects of $Q_{\text{net}}^{\prime }$ in the eastern equatorial Pacific might alter the zonal distribution of $C\partial T_{\mathit{O}}^{\prime }/\partial t$ in moderate El Niños by partly offsetting the local modest positive effects of ocean heat transport anomalies, favoring more positive SSTAs to be located in the central equatorial Pacific.

The $Q_{\text{res}}^{\prime }$ in extreme El Niños is negligible but appears to be another contributor to the zonal SSTA pattern formation in moderate El Niños by producing negative (positive) effects in the eastern (central) Pacific, favoring more positive SSTAs in the central equatorial Pacific. Thus, the role of oceanic subgrid-scale processes, which are beyond the scope of this study, should be paid more attention to in shaping the zonal SSTA pattern of moderate El Niños.

c. Discrepant atmospheric adjustments involved in $Q_{\text{net}}^{\prime }$ between extreme and moderate El Niños

Figure 7 displays the spatial patterns of $Q_{\text{SW}}^{\prime }$, $Q_{\mathit{E}}^{\prime }$, and the sum of both in the developing phase for the two types of El Niño. It is shown that the sum of $Q_{\mathit{E}}^{\prime }$ and $Q_{\text{SW}}^{\prime }$ matches well with the spatial patterns of $Q_{\text{net}}^{\prime }$, both in extreme and moderate El Niños, with the spatial correlations both exceeding 0.98 (Figs. 7c,f). This indicates that the $Q_{\mathit{E}}^{\prime }$ and $Q_{\text{SW}}^{\prime }$ terms dominate $Q_{\text{net}}^{\prime }$, while the $Q_{\mathit{H}}^{\prime }$ and $Q_{\text{LW}}^{\prime }$ terms (not shown) are negligible. The negative $Q_{\text{SW}}^{\prime }$, with their damping centers being located in the central equatorial Pacific in response to the atmospheric deep convection anomalies, extend to the eastern equatorial Pacific in extreme El Niños (Fig. 7a), but are confined to the central equatorial Pacific west of 180° and the north of the eastern Pacific in moderate El Niños (Fig. 7d). The $Q_{\mathit{E}}^{\prime }$ exhibits a zonal dipole pattern both in extreme and moderate El Niños, with weak (strong) positive (negative) anomalies in the central (eastern) equatorial Pacific (Figs. 7b,e). The sum of $Q_{\text{SW}}^{\prime }$ and $Q_{\mathit{E}}^{\prime }$ shows that the positive effects of $Q_{\mathit{E}}^{\prime }$ in the central equatorial Pacific are totally offset by the local negative effects of $Q_{\text{SW}}^{\prime }$, and the negative effects of $Q_{\mathit{E}}^{\prime }$ dominate the damping role of $Q_{\text{net}}^{\prime }$ in the eastern equatorial Pacific, leading to relatively weak damping effects in the central equatorial Pacific and strong damping effects in the eastern equatorial Pacific, both in extreme and moderate El Niños (Figs. 7c,f). Therefore, the $Q_{\mathit{E}}^{\prime }$ plays a dominant role both in the “larger warming gets more damping” zonal paradigm of $Q_{\text{net}}^{\prime }$ in extreme El Niños and in the zonal deviation between the positive SSTA center and the negative $Q_{\text{net}}^{\prime }$ center in moderate El Niños.

View Full Size

Spatial patterns of (a) surface net shortwave radiation anomalies, (b) surface latent heat flux anomalies, and (c) the sum of the two in extreme El Niños. The black contours in (a) and (c) are the spatial patterns of precipitation anomalies (units: °C, with an interval of 0.5 mm day⁻¹; zero contour thickened and negative dashed) and surface net heat flux anomalies (units: W m⁻², with an interval of 7.5 W m⁻²; zero contour thickened and negative dashed), respectively. (d)–(f) As in (a)–(c), but for moderate El Niños. Stippling indicates that the compositions of shaded values are significant at the 95% confidence level based on the Student’s t test.

Citation: Journal of Climate 34, 13; 10.1175/JCLI-D-20-0757.1

• Download Figure
• Download figure as PowerPoint slide

The factors contributing to $Q_{\mathit{E}}^{\prime }$ are further compared between extreme and moderate El Niños (Fig. 8). The reconstructed spatial patterns and magnitudes of $Q_{\mathit{E}}^{\prime }$ in extreme and moderate El Niños are almost identical compared with their original counterparts (Figs. 7b,e), with the spatial correlations both exceeding 0.97, indicating that the decomposition of $Q_{\mathit{E}}^{\prime }$ based on Eq. (6) is reasonable. The oceanic response represented by $Q_{\text{EO}}^{\prime }$ plays a negative role both in extreme and moderate El Niños (Figs. 8b,g). Regarding the atmospheric adjustments, the $Q_{\text{EW}}^{\prime }$ is totally negative in the eastern equatorial Pacific in moderate El Niños, but involves both positive and negative effects in extreme El Niños (Figs. 8c,h); the $Q_{\text{ERH}}^{\prime }$ term appears to play a critical role in the damping effects of $Q_{\mathit{E}}^{\prime }$ in the eastern equatorial Pacific both in extreme and moderate El Niños (Figs. 8d,i); and the $Q_{\text{E}\Delta \text{T}}^{\prime }$ term plays another important role for the damping effects in the eastern equatorial Pacific in extreme El Niños with local robust positive SSTAs, but is negligible in moderate El Niños with weak SSTAs (Figs. 8e,j).

View Full Size

Spatial patterns of the (a) reconstructed surface latent heat flux anomalies based on Eq. (6) and (b)–(e) each factor involved in the surface latent heat flux anomalies in extreme El Niños based on Eqs. (7)–(10): (b) the Newtonian cooling effect, and the atmospheric forcing effect due to anomalies in (c) surface wind speed, (d) relative humidity, and (e) surface stability. Contours in (a)–(e) are the spatial patterns of the original surface latent heat flux anomalies (units: W m⁻², with an interval of 7.5 W m⁻²; zero contour thickened and negative dashed), the SSTAs (units: °C, with an interval of 0.2°C; zero contour thickened and negative dashed), the surface wind speed anomalies (units: m s⁻¹, with an interval of 0.15 m s⁻¹; zero contour thickened and negative dashed), the relative humidity anomalies (with an interval of 7.5 × 10⁻³; zero contour thickened and negative dashed), and the surface stability anomalies (units: °C, with an interval of 0.15°C; zero contour thickened and negative dashed), respectively. (f)–(j) As in (a)–(e), but for moderate El Niños. Stippling indicates that the compositions of shaded values are significant at the 95% confidence level based on the Student’s t test.

Citation: Journal of Climate 34, 13; 10.1175/JCLI-D-20-0757.1

• Download Figure
• Download figure as PowerPoint slide

The discrepant effects of $Q_{\text{EW}}^{\prime }$ in the eastern equatorial Pacific between extreme and moderate El Niños could be due to local different positive WES feedback processes. In extreme El Niños, the robust positive SSTAs in the eastern equatorial Pacific trigger local deep convections (Fig. 7a, contours). The convective heating causes surface convergent anomalies in the eastern equatorial Pacific, including the intrusion of strong westerly wind anomalies from the central to the eastern equatorial Pacific (Fig. 9a; Xie et al. 2018) and the convergence of meridional wind anomalies to the equator (Fig. 9b). The intrusion of westerly anomalies weakens the background easterly winds and lowers the surface evaporation, contributing positively to the growth of warm SSTAs in the eastern equatorial Pacific (Fig. 9a), while the convergence of meridional wind anomalies weakens (enhances) the background cross-equatorial southerly winds and increases (decreases) the SSTAs north (south) of the equator (Fig. 9b). The positive and negative effects of $Q_{\text{EW}}^{\prime }$ largely counterbalance each other in the eastern equatorial Pacific, leading to relatively small negative effects on the growth of SSTAs (Fig. 8c). In moderate El Niños, however, the relatively weak positive SSTAs in the eastern Pacific cannot trigger local deep convections due to too cold background SST, but could be sufficient enough to trigger deep convections in the central equatorial Pacific and the climatological ITCZ region north of the eastern Pacific where the background SSTs are already high (Fig. 7d, contours), thus causing westerly anomalies confined to the central-western Pacific and cross-equatorial southeasterly anomalies in the eastern equatorial Pacific (Figs. 9c,d). The SSTA-induced southeasterly anomalies can feed back to the further development of SSTAs by enhancing the background southeasterlies and evaporation through the WES feedback, which produce prominent damping effects on the subsequent growth of SSTAs in the eastern equatorial Pacific and favor larger SSTAs to be located in the central equatorial Pacific.

View Full Size

Spatial patterns of the atmospheric forcing effect due to anomalies in (a) surface zonal wind speed and (b) meridional wind speed in extreme El Niños. Contours in (a) and (b) are the surface zonal wind anomalies and meridional wind anomalies (units: m s⁻¹, with an interval of 0.4 m s⁻¹; zero contour thickened and negative dashed), respectively. Vectors in (a) and (b) are the surface wind vector anomalies (units: m s⁻¹). (c),(d) As in (a) and (b), but for moderate El Niños. Note that the interval of contours in (c) and (d) is 0.2 m s⁻¹, which is different from that in (a) and (b). Stippling indicates that the compositions of shaded values are significant at the 95% confidence level based on the Student’s t test.

Citation: Journal of Climate 34, 13; 10.1175/JCLI-D-20-0757.1

• Download Figure
• Download figure as PowerPoint slide

The damping effects of $Q_{\text{ERH}}^{\prime }$ in the eastern equatorial Pacific could be attributable to the local negative feedback between SST and relative humidity. To verify such feedback, we define a relative humidity–SST feedback index (RSFI) by regressing the monthly anomalies of surface relative humidity onto the SSTAs. As shown in Fig. 10a, prominent negative RSFI values appear in the eastern equatorial Pacific. This indicates that the positive SSTAs in the eastern equatorial Pacific during El Niño events will reduce the local relative humidity, which further suppresses the growth of local SSTAs by inducing negative $Q_{\mathit{E}}^{\prime }$ (Figs. 8d,i). Such inherent negative feedback could be due to local strong vertical mixing between the boundary layer with relatively high relative humidity and the upper free atmosphere with relatively low relative humidity (Fig. 10c, contours) that is induced by positive SSTAs. The positive SSTAs increase the production of vertical mixing in the eastern Pacific boundary layer where the stratocumulus prevails (Wood 2012), thus enhancing the entrainment of upper-level dry air at the stratus cloud top, which tends to desiccate the whole boundary layer (Scott et al. 2020) and raise the boundary layer height. Indeed, the equatorial RSFIs are negative from the surface to the top of the boundary layer (which is also the stratus cloud top) where the climatological relative humidity is the largest due to the vertical mixing, but are positive in the upper free atmosphere (Fig. 10c, shaded). Moreover, there are positive feedbacks between the monthly anomalies of boundary layer height and SST in the eastern equatorial Pacific (BHFI, Fig. 10b), further verifying a stronger vertical mixing between the boundary layer and the free atmosphere that helps to reduce the surface relative humidity during El Niño events (Deser and Wallace 1990; Ham et al. 2018). Therefore, no matter which type of El Niño occurs, the inherent negative relative humidity–SST feedback helps to confine the damping effects of $Q_{\mathit{E}}^{\prime }$ to the eastern equatorial Pacific, contributing to both the “larger warming gets more damping” zonal paradigm in extreme El Niños and the more SSTAs in the central equatorial Pacific in moderate El Niños.

View Full Size

Spatial patterns of (a) relative humidity–SST feedback index (RSFI) and (b) boundary layer height–SST feedback index (BHFI). (c) Vertical distribution of equatorial (2.5°S–2.5°N) RSFI in the eastern Pacific. Contours in (c) denote the climatological relative humidity. Stippling indicates that the regressions are significant at the 95% confidence level based on the Student’s t test.

Citation: Journal of Climate 34, 13; 10.1175/JCLI-D-20-0757.1

• Download Figure
• Download figure as PowerPoint slide

Figure 11 quantifies the major flux anomalies during the developing phase of El Niño both in the eastern (2.5°S–2.5°N, 140°–90°W) and central (2.5°S–2.5°N, 180°–140°W) equatorial Pacific, as well as their differences. In extreme El Niños with the larger SSTAs in the eastern equatorial Pacific, $Q_{\text{net}}^{\prime }$ is mainly contributed by both $Q_{\text{SW}}^{\prime }$ and $Q_{\mathit{E}}^{\prime }$. The former contributes more damping effects in the central equatorial Pacific, which are partly offset by the local positive effects of $Q_{\text{EW}}^{\prime }$ involved in $Q_{\mathit{E}}^{\prime }$, while the latter plays a dominant damping role in the eastern equatorial Pacific, which is mainly contributed by $Q_{\text{EO}}^{\prime }$, $Q_{\text{ERH}}^{\prime }$, and $Q_{\text{E}\Delta \text{T}}^{\prime }$. In moderate El Niños with the larger SSTAs in the central equatorial Pacific, the zonal deviation between the positive SSTA center and the negative $Q_{\text{net}}^{\prime }$ center is mainly caused by more damping effects of $Q_{\mathit{E}}^{\prime }$ in the eastern equatorial Pacific, which are mainly contributed by $Q_{\text{EO}}^{\prime }$, $Q_{\text{EW}}^{\prime }$, and $Q_{\text{ERH}}^{\prime }$. Thus, apart from the oceanic response ($Q_{\text{EO}}^{\prime }$), it appears that the positive WES feedback and the negative relative humidity–SST feedback in the eastern equatorial Pacific are the two major atmospheric adjustments that lead to the zonal deviation between the positive SSTA center in the central Pacific and the negative $Q_{\text{net}}^{\prime }$ center in the eastern Pacific, favoring the largest SSTAs being confined to the central equatorial Pacific in moderate El Niños.

View Full Size

The major heat flux anomalies during the developing phase of (a) extreme and (b) moderate El Niños. The red, blue, and black bars denote the regional-mean values in the eastern equatorial Pacific (2.5°S–2.5°N, 140°–90°W), the central equatorial Pacific (2.5°S–2.5°N, 180°–140°W), and their differences (eastern Pacific minus central Pacific). The $Q_{\text{net}}^{\prime }$, $Q_{\text{SW}}^{\prime }$, $Q_{\mathit{E}}^{\prime }$, $Q_{\text{EO}}^{\prime }$, $Q_{\text{EW}}^{\prime }$, $Q_{\text{ERH}}^{\prime }$, and $Q_{\text{E}\Delta \text{T}}^{\prime }$ terms denote the surface net heat flux anomalies, the surface net shortwave radiation anomalies, the surface latent heat flux anomalies, the Newtonian cooling effect, and the atmospheric adjustments due to anomalies in surface wind speed, relative humidity, and surface stability, respectively. Note that the values on the y axis are different between (a) and (b).

Citation: Journal of Climate 34, 13; 10.1175/JCLI-D-20-0757.1

• Download Figure
• Download figure as PowerPoint slide