1. Introduction
The Pacific meridional mode (PMM; Chiang and Vimont 2004) is a dominant mode of ocean–atmospheric variation in the subtropical North Pacific, which is characterized by warm sea surface temperature (SST) and southwesterly surface wind anomalies extending from Baja California to the central-western equatorial Pacific during its positive phase (Amaya 2019). It has been documented that the PMM is an important modulator for the development of El Niño–Southern Oscillation (ENSO) phenomenon. Particularly, Chang et al. (2007) reported that 12 out of 17 ENSO events during 1960–2000 were preceded by spring PMM events. Based on model experimental analysis, Zhang et al. (2009) showed that the PMM-induced zonal wind anomalies were able to evoke oceanic waves in the tropics and promote ENSO development. Focusing on the mechanism for the impact of the PMM, Vimont et al. (2009) and Alexander et al. (2010) pointed out that the PMM-related variability first generated in the subtropical Pacific could continue to amplify and propagate to the tropics through the wind–evaporation–SST (WES) feedback (Xie and Philander 1994). The PMM also exerts an enormous dynamic impact on charging or discharging subsurface heat content of the equatorial ocean to support the evolution of an ENSO event, which was defined as “trade wind charging” (Anderson et al. 2013; Chakravorty et al. 2020). Recently, another mechanism named “summer deep convection” was proposed by Amaya et al. (2019); they found that the PMM-related SST anomalies could generate anomalous deep convection during the boreal summer and fall when the mean intertropical convergence zone (ITCZ) moved northward along its seasonal cycle, which produced a Gill-like (Gill 1980) atmospheric response with zonal wind anomalies over the equator, initiating equatorial oceanic waves and ENSO events.
The PMM is generally considered a contributor to the generation and development of central Pacific (CP) ENSO events (Yu and Kim 2011; Yu et al. 2011; Capotondi and Sardeshmukh 2015). Some studies attributed the increasing frequency of CP ENSO events after 1990 to the intensification of the PMM (Yu et al. 2012; Di Lorenzo et al. 2015). Besides, the PMM has been suggested to act as a bridge in the intensified influence of the tropical North Atlantic on ENSO over the past decades (Wang et al. 2017; Park et al. 2019). The strengthening impact of the PMM is believed to be associated with the change in background mean state (Amaya 2019). Wang et al. (2013) argued that the apparent warming over the subtropical and tropical western Pacific under global warming reinforced the link between ENSO and the precursors over the North Pacific including the PMM. Yu et al. (2015) found that the phase change in the Atlantic multidecadal oscillation (AMO) around 1990 led to enhancement of background trade wind, favoring stronger WES feedback and intensifying the PMM. The location of the ITCZ is another factor that can modulate the effectiveness of the PMM’s equatorward propagation (Zhang et al. 2014b; Martinez-Villalobos and Vimont 2016). Using reanalysis data starting from 1948, Park et al. (2021) demonstrated that the WES feedback became stronger when the northern portion of the ITCZ was intensified.
While those studies diagnosed the strengthening of the PMM from a historical perspective or using idealized experiments, the long-term change in the PMM and its impact on ENSO under greenhouse warming need more careful consideration. On the other hand, it is hard to know whether the increased magnitude and impact of the PMM in observations are caused by a natural decadal variation or anthropogenic global warming due to the short data record. To address the uncertainty, several studies tried to evaluate the long-term change in the PMM in future projections using a coupled climate model (Liguori and Di Lorenzo 2018; Sanchez et al. 2019). It was shown that the PMM variability becomes more energetic in the warmer mean state of the RCP8.5 projection. However, the conclusion based on only one climate model, the Community Earth System Model (CESM; Hurrell et al. 2013), may be less convincing. In this context, model simulations produced by different climate models available from the Coupled Model Intercomparison Project (CMIP) provide us an opportunity for more comprehensive investigation.
Here, we use model simulations from phase 6 of CMIP (CMIP6; Eyring et al. 2016) to examine the change in the PMM in response to greenhouse warming. Recently, using a large number of CMIP5 and CMIP6 models, Jia et al. (2021) found that the North Pacific impact on ENSO was likely to be enhanced. The model selection in Jia et al. (2021) was conducted mainly based on the features of the PMM, without considering the model spread in the PMM–ENSO relationship. Prior studies with CMIP models revealed a large spread in the simulations of not only the basic features of the PMM, but also its relationship with ENSO (Lin et al. 2015; Wang et al. 2019; Park et al. 2021; Jia et al. 2021). In this study, we aim to further understand the diversity in the CMIP6 models’ representation of the PMM and especially its impact on ENSO before performing future projections based on the best performing models.
The present study aims to answer three scientific questions: 1) How well do the CMIP6 models simulate the PMM and what are the factors responsible for the intermodel uncertainty? 2) Is the strengthening PMM caused by anthropogenic warming? 3) How and why will the PMM and its impact on ENSO change in a warmer climate state on the basis of CMIP6 future projections? This paper is organized as follows. In section 2, we describe the observational datasets and CMIP6 model outputs, and introduce the analysis methods applied in this study. Observed changes in PMM strength and its impact on ENSO are presented in section 3. In section 4, we evaluate the performance of CMIP6 models in simulating the PMM. Using the models with good simulations of the PMM, we present future projections of the PMM and related mechanism in section 5. A summary and discussion are given in section 6.
2. Data and methods
a. Observational datasets
The observed SST data analyzed in this study is from the National Oceanic and Atmospheric Administration (NOAA) Extended Reconstructed Sea Surface Temperature dataset version 5 (ERSSTv5; Huang et al. 2017), with a resolution of 2° × 2°. Surface winds are obtained from the National Centers for Environmental Prediction–National Center for Atmospheric Research reanalysis product (NCEP–NCAR; Kalnay et al. 1996), available on a global grid of 2.5°. Both datasets are available from 1949 to 2018. Long-term linear trends are removed before subsequent calculation and analysis.
b. CMIP6 model outputs
Monthly mean outputs from the historical and shared socioeconomic pathway (SSP) 585 (SSP585) simulations produced by 29 CMIP6 models (Table 1) are analyzed to investigate the future projections of the PMM. The CMIP6 datasets are archived and freely accessible online (https://esgf-node.llnl.gov/search/cmip6/). The variables analyzed in this study include SST, surface wind, precipitation, and latent heat flux. To demonstrate the projected changes in the PMM, we choose two 50-yr periods to represent the historical (1955–2004) and future warming (2050–99) scenarios. All CMIP6 model outputs have been interpolated to the resolution of 2.5° × 2.5° before further analysis. Linear trends are removed from both historical and SSP585 simulation fields.
List of CMIP6 models analyzed and historical PMM–ENSO correlation coefficients. The models that successfully simulate positive PMM–ENSO correlation coefficient are selected to conduct the analysis of future projection (denoted in boldface). The models denoted in italics did not produce SSP585 simulations and thus are not used in spite of their positive PMM–ENSO relationship.
c. Definition of PMM indices
The PMM is calculated using a method similar to that defined in Chiang and Vimont (2004), that is, conducting maximum covariation analysis (MCA; Newman and Sardeshmukh 1995) on monthly SST and surface wind anomalies over the domain of 20°S–30°N, 175°E–95°W. ENSO-related variability is not removed, so the PMM here is the second mode of the MCA after ENSO. Two PMM indices are obtained by the MCA operation, namely, the SST expansion coefficient (PMMsst hereafter) and wind expansion coefficient (PMMwind). To depict SST variation, we mainly pay attention to the PMMsst index and then obtain the PMM patterns by regressing both SST and surface wind anomalies on the standardized PMMsst index.
d. Significance test
3. Intensified strength of observed PMM and its impact on ENSO
We first examine the trends of PMM strength and impact on ENSO in observations. Figure 1a displays the PMM patterns of SST and surface wind anomalies, resembling the results in Chiang and Vimont (2004), which is the first MCA mode with ENSO-related variability removed in advance. Considering the high correlation between the PMMsst and PMMwind indices (0.6 for 840 samples; Fig. 1b), we define the strength of the PMM as the 21-yr running standard deviation of the PMMsst index (Figs. 1c,d), which was also adopted in previous studies (e.g., Sanchez et al. 2019). In both monthly (Fig. 1c) and March–May (MAM; peak season of the PMM) mean (Fig. 1d) time series, PMM strength shows an increasing trend (significant at the 99% confidence level) from 1949 to the present. Correspondingly, similar increasing trends have also been observed in the 21-yr running correlation coefficients of the winter [December (0)–February (+1) (DJF)] Niño-4 (5°S–5°N, 160°E–150°W; Fig. 1e) and Niño-3 (5°S–5°N, 150°–90°W; Fig. 1f) indices with the preceded spring (MAM) PMMsst index, implying that the relationship between the PMM and ENSO became stronger over the past decades. Note that the negative values of correlation coefficient between the PMM and Niño-3 index during the 1960s were insignificant; thus, the precursor role of the PMM in ENSO development should not be considered. This result reaffirms the intensified PMM strength and impact on ENSO, as reported previously (Liguori and Di Lorenzo 2018; Sanchez et al. 2019). In the following, we employ the CMIP6 model simulations to analyze the projected changes in the PMM and its impact on ENSO.
(a) PMM patterns of SST anomalies (shading; K) and surface wind anomalies (vectors; m s−1, shown only for values exceeding the 95% confidence level) as regressions of SST and 1000-hPa wind anomalies onto the SST expansion coefficient of MCA second mode. Stippling indicates the SST anomalies are significant at the 95% confidence level. (b) Time series of the SST (red; PMMsst index) and wind (blue; PMMwind index) expansion coefficients. 21-yr running standard deviation of (c) monthly PMMsst index and (d) spring (March to May) PMMsst index. 21-yr running correlation between spring PMMsst index and winter [December (0) to February (+1)] (e) Niño-4 index and (f) Niño-3 index.
Citation: Journal of Climate 35, 15; 10.1175/JCLI-D-21-0683.1
4. Performances of CMIP6 models in PMM simulations
For CMIP6 models, we first examine their performances in simulating the PMM and its connection with the subsequent ENSO. Figure 2 shows the PMM patterns of SST and surface wind anomalies in the historical simulations of 29 CMIP6 models, together with the observed ones for comparison. The meridional dipole pattern of SST anomalies, especially the southwestward extending warm center from Baja California overlayed by the southwesterly wind anomalies, can be generally captured by almost all CMIP6 models in spite of some differences in magnitude and location. In particular, several models including BCC-CSM2-MR, CanESM5, FGOALS-f3-L, GISS-E2-1-G, IPSL-CM6A-LR, MCM-UA-1-0, and NorCPM1 mix the signals of the north PMM and south PMM (the counterpart of the north PMM in the Southern Hemisphere, with similar nature but slightly different zonal and meridional locations; Zhang et al. 2014a), showing subtropical warming centers in both the North Pacific and South Pacific. Moreover, the northern subtropical warming center in NorESM2-LM is excessively broad and strong in comparison to observations and the other models.
PMM patterns of SST anomalies (shading; K) and surface wind anomalies (vectors; m s−1) in historical simulations of 29 CMIP6 models and their MME mean. Stippling indicates the SST anomalies are significant at the 95% confidence level. Value above each panel indicates the corresponding explained variance of the PMM pattern.
Citation: Journal of Climate 35, 15; 10.1175/JCLI-D-21-0683.1
To quantitatively evaluate the PMM patterns produced by the 29 models, we display the spatial correlation coefficients and standard deviations of SST anomalies in the subtropical–tropical Pacific (20°S–30°N, 175°E–95°W; the region for MCA) in a Taylor diagram. Figure 3 shows that the spatial correlation coefficients are all above 0.6, indicating that the CMIP6 models can reproduce the observed PMM patterns quite well, although the models with macroscopic biases mentioned above are relatively weakly correlated with the observed. Except for the apparent overestimate by NorESM2-LM and MIROC-ES2L and underestimate by GISS-E2-1-H, most models simulate the PMM magnitude similar to the observations. Note that the spatial correlation coefficient is higher for the MME mean than for each individual model, implying that the result is less convincing if only one model is considered.
Taylor diagram of PMM SST anomalies in the tropical–subtropical Pacific (20°S–30°N, 175°E–95°W) presented in Fig. 2, with respect to observations. The blue dot represents the MME mean.
Citation: Journal of Climate 35, 15; 10.1175/JCLI-D-21-0683.1
We now assess the connection of the PMM with subsequent ENSO in the historical simulations of the CMIP6 models. Figure 4 displays the regression maps of the SST anomalies in winter (DJF) on spring (MAM) standardized PMMsst index for observation and each of the 29 models. In the reanalysis, SST warming appears in the central equatorial Pacific in winter if positive PMM events occur in the preceding spring, and the PMM–ENSO correlation coefficient is significant at the 99% confidence level, suggesting that the PMM contributes to the development of ENSO. Such connection is captured by more than half of the CMIP6 models, in which, however, the SST anomalies in response to spring PMM occur not only in the central equatorial Pacific but also in the eastern equatorial Pacific attributed possibly to the deficiency in simulating CP ENSO (Wang et al. 2019). The corresponding correlation coefficients between the MAM PMMsst index and the DJF Niño-4 index (PMM–ENSO correlation coefficient hereafter) of all the 29 models, shown by the bar plots in Fig. 4, range from below −0.2 (BCC-CSM2-MR) to above 0.6 (MIROC-ES2L). It is noteworthy that the PMM–ENSO correlation coefficients in six models (out of the 29 models) are negative. As a result, the MME-mean PMM–ENSO correlation coefficient is slightly lower than the 95% confidence level, and the intermodel spread is comparable to the MME mean although 16 out of the 29 models simulate significant PMM–ENSO correlation coefficients (exceeding the 95% confidence level). A question to ask here is what causes the large spread in the connection of the PMM with ENSO among the CMIP6 models? To address this question, we carry out a series of comparison between 10 models with the highest PMM–ENSO correlation coefficients (HC group) and 10 models with the lowest coefficients (LC group).
(top) Regressions of winter SST anomalies (K) onto the spring standardized PMMsst index. Stippling indicates the SST anomalies are significant at the 95% confidence level. (bottom) Correlation coefficients between the spring PMMsst index and the following winter Niño-4 index. The red bar presents the observed result, blue bars show the results in the historical simulations of the 29 CMIP6 models, and the orange bar is the MME mean. Error bars denote the 97.5% and 2.5% confidence bounds measured by a 10 000-resampling bootstrap method. Black dashed lines represent the 95% and 99% confidence levels based on Student’s t test.
Citation: Journal of Climate 35, 15; 10.1175/JCLI-D-21-0683.1
Figure 5 shows the ensemble mean of seasonal evolution of SST anomalies regressed upon the preceding spring PMMsst index for the HC and LC groups, and their differences. In spring (MAM), PMM-like patterns appear in both HC and LC groups with highly similar amplitude and distribution (Figs. 5a,e) and slight differences in SST cooling in the eastern tropical Pacific (Fig. 5i). Despite the similarity during spring, a different evolution of SST anomalies emerges in the tropical Pacific in the following seasons (Figs. 5j–l). As the SST anomalies in the extratropics decline gradually, a robust El Niño–like pattern appears subsequently in the HC group (Figs. 5b–d), well capturing the so-called seasonal footprinting mechanism (SFM; Vimont et al. 2003). In contrast, no SST warming can be seen in the tropical Pacific for the LC group (Figs. 5f–h). Comparing the PMM-related SST anomalies suggests that different performances in the PMM–ENSO connection between the HC and LC groups is not a result of the contrast in PMM strength during spring.
Seasonal [March (0) to May (0), June (0) to August (0), September (0) to November (0), and December (0) to February (+1)] regressions of SST anomalies (K) in historical simulations onto the spring PMMsst index: (a)–(d) HC group and (e)–(h) LC group, with stippling in (a)–(h) indicating SST anomalies are significant at the 95% confidence level. (i)–(l) The differences between (a)–(d) and (e)–(h), with stippling indicating the difference is significant at the 95% confidence level based on a 10 000-resampling bootstrap method.
Citation: Journal of Climate 35, 15; 10.1175/JCLI-D-21-0683.1
We further compare differences in the physical processes of the SFM between the HC and LC groups to explore the factors responsible for model diversity. Differences in the seasonal evolution of PMM-related precipitation, surface wind, and latent heat flux anomalies are displayed in Fig. 6. Although the PMM-related SST anomalies during spring are comparable in the two groups, larger precipitation and westerly wind anomalies in the HC group occur over the western subtropical Pacific. Accordingly, the positive latent heat flux anomalies that ensure the equatorward expansion of SST warming are also significantly larger in the HC group than in the LC group. The above results indicate that the efficiency of the SFM in the HC group might be higher and lead to a stronger relationship between the PMM and ENSO compared to the LC group. To test this hypothesis, we draw intermodel scatterplots of the correlation coefficients between the PMMsst index in MAM and the Niño-4 index in DJF versus the springtime regional mean PMM-related precipitation anomalies (Fig. 7a) and latent heat flux anomalies (Fig. 7b) over the western tropical Pacific (box in Figs. 6a, b; 0°–15°N, 130°E–140°W). As expected, the two pairs of factors are both highly correlated, exceeding the 99% confidence level. More specifically, the correlation coefficient between the two factors in Fig. 7a (PMM–ENSO correlation coefficients and regional mean precipitation anomalies) is 0.59, and the one in Fig. 7b (PMM–ENSO correlation coefficients and regional mean latent heat flux anomalies) is 0.44. Combined with the composite maps in Fig. 6, this result further suggests that the intermodel spread of the PMM–ENSO connection is likely due to the diversity of the SFM efficiency among the CMIP6 models.
(a)–(d) As in Figs. 5i,j, but for precipitation anomalies (shading; mm day−1; stippling) and surface wind anomalies (vectors; m s−1; shown only for values exceeding the 95% confidence level). (e)–(f) As in Figs. 5i,j, but for latent heat flux anomalies (W m−2).
Citation: Journal of Climate 35, 15; 10.1175/JCLI-D-21-0683.1
Scatterplots of the correlation coefficients between spring PMMsst index and winter Niño-4 index vs the regional mean PMM-related (a) precipitation anomalies and (b) latent heat flux anomalies over the western tropical Pacific (box in Figs. 6a,b; 0°–15°N, 130°E–140°W) in spring. Red dots denote the HC models, blue dots denote the LC models, and black dots denote the rest of the models, which are not the HC and LC models. Value at the top right of each panel indicates the corresponding correlation coefficient and P value.
Citation: Journal of Climate 35, 15; 10.1175/JCLI-D-21-0683.1
Previous studies demonstrated that the SFM efficiency is mainly determined by the intensity of the WES feedback (Vimont et al. 2009; Park et al. 2021). Two processes in the WES feedback may modulate the efficiency: 1) anomalous atmospheric convection in response to SST anomalies, and 2) the latent heat flux change due to wind speed anomalies (Xie and Philander 1994). In terms of process 1, Figs. 8a–c show the MAM regression maps of precipitation anomalies in the HC and LC groups against the regional mean SST anomalies in the extratropics (box in Figs. 8a,b; 10°–30°N, 160°E–110°W) and the corresponding differences. It is clear that the same magnitude of SST anomalies can lead to larger precipitation anomalies for the HC group, and that the differences in the western tropical Pacific are statistically significant at the 95% confidence level. The significantly stronger precipitation response here implies a stronger surface atmospheric circulation and thus larger zonal wind anomalies along the equator, which ultimately would lead to a stronger impact on the development of ENSO in the HC group.
(a)–(c) Regressions of precipitation anomalies onto the regional mean SST anomalies in the subtropical Pacific [box in (a) and (b), 10°–30°N, 160°E–110°W] (mm day−1 K−1) in spring. (d)–(f) The WES parameter [W m−2 (m s−1)−1] in spring. (a),(d) The results for the HC group; (b),(e) the results for the LC group; (c),(f) their differences. Stippling indicates the difference is significant at the 95% confidence level.
Citation: Journal of Climate 35, 15; 10.1175/JCLI-D-21-0683.1
In short, some of the 29 CMIP6 models considered in the present study can properly simulate the spatial characteristics of the PMM and its connection with ENSO. However, large intermodel uncertainties exist due to the diversity of SFM efficiency, which can be further attributed to the variation of atmospheric convection response to SST anomalies in the WES feedback. Next, we will use the models with better ability in simulating the PMM to investigate potential changes in the PMM in a warmer mean state.
5. Projected change in PMM strength and its impact on ENSO
Among the 29 models analyzed in this study, 23 models simulated positive PMM–ENSO correlation coefficients in the historical simulation. However, seven of the models did not provide SSP585 simulations, so we can only use the 16 models (denoted in bold in Table 1) to conduct the analysis of future projection. Figure 9 compares the PMM patterns of SST anomalies between the historical period (1955–2004) and future climate change period (2050–99). It shows no essential differences in general. Particularly, the spatial correlation coefficients between the PMM patterns in the historical and SSP585 simulations are above 0.65 for all the models and above 0.8 for 11 models; and the MME-mean correlation coefficient is as high as about 0.9 (Fig. 10a). These high correlation coefficients suggest that the intrinsic features of the PMM are still robust in the warming climate. Yet, striking equatorward expansions and positive differences can be seen in the SSP585 PMM patterns compared with the historical ones, indicating that the ocean–atmospheric coupling related to the PMM may even become stronger under future global warming.
PMM patterns of SST anomalies in SSP585 simulations (shading; K) and the differences from those in the historical simulations (contours; −0.5 to 0.5 with an interval of 0.1) of the 16 selected CMIP6 models. Stippling indicates the PMM-related SST anomalies in the SSP585 simulations are significant at the 95% confidence level. Value above each panel indicates the corresponding explained variance of the PMM pattern.
Citation: Journal of Climate 35, 15; 10.1175/JCLI-D-21-0683.1
(a) Pattern correlation coefficients over (20°S–30°N, 175°E–95°W) of the PMM-related SST anomalies between the SSP585 simulations and historical simulations. (b) Ratios of the variance of the historical-based PROJsst index in the SSP585 simulations to that in the historical simulations. (c) As in (b), but for the SSP585-based PROJsst index. Blue bars show the results from the 16 selected models, orange bars are the MME means, and error bars denote the 97.5% and 2.5% confidence bounds measured by a 10 000-resampling bootstrap method. Dashed lines in (b) and (c) show the bounds of 1, which mean that the variance of the historical PROJsst is equal to the that in the SPS585 simulation.
Citation: Journal of Climate 35, 15; 10.1175/JCLI-D-21-0683.1
Figure 11 displays the differences in the regression maps of winter SST anomalies between the SSP585 and historical simulations on the MAM PMMsst index, for the 16 models. Except for two models that simulate a weaker PMM in a warming climate (CNRM-CM6-1 and NESM3), the winter SST anomalies in the equatorial Pacific in response to the PMM are relatively larger in the SSP585 scenario. Specifically, the intensified linkage is most significant in EC-Earth3 and EC-Earth3-Veg, matching the most pronounced enhancement of PMM amplitude in these two models. We also compare the PMM–ENSO correlation coefficient between the SSP585 and historical simulations (bar plots in Fig. 11). In the future period, 12 of the 16 models simulate a larger positive relationship between the PMM and ENSO, compared to the historical simulations. Among the four remaining models, three models still simulate significant PMM–ENSO correlation coefficients at the 99% confidence level. The increase in the MME mean is close to 0.15, roughly 40% of the historical MME-mean PMM–ENSO correlation coefficient (0.36). The future MME mean is 0.49, significantly greater than the historical one at the 90% confidence level. Overall, the above results present an intermodel consensus that both the PMM strength and its impact on ENSO are likely to become stronger in a warming climate.
(top) Winter SST anomalies related to previous spring PMM in the SSP585 simulations (contours; K) and projected changes (shading; K; SSP585 minus historical). (bottom) Correlation coefficients between the spring PMMsst index and the following winter Niño-4 index. Light blue (individual) and yellow (MME mean) bars show the historical results, deep blue (individual) and orange (MME mean) bars show the SSP585 results, and red (increase) and green (decrease) bars indicate the differences between the SSP585 and historical results. Error bars in the MME-mean results denote the 95% and 5% confidence bounds measured by a 10 000-resampling bootstrap method. Black dashed lines represent the 95% and 99% confidence levels for the blue bars based on Student’s t test.
Citation: Journal of Climate 35, 15; 10.1175/JCLI-D-21-0683.1
The physical processes responsible for the potential strengthening of the PMM in future projections are explored next. We examine the changes in the SFM associated with the spring PMM. Figure 12 shows the seasonal differences in the SST anomalies related to the MAM PMMsst index. In spring, large SST warming prevails in the active zone of the PMM, which is followed by a strong warming in the central equatorial Pacific in summer. In fall and winter, the SST difference at the equator resembles an El Niño–like pattern, indicating a stronger SFM in the future period. The stronger SFM is even more pronounced if considering only the 12 models with high PMM–ENSO correlation coefficients (Figs. 12e–h). The differences in the changes in PMM-related convection and atmospheric circulation during spring and summer are also statistically significant, especially over the western tropical Pacific. Particularly, large precipitation anomalies are coupled with the southwesterly wind change in the vicinity of the equator, which is essential for the intensified impact of the PMM on ENSO development (Figs. 13a–d). Further evidence for the SFM enhancement in the future period is the robust positive difference in latent heat flux anomalies to the southwest of the SST anomalies in spring (Figs. 13e–h), ensuring more robust equatorward propagation of the PMM and thus greater SFM efficiency. Since the SFM is projected to strengthen, we then diagnose the projected changes in the WES feedback, focusing on two processes: 1) anomalous atmospheric convection in response to SST anomalies and 2) latent heat flux change due to wind speed anomalies, as discussed in section 4.
Seasonal differences in SST anomalies (K) related to spring PMM between the SSP585 simulations and historical simulations. (left) The results from all 16 selected models, and (right) the results from the 12 models with increased correlation coefficients between the spring PMMsst index and the winter Niño-4 index. Stippling indicates the difference is significant at the 95% level based on a 10 000-resampling bootstrap method.
Citation: Journal of Climate 35, 15; 10.1175/JCLI-D-21-0683.1
(a)–(d) As in Figs. 12a–d, but for precipitation anomalies (shading; mm day−1; stippling) and surface wind anomalies (vectors; m s−1; shown only the values exceeding the 95% confidence level). (e)–(h) As in Figs. 12a–d, but for latent heat flux anomalies (W m−2).
Citation: Journal of Climate 35, 15; 10.1175/JCLI-D-21-0683.1
The spatial distributions of simultaneous PMM-related precipitation anomalies in spring are broadly similar over the subtropical–tropical Pacific between the historical and SSP585 simulations, that is, the precipitation anomalies show a response with an apparent meridional gradience in both scenarios (Figs. 14a,b,d,e). However, compared with the historical simulations, the amplitude of precipitation anomalies indeed becomes larger in the future period, though not by much and only over a narrow region (Figs. 14c,f). The contrasts here indicate that the projected strengthening of the SFM may be partially due to the change in process 1 (see the previous paragraph). Figure 15 exhibits the patterns of the springtime WESp in the historical and SSP585 simulations and their differences. WESp is expected to generate a change in significantly larger magnitude under the SSP585 greenhouse gas forcing, especially over the region where the PMM is most active.
Regressions of spring precipitation anomalies onto the regional mean SST anomalies in the subtropical Pacific [box in (a)–(e), 10°–30°N, 160°E–110°W] (mm day−1 K−1) for (a),(d) historical simulations; (b),(e) SSP585 simulations; and (c),(f) projected changes (SSP585 minus historical). (a)–(c) The results from all 16 selected models and (d)–(f) the results from the 12 models with increased correlation coefficients between the spring PMMsst index and the winter Niño-4 index. Stippling indicates the difference is significant at the 95% confidence level.
Citation: Journal of Climate 35, 15; 10.1175/JCLI-D-21-0683.1
As in Fig. 14, but for the WES parameter [W m−2 (m s−1)−1]. Box denotes the active region of the PMM (5°–25°N, 150°E–120°W).
Citation: Journal of Climate 35, 15; 10.1175/JCLI-D-21-0683.1
To illustrate the temporal evolution, we compute the regionally averaged WESp over the PMM hotspot (5°–25°N, 150°E–120°W; box in Fig. 15) from 1950 to 2100 spanning both historical and SSP585 simulations. The time series of WESp displays a continuous increasing trend that accelerates after 2010 (Fig. 16), which should be caused by the warmer SST and thus greater mean heat flux loss. The MME-mean WESp in the twentieth century fluctuates around 14.3, whereas the value increases to more than 15.5 by the end of the twenty-first century. The 50-yr mean WESp during the period of 2050–99 is significantly greater than that during 1955–2004 at the 95% confidence level, which indicates a higher sensitivity of latent heat flux to zonal wind speed in future warming climate. In other words, the potential enhancement of PMM magnitude and its impact on ENSO under global warming seems to be attributed to the stronger efficiency of process 2 described above.
Long-term change in the WES parameter [W m−2 (m s−1)−1] averaged over 5°–25°N, 150°E–120°W (box in Fig. 15) from 1949 to 2099 in the CMIP6 model simulations. Black lines are for the MME mean, solid color lines denote 50-yr mean in the historical period (1955–2004; blue) and future period (2050–99; red), and shading indicates the confidence intervals as the 97.5% and 2.5% confidence bounds measured by a 10 000-resampling bootstrap method. (a) The results from all 16 selected models and (b) the results from the 12 models with increased correlation coefficients between the spring PMMsst index and the winter Niño-4 index.
Citation: Journal of Climate 35, 15; 10.1175/JCLI-D-21-0683.1
6. Summary and discussion
The PMM is an important ocean–atmospheric process in the subtropical North Pacific, which modulates the development of ENSO by inducing zonal wind anomalies over the equator. It was reported that the PMM had become more active during the past decades, leading to more frequent CP ENSO events after 1990 (Yu and Kim 2011). The intensified PMM and its impact on ENSO were suggested to be associated with the change in the mean climate state (Wang et al. 2013; Yu et al. 2015). However, it is unclear whether this change is an interdecadal variability or the result of anthropogenic warming. It was also argued that the amplifying trend is projected to continue throughout the twenty-first century under a strong greenhouse warming scenario, though the authors used only a single model (Liguori and Di Lorenzo 2018; Sanchez et al. 2019).
This study examined the projections of the PMM and its connection with ENSO under greenhouse gas forcing using CMIP6 models. A total of 29 models were analyzed, and most of them can well capture the characteristics of the PMM patterns, which are comparable to observations. However, these 29 models have varying performances in simulating the impact of the PMM on ENSO and relevant SFM processes. Further analysis demonstrated that the intermodel uncertainty of the PMM–ENSO relationship among the CMIP6 models is primarily due to the spread in the WES feedback amplitude, especially due to different atmospheric convection responses to the subtropical SST anomalies related to the PMM. Previous studies showed that precipitation response to SST anomalies is stronger in the high latitudes as the mean latitudinal position of the Pacific ITCZ moves poleward (e.g., Park et al. 2021). Hence, we examined the differences in mean precipitation between the HC and LC groups (Fig. S1a in the online supplemental material). It is shown that the mean precipitation is more intense over the northern part and weaker over the southern part of the ITCZ in the HC group, implying a poleward shift of the mean ITCZ relative to the LC group. However, these differences are only moderately significant, and so are the regression maps of mean precipitation across the 29 models on PMM–ENSO correlation coefficients (Fig. S1b). The SST mean state may also modulate the convection activity, that is, warmer SSTs would ensure stronger convection sensitivity (Zhong et al. 2017). However, the differences between the mean SSTs of the HC and LC groups are not significant, similar to precipitation (Fig. S2). Therefore, we suspect that different sensitivity of atmospheric response to SST anomalies among the CMIP6 models should be attributed mainly to different thresholds of convection rather than to different mean states.
Based on the assessment of the historical simulations, we employed 16 models with the best representation of the PMM and its impact on ENSO, using their SSP585 simulations to explore the potential changes due to anthropogenic warming. Results showed that more than 10 models simulate stronger PMM in a warming mean state. For the PMM impact on ENSO, 12 of the 16 models produce enhancement of the PMM–ENSO correlation. Among the four models that have a weaker PMM–ENSO relationship, three models still simulate significant PMM–ENSO correlation coefficients at the 99% confidence level in the future period. The analysis indicated that both the PMM and its impact on ENSO would probably intensify under global warming. Further diagnosis demonstrated that WESp, which can estimate the sensitivity of latent heat flux to zonal wind speed, is expected to experience an increasing trend throughout the twenty-first century in the SSP585 simulations; this should be the main reason for the potential enhancement of the PMM impact on ENSO. A further examination showed that the increase in WESp is associated with a larger amount of mean latent heat flux in the warming climate (Fig. S3). In addition, the atmospheric convection response to the SST anomalies in the subtropical Pacific shows a slight but significant intensification, which may also contribute to the stronger PMM–ENSO relationship in future projections. To sum up, the CMIP6 multimodel results suggest that anthropogenic warming may lead to strengthening amplitude of the PMM and its impact on ENSO, supporting previous results based on a single model (Liguori and Di Lorenzo 2018; Sanchez et al. 2019), and a recent study based on CMIP5 and CMIP6 models (Jia et al. 2021). Based on the expectation of a stronger PMM impact, more attention should be paid to the PMM in ENSO prediction, as pointed out previously (Larson and Kirtman 2014; Lu et al. 2017; Ma et al. 2017).
Since the PMM can be driven by the North Pacific Oscillation (Walker and Bliss 1932; Rogers 1981; Linkin and Nigam 2008), tropical North Atlantic SST anomalies (Ham et al. 2013), and anomalous Kuroshio Extension (Zhang et al. 2010; Joh and Di Lorenzo 2019), the changes in these variabilities may also be conducive to the potential strengthening of the PMM (Park et al. 2019; Chen and Yu 2020a,b). This issue should be addressed in further investigation. Moreover, it is of interest to understand how the south PMM (Zhang et al. 2014) change is related to anthropogenic forcing since it was also suggested as an important modulator of ENSO development (Sun et al. 2013; Min et al. 2017).
Acknowledgments.
The authors greatly appreciate the useful comments and suggestions by the editor and three anonymous reviewers who provided thorough reviews of our paper. This research was jointly supported by the National Natural Science Foundation of China (41690123, 41690120, and 41731173), the National Key Research and Development Program of China (2019YFA0606701 and 2020YFA0608803), the National Postdoctoral Program of Innovative Talents (BX2021324), the Guangdong Major Project of Basic and Applied Basic Research (Grant 2020B0301030004), the Guangdong Province Key Laboratory for Climate Change and Natural Disaster Studies (Grant 2020B1212060025), and the Jiangsu Collaborative Innovation Center for Climate Change.
REFERENCES
Alexander, M. A., D. J. Vimont, P. Chang, and J. D. Scott, 2010: The impact of extratropical atmospheric variability on ENSO: Testing the seasonal footprinting mechanism using coupled model experiments. J. Climate, 23, 2885–2901, https://doi.org/10.1175/2010JCLI3205.1.
Amaya, D. J., 2019: The Pacific meridional mode and ENSO: A review. Curr. Climate Change Rep., 5, 296–307, https://doi.org/10.1007/s40641-019-00142-x.
Amaya, D. J., Y. Kosaka, W. Zhou, Y. Zhang, S. P. Xie, and A. J. Miller, 2019: The North Pacific pacemaker effect on historical ENSO and its mechanisms. J. Climate, 32, 7643–7661, https://doi.org/10.1175/JCLI-D-19-0040.1.
Anderson, B. T., R. C. Perez, and A. Karspeck, 2013: Triggering of El Niño onset through trade wind-induced charging of the equatorial Pacific. Geophys. Res. Lett., 40, 1212–1216, https://doi.org/10.1002/grl.50200.
Capotondi, A., and P. D. Sardeshmukh, 2015: Optimal precursors of different types of ENSO events. Geophys. Res. Lett., 42, 9952–9960, https://doi.org/10.1002/2015GL066171.
Chakravorty, S., R. C. Perez, B. T. Anderson, B. S. Giese, S. M. Larson, and V. Pivotti, 2020: Testing the trade wind charging mechanism and its influence on ENSO variability. J. Climate, 33, 7391–7411, https://doi.org/10.1175/JCLI-D-19-0727.1.
Chang, P., L. Zhang, R. Saravanan, D. J. Vimont, J. C. H. Chiang, L. Ji, H. Seidel, and M. K. Tippett, 2007: Pacific meridional mode and El Niño–Southern oscillation. Geophys. Res. Lett., 34, L16608, https://doi.org/10.1029/2007GL030302.
Chen, S., and B. Yu, 2020a: The seasonal footprinting mechanism in large ensemble simulations of the second generation Canadian earth system model: Uncertainty due to internal climate variability. Climate Dyn., 55, 2523–2541, https://doi.org/10.1007/s00382-020-05396-y.
Chen, S., and B. Yu, 2020b: Projection of winter NPO-following winter ENSO connection in a warming climate: Uncertainty due to internal climate variability. Climatic Change, 162, 723–740, https://doi.org/10.1007/s10584-020-02778-3.
Chiang, J. C. H., and D. J. Vimont, 2004: Analogous Pacific and Atlantic meridional modes of tropical atmosphere–ocean variability. J. Climate, 17, 4143–4158, https://doi.org/10.1175/JCLI4953.1.
Di Lorenzo, E., G. Liguori, N. Schneider, J. C. Furtado, B. T. Anderson, and M. A. Alexander, 2015: ENSO and meridional modes: A null hypothesis for Pacific climate variability. Geophys. Res. Lett., 42, 9440–9448, https://doi.org/10.1002/2015GL066281.
Eyring, V., S. Bony, G. A. Meehl, C. A. Senior, B. Stevens, R. J. Stouffer, and K. E. Taylor, 2016: Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization. Geosci. Model Dev., 9, 1937–1958, https://doi.org/10.5194/gmd-9-1937-2016.
Fan, H., B. Huang, S. Yang, and W. Dong, 2021: Influence of the Pacific meridional mode on ENSO evolution and predictability: Asymmetric modulation and ocean preconditioning. J. Climate, 34, 1881–1901, https://doi.org/10.1175/JCLI-D-20-0109.1.
Gill, A. E., 1980: Some simple solutions for heat‐induced tropical circulation. Quart. J. Roy. Meteor. Soc., 106, 447–462, https://doi.org/10.1002/qj.49710644905.
Ham, Y. G., J. S. Kug, J. Y. Park, and F. F. Jin, 2013: Sea surface temperature in the north tropical Atlantic as a trigger for El Niño/Southern Oscillation events. Nat. Geosci., 6, 112–116, https://doi.org/10.1038/ngeo1686.
Huang, B., and Coauthors, 2017: Extended reconstructed sea surface temperature, version 5 (ERSSTv5): Upgrades, validations, and intercomparisons. J. Climate, 30, 8179–8205, https://doi.org/10.1175/JCLI-D-16-0836.1.
Hurrell, J. W., and Coauthors, 2013: The Community Earth System Model: A framework for collaborative research. Bull. Amer. Meteor. Soc., 94, 1339–1360, https://doi.org/10.1175/BAMS-D-12-00121.1.
Jia, F., W. Cai, B. Gan, L. Wu, and E. Di Lorenzo, 2021: Enhanced North Pacific impact on El Niño/Southern Oscillation under greenhouse warming. Nat. Climate Change, 11, 840–847, https://doi.org/10.1038/s41558-021-01139-x.
Joh, Y., and E. Di Lorenzo, 2019: Interactions between Kuroshio Extension and central tropical Pacific lead to preferred decadal-timescale oscillations in Pacific climate. Sci. Rep., 9, 13558, https://doi.org/10.1038/s41598-019-49927-y.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437–471, https://doi.org/10.1175/1520-0477(1996)077<0437:TNYRP>2.0.CO;2.
Larson, S. M., and B. P. Kirtman, 2014: The Pacific meridional mode as an ENSO precursor and predictor in the North American multimodel ensemble. J. Climate, 27, 7018–7032, https://doi.org/10.1175/JCLI-D-14-00055.1.
Liguori, G., and E. Di Lorenzo, 2018: Meridional modes and increasing Pacific decadal variability under anthropogenic forcing. Geophys. Res. Lett., 45, 983–991, https://doi.org/10.1002/2017GL076548.
Lin, C.-Y., J.-Y. Yu, and H.-H. Hsu, 2015: CMIP5 model simulations of the Pacific meridional mode and its connection to the two types of ENSO. Int. J. Climatol., 35, 2352–2358, https://doi.org/10.1002/joc.4130.
Linkin, M. E., and S. Nigam, 2008: The North Pacific Oscillation–west Pacific teleconnection pattern: Mature-phase structure and winter impacts. J. Climate, 21, 1979–1997, https://doi.org/10.1175/2007JCLI2048.1.
Lu, F., Z. Liu, Y. Liu, S. Zhang, and R. Jacob, 2017: Understanding the control of extratropical atmospheric variability on ENSO using a coupled data assimilation approach. Climate Dyn., 48, 3139–3160, https://doi.org/10.1007/s00382-016-3256-7.
Ma, J., S. P. Xie, and H. Xu, 2017: Contributions of the North Pacific meridional mode to ensemble spread of ENSO prediction. J. Climate, 30, 9167–9181, https://doi.org/10.1175/JCLI-D-17-0182.1.
Martinez-Villalobos, C., and D. J. Vimont, 2016: The role of the mean state in meridional mode structure and growth. J. Climate, 29, 3907–3921, https://doi.org/10.1175/JCLI-D-15-0542.1.
Min, Q., J. Su, and R. Zhang, 2017: Impact of the South and North Pacific meridional modes on the El Niño–Southern Oscillation: Observational analysis and comparison. J. Climate, 30, 1705–1720, https://doi.org/10.1175/JCLI-D-16-0063.1.
Newman, M., and P. D. Sardeshmukh, 1995: A caveat concerning singular value decomposition. J. Climate, 8, 352–360, https://doi.org/10.1175/1520-0442(1995)008<0352:ACCSVD>2.0.CO;2.
Park, J. H., T. Li, S. W. Yeh, and H. Kim, 2019: Effect of recent Atlantic warming in strengthening Atlantic–Pacific teleconnection on interannual timescale via enhanced connection with the Pacific meridional mode. Climate Dyn., 53, 371–387, https://doi.org/10.1007/s00382-018-4591-7.
Park, J. H., M. K. Sung, Y. M. Yang, J. Zhao, S. Il An, and J. S. Kug, 2021: Role of the climatological intertropical convergence zone in the seasonal footprinting mechanism of El Niño–Southern Oscillation. J. Climate, 34, 5243–5256, https://doi.org/10.1175/JCLI-D-20-0809.1.
Rogers, J. C., 1981: The North Pacific Oscillation. J. Climatol., 1, 39–57, https://doi.org/10.1002/joc.3370010106.
Sanchez, S. C., D. J. Amaya, A. J. Miller, S. P. Xie, and C. D. Charles, 2019: The Pacific meridional mode over the last millennium. Climate Dyn., 53, 3547–3560, https://doi.org/10.1007/s00382-019-04740-1.
Sun, M. N., Z. Y. Guan, P. B. Zhang, and S. Y. Cao, 2013: Principal modes of the South Pacific SSTA in June, July and August and their relations to ENSO and SAM. J. Trop. Meteor., 19, 154–161.
Vimont, D. J., 2010: Transient growth of thermodynamically coupled variations in the tropics under an equatorially symmetric mean state. J. Climate, 23, 5771–5789, https://doi.org/10.1175/2010JCLI3532.1.
Vimont, D. J., J. M. Wallace, and D. S. Battisti, 2003: The seasonal footprinting mechanism in the Pacific: Implications for ENSO. J. Climate, 16, 2668–2675, https://doi.org/10.1175/1520-0442(2003)016<2668:TSFMIT>2.0.CO;2.
Vimont, D. J., M. Alexander, and A. Fontaine, 2009: Midlatitude excitation of tropical variability in the Pacific: The role of thermodynamic coupling and seasonality. J. Climate, 22, 518–534, https://doi.org/10.1175/2008JCLI2220.1.
Walker, G. T., and E. W. Bliss, 1932: World weather V. Mem. Roy. Meteor. Soc., 4, 53–84.
Wang, L., J. Y. Yu, and H. Paek, 2017: Enhanced biennial variability in the Pacific due to Atlantic capacitor effect. Nat. Commun., 8 14887, https://doi.org/10.1038/ncomms14887.
Wang, S. Y., M. L’Heureux, and J. H. Yoon, 2013: Are greenhouse gases changing ENSO precursors in the western North Pacific? J. Climate, 26, 6309–6322, https://doi.org/10.1175/JCLI-D-12-00360.1.
Wang, X., M. Chen, C. Wang, S. W. Yeh, and W. Tan, 2019: Evaluation of performance of CMIP5 models in simulating the North Pacific Oscillation and El Niño Modoki. Climate Dyn., 52, 1383–1394, https://doi.org/10.1007/s00382-018-4196-1.
Xie, S., and S. G. H. Philander, 1994: A coupled ocean–atmosphere model of relevance to the ITCZ in the eastern Pacific. Tellus, 46A, 340–350, https://doi.org/10.3402/tellusa.v46i4.15484.
Yu, J.-Y., and S. T. Kim, 2011: Relationships between extratropical sea level pressure variations and the central Pacific and eastern Pacific types of ENSO. J. Climate, 24, 708–720, https://doi.org/10.1175/2010JCLI3688.1.
Yu, J.-Y., H. Y. Kao, T. Lee, and S. T. Kim, 2011: Subsurface ocean temperature indices for central-Pacific and eastern-Pacific types of El Niño and La Niña events. Theor. Appl. Climatol., 103, 337–344, https://doi.org/10.1007/s00704-010-0307-6.
Yu, J.-Y., M.-M. Lu, and S. T. Kim, 2012: A change in the relationship between tropical central Pacific SST variability and the extratropical atmosphere around 1990. Environ. Res. Lett., 7, 034025, https://doi.org/10.1088/1748-9326/7/3/034025.
Yu, J.-Y., P. Kao, H. Paek, H.-H. Hsu, C. Hung, M.-M. Lu, and S.-I. An, 2015: Linking emergence of the central Pacific El Niño to the Atlantic multidecadal oscillation. J. Climate, 28, 651–662, https://doi.org/10.1175/JCLI-D-14-00347.1.
Zhang, H., A. Clement, and P. Di Nezio, 2014a: The South Pacific meridional mode: A mechanism for ENSO-like variability. J. Climate, 27, 769–783, https://doi.org/10.1175/JCLI-D-13-00082.1.
Zhang, H., C. Deser, A. Clement, and R. Tomas, 2014b: Equatorial signatures of the Pacific meridional modes: Dependence on mean climate state. Geophys. Res. Lett., 41, 568–574, https://doi.org/10.1002/2013GL058842.
Zhang, L., P. Chang, and L. Ji, 2009: Linking the Pacific meridional mode to ENSO: Coupled model analysis. J. Climate, 22, 3488–3505, https://doi.org/10.1175/2008JCLI2473.1.
Zhang, Y. C., L. F. Zhang, and Y. Luo, 2010: The coupled mode between the Kuroshio region marine heating anomaly and the North Pacific atmospheric circulation in wintertime. J. Trop. Meteor., 16, 51–58, https://doi.org/10.3969/j.issn.1006-8775.2010.01.007.
Zhong, W., X. T. Zheng, and W. Cai, 2017: A decadal tropical Pacific condition unfavorable to central Pacific El Niño. Geophys. Res. Lett., 44, 7919–7926, https://doi.org/10.1002/2017GL073846.
