El Niño-Southern Oscillation (ENSO) is the strongest interannual variability mode in global climate system, which can affect not only temperature and precipitation of tropical Pacific, but also remote climate via large-scale teleconnective processes (Alexander et al., 2002). The response of ENSO to external forcing has long been a hot topic for climate research. A great number of studies have paid attention to the behavior of ENSO in warm climate (Timmermann et al., 1999; Collins, 2000; Merryfield, 2006; Yeh et al., 2009; Collins et al., 2010). Early studies suggested that there seemed no significant changes for ENSO with increasing greenhouse gas (Knutson et al., 1997), however, Timmermann et al. (1999) and Collins (2000) showed that ENSO might be amplified due to the thermocline adjustment, and frequency of ENSO would be increased because of enlarged meridional temperature gradient (MTG) between equatorial and off-equatorial region (Collins, 2000). On the basis of the analysis of the Intergovernmental Panel on Climate Change (IPCC) model results, Merryfield (2006) and Collins et al. (2010) indicated that the IPCC models did not display a consistent change of ENSO under double CO2 forcing, and the variance change in each model was broadly significant despite the large extent of internal variability, which suggested that the response of ENSO to global warming is complicated and influenced by many factors (Collins et al., 2010).
Recently, more and more water-hosing studies became to address the impact of freshening over the North Atlantic on ENSO variability, which aimed to investigate the response of ENSO to a potential collapse of the Atlantic meridional overturning circulation (AMOC). Using an intermediate-complexity coupled model and an intermediate ENSO model, Timmermann et al. (2005) suggested that ENSO would be weakened due to the deepening of tropical thermocline induced by a shutdown of the AMOC, indicating the importance of ocean dynamics. However, based on the fully-coupled climate model results, Dong and Sutton (2007) and Timmermann et al. (2007) found that ENSO could be enhanced by the competitive interaction between ENSO and equatorial annual cycle, showing the critical role of the ‘frequency entrainment effect’ (i.e. a weakened annual cycle tends to change the climate variability from a forced annual cycle to a free ENSO oscillation, and therefore amplifies ENSO) (Liu, 2002).
In addition to the North Atlantic water-hosing experiments, the Southern Ocean (SO) freshening cases have also been conducted in the past few years to explore the impact of Antarctic ice-melting induced by elevation of greenhouse gas on global climate system (Stouffer et al., 2007; Swingedouw et al., 2008; Ma and Wu, 2011). However, there are still lack of discussions on the modulation of ENSO by freshwater perturbation over the SO. This is the major focus of our present article.
Model and sensitivity experiment
We use the Fast Ocean–atmosphere Model (FOAM) version 1.5, a fully-coupled global model developed jointly at University of Wisconsin and the Argonne National Laboratory. The atmospheric model is a parallel version of the NCAR Community Climate Model version 2 (CCM2) but the atmospheric physics is replaced with those of CCM3. The ocean model was developed following the Geophysical Fluid Dynamics Laboratory (GFDL) Modular Ocean Model (MOM). The FOAM used here has an atmospheric resolution of R15 with 18 vertical levels, and an oceanic resolution of 1.4° latitude × 2.8° longitude with 32 vertical levels. The coupled model has a thermodynamic sea ice component. Without flux adjustment, the fully-coupled control simulation has been integrated for over 2000 years, without apparent climate shift. FOAM captures major features of the observed climatology (Jacob, 1997), and also produces reasonable climate variability, such as ENSO (Liu et al., 2000) and decadal climate variability (Wu and Liu, 2003; Wu et al., 2003; Wu and Liu, 2005).
The SO freshening experiment is conducted as 1.0 Sv (Sverdrup; 1 Sv = 1.0 × 106 m3 s−1) freshwater flux is added uniformly to the south of 60°S as a virtual salt flux, and fixed for 400 model years. This experiment is named as the SOWH (Southern Ocean Water-Hosing) experiment. A parallel experiment starting from the same initial condition is also conducted for 400 years with no freshwater anomaly added and this control run is named as CTRL. Differences between these two experiments are taken as modeling responses.
Changes of the mean climate
Freshwater forcing over the SO triggers a series of ocean–atmosphere coupled responses, which has been described in detail by Ma and Wu (2011). In particular, La Niña-like sea surface temperature (SST) anomalies prevail in tropical Pacific (Figure 1(a)), which leads to shoaling and deepening of thermocline in the eastern and western Pacific respectively, and thus an enhancement of the zonal tilt of tropical thermocline (Figure 1(b)). Besides, La Niña-like SST anomalies lead to acceleration of surface trade winds as well as strengthened Walker Circulation (not shown).
Changes of ENSO variability
Numerous studies have shown that changes of mean state, which play an important role in determining the stability of tropical ocean–atmosphere coupled system (Fedorov and Philander, 2001), can significantly influence the amplitude and frequency of ENSO (Guilyardi, 2006; Collins et al., 2010). To qualitatively access the change of ENSO, we first calculate the change of interannual standard deviation (ISTD) in tropical Pacific. As described in Figure 1(c), ISTD in tropical Pacific is broadly enlarged with the maximum value rising up to more than 0.05 °C, which is statistically significant despite the large extent of internal variability, implying that interannual variability is more active in the SOWH experiment (Wittenberg, 2009; Stevenson et al., 2012).
To further investigate the change of ENSO in response to freshwater forcing in the SO, we conduct an Empirical Orthogonal Function (EOF) analysis of tropical SST in the last 100 years, during which the evolution of SST is basically stable without remarkable linear trend. The leading EOF mode in the CTRL exhibits distinct anomalies with the maximum of 0.3 °C in the center-eastern tropical Pacific (Figure 2(a)). This mode broadly captures major characteristic of ENSO, and is similar to other climate model results with anomalies extending too westward to the warm pool region (Kug et al., 2010). In the SOWH experiment, ENSO pattern is still robust in the leading mode (Figure 2(b)). Moreover, the amplitude of leading principal component (PC1) appears to be much larger than that in the CTRL (Figure 2(c) and (d)), suggesting an enhancement of ENSO, which is consistent with the change of ISTD. The power spectrum of PC1 in the CTRL shows a significant peak at 2–3 years and a weaker peak around 4–5 years (Figure 2(e)). In the sensitivity experiment, the peak at 2–3 years is strengthened, and more importantly, a new peak around 7–8 years becomes prominent, making the spectrum more complicated. The result of spectrum analysis shows that freshwater forcing over the SO can intensify ENSO and tend to shift ENSO frequency toward a lower band.
The amplified ENSO can be related to the change of equatorial thermocline. Enlarged zonal tilt of thermocline promotes more cold water from subsurface to surface ocean in the eastern Pacific, which tends to weaken vertical stratification and thus decreases the stability of tropical ocean–atmosphere coupled system, and eventually intensifies ENSO variability (Fedorov and Philander, 2001; Collins et al., 2010). Besides, the zonal-mean equatorial thermocline depth rises up by 2–5 m (Figure 3(a)), which may also strengthen ENSO according to the ‘dynamic thermostat’ theory (Sun and Liu, 1996). However, it should be noted that the mean thermal stratification is decreased in the SOWH experiment (Figure 1(b)), which partly offsets the effect of thermocline shoaling, and finally leads to a small net change of ENSO amplitude (Figures 1(c) and 2(d)).
As ENSO can be considered as the outcome of large-scale tropical ocean–atmosphere interaction, the change of ENSO does reflect variation of tropical air–sea coupling strength (Timmermann et al., 1999; Fedorov and Philander, 2001). To reveal the change of air–sea coupling strength induced by modulation of background climatology, we calculate atmospheric and oceanic sensitivities in the CTRL and SOWH experiment, respectively (Figure 3(b) and (c)), following the method proposed by Timmermann et al. (1999). Atmospheric sensitivity substantially depicts the response of wind stress to SST perturbation, while oceanic sensitivity reflects the response of SST to wind stress alteration (a detailed explanation of the methodology for calculating atmospheric and oceanic sensitivity is provided in the caption of Figure 3). It can be explicitly seen that oceanic sensitivity is significantly increased in the SOWH experiment, while atmospheric sensitivity is reduced, suggesting a primary effect of the shallower thermocline. The growth of oceanic sensitivity indicates that ocean plays a more active role in air–sea coupled system after freshwater-hosing in the SO.
As for the modulation of ENSO frequency, previous studies have proposed that change of equatorial wave speed and upper-ocean MTG between equatorial and off-equatorial region can both influence the transition of ENSO frequency (Collins, 2000; Merryfield, 2006; Deng et al., 2010). The former hypothesis emphasizes the impact of equatorial Kelvin wave based on the delay oscillator theory (Suarez and Schopf, 1988), while the latter argument highlights the efficiency of recharge and discharge processes according to the recharge-discharge oscillator theory (Jin, 1997). Recently, Deng et al. (2010) suggested that in the IPCC models, the change of upper-ocean MTG may lead to shift of ENSO frequency under greenhouse warming scenario.
In the opinion of the recharge-discharge oscillator theory, ENSO cycle can be sustained by continuous recharge (warm water in the upper ocean flows from off-equatorial region toward equator) and discharge (equatorial warm water in the upper ocean flows toward either side of equator) processes. In such sense, when the MTG is enlarged/reduced, there will be more/less heat transfer between equatorial and off-equatorial region in unit time, and the efficiency of recharge-discharge processes will be increased/decreased. It can be seen that the upper-ocean MTG in both CTRL and SOWH experiment is basically negative, although in the eastern Pacific (east of 100°W) its sign has been changed (Figure 4). Moreover, MTG in the SOWH experiment generally lies above that in the CTRL, implying the decrease of MTG in the SO-freshening case. We further examine the zonal-mean upper-ocean MTG in tropical Pacific, and find zonal-mean gradient has reduced by 27% in the SOWH case. Hence, we can conclude that the MTG does decrease after freshwater perturbation, which contributes to vary the frequency of ENSO.
Summary and discussion
The impact of freshening over the SO on ENSO is investigated in a fully-coupled model. The modeling results indicate that both the mean climate of tropic Pacific and ENSO are significantly influenced by freshwater forcing over the SO. The change of mean climate in tropical Pacific is characterized by La Niña-like SST anomalies and an enhancement of zonal tilt of equat-orial thermocline. It is found that freshwater perturbation in the SO could lead to an amplification of ENSO and a shift of ENSO frequency to a lower band. The intensification of ENSO variability can be attributed to the strengthened zonal tilt of equatorial thermocline and an overall lift of thermocline especially in the center-eastern Pacific. The shift of ENSO frequency is primarily due to the weakened upper-ocean MTG between equatorial and off-equatorial region, which reduces efficiency of the recharge-discharge processes, and eventually slows the frequency of ENSO.
One important factor that may affect ENSO is the annual cycle. For the North Atlantic water-hosing studies, the interaction between seasonal cycle and ENSO appears to be responsible for the enhancement of ENSO (Dong and Sutton, 2007; Timmermann et al., 2007). However, in our SOWH experiment, there seems little change of seasonal cycle (not shown), so the interplay between ENSO and seasonal cycle cannot explain the variation of ENSO behavior, and thermocline feedback is more useful. Therefore, which factor determines the change of ENSO is contingent and strongly depends on the characteristic of external forcing.
The strengthened ENSO may indicate a positive feedback induced by freshwater forcing (Reason, 1992; Zhang and Busalacchi, 2009). Zhang and Busalacchi (2009) showed that a non-local response of freshwater flux to the SST anomalies associated with ENSO could in turn intensify ENSO variability by influencing oceanic stratification and upper-ocean stability. In the SOWH case, freshwater forcing over the tropical Pacific originated from the SO will intensify oceanic stratification, so the mechanism above may work in FOAM. Although thermocline feedback dominates the change of ENSO amplitude, freshwater flux-induced ENSO variability may also be important.
In our study, though modulation of MTG contributes to the shift of ENSO frequency, a change of equatorial wave speed can also influence the frequency of ENSO. Merryfield (2006) documented that in the IPCC models, acceleration of equatorial Kelvin wave by enhanced stratification under global warming helped to speed up the ENSO cycle. In order to get a comprehensive understanding of the transition of ENSO frequency, it is necessary to investigate the change of equatorial wave, and further discuss the relative roles of equatorial wave and MTG in controlling the change of ENSO period.
This work is jointly supported by the Chinese National Science Foundation Outstanding Youth Program (NSFC 40788002) and NSFC Creative Research Group Project (NSFC 40921004). Discussions with Drs Axel Timmermann, Wenju Cai and De-zheng Sun are greatly appreciated. All calculations were carried on SGI supercomputer at Ocean University of China.