Impact of the Atlantic Ocean on the multidecadal fluctuation of El Niño–Southern Oscillation–South Asian monsoon relationship in a coupled general circulation model

Authors

  • Wei Chen,

    1. State Key Laboratory of Numerical Modelling for Atmospheric Sciences and Geophysical Fluid Dynamics and Center for Monsoon System Research, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, China
    Search for more papers by this author
  • Buwen Dong,

    1. National Centre for Atmospheric Science, Department of Meteorology, University of Reading, Reading, UK
    Search for more papers by this author
  • Riyu Lu

    1. State Key Laboratory of Numerical Modelling for Atmospheric Sciences and Geophysical Fluid Dynamics and Center for Monsoon System Research, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, China
    Search for more papers by this author

Abstract

[1] The multidecadal variability of El Niño–Southern Oscillation (ENSO)–South Asian monsoon relationship is elucidated in a 1000 year control simulation of a coupled general circulation model. The results indicate that the Atlantic Multidecadal Oscillation (AMO), resulting from the natural fluctuation of the Atlantic Meridional Overturning Circulation (AMOC), plays an important role in modulating the multidecadal variation of the ENSO-monsoon relationship. The sea surface temperature anomalies associated with the AMO induce not only significant climate impact in the Atlantic but also the coupled feedbacks in the tropical Pacific regions. The remote responses in the Pacific Ocean to a positive phase of the AMO which is resulted from enhanced AMOC in the model simulation and are characterized by statistically significant warming in the North Pacific and in the western tropical Pacific, a relaxation of tropical easterly trades in the central and eastern tropical Pacific, and a deeper thermocline in the eastern tropical Pacific. These changes in mean states lead to a reduction of ENSO variability and therefore a weakening of the ENSO-monsoon relationship. This study suggests a nonlocal mechanism for the low-frequency fluctuation of the ENSO-monsoon relationship, although the AMO explains only a fraction of the ENSO–South Asian monsoon variation on decadal-multidecadal timescale. Given the multidecadal variation of the AMOC and therefore of the AMO exhibit decadal predictability, this study highlights the possibility that a part of the change of climate variability in the Pacific Ocean and its teleconnection may be predictable.

1. Introduction

[2] Interaction between the El Niño–Southern Oscillation (ENSO) and South Asian monsoon is one of the most dominant coupled phenomena in the climate system, a drier than normal South Asian summer monsoon preceding peak El Niño condition [Elliot and Angell, 1987; Webster and Yang, 1992]. This interaction fluctuates on a decadal-multidecadal timescale and it has weakened substantially since the late 1970s [Krishna Kumar et al., 1999]. The mechanism behind this weakened ENSO-monsoon relationship has been a subject of considerable attention because of its impact on South Asian summer monsoon predictability. Various hypotheses have been proposed, including the changes of atmospheric circulation in the North Pacific [Kinter et al., 2002], the variability of Pacific Decadal Oscillation [Krishinan and Sugi, 2003], and the tropical South Atlantic sea surface temperature (SST) anomalies concurring with the ENSO [Kucharski et al., 2007].

[3] Furthermore, some recent studies suggested that the enhanced ENSO variability, which is associated with the change of background mean states, may concur with the intensification of the ENSO-monsoon relationship. Lu et al. [2008] suggested that an enhanced ENSO variability could intensify the ENSO-monsoon relationship in the background of a substantially weakened AMOC. On the other hand, Turner et al. [2005] indicated an enhanced ENSO variability and ENSO-monsoon interaction in a version of a coupled general circulation model (CGCM) that uses equatorial flux adjustments and suggested that this enhancement is due to the improvement of mean states in the Indo-Pacific oceans. Actually, the amplitude of ENSO in observations also varies on decadal-multidecadal timescale, and this variation is related to the changes of the background mean states of the tropical Pacific [e.g., Gu and Philander, 1995, 1997; Wang and An, 2001, 2002]. It is known that the basic states of the ocean thermal structure can influence the ENSO growth and modulate the ENSO variance [Zebiak and Cane, 1987; Kirtman and Schopf, 1998]. However, the cause of the changes in the mean states remains unknown.

[4] Several studies [e.g., Dong and Sutton, 2002; Zhang and Delworth, 2005; Timmermann et al., 2005; Dong et al., 2006; Sutton and Hodson, 2007] have demonstrated the influence of the Atlantic Ocean on the mean states in the tropical Pacific. The observed SSTs in the Atlantic exhibit an oscillation with a period of 60–100 years on the multidecadal timescale, which is referred to as the AMO [Delworth and Mann, 2000; Kerr, 2000], and is thought to be related to multidecadal fluctuation of the AMOC [Delworth and Mann, 2000; Knight et al., 2005]. Coupled modeling studies suggested that a negative phase of the AMO (or a weakened AMOC) induces cold SST anomalies in the North Atlantic and warm anomalies in the South Atlantic [e.g., Delworth et al., 1993; Dong and Sutton, 2002, 2007; Timmermann et al., 2005, 2007; Zhang and Delworth, 2005]. A wide range of regional climate signals have been linked to the AMO, including important changes in Atlantic cyclone activity [e.g., Goldenberg et al., 2001; Wang et al., 2008] and North American and European summer climate [Sutton and Hodson, 2005]. There is further evidence that the variations in Atlantic SST associated with the AMO not only restrict in the Atlantic basin but also may influence the climatology of the Asian summer monsoon [Zhang and Delworth, 2005; Lu et al., 2006; Lu and Dong, 2008], and mean states of the tropical Pacific [Zhang and Delworth, 2005; Dong et al., 2006; Timmermann et al., 2007] and therefore modulate the interannual variability of ENSO [e.g., Dong and Sutton, 2007; Timmermann et al., 2007].

[5] Furthermore, as previously mentioned, a substantially weakened AMOC intensifies the ENSO–South Asian monsoon relationship [Lu et al., 2008]. In these studies, the AMOC is substantially weakened through water-hosing experiments in which an extra freshwater flux is artificially applied in the North Atlantic. Using a coupled model simulation without external forcing, Knight et al. [2005] demonstrated that the multidecadal fluctuation of the Atlantic SSTs associated with the fluctuation of the AMOC is similar to the observed AMO in both the SST pattern and the magnitude. Is there a low-frequency variation of the ENSO-monsoon interaction in this simulation? If so, what factors affect this low-frequency variability? Does the AMO play a role? These questions will be investigated in this study by using the same model simulation as Knight et al. [2005].

[6] The paper is organized as follows: Section 2 briefly describes the model experiment and observational data used. Section 3 presents the validation of the simulated mean climate and ENSO-monsoon relationship. Section 4 investigates the multidecadal fluctuation of ENSO variability and the ENSO-monsoon interaction. Section 5 elucidates the processes that are responsible for the impact of the Atlantic Ocean on the multidecadal modulation of ENSO variability and the ENSO-monsoon relationship. Section 6 provides concluding remarks.

2. Model and Experiment

[7] The model we used is a version of the UK Hadley Centre CGCM known as HadCM3 [Gordon et al., 2000]. The atmospheric component of this model is run with a horizontal latitude-longitude grid spacing of 2.5° × 3.75° and 19 vertical levels using a hybrid vertical coordinate. A detailed description of the atmospheric model formulation and its performance in a simulation forced with observed SSTs is described in the work of Pope et al. [2000]. The oceanic component of the model is a 20 level version of the model used in the work of Cox et al. [1999] on a 1.25° × 1.25° latitude-longitude grid. The vertical levels are distributed to provide enhanced resolution near the ocean surface. The two components are coupled once a day. Heat and water fluxes are conserved exactly in the transfer between their different grids. The coupled model uses preindustrial atmospheric trace gas concentrations, and incoming solar radiation provides the only external forcing; the model has been run without the use of flux adjustments for more than 1000 years without appreciable drift in the model's climate [Gordon et al., 2000]. In this study, we analyze the results in 1000 year integration of the model.

[8] The observed data used in this study to evaluate model performance include the monthly precipitation data from 1979 to 2005 derived by the Global Precipitation Climatology Project [Adler et al., 2003] and the monthly wind data from European Centre for Medium-Range Weather Forecasts reanalysis data set (ERA-40), covering the period from 1957 to 2002. In addition, the All-Indian Rainfall (AIR) data set [Parthasarathy et al., 1994] is used to provide one measure of South Asian summer monsoon variability. This data set is based on rain-gauge data, and the detailed information of rain-gauge stations can be found through the website: http://www.tropmet.res.in.

3. Mean Climate and ENSO-South Asian Summer Monsoon Relationship in the Model Simulation

[9] The basic states play an important role in the ENSO-monsoon relationship [e.g., Turner et al., 2005]. Thus, we evaluate the model's ability to reproduce climatological features in summer in this section. Figure 1 shows the climatological mean 850 hPa wind and precipitation for observations and 1000 year simulation in summer (June, July, and August (JJA)). Compared to observations, the model simulates well the main features of the anticyclonic circulation over the Indian Ocean, which is characterized by the cross-equatorial Somali jet in the western Indian Ocean and the southwesterly mean Indian monsoon flow over the Indian subcontinent. The basic features of precipitation in the Indian monsoon region are reasonably reproduced, including the two precipitation centers over the western Indian Ocean and the Bay of Bengal. However, the precipitation over the Maritime Continent is stronger in the simulation. In the Pacific Basin, the model reconstructs easterly trades and the subtropical anticyclone over the western North Pacific. Also, the Intertropical Convergence Zone (ITCZ) north of the equator is well simulated as well as the rain band over East Asia and the western Pacific.

Figure 1.

Climatological distribution of the 850 hPa winds (vector) and precipitation (shading) in JJA: (a) observations and (b) model simulation. Unit is in mm d−1 for precipitation and m s−1 for winds.

[10] In this study, we take the area-averaged SST anomalies in the Niño 3 region (5°S–5°N, 150°W–90°W) as an ENSO index [AchutaRao and Sperber, 2002; Collins et al., 2001]. Figure 2a shows the lead-lag correlation coefficients between JJA mean AIR and seasonal mean Niño 3 index. The significant inverse correlation between the AIR and the Niño 3 index in the model is consistent with that based on observations. This inverse correlation means that the warm SST anomalies in the eastern tropical Pacific are associated with weak monsoon rainfall in the Indian peninsula. Although the time evolution of the ENSO-monsoon relationship in the simulation differs somewhat from observations and the model overestimates the relationship when ENSO leads and underestimates the relationship when monsoon leads [e.g., Turner et al., 2005], the negative simultaneous correlation between ENSO and monsoon in JJA is well captured by the model. Another measure of South Asian monsoon variability is a dynamical monsoon index (DMI), defined as the vertical shear of the horizontal mean zonal wind [u(850 hPa) – u(200 hPa)] in the region (40°E–110°E, 5°N–20°N), which was first suggested by Webster and Yang [1992]. Figure 2b shows the lead-lag correlation coefficients between JJA mean DMI and seasonal mean Niño 3 index. The model also reproduces the dynamically weak Indian monsoon associated with El Niño.

Figure 2.

(a) Lead-lag correlation coefficients between JJA mean AIR and seasonal mean Niño 3 index in the 1000 year simulation (solid line) and in observations (dashed line). (b) Same as Figure 2a but for JJA mean DMI instead of AIR.

[11] To investigate the spatial pattern of climate variables associated with ENSO in the 1000 year simulation, JJA mean SST, precipitation, and 850 hPa wind are regressed onto the simultaneous Niño 3 index; the results are shown in Figure 3. The strong warm SST anomalies occur in the central and eastern tropical Pacific, despite the fact that the SST anomalies associated with El Niño extend too far westward into the warm pool [Collins et al., 2001]. There are also warmer, although weak, SSTs in the Indian Ocean influenced by the warm events in the tropical Pacific (Figure 3a). Meanwhile, the strong westerly anomalies in the equatorial Pacific correspond to the warm SST anomalies there. The easterly anomalies in the tropical Indian Ocean weaken the climatological mean flow, and the anticyclone anomaly appears over the Bay of Bengal (Figure 3b). These circulation anomalies give arise to the small but significant negative precipitation anomalies over the Indian monsoon region (Figure 3c). Compared with observations (not shown), the spatial pattern of the inversed ENSO-monsoon relationship is reasonably captured by this model.

Figure 3.

(a) Regression of SST, (b) 850 hPa wind, and (c) precipitation onto the Niño 3 index. Shading indicates the regions where the anomalies are significant at the 95% confidence level using the F-test.

[12] The interannual variability of ENSO and the South Asian summer monsoon is evaluated by calculating standard deviations of Niño 3, AIR, and DMI indices (Table 1). The intensity of ENSO and monsoon measured by DMI in the control experiment is similar to observations, but the standard deviation of the AIR index is underestimated in the model, indicating the model's deficiency in simulating the regional precipitation variability even if it can simulate the correct magnitude of the large-scale circulation variability.

Table 1. Standard Deviation of the JJA Mean Niño 3, AIR, and DMI in Observations and 1000 Year Control Simulation
 Niño 3AIRDMI
Observation0.732.421.38
CNTL0.750.871.34

4. Multidecadal Fluctuations of ENSO Variability and ENSO-Monsoon Interaction

[13] To investigate the multidecadal modulation of the ENSO-monsoon relationship, a 41 year running mean is used to depict the low-frequency oscillation, and a 41 year sliding window is used to show the low-frequency fluctuation of correlation and interannual variability. It is worth pointing out that similar results and conclusions are obtained by using the 21 year, 31 year, and 51 year running mean or sliding window.

[14] Figure 4a shows the correlation between JJA mean DMI and Niño 3 index with a 41 year sliding window. The correlation exhibits a clear decadal-multidecadal fluctuation ranging from –0.63 to 0.01. This running correlation is significant over many epochs that are about half of the whole 1000 year period. The power spectrum of the correlation time series shows two significant peaks: around 60 years and 70–120 years (Figure 4b), indicating that the multidecadal fluctuation of the ENSO-monsoon relationship has dominant periods. The correlation coefficient for the whole 1000 years is −0.31, significant at the 99.9% level using the t-test.

Figure 4.

(a) 41 year sliding correlation coefficients between JJA mean DMI and Niño 3 index (thick line). The dashed straight line represents the 95% confidence level of the running correlation coefficients. Also shown is the standard deviation of the annual mean Niño 3 index with a 41 year sliding window (thin line). The correlation coefficient between the two series is −0.37, significant at the 99% confidence level with the number of independent samples being 55 using the t-test. (b) The power spectrum of correlation coefficients between JJA mean DMI and Niño 3 index with a 41 year sliding window (the thick line in Figure 4a). (c) Same as Figure 4b but for the standard deviation of the Niño 3 index with a 41 year sliding window (the thin line in Figure 4a).

[15] The intensity of ENSO varies on the multidecadal timescale as well. The standard deviation of annual mean Niño 3 index with a 41 year sliding window is also plotted in Figure 4a, and its power spectrum also indicates two significant peaks at the multidecadal timescale (Figure 4c), similar to the power peak of the ENSO-monsoon correlation. As Figure 4a indicates, the two time series are in principle reversely correlated on the multidecadal timescale. The correlation coefficient between the two is −0.37, significant at the 99% level with the number of independent samples being 55 using the t-test, indicating that the fluctuation of the ENSO–South Asian monsoon relationship concurs with the ENSO intensity on the multidecadal timescale. The negative correlation between them indicates that strong ENSO amplitude enhances the ENSO-monsoon teleconnection, which is consistent with previous studies [Turner et al., 2005; Lu et al., 2008].

[16] The power spectrum shown in Figures 4b and 4c indicates a preferred multidecadal fluctuation of 60–100 years in both the ENSO-monsoon relationship and ENSO variability. This timescale is close to the multidecadal timescale of Atlantic SST variability associated with the AMO that resulted from naturally occurring AMOC fluctuation identified by Knight et al. [2005]. Are the AMO, the fluctuation of ENSO variability, and therefore ENSO-monsoon interaction related to each other?

[17] The time series of the AMO index (defined as the averaged SST anomalies in the region 75°W–7.5°W, 0°–60°N) in the 1000 year simulation is shown in Figure 5a. Also shown is the time series of AMOC index in the 1000 year simulation, which is consistent with the multidecadal fluctuation of AMO (Figure 5a). The correlation coefficient between AMO and AMOC is 0.8, which indicates the modulation of naturally occurring AMOC to the multidecadal variability of AMO. The power spectrum of the AMO index exhibits significant periods at 60 years and 80–120 years (Figure 5c). The multidecadal fluctuation of the AMO index is basically consistent with the ENSO-monsoon relationship; the correlation coefficient between them is 0.26, significant at the 99% level with the number of independent samples being 55 using the t-test (Figure 5b). The positive correlation suggests that warm phases of the AMO concur with weakened ENSO-monsoon relationship. This result is consistent with the work of Lu et al. [2008], who indicated that an artificially substantially weakened AMOC, which may be considered as the extreme case of cold phase of the AMO, intensifies the ENSO-monsoon relationship. The results indicate that the multidecadal fluctuation of the ENSO-monsoon relationship might be modulated by the variation of the AMO resulting from the naturally occurring AMOC fluctuation in the HadCM3 control simulation. In addition, the correlation coefficient between AMO and ENSO variability is −0.21. The negative correlation indicates that the weak (strong) ENSO variability coincides with the positive (negative) phase of the AMO, consistent with previous results [Dong et al., 2006; Dong and Sutton, 2007].

Figure 5.

(a) The thick (thin) line is the annual mean AMOC (AMO) index with a 41 year running mean smoothing. The AMOC index (Sv) is defined as the averaged meridional overturning circulation over the latitude band (30°N–60°N) at the depth of 996 m. The AMO index (°C) is defined as the average SST anomalies in the region 75°W–7.5°W, 0°N–60°N. The correlation coefficient between the two series is 0.8, significant at the 99% confidence level with the number of independent samples being 40 using the t-test. (b) The thick line is the same as that shown in Figure 4a, and the thin line is the AMO index (same as shown in Figure 5a). The correlation coefficient between the two series is 0.26, significant at the 95% confidence level with the number of independent samples being 55 using the t-test. (c) The power spectrum of the AMO index (the thin line in Figure 5a or 5b).

[18] The spatial pattern of the inverse relationship between ENSO and South Asian monsoon in positive (warm) and negative (cold) phase of the AMO is investigated. A period in which the AMO index is greater (less) than 0.8 (−0.8) standard deviation from the time mean has been defined as a positive (negative) phase of the AMO and is combined into two records for both AMO positive and negative phases, respectively. We perform the composites of precipitation and 850 hPa wind anomalies associated with the strong ENSO events for positive and negative records, and the results are illustrated in Figure 6. The major features of precipitation and lower-tropospheric circulation patterns in positive and negative phases of the AMO are similar to those shown in Figure 3. However, in the negative phase of the AMO, the strong ENSO-monsoon relationship is indicated by the significant large negative precipitation anomalies over the Indian subcontinent (Figure 6b) than that in the positive phase of the AMO (Figure 6a). This large decrease in precipitation over India associated with El Niño in negative phase of the AMO is related to larger zonal wind divergence on the west side of the Indian peninsula (Figure 6d) than that for the positive phase of the AMO (Figure 6c). In addition, the precipitation anomalies in the western tropical Pacific for the negative phase of AMO is also stronger than that for the positive phase of AMO. These strong tropical responses indicate the stronger ENSO for the negative phase of AMO.

Figure 6.

Composites of precipitation and 850 hPa wind anomalies related to the strong ENSO for (a, c) positive and (b, d) negative phases of the AMO. AMO positive (negative) phase is defined as the period that the AMO index is greater (less) than 0.8 (−0.8) standard deviation from the time mean. The ENSO is defined as the year that the Niño 3 index is greater than 1.5 standard deviation. Shading indicates the regions where anomalies are significant at the 95% confidence level using the t-test.

5. Mechanisms of Multidecadal Fluctuation of ENSO-Monsoon Interaction

[19] The results in previous section indicate multidecadal fluctuations of the ENSO-monsoon relationship and ENSO variability with a preferred timescale of 70–120 years. What processes contribute to this multidecadal fluctuation of ENSO variability and the ENSO-monsoon relationship? Shown in Figure 7 are SST, precipitation, lower-tropospheric wind, and stream-function changes associated with the AMO. In the Atlantic, associated with the positive phase of the AMO are warmer SSTs in the North Atlantic and colder SSTs in the South Atlantic. This interhemispheric asymmetry leads to a strong cross-equatorial gradient of SST anomalies and a northward shift of the ITCZ. The latter induces an increase of precipitation north of the equator and a decrease of precipitation south of the equator. Corresponding to the SST and precipitation anomalies are the stronger trades in the tropical South Atlantic and weaker trades in the tropical North Atlantic [e.g., Knight et al., 2006]. A pair of cyclones is formed in the northeastern and southeastern tropical Pacific, consistent with a Gill-type response to the heating anomalies slightly north of the tropical Atlantic [Heckley and Gill, 1984; Lee et al., 2009; Wang et al., 2010]. As indicated in Figure 7, SST and circulation anomalies are not confined in the Atlantic sector but extend to the Pacific. The remote SST anomalies are characterized by warm SST anomalies in the North and northeast Pacific and the western and central tropical Pacific [e.g., Knight et al., 2005]. This is associated with the weakened easterly trades in the tropical Pacific, similar to the remote responses demonstrated in an atmospheric model with prescribed SST in the Atlantic [Sutton and Hodson, 2007] and the coupled feedbacks that further shape the remote responses [Dong et al., 2006]. The strong anomalous westerly in the eastern equatorial Pacific is due to a Gill-type response excited by the anomalous diabatic heating in the tropical Atlantic [Dong et al., 2006]. The anomalous westerly weakens the easterly trades and reduces upwelling on the eastern side of the tropical Pacific.

Figure 7.

Regression of (a) SST, (b) precipitation, (c) 850 hPa wind, and (d) 850 hPa stream function onto the 41 year running mean AMO index. Shading indicates the regions where anomalies are significant at the 95% confidence level using the F-test.

[20] Besides the changes of SST and lower-tropospheric circulation in the tropical Pacific, the thermocline structure in the eastern Pacific basin is also changed. The depth of 20°C isothermal line (D20) in the ocean is used to represent the thermocline and regressed onto the AMO index (Figure 8a). It shows that the positive phase of the AMO is associated with deeper thermocline in the eastern equatorial Pacific. The thermocline structure changes associated with the multidecadal fluctuation in the ENSO-monsoon relationship is given in Figure 8b. The positive value in Figure 8b indicates that the shallow thermocline in the central and eastern Pacific is associated with an enhanced ENSO-monsoon relationship because ENSO and monsoon are negatively correlated with the strong negative correlation indicating a strong relationship. The similarity of thermocline anomalies in Figures 8a and 8b suggests that the positive phase of the AMO and weakened ENSO-monsoon relationship are both related to the deeper thermocline in the central and eastern equatorial Pacific. In addition, the regression pattern of D20 onto standard deviation of Niño 3 index (Figure 8c) indicates that enhanced ENSO variability is associated with shallow thermocline in the tropical Pacific. These changes in the equatorial Pacific thermocline associated the strong ENSO variability are consistent with the work of Zebiak and Cane [1987], indicating the shallower equatorial thermocline is associated with the stronger vertical temperature gradient (strong thermocline feedback) in the tropical Pacific and stronger influence on SSTs by oceanic Kelvin wave. The more active Bjerknes feedback leads to stronger coupled instability for ENSO to grow and therefore stronger ENSO variance. The present results suggest that the AMO plays a role in the multidecadal fluctuations of ENSO variability and ENSO-monsoon interaction. Warm AMO phase is associated with deeper thermocline in the central and eastern tropical Pacific. The deeper equatorial thermocline weakens the coupled instability and reduces ENSO variability, which is related to the weak ENSO–South Asian monsoon relationship.

Figure 8.

Regression of D20 (a) onto the 41 year running mean AMO index, (b) onto correlation coefficients between DMI and Niño 3 with a 41 year sliding window, and (c) onto standard deviation of Niño 3 index with a 41 year sliding window. The D20 represents the depth of 20°C isothermal line in the ocean. Shading indicates the regions where anomalies are significant at the 95% confidence level using the F-test.

[21] We have shown that the ENSO-monsoon relationship concurs with ENSO variability on the multidecadal timescale in the previous section. Why are they strongly correlated when ENSO variability is strong? The composites of precipitation, 850 hPa wind, and SST anomalies for the strong and weak ENSO are shown in Figure 9. The precipitation and wind anomalies associated with the strong ENSO are stronger than those associated with the weak ENSO, and the magnitudes of both precipitation and wind anomalies are not linearly scaled with the amplitude of ENSO. These results indicate the nonlinearity of atmospheric responses to ENSO amplitude and suggest the strong ENSO is associated with large changes of monsoon. The fact that the correlation coefficient between Niño 3 index and DMI index is −0.51 for strong ENSO while it is −0.23 for the weak ENSO and a scatter diagram of Niño 3 index and DMI index (not shown) further demonstrate the nonlinearity of the ENSO-monsoon relationship with respect to the amplitude of ENSO.

Figure 9.

Composites of precipitation, 850 hPa wind, and SST anomalies for (a, c, e) strong and (b, d, f) weak ENSO. Strong (weak) ENSO is defined as the year that the Niño 3 index is greater than 1.2 (less than 1.2 but greater than 0.5) standard deviation from the time mean. Shading indicates the regions where anomalies are significant at the 99% confidence level using the t-test.

6. Conclusion and Remarks

[22] In this study, we have investigated the multidecadal variability of ENSO–South Asian monsoon relationship in a long-term control simulation by a CGCM and shown that the AMO (the low-frequency fluctuation of SSTs in the Atlantic Ocean) might play an important role. The SST anomalies associated with the AMO can influence the basic states not only in the Atlantic, but also in the Pacific. The remote response of thermocline in the tropical Pacific to the AMO bears a similarity to the change of thermocline depth associated with the low frequency of ENSO variability. Therefore, the AMO can affect the ENSO-monsoon relationship through modulating the multidecadal fluctuation of ENSO intensity. The teleconnective processes from AMO to the ENSO-monsoon relationship is shown as a schematic diagram in Figure 10. The key processes involved in AMO modulating the ENSO-monsoon relationship can be summarized as follows:

Figure 10.

The schematic diagram of the major elements of the mechanism by which the AMO modulates the ENSO–South Asian monsoon relationship in HadCM3. Dark (light) areas in the Atlantic are regions of positive (negative) SST and precipitation anomalies. The curved arrow means the anomalous cyclonic circulation. The grey arrow is the easterly trades and the dashed arrow is the westerly anomalies. The dash line is the climatological thermocline and the solid line is the thermocline related to the positive phase of AMO.

[23] 1. The warm AMO in the Atlantic is associated with an anomalous cross-equatorial SST gradient with positive SST anomalies in the North Atlantic and negative SST anomalies in the South Atlantic. The impact of the AMO is not confined in the Atlantic Basin, but extends to the Pacific. The remote SST anomalies are characterized by warm SST anomalies in the north and western tropical Pacific. This is associated with the weakened easterly trades in the central and eastern tropical Pacific.

[24] 2. The anomalous westerly winds on eastern side of the tropical Pacific reduce the upwelling that leads to a deep thermocline in the central and eastern tropical Pacific. The deep equatorial thermocline weakens the coupled instability and reduces ENSO variability.

[25] 3. The weakened ENSO variability leads to the weakened ENSO–South Asian monsoon relationship. The weak ENSO is associated with small monsoon changes through the nonlinear atmospheric response with respect to the sign and magnitude of SST anomalies associated with ENSO.

[26] 4. There are clear contrasts of both precipitation and lower-tropospheric circulation anomalies associated with ENSO for AMO positive phase against those for AMO negative phase. The large changes of precipitation over the Indian subcontinent associated with ENSO during AMO negative phases imply more intensive drought and flooding condition associated with weakening AMOC.

[27] This work suggests that the fluctuation of the AMO can modulate the ENSO-monsoon relationship on the decadal-multidecadal timescale, which provides a nonlocal mechanism to understand the ENSO-monsoon relationship. The modulation of the AMO on the ENSO-monsoon teleconnection is also seen in observations. Figure 11 indicates that the strong ENSO-monsoon relationship (i.e., negative anomaly of correlation coefficients) is related to the negative phase of the AMO during the most periods. The correlation coefficient between the two time series shown in Figure 11 is 0.44, being significant at the 95% confidence level by the t-test. Therefore, the observational evidence supports the major conclusion drawn by using model results in this study. In addition, the correlation between the standard deviation of the Niño 3 index and the correlation between the AIR and the Niño 3 index with a 21 year sliding window is −0.33 in the past 134 years. This result indicates that the enhanced ENSO variability intensifies the strong ENSO-monsoon relationship. On the other hand, in recent decades, the enhanced ENSO variability coincides with the weakened ENSO-monsoon relationship [e.g., Krishna Kumar et al., 1999].Thus, the weakened ENSO-monsoon teleconnection in recent decades in observations might be caused by other factors such as the decadal change in SST pattern associated with ENSO [e.g., Krishna Kumar et al., 2006] or the decadal fluctuation of interannual SST anomalies in south equatorial Atlantic concurring with ENSO [e.g., Kucharski et al., 2007] rather than ENSO intensity.

Figure 11.

The normalized correlation coefficients between JJA mean AIR and Niño 3 index with a 21 year sliding window in observations (thick line). Also shown is the normalized 21 year running mean AMO index based on observations (thin line). The AMO index is defined as the average SST anomalies in the region 75°W–7.5°W, 0°N–60°N. The correlation coefficient between the two series is 0.44, significant at the 95% confidence using t-test.

[28] The results presented in this study are obtained by using one model. It is likely that remote responses to the AMO associated with fluctuation of the AMOC may vary, to some extent, with the model [e.g., Van Oldenborgh et al., 2009]. Similar analysis of other models would be helpful to test whether the findings identified in this paper are robust features of the remote response to the AMO fluctuation associated with the AMOC.

Acknowledgments

[29] We thank anonymous reviewers for their constructive comments. This article was supported by the National Natural Science Foundation of China (grants 40821092 and 40725016). Financial support by the EU FP6 Integrated Project ENSEMBLES (contract 505539) is gratefully acknowledged. B.D. is supported by the UK National Centre for Atmospheric Sciences-Climate.

Ancillary

Advertisement