Modelling the influence of North Atlantic multidecadal warmth on the Indian summer rainfall



[1] Ensemble experiments with an atmospheric general circulation model reveal that a positive (warm) ocean phase of the Atlantic Multidecadal Oscillation (AMO) increases Indian summer rainfall. The intensification is driven by extratropical North Atlantic warmth, with some cancellation associated with monsoon weakening in response to tropical North Atlantic warmth. Mechanistically, warm extratropical North Atlantic SSTs increase local rainfall, inducing an arching extratropical wavetrain response. The latter leads to intensified northern subsidence of monsoon mean meridional streamflow as well as widespread low surface pressure over North Africa, the Middle East and the western Indian Ocean contributing to a strengthened Indian monsoon trough and increased monsoon rainfall. Warm tropical North Atlantic SSTs primarily increase local tropical Atlantic rainfall that induces a tropically-confined response consisting of low level easterly wind anomalies over the Indian Ocean and dynamically induced subsident drying over India.

1. Introduction

[2] The Atlantic Multidecadal Oscillation (AMO) is a leading large scale pattern of multi-decadal variability in North Atlantic sea surface temperatures (SST). Observational and model studies suggest a potential impact of the AMO on regional climate including Sahel and Brazilian rainfall, Asian summer and winter monsoons, both the summer and winter climates of North America and Europe, and the intensity and frequency of Atlantic hurricanes [e.g., Folland et al., 2001; Rowell et al., 1995; Enfield et al., 2001; Goldenberg et al., 2001; McCabe et al., 2004; Sutton and Hodson, 2007, hereinafter referred to as SH07; Knight et al., 2006; Lu et al., 2006; Li and Bates, 2007]. In addition, a modulation of the low-frequency variability of El Niño/Southern Oscillation by the AMO was proposed by Dong et al. [2006].

[3] Several recent studies have focused on the influence of the AMO on Indian summer rainfall but with non conclusive results. The analysis of observational data by Zhang and Delworth [2006] (hereinafter referred to as ZD06) reveals a coherent relationship between AMO and Indian summer rainfall, though no mechanism was proposed in their investigation. Based on interannual and interdecadal correlation between the AMO, the summer North Atlantic Oscillation (NAO) [Portis et al., 2001], and Indian rainfall, Goswami et al. [2006] (hereinafter referred to as G06) proposed the following mechanism: a positive-phase AMO causes a south-north dipolar pattern, which is analogous to the summer NAO; this circulation pattern leads to a warming of the Eurasia continent, intensifies the thermal contrast between the Indian sub-continent and the tropical Indian Ocean and consequently the Indian summer monsoon. The efficacy of such a causal chain has yet to be established.

[4] The question of mechanisms continues to be difficult to answer in part because various model studies yield inconsistent results about the AMO influence. ZD06 forced the GFDL (Geophysical Fluid Dynamics Laboratory) coupled atmosphere ocean general circulation model (AOGCM) CM2.1 with a positive phase AMO flux anomaly and obtained enhanced rainfall in summer (JJAS). In contrast, SH07 forced the atmospheric general circulation model (AGCM) of the Hadley Centre, HadAM3, with idealized 4-times-amplified AMO SST anomalies. They obtained weakly suppressed rainfall in southern India and increased rainfall in northeastern India in the summer (JJA) (SH07, Figure 3h). Knight et al. [2006] investigated a 1400-year control integration with the AOGCM, HadCM3, that captures well the AMO, and found the following response: A positive-phase AMO corresponds to enhanced summer (JJA) rainfall in northern India, a weakened rainfall in western-central India and the northern Bay of Bengal, as well enhanced autumn (SON) rainfall in most areas of India. Finally, Lu et al. [2006] conducted a different simulation, where the HadAM3 is coupled with the ocean except over the Atlantic where specified SSTs representing three-times amplified AMO warming or cooling. They obtained no significant response in summer (JJA) but enhanced autumn (SON) rainfall.

[5] The discrepancy among the model results and the limitation of a short instrument record length motivated the present study to further document the Indian summer monsoon sensitivity to North Atlantic SST forcing, and assess candidate mechanisms.

2. Model Description and Experiment Design

[6] The AGCM is an earlier version of the National Centers for Environment Prediction's (NCEP's) AGCM for seasonal prediction [Kanamitsu et al., 2002]. It is a global spectral model with a T42 horizontal truncation and 28 sigma vertical levels. We conducted four sets of ensembles with 60 members each, starting from different initial fields from the NCEP-NCAR (National Centers for Atmospheric Research) reanalysis of 00UTC, 1–3 September 1980–1999 [Kalnay et al., 1996] and being integrated for 13 months until the following September. Ensemble one, referred to as “Control” ensemble, is based on experiments with seasonally-evolving climatological SST. Ensemble two, referred to as “AMO” ensemble, is based on experiments in which a positive AMO SST anomaly (SSTA) is added to the climatological SST. This anomaly is derived by calculating the SST difference between AMO warm-phase (1935–1955) and cold-phase (1970–1990). The resulting SST anomaly magnitudes are roughly twice as large as the observed AMO SSTA during a typical warm phase ([Enfield et al., 2001] (Figure 1b) and [Dong et al., 2006] (Figure 1a)) (see also Figure S1). Ensemble three, referred to as “AMO_trop”, is the same as the AMO ensemble, but only the tropical part of the AMO SSTA is used. The last ensemble, referred to as “AMO_extratrop”, is forced only with the extratropical part of the SSTA. The purpose of the last two ensembles is to investigate the relative contribution of the tropical SSTA, particularly that over the ITCZ area, and the extratropical SSTA to the total response. The individual responses are calculated as differences between the SSTA experiments (AMO, AMO_trop, or AMO_extrop) and the control ensemble. The latitudinal separation between the tropical and the extratropical domains is similar to SH07 (Figure S1).

Figure 1.

(a) Comparison of the AIR index between the Control ensemble, the AMO ensemble, and the observations. (b) Monthly AIR response to the positive AMO SST anomaly, (top) total, (middle) tropical, and (bottom) extratropical contribution. The July–September monthly responses to the AMO and AMO_extratrop are significant at least at the 95% level, but not significant for AMO_trop. Unit: mm day−1.

3. Results

3.1. Validation of the Model's Internal Variability

[7] First, we investigated whether the observed links between the Indian summer rainfall, the so-called summer NAO, and the land-sea thermal contrast are realistically reproduced by the AGCM. Following G06, we determined the All Indian Rainfall index (AIR) of the AGCM by averaging over all 39 model grid points that cover India. The land-sea thermal contrast is represented by the meridional gradient of the Tropospheric Temperature (ΔTT). A summer NAO index is defined based on 1000-hPa heights (Figure S2). We found that the model's AIR is significantly correlated with the land-sea thermal contrast (ΔTT), in agreement with the observations. However, the statistical significance of the correlation between AIR and the NAO index is only marginal.

[8] In order to evaluate the model's precipitation, we use the global monthly precipitation dataset from the NCEP [Xie and Arkin, 1997]. This observational dataset is gridded at 2.5° latitude by 2.5° longitude, and spans from 1979–2002. The observed AIR index is determined as an area-mean of 50 grid points covering India. Figure 1a (also Figure S3) shows that the AGCM reproduces well the observed seasonality of AIR, though with smaller amplitude. The dry bias of the control run's simulated Indian monsoon rainfall is a feature common to many AGCMs, and is related in part to the coarse spatial resolution that reduces heavy orographic rainfall along the Ghats Mountains, and it hinders the ability to generate intense tropical depressions influencing eastern India [e.g., Gadgil and Sajan, 1998].

3.2. Modeled Surface Response

[9] Figure 1a shows the curve of the seasonal cycle of rainfall in the simulations forced with the AMO warm SST anomalies. An increase of AIR from mid-summer to autumn (July–November) is evident, with monthly totals 20% above climatology during July and August. Figure 2a shows the spatial pattern of rainfall response, and reveals that the principal increase occurs over western-central India, though virtually all of the sub-continent experiences greater rainfall. An almost equal reduction of rainfall occurs over the Bay of Bengal, resulting in a meridional dipole that is suggestive of a northward shift of the climatological ITCZ. These responses are in qualitative agreement with G06 and ZD06, but their intensity is greater than the responses in the uncoupled or coupled simulations of the Hadley Centre's models. Our study also reveals that a positive AMO state induces enhanced rainfall over the Sahel and the tropical Atlantic as well as a modest northward shift of the Atlantic ITCZ, which are in agreement with SH07.

Figure 2.

July–September mean response of (top) precipitation (mm/day), (middle) surface air temperature (°C), and (bottom) surface air pressure (hPa) to the total, the tropical, and the extratropical positive AMO SST anomalies. In the middle and lower panels, contour interval is 0.1, with positive (negative) values in red (blue) and the contour 0 in thick line. Shading indicates significant at the least at the 95% level.

[10] Surface temperature (Ts) responses (Figure 2b) in the tropics are consistent with the rainfall response suggestive of the radiative effects resulting from anomalous cloud cover in wet and dry regions: areas with more (less) rainfall become cooler (warmer). This is especially evident in the Sahel and the Indian monsoon area. Most of the extratropical Eurasian continent is warmer, in agreement with SH07.

[11] Significant negative surface pressure (Ps) responses can be found over the North Atlantic, northern Africa, the Middle East and the western India Ocean (Figure 2c). As will be shown further, the intensified monsoon trough over the Indian Ocean is associated with increased southwesterly low level winds over the Arabian Sea and western Indian Ocean that appears to be the immediate cause for the increased Indian monsoon rains.

[12] The investigation of the relative contributions of the warm AMO_trop (Figures 2d2f) and the warm AMO_extratrop SSTA (Figures 2g2i) indicates that both exert significant downstream influences over Europe, North Africa, the Middle East and India. Regarding Indian rainfall, the AMO_extratrop effect dominates the overall response seen in the AMO simulations (Figures 1b, 2g). It is also noteworthy that the eastern Indian and Bay of Bengal rainfall response to warm tropical North Atlantic SSTs is mostly opposite in sign to that induced by the warm extratropical North Atlantic, though somewhat weaker in amplitude. Such cancellation may partly account for the lack of consistency in previous model simulations of the monsoon response to the AMO because the results would depend on details of imposed SST forcings and the model sensitivity.

3.3. Atmospheric Circulation Response

[13] A better physical understanding of the teleconnection processes linking the Atlantic basin with Indian monsoon rainfall is gained from analysis of upper tropospheric features of the AGCM responses. Figures 3a, 3d, and 3g show the 200 hPa geopotential height anomalies and wave activity flux [Takaya and Nakamura, 2001] in the three experiments. Each experiment is characterized by a distinct wavetrain of alternate positive and negative polarity centers extending from the Atlantic across Asia. The upper-level convergence implied by the negative 200 hPa height anomalies from the Caspian Sea to central Asia in Figure 3g favors the intensification of the northern subsidence branch of the monsoon meridional streamflow (Figure S4). This may act as one path for the extratropical signals to reach the monsoon area. The smaller response in the AMO_trop experiments illustrates that the tropical Atlantic SSTA is less efficient in producing extratropical height anomalies over Northern Europe.

Figure 3.

As Figure 2, but for (top) 200-hPa geopotential height (gpm), (middle) 300-hPa temperature (K), and (bottom) 1000-hPa horizontal wind vector (mS−1). Shading indicates significance at the 95% level. In Figures 3a, 3d, and 3g, thick solid line indicates the route of the Rossby wave-train, and the arrows represent the anomalous wave activity flux. The sign “+” and “−” in Figures 3b, 3e, and 3h indicates warming and cooling, respectively. In Figures 3c, 3f, and 3i, only regions with at least one component of the wind significant at the 95% level are shown.

[14] Over the Atlantic, height anomalies have baroclinic vertical structure in which surface low pressure anomalies are overlain by positive 200 hPa height anomalies. Diagnostic model experiments using a steady state linear baroclinic model [Peng and Whitaker, 1999] forced with specified Atlantic atmospheric heating (Figure S5) confirm that this baroclinic structure is consistent with the tropospheric response to tropospheric diabatic heating over the warm SST anomalies (either tropical or midlatitude). In the diagnostic simulations (Figure S5), the extratropical heating is also more efficient in producing the extratropical atmospheric response than the tropical heating, and a downstream wave propagation from the Atlantic is also evident. The overall wavetrain, its vertical structure, and its dispersive characteristics that are poleward across Europe and Asia and equatorward across the Middle East are thus attributable to heating-induced solution predicted by linear wave theory [Gill, 1980].

[15] We also diagnose the upper tropospheric temperature response in the AGCM simulations (Figures 3b, 3e, and 3h). This is motivated by the empirical relation between the observed seasonal march of Indian monsoon rains and the tropospheric meridional temperature contrast between the Indian Ocean and Asia (G06). Over the monsoon region, a meridional pattern of alternate warm and cold tropospheric anomalies is seen in the AMO and the AMO_extratrop runs. The anomalies, though weak, represent an enhanced thermal gradient which could influence monsoon intensity. Similar features emerge in the vertically averaged TT (Figure S6) and Ts (Figure 2b).

[16] Finally, the 1000-hPa wind anomalies (Figures 3c, 3f, and 3i) reveal another contributing factor shaping the Indian monsoon rainfall response to Atlantic SSTs. The AMO and AMO_extratrop runs reveal a strengthened low-level southwesterly jet over coastal Africa, the Arabian Sea, and southern India. The implied anomalous lower-level convergence over the western Indian Peninsula is consistent with the increased rainfall in those areas. In contrast, the low level wind response in the AMO_trop simulations, characterized by exclusive easterly wind anomalies over the Indian Ocean (Figure 3f), reveals a weakened monsoon circulation. This is consistent with a general reduction in rainfall over the monsoon region including the Bay of Bengal in response to warm tropical Atlantic. Diagnostic experiments with the linear baroclinic model forced with a tropical Atlantic heat source are able to replicate such a weakened monsoonal circulation and also reveal enhanced Indian monsoon subsidence (Figure S7).

[17] The investigation of the relative contribution of tropical and extratropical SSTA illustrates that the circulation features favorable to increased Indian monsoon rainfall are solely attributable to warm extratropical Atlantic waters. The tropical Atlantic warming is also shown to be effective in driving Indian monsoon rainfall, though mostly in opposition to effects driven from warm far North Atlantic waters. Finally, note also that the height response to the total positive AMO SSTA is smaller than the summed responses of AMO_extratrop and AMO_trop (Figures 3f and 3i), implying a nonlinear interaction between the tropical and the extratropical SST forcing.

4. Summary Remarks

[18] In summary, we have highlighted several candidate factors apparently driving the Indian monsoon rainfall response to a warm polarity of the AMO. A significant increase in Indian summer monsoon rainfall is found in our AGCM simulations forced with a double amplitude AMO warm phase SST condition. The increased rainfall in these so-called AMO experiments is largely the results of warm extratropical Atlantic SST effects. These effects appear to be manifest in two manners: one is via an increased meridional gradient of the tropospheric temperatures over the greater monsoon region, and the other is via an intensification of the monsoonal low-level jet. The former is linked to a Rossby wave-like response in the troposphere, while the latter is tied to the surface pressure response to extratropical Atlantic forcing. The relative role of these, and the importance of other factors, has not been determined here, though we note that no homogeneous warming of Eurasia, nor an NAO-like atmospheric response, is found in our runs, contrary to the theory of G06. In addition, there are substantial height anomalies and wave activity flux divergence at the exit of the Atlantic jet (cf. Figures 3g and S8). Thus, the mechanism proposed by Ding and Wang [2005] about the mid-latitudinal North Atlantic atmospheric circulation influence on the Indian summer monsoon may have worked.

[19] We also note that warm tropical Atlantic SSTs also influence the Indian monsoon rains, largely via a weakening of the Indian monsoon southwesterly jet and increased subsidence over greater India. Whereas this drying effect is overwhelmed by the wet signal induced by the far North Atlantic warm waters in our experiments, it is evident that the net effect of Atlantic SST forcing on India could depend sensitively on the relative warmth of the tropics and extratropics.


[20] SH is grateful for one reviewer and editor Mark New, whose insightful scientific remarks and helpful writing suggestions lead to a significant improvement of the revised manuscript; also thanks Drs. Bueh Chowlu, Feng Xue, and Yongqi Gao for helpful discussions. This study was jointly supported by “One Hundred Talent Plan” of the Chinese Academy of Sciences and the NSFC projects with funding numbers 40775053 and 90711004.