The global monthly SST data was obtained from the Hadley Centre Sea Ice and Sea Surface Temperature (HadISST; Rayner et al., 2003) dataset with a horizontal resolution of 1° × 1°. Monthly three-dimensional wind datasets, which are available at a horizontal resolution of 2.5° × 2.5° and extend from 1000 to 10 hPa with 17 vertical pressure levels, were extracted from the United States National Center for Environmental Prediction–National Centers for Atmospheric Research (NCEP-NCAR) reanalysis products (Kalnay et al., 1996). Monthly rainfall data for all India were provided by the Indian Institute of Tropical Meteorology–Indian regional/subdivisional monthly rainfall (IITM-IMR; Parthasarathy et al., 1994) dataset. The anomaly in each variable was obtained by subtracting the climatological value from 1951 to 2016. All the monthly mean variables were linearly detrended and then smoothed by taking three-month running averages to exclude the subseasonal variability.

Following the criterion of the NOAA, an El Niño event is defined by a three-month running Niño3.4 (5°S–5°N, 120°W–170°W) SSTA index greater than or equal to +0.5 K for at least five consecutive overlapping time periods. Because we mainly focused on moderate and stronger El Niño events, a total of 15 El Niño events were identified during the time period 1951–2016 (Table 1) based on peak values of the Niño3.4 SSTA greater than +1.0 K. We then determined the type of these selected El Niño events based on the consensus of three identification methods, including the Niño3/El Niño Modoki index (EMI) method of Ashok et al. (2007), the Niño3/Niño4 method of Yeh et al. (2009) and the EP/CP-index method of Kao and Yu (2009). Using the Niño3/EMI method, El Niño events were classified as central (eastern) Pacific types when the DJF mean EMI value was greater (less) than that of the Niño3 (5°S–5°N, 90°W–150°W) SSTA index. The EMI is defined as: EMI = [ SSTA ] c −   0.5 [ SSTA ] E − 0.5 [ SSTA ] W   (1) where the square brackets with a subscript represent the area-averaged SSTA over the central (10°S–10°N, 165°E–140°W), eastern (15°S–5°N, 110°W–70°W) and western (10°S–20°N, 125°E–145°E) tropical Pacific, respectively.

With the Niño3/Niño4 method, El Niño events are classified as CP (EP) types when the DJF-averaged values of the Niño4 (5°S–5°N, 160°E–150°W) index are greater (less) than those of the Niño3 index.

Kao and Yu (2009) applied a combined regression–empirical orthogonal function (EOF) analysis to identify the EP and CP types of El Niño events and referred to this as the EP/CP-index method. In this method, El Niño events are classified as CP types when the DJF-averaged principal components of the EOF for the residual SSTA after removing the anomalies regressed with the Niño1+2 (0°–10°S, 80°W–90°W) SST index are greater than their counterparts after removing the anomalies regressed with the Niño4 index, and vice versa for the EP events. According to the majority consensus shown in Table 1, eight of the 15 major El Niño events were of the EP type and seven were of the CP type events.

To validate the rationality of the classification method for the two types of El Niño event used in this study, Figure 1 shows the temporal variation of the equatorial Indo-Pacific Ocean SSTA averaged between 5°S and 5°N associated with the EP and CP El Niño events. The center of the warming EP El Niño-related SSTA is observed in the eastern-central equatorial Pacific, with the warmest SSTA of +2.0 K to the east of 120°W (Figure 1A). By contrast, the significant anomalous SST warming during CP El Niño events shifts westward to the west of 150°W and its corresponding center with a maximum positive SSTA of +1.0 K is much weaker than that of the EP El Niño events (Figure 1B). Based on a significant SSTA greater than +0.5 K, a warm SSTA related to an EP (CP) El Niño event develops in April (June) of the developing year, peaks in winter and subsequently decays after the following June (April), indicating that the positive SSTA starts later and ends earlier during CP El Niño events, with a duration four months shorter and a decay rate much faster than for EP El Niño events.

Differences also exist in the strength and timing of the warming SSTA responses in the equatorial Indian Ocean to the two types of El Niño event. Anomalous equatorial Indian Ocean SST warming associated with EP El Niño events is stronger and lasts longer than that of CP El Niño events (Figure 1), highlighting the stronger IOBM and the subsequent air–sea interactions during EP El Niño events. This result indicates that EP El Niño events are followed by the IOBM until the following summer, but the warming SSTA related to CP El Niño events only lasts until the following spring and is insignificant in summer, consistent with the results of Tao et al. (2014). The classification method therefore effectively reflects the differences between EP and CP El Niño events, especially their different impacts on the SSTA in the equatorial Indian Ocean.

To investigate the different impacts of EP and CP El Niño events on rainfall anomalies over India, Figure 2 shows composites of the all-Indian rainfall anomalies during the decaying years of the two types of El Niño event. In the climatology, the ISMR in JJAS contributes as much as 75% to the annual rainfall over India, consistent with the results reported by Parthasarathy et al. (1994). It is clear that the larger standard deviations of the ISMR occur in JJAS (Figure 2, error bars), indicating that the ISMR also shows the most pronounced interannual variation. For the decaying years of EP El Niño, the ISMR shows a deficit in early summer (June–July), a surplus in late summer (August–September) and peaks in September, consistent with previous studies (Park et al., 2010; Lau and Nath, 2012; Lü and Zheng, 2017). By contrast, the ISMR is shows a greater surplus during the following JJAS of CP El Niño events and the peak in the ISMR anomalies occurs a month earlier than the peak in EP El Niño events.

The variations in the ISMR can be very different during the following JJAS, with suppressed (enhanced) rainfall in early summer for EP (CP) El Niño events (Figure 2). The opposite signs of the early summer rainfall over India associated with two the types of El Niño event imply that the corresponding onset time of the ISM also shows contrasting differences, with a late (early) onset of the ISM for EP (CP) El Niño events. The responses of the ISMR to the two types of El Niño event are significantly different in early summer. The summer in El Niño decaying years is therefore divided into early (June–July, JJ) and late (August–September, AS) summer in the following analysis to explore the anomalous atmospheric circulation and SST and the related physical mechanisms.

After inspecting the equatorial Indo-Pacific Ocean SSTAs related to the two types of El Niño event and their different impacts on the following ISMR, we investigated the associated atmospheric circulation anomalies over the tropics. Figure 3 shows the EP and CP El Niño composites of the 850 hPa winds and the relative vorticity anomalies in early and late summer. During the decaying years of EP El Niño events, an antisymmetric wind structure over the TIO is observed in early summer, with anomalous northeasterly flow to the north of the equator and anomalous northwesterly flow to the south of the equator (Figure 3A). Such an antisymmetric wind pattern is forced by the sustained warming of the SST over the SWIO associated with the El Niño-induced downwelling oceanic Rossby waves (Masumoto and Meyers, 1998; Du et al., 2009; Wu and Yeh, 2010). Anomalous northeasterly winds over the western NIO (WNIO) inhibit the southwest summer monsoon and reduce surface evaporation, thus delaying the onset of the ISM, leading to prominent warming of the SST over the NIO via the positive WES feedback (Xie and Philander, 1994). As a result of the westward expansion of the strong lower level easterlies over the NIO, less moisture is transported to South Asia and weak negative relative vorticity anomalies appear over the South Asian subcontinent (Figure 3A), which also reduce early summer rainfall over India (Figure 2). In late summer, however, the antisymmetric wind pattern is weakened and shifted eastward; weak westerly anomalies are observed over the WNIO (Figure 3B), indicating the start of the ISM. These warm and humid southwesterly winds substantially enhance the supply of moisture to South Asia and lead to stronger moisture convergence over India and significantly positive relative vorticity anomalies in situ, resulting in excess rainfall in this sector during the following AS season of the EP El Niño events.

During the early summer of the CP El Niño events, the lower level atmospheric circulation anomalies are similar to those for the late summer EP El Niño events, with an eastward shift of the weakened antisymmetric wind pattern over the TIO (Figure 3C). Significantly anomalous southwesterly winds over the WNIO cause the easterly anomalies to retreat eastward to the eastern Arabian Sea and then transport more moisture to South Asia. Significantly positive relative vorticity anomalies are also found over the south of the Indian subcontinent, leading to enhancement of the ISMR. This indicates that the ISM starts in JJ season when the onset time for the CP El Niño is about two months earlier than that for the EP El Niño. Two months later, anomalous southwesterly flow over the WNIO is gradually intensified and is more significant in AS season, along with the disappearance of the antisymmetric wind pattern over the TIO (Figure 3D). Over India, significant convergence of the relative vorticity anomalies and the increase in the amount of moisture transported by the anomalous southwesterly winds give rise to more rainfall in situ (Figure 2).

The velocity potential and divergence wind can be used to reflect the large-scale features of divergent motion (Tanaka et al., 2004). Figure 4 shows the differences between the velocity potential and divergence wind anomalies between 200 and 850 hPa in early and late summer related to EP and CP El Niño events. For the EP El Niño events, the most significant and strongest upper level convergence anomalies are centered over the western equatorial Pacific in early summer (Figure 4A), collocating with significantly negative local SSTAs (Figure 1A), whereas insignificant upper level divergence anomalies are located over the TIO (Figure 4A), along with the lower level antisymmetric wind pattern. In late summer, however, these upper level divergence anomalies are enhanced and are centered over the eastern TIO (Figure 4B). This divergent circulation indicates ascending motion over the TIO, which favors excess rainfall over India in AS season. By contrast, for CP El Niño events, the significant upper level divergence anomalies (and, by implication, ascending motion) in early summer are centered over the western TIO (Figure 4C), whereas the anomalies in late summer extend into the maritime continent and are progressively intensified (Figure 4D). Compared with the EP El Niño events, the most prominent difference in the upper level circulation related to CP El Niño events is the strong divergent anomalies over South Asia, which highlight the stronger ascending motion and higher rainfall during the following summer of CP El Niño events, especially in early summer. These results confirm the late onset of the ISM and lower ISMR in early summer for EP El Niño events, but a normal onset of the ISM and higher ISMR for CP El Niño events.

To further identify the possible connection between the ISMR and the local meridional cells associated with EP and CP El Niño events, Figure 5 shows composites of the latitude–height cross-sections of the wind and geopotential height anomalies averaged over India between 65 and 85°E. An anomalous clockwise circulation in early summer is observed over the Indian subcontinent for EP El Niño events (Figure 5A), with significant anomalous descending (ascending) motion over northern India (the TIO), suppressing convective activity over India and leading to less rainfall, despite the slightly converging anomalies of the water vapor flux (figure not shown). The upper level southerly wind and southward geopotential height gradient anomalies over India are also related to the anomalous southwesterly winds, implying a delay and weakening of the ISM. Two months later, the significant anomalous sinking over northern India disappears and changes to ascending motion in the mid-troposphere from 300 to 500 hPa (Figure 5B). The inverted wind direction and geopotential height gradient indicate the occurrence of upper level northeasterly wind anomalies as well as the onset of the ISM, which both favor surplus rainfall in India in late summer.

By contrast, for CP El Niño events, significant anomalous ascending motion over India and the TIO are present in both early and late summer, along with upper level northerly wind anomalies, small lower level southerly wind anomalies (Figures 5C,D) and strong water vapor flux convergence anomalies (figure not shown). These types of circulation contribute to the onset of the ISM. The anomalous lower level southwesterlies with local convergence and ascending anomalies in late summer are stronger than those in early summer (Figures 5C,D). As a result, they can transport more moisture to India, enhancing the ISMR.

We have shown that the two types of El Niño event have different influences on the timing of the onset of the following ISM and ISMR. This is clearly due to the associated atmospheric circulation anomalies over the TIO in early summer. For EP El Niño events, India is influenced by anomalous easterlies related to the lower level antisymmetric wind pattern over the TIO, which delay the onset of the ISM and then suppress the transport of moisture to India, resulting in less-than-normal ISMR in early summer. By contrast, for the CP El Niño events, the enhanced southwesterly flow over the WNIO and weakened antisymmetric wind pattern over the TIO facilitate the onset of the ISM in early summer and also transport more moisture to South Asia. There is anomalous ascending motion over India, which boosts precipitation processes, leading to an increase in the ISMR.

Previous studies have shown that El Niño events can modulate the ISM via the subsequent warming of the SST in the TIO (Alexander et al., 2002; Lau and Nath, 2003; Du et al., 2009). We also found remarkable differences in the strength and timing of the warming SSTA in the equatorial Indian Ocean in response to the two types of El Niño event (see Figure 1). We therefore further investigated the importance of the El Niño-associated SSTA patterns in the TIO in determining the atmospheric circulations in the TIO during the following summer of El Niño years (Figure 6).

During the early summer following EP El Niño events, the SSTA is characterized by the significantly warmer-than-normal temperature in the TIO, especially in the NIO (Figure 6A). This basin warming, which is the so-called second NIO SST warming (Du et al., 2009), can sustain an antisymmetric wind structure over the TIO with northeasterly (northwesterly) wind anomalies north (south) of the equator (see Figure 3A) via positive WES feedback. The significant southwestward SST gradient in the WNIO corresponds to the maintenance of an anomalous northeasterly flow, which, in turn, delays the onset of the ISM. Two months later, warming of the SST in the NIO gradually weakens in late summer (Figure 6B) and the positive SSTA in the WNIO decays faster and earlier than that in the eastern NIO (ENIO), corresponding to the weakening and eastward movement of the antisymmetric wind pattern over the TIO. This result indicates that the anomalous southwesterly flow transports more moisture to India and leads to excessive ISMR in late summer following EP El Niño events.

The largest difference in the SSTA pattern related to CP El Niño events compared with EP El Niño events is a significant cooling of the SSTA in the western TIO, highlighting a key role of the negative SSTA in the different responses of the ISMR to the two types of El Niño event. In early summer, significant warming of the SST is mainly concentrated in the NIO and the southeastern TIO, whereas a negative SSTA center is found along the eastern coast of Africa (Figure 6C). Both the warming SSTA in the Arabian Sea and the cooling SSTA in the western TIO form a positive meridional SSTA gradient in the WNIO, which can excite local southwesterly wind anomalies (see Figure 3C), contributing to the onset of the ISM. In AS season, the negative SSTA center along the coast of Africa is intensified and expanded, although warming of the SST in other regions is weakened (Figure 6D). The increased meridional SSTA gradient in the WNIO therefore continues to strengthen the southwesterly anomalies, which favors the increased transport of moisture to India and enhances the ISMR in late summer.

To further distinguish the relative role of the SSTA, we examined the El Niño composites for the monthly evolution of the SSTA in some key regions (Figure 7). It is clear that there are remarkably different responses of the SSTA in the NIO to different types of El Niño events. The warming of the SST over the NIO related to EP El Niño events (0–20°N, 50–100°E) is characterized by double peaks in the developing November and the following June, whereas the warming of the SST over the NIO related to CP El Niño events increases slowly with time and shows a single peak in the following March (Figure 7A). This result for CP El Niño events is inconsistent with the double warming of the NIO associated with El Niño events reported by Du et al. (2009).

The warming of the SST over the NIO in the following summer also shows a non-uniform zonal distribution in the two types of events (Figure 6). Significant double warming peaks are observed over the WNIO (0–20°N, 50–80°E) during EP El Niño events, but only a single warming peak during CP El Niño events (Figure 7B), consistent with the results for the NIO shown in Figure 7A. However, the responses of the SSTA over the ENIO (0–20°N, 80–100°E) to the two types of El Niño events show few differences, with a single warming peak in the following March (Figure 7C). These results indicate that the warming SSTA in the WNIO makes a major contribution to the warming pattern of the SST in the NIO, whereas the counterpart in the ENIO has little effect.

Significantly different responses of the SSTA in the WNIO can result in different meridional SSTA gradients between the Arabian Sea (15–25°N, 55–75°E) and the western TIO (10°S–10°N, 40–60°E). As a result of the double warming peaks in the NIO associated with EP El Niño events, the negative meridional SSTA gradient persists from the winter to the following early summer (Figure 7D), which favors the maintenance of the northeasterly wind anomalies and leads to a delay in the onset of the ISM, suppressing ISMR in early summer. Until late summer, the meridional SSTA gradient changes from negative to slightly positive, heralding the onset of the ISM and an increase in the ISMR. By contrast, when the single warming peak related to CP El Niño events decays in early summer, the meridional SSTA gradient becomes positive and then remains significantly positive from the following May to September (Figure 7D), which corresponds to the anomalous southwesterly wind in the WNIO. This wind anomaly can contribute to the onset and intensification of the ISM, resulting in the enhanced ISMR. These analyses show that the different SSTAs in the WNIO associated with the two types of El Niño event are primarily attributed to the differences in the early summer ISMR anomalies between the EP and CP El Niño events.

We have considered the remarkable differences in the anomalous SST and atmospheric circulation patterns over the NIO associated with the two types of El Niño event. To investigate the possible physical mechanism of how the different SSTAs in the NIO might modulate the circulation related to the ISMR, we examined the monthly evolution of three groups of the ISM strength index in decaying years of the EP and CP El Niño events (Figure 8). The Indian monsoon index (IMI; Wang and Fan, 1999) is used to indicate the difference in the 850 hPa zonal wind between the southern Arabian Sea (5–15°N, 40–80°E) and the northern Indian continent (20–30°N, 70–90°E). The Indian rainfall index (IRI; Shukla and Paolino, 1983) refers to the standardized Indian rainfall, similar to the ISMR (shown in Figure 2) in this study. The Webster–Yang index (Webster and Yang, 1992) denotes the zonal wind shear between 850 and 200 hPa over the NIO (0–20°N, 40–110°E). The changes in these three indices from negative to positive represent the onset of the ISM. The IMI and the Webster–Yang index lead and lag the IRI, respectively (Figure 8), indicating that the onset of the ISM manifests first in the lower level zonal wind field, then in the rainfall over India and finally in the vertical shear of the zonal wind over the NIO.

For EP El Niño events, the second NIO SST warming peak in the following June (Figure 7A) induces the lower level antisymmetric wind over the TIO and its resultant northeasterly anomalies over the Arabian Sea inhibit the onset of the ISM, leading to a negative value of the simultaneous IMI. This lower level circulation decreases the transport of moisture and then gives rise to the negative IRI anomalies in early summer, which suppress the release of the heat of condensation over South Asia, resulting in negative zonal wind shear over the NIO in the following August (Figure 8A). By contrast, the single SST warming induced by CP El Niño events, which peaks in the following March, decays quickly during the late spring and summer (Figure 7A). The early summer SST cooling in the western TIO (Figure 6C) directly causes the significantly positive meridional SSTA gradient in the WNIO, which induces lower level southwesterly wind anomalies (Figure 3C), corresponding to the positive IMI in the following June and the onset of the ISM (Figure 8B). This lower level circulation can cause an eastward shift in the decreased lower level antisymmetric wind over the TIO, which favors the onset of the ISM.

A comparison between EP and CP El Niño events suggests that the time of onset of the ISM circulation and the ISMR for CP El Niño events is about two months earlier than that of EP El Niño events. Anomalous ascending motion over India also enhances the positive IRI anomalies and the resultant release of the heat of condensation in early summer, leading to positive zonal wind shear over the NIO in the following July (Figure 8B). These results confirm that the different SSTA gradients in the WNIO attributed to the diversity of El Niño events can affect the time of onset of the ISM circulation in the early summer following the El Niño event and can therefore result in a deficit (surplus) in the ISMR associated with EP (CP) El Niño events.