Introduction

The Indian Summer Monsoon (ISM) is well known for its quasi-biennial character: the likelihood of more than two consecutive drought or flood years is low1. Early studies noted that this inter-annual variability is tightly coupled to the Southern Oscillation (SO)2, and subsequent work established the broader tropospheric biennial oscillation as a key bridge between the SO and ISM rainfall3. Because low-frequency oscillations can confer longer predictability to the tropical atmosphere4, considerable effort has gone into understanding how El Niño–Southern Oscillation (ENSO) modulates monsoon rainfall.

Evidence for a robust ENSO–monsoon link spans both observations and modelling. Shukla and Paolino5 argued that SO indices could serve as long-range predictors of ISM droughts, while Meehl6 showed that extreme SO events systematically coincide with drought and flood years over Asia. Large sea surface temperature (SST) anomalies in both the Indian and Pacific Oceans display a clear biennial signature linked to ENSO phases7. Mechanistically, Niño-3 SST anomalies that develop in the preceding winter can pre-condition the following summer monsoon via the Walker circulation8,9. The importance of this remote control is underscored by modelling experiments showing that land-surface feedbacks alone cannot overturn ENSO’s influence on monsoon circulation anomalies10. Recent palaeoclimate reconstructions confirm that ENSO has remained the dominant driver of ISM inter-annual variability for at least the past five centuries11, and new analyses highlight how equatorial Rossby-wave adjustments and Walker-circulation changes modulate the ISM period during ENSO events12.

However, beyond the influence of ENSO, Indian Ocean (IO) processes are equally crucial for a more complete understanding of monsoon variability. The cross-equatorial ocean heat transport from the northern to the southern IO is a crucial component of the ISM. The southern hemisphere atmosphere transports latent heat during the northern summer monsoon13. Moreover, the moisture evaporated from southern IO is carried to the north across the equator, which in turn precipitates. In a later study, Hastenrath and Greischar14 found that this southward heat transport to the southern IO a nd northward cross-equatorial latent heat export balance each other. The ISM is argued to be a self-regulated coupled ocean-atmosphere system15,16. Self-regulation means that the inter-annual fluctuations in the ISM rainfall are modulated by the coupling between the ocean and the atmosphere over the IO. At the core of this argument is the southward cross-equatorial Ekman transport from the North Indian Ocean (NIO) to the south, driven by the low-level southwesterly low-level monsoon winds. According to Webster et al.15, an anomalously strong monsoon results in a stronger cross-equatorial ocean heat transport (OHT), which in turn would lead to an anomalous oceanic cooling (warming) in the north (south). This can lead to a weakening of the meridional temperature gradient in the following summer, leading to an anomalously weak monsoon. At the peak of the ISM season, the magnitude of OHT over the IO is about 1.5 to 2 PW15. The seasonal reversal of the Ekman transport is essential to maintaining the Indian Ocean’s heat balance, as demonstrated by Loschnigg and Webster16 and Wacongne and Pacanowski17. Webster et al.15 modified Meehl18 biennial monsoon theory by incorporating OHT and Indian Ocean Dipole(IOD). The IO southward heat transport during boreal summer can be explained by the Ekman flux19. According to hydrographic measurements, the IO transports heat of about 1.5 petawatts (PW) at 32°S20. IO meridional heat transport demonstrates considerable variability across different time scales, including both shorter-term variations, such as intraseasonal fluctuations spanning approximately 10 days, and longer-term variations that occur over interannual timescales, lasting several years21.

Unlike in the Pacific, convection does not accompany the warm SST in the IO in the spring and early summer. It has been demonstrated that the main mechanism controlling the SST in the NIO and modifying the heat balance is cross-equatorial heat transport16. Lee and Marotzke22 investigated IO meridional overturning heat transport and found that seasonal reversal of transport occurs near 10°S and 10°N between two monsoon seasons. Lutsko et al.23 analysed the impact of OHT over ISM using an atmospheric model and OHT delineation is achieved by coupling a slab ocean model. Their sensitivity experiment without OHT gives a weak monsoon, and on the other hand, a gradual increase of OHT in the experiments improved monsoon performance in the model.

Internal processes introduce an additional atmosphere-driven source of year-to-year variability. The quasi-biennial oscillation (QBO) emerges from the interplay between intraseasonal oscillations and the annual cycle24, and numerical experiments suggest that such internal dynamics can modulate mean monsoon circulation even under fixed boundary conditions25,26. Although the seasonal-mean monsoon rainfall contains a strong intraseasonal component, its net contribution tends to cancel out in pronounced flood or drought years27. Nevertheless, the persistence of SST anomalies remains essential for maintaining the biennial mechanism that links successive seasons28.

The IOD introduces additional complexity to this system, as demonstrated by Saji et al.29. Positive IOD events, characterized by anomalous cooling (warming) in the eastern (western) equatorial Indian Ocean, can enhance monsoon rainfall independent of ENSO forcing. However, as shown by Ashok et al.30, the interplay between IOD and ENSO creates nonlinear interactions that complicate simple attribution of monsoon variability. Recent work by Pillai et al.31 has further revealed how North Tropical Atlantic SST anomalies can influence monsoon behaviour through teleconnections to the Indian Ocean, highlighting the multi-basin nature of monsoon drivers.

Despite the central role attributed to OHT in the self-regulation hypothesis, a rigorous evaluation based on observations is still lacking; most supporting studies rely on idealised or slab-ocean models. Here we quantify Indian Ocean OHT using contemporary reanalysis data and assess its relationship to ISM interannual variability with a focus on its biennial nature. Specifically, we test the hypothesis that OHT acts as a dominant internal regulator of ISM variability, potentially rivalling or complementing the role of ENSO as a remote forcing agent. By disentangling the contributions of OHT, ENSO, and internal atmospheric variability, we aim to provide a more comprehensive understanding of the physical processes governing monsoon fluctuations.

Results

Summer cross-equatorial Indian Ocean heat transport

The meridional OHT is calculated using Eq: 1 as the integral product of the potential temperature and the meridional velocity, following Chirokova and Webster21. The southward cross-equatorial heat transport in the IO is stronger in the boreal summer. Fig. 1 a shows the OHT as a function of latitude during June-September (JJAS). The OHT is mostly southward across all latitude bands over the tropical IO during JJAS, with a maximum value of slightly more than 1.5 PW between the equator and 10°S. The OHT values between 20°S and 20°N, range from 0.5 to − 1.6 PW. These values are in line with the estimates of Webster et al.15 and Chirokova and Webster21. Fig. 1b shows the OHT time series averaged over 5°S and 5°N for the period of 1980-2018. The time series reveals that mean cross-equatorial heat transport is southward in the boreal summer season in the IO.

Fig. 1
figure 1

(a) June-September (JJAS) meridional ocean heat transport (OHT) climatology over the Indian Ocean, (b) Time series of JJAS OHT averaged over 5°S−5°N. The climatology is computed for the 1980–2018 period. Units are in petawatts (PW).

Full size image

Understanding the factors contribute to the variability of Indian Ocean meridional heat transport

The longitude-depth cross-section of potential temperature (\(\theta\)), meridional velocity (v) of ocean water, and their product are shown in Fig. 2. We construct a time series by averaging the OHT between 5°S and 5°N. The analysis shown in Fig. 1 suggests that maximum OHT values are in between 5°S and 10°S. The highest values of \(\theta\) and v are observed within the top 100 m, and thus, maximum OHT must also be confined to this layer. This is clear from the product of v and \(\theta\).

To understand the interannual variability in OHT, we performed an empirical orthogonal function (EOF) analysis on the longitude-height section of the JJAS mean \(\theta\), v and product of \(\theta\) and v (\(\theta\) x \(v\)) anomalies, averaged over 5°S−5°N (Fig. 3). A dipole-like pattern is observed in the EOF1 of \(\theta\) anomalies over depth. The EOF1 of the v and \(\theta\) x \(v\) also shows a similar EOF1 mode, which suggests that v controls the variability in the OHT. The leading mode of \(\theta\) captures 41.9% variance, which indicating that this mode explains bulk of the interannual variability in the sea water temperature. Further, the opposing sign of the variability in the east and west of the IO is indicative of IOD-like variability. However, the EOF1 patterns of v and \(\theta\) x \(v\) do not have a coherent structure like that of \(\theta\) in the region west of 80°E. By “coherent structure”, we refer to a spatially continuous and organized pattern, while the EOF1 of v and \(\theta\) x \(v\) appears fragmented and lacks such spatial organization. This suggests that interannual anomalies in meridional velocity and OHT are weaker and less spatially organized in this region. We interpret this lack of coherent structure in the EOF1 of \(\theta\) x \(v\) as an indication of weaker and less organized interannual variability in OHT in the IO, particularly west of 80°E, noting that \(\theta\) and v are the constituent variables used in calculating OHT. These results raise a question on the supposed role of the OHT in contributing to the interannual variability of ISM as argued by Webster et al.15 and Chirokova and Webster21.

Fig. 2
figure 2

Longitude-depth cross-section of JJAS mean (a) potential temperature (\(\theta\), units: K), (b) meridional velocity (v, units: m s−1), and (c) product of \(\theta\) and v (\(\theta\) x v units: K m s−1) averaged over 5°S−5°N.

Full size image
Fig. 3
figure 3

(a), (c), (e) – EOF 1 of potential temperature ( \(\theta\)), meridional velocity (v) and product of \(\theta\) and v respectively; (b), (d), (f)- PC1 of \(\theta\), v and product of \(\theta\) and v respectively for JJAS, averaged over 5°S−5°N.

Full size image

The role of summer Indian Ocean heat transport in the subsequent year monsoon and SST pattern

The strong annual cycle of cross-equatorial heat transport evident in our analysis is also supported by previous studies21,22. The southward heat transport during boreal summer can modulate Northern IO heat balance and thereby regulate SST in the IO16. Earlier studies6,32 associate the interannual variability of the monsoon with the intensity of the SST prior to the boreal summer. Meehl’s observations have been substantiated by several studies33,34,35, demonstrating a substantial correlation between Indian Ocean SST anomalies in winter and spring and the strength of the subsequent monsoon. To understand the effect of OHT on ISM variability, a linear regression of JJAS mean OHT timeseries averaged over 5°S−5°N is performed on IO SST, wind vectors at 850 hPa, and precipitation anomalies in the same year (year 0) and following year (year +1). This analysis is restricted to 1998-2018 due to constrained TRMM precipitation data availability. For consistency the regression on SST and wind is also done on same period. The argument of OHT playing a central role in regulating ISM is based on its plausible influence on IO SST variability. The regression of SST for different seasons with the JJAS mean OHT is shown in Fig. 4. The regression pattern shows a muted relationship between OHT and SST except for a positive regression pattern in the eastern part of the South Indian Ocean (SIO) for September – November (SON) and December – February (DJF) seasons of year 0 (Fig. 4a and b). However, none of the positive regression values are statistically significant. The regression coefficients become weaker for the March – May (MAM) and June – August (JJA) SSTs of the year+1 (Fig. 4 c and d). Statistically significant negative regression pattern in the southern central IO during MAM(+1) and JJAS(+1) is not consistent with the idea that increased southward OHT causes an increase in SST. This weakening of the OHT influence on SST beyond six months casts more doubt on the role of OHT in regulating the interannual variability of ISM. A similar regression of OHT on wind vectors and precipitation was also done (Fig. 5). Most of these regression coefficients were also not statistically significant. The effect of JJAS mean OHT on precipitation and low-level winds weakened substantially by the following boreal spring and summer (Fig. 5c and d), similar to that in the case of SSTs. The OHT time series is also regressed onto the volume-mean temperature anomalies over the SIO for the same seasons (Figure S1), and this reveals a statistically insignificant relationship between JJAS OHT and seasonal volume-mean temperature anomalies.

Gupta et al.36 recently utilized Transfer Entropy (TE) to examine the exchange of information between pollutants and meteorological variables. Srujan et al.37 demonstrated a causal link between Rossby wave variance in the west Pacific and the development of downstream low-pressure systems (LPS) over the Bay of Bengal using TE. A high TE value reflects a strong transfer of information. TE reveals the flow of information from the memory of one variable to another36 and captures both the memory and the direction of information transfer, offering an advantage over mutual information. To understand the causal relationship between cross-equatorial OHT and SST, TE is computed between the OHT time series (averaged over 5°S−5°N) and the SST across the SIO at varying time lags (Fig. 6). To investigate the potential predictive relationship between OHT and SST, TE is determined by holding the OHT time series constant while shifting the SST time series from the first month (t0) to the 24th month (t+24). The TE values between OHT and SST are relatively small and this indicates that the influence of cross-equatorial OHT is less significant in the evolution of SST over SIO. Srujan et al.37 reported large TE values on the order of \(10^9\) between mean sea level pressure and Rossby-filtered outgoing longwave radiation, whereas our analysis shows significantly smaller TE values between OHT and SST. There is evidence of a non-zero causal relationship between OHT and SST; however, the values are extremely small, making them insufficient to interpret as a meaningful relationship. For the calculation of OHT, we are integrating ocean temperature and meridional velocity over the volume of the entire IO, and it gives apparently high values of 1 to 2 PW. However, at individual locations, the OHT value might be too small to make significant changes.

The opposing relationship between OHT and SST contradicts the role of OHT in modulation of the heat balance and the regulation of amplitude of SST as proposed by previous studies15,16,21. This further indicates that the interannual variability in ISM is not regulated by the ocean-atmosphere coupling over the IO as envisaged by Webster et al.15. Adding to this Lutsko et al.23 examined the relationship between OHT and the monsoon within the same season, rather than focusing on interannual variability. Their idealized simulations showed that an increase in OHT led to a decrease in SST south of the continent during summer. This, in turn, increased the moist static energy over the continent, enhancing the land-ocean thermal contrast and ultimately resulting in stronger precipitation within the same year. These findings highlight the crucial role of strong southward OHT in creating and maintaining conditions conducive to a robust monsoon, but not to the interannual variability. The discrepancy between these model-based findings and our observational results may stem from several factors. Idealized or slab ocean models often exaggerate certain feedbacks due to their simplified physical representations. Model biases, especially in the simulation of SST–OHT interactions or vertical ocean structure, may further complicate the interpretation. Additionally, spatial averaging or coarse resolution can obscure key regional processes that are more accurately captured in observational datasets.

To understand how Hadley cell circulation responds to cross-equatorial OHT, we have applied similar linear regression of OHT time series into the zonal mean meridional stream function anomalies for different seasons in year 0 and year +1(Fig. 7). Hadley cell can be explained by the zonal mean of meridional streamfunction. The less statistically significant regression coefficients between OHT and streamfunction in all seasons indicates OHT influence is not conveyed into the mean circulation over the IO. This also leads to the assumption that OHT is not regulating the interannual variability of ISM.

Fig. 4
figure 4

The linear regression of JJAS mean OHT averaged over 5°S−5°N on SST anomalies averaged for (a) SON, (b) DJF, (c) (MAM(+1) and (d) JJA(+1) from 1998-2018. The stippling indicates statistically significant regression coefficients at 95% confidence level as revealed by a two-tailed Student’s t-test. To maintain consistency with our interpretation and to avoid confusion regarding the sign of heat transport, we have multiplied all regression coefficients by –1.

Full size image
Fig. 5
figure 5

The linear regression of JJAS mean OHT averaged over 5°S−5°N on precipitation and 850 hPa wind anomalies averaged for (a) SON, (b) DJF, (c) MAM(+1) and (d) JJA(+1) from 1998-2018. The stippling indicates statistically significant regression coefficients at 95% confidence level as revealed by a two-tailed Student’s t-test. To maintain consistency with our interpretation and to avoid confusion regarding the sign of heat transport, we have multiplied all regression coefficients by –1.

Full size image
Fig. 6
figure 6

The Transfer Entropy between the OHT time series (averaged over 5°S−5°N) and the SST across the southern Indian Ocean (40°S–EQ, 40°E–120°E) at varying time lags from t0 to t+24.

Full size image
Fig. 7
figure 7

The linear regression of JJAS mean OHT averaged over 5°S−5°N on zonal mean meridional streamfunction anomalies averaged for (a) SON, (b) DJF, (c) MAM(+1) and (d) JJA(+1) from 1998-2018. The stippling indicates statistically significant regression coefficients at 95% confidence level as revealed by a two-tailed Student’s t-test.

Full size image

The boreal summer OHT during drought and flood years of ISM

Although our results show that OHT does not play any role in the interannual variability in ISM, it is interesting to examine the differences in the OHT in contrasting monsoon seasons. Here we identify eight cases of flood and drought monsoon years during the 1980 – 2018 period and do a difference of flood and drought composites from the normal JJAS mean OHT composite. The flood (drought) years of ISM are identified as the years in which JJAS mean precipitation over central India exceeds (falls below) 10% of climatological mean precipitation38 (Fig. 8). The remaining years are considered as normal monsoon years. The composite mean time series of JJAS mean OHT over the IO for normal, flood, and drought years looks similar (Fig. 9a). The plots showing the differences of flood and drought composites from the normal composite are shown in Fig. 9b. This analysis also suggests that the difference in OHT between flood and drought years is weak. The maximum magnitude of the OHT is about 1.5 PW, whereas the maximum difference in the OHT between normal and drought years is about 0.06 PW. Similar differences are seen between normal and flood years as well. The difference in OHT between flood and drought years averaged between equator and 30°S is found to be about 0.012 PW, which is about 1.7% of the OHT in normal monsoon years. Thus, it is clear that the OHT difference between two extremes of ISM is much smaller in magnitude. While these analyses show that the OHT does not play a role in the interannual variability of ISM, the muted difference in OHT between flood and drought monsoons is an interesting phenomenon that warrants further investigation. The small OHT differences observed in flood/drought composites may be due to the dominant influence of ENSO on ISM variability, which could mask subtler oceanic signals39,40. Additionally, compensating atmospheric feedbacks, such as changes in wind and cloud patterns, may dampen responses to OHT anomalies41.

The cross-equatorial OHT in the IO during summer is an Ekman response to the strong southwesterly monsoon winds. Therefore, the changes in the low-level monsoon flow can help us understand the OHT variability. Flood, drought and normal composites for monsoon low-level winds over the IO are shown in Fig. 10. The strong low-level flow over the Arabian Sea, the monsoon low-level jet, is a characteristic feature of ISM (Fig. 10a). The maximum wind speed in the normal composite is about 18 m s−1. This strong low-level southwesterly flow is responsible for the cross-equatorial OHT in the IO. The difference in the low-level winds between the drought and normal years shows a weakening of the wind speeds over the Arabian Sea by about 1 m s−1 (Fig. 10 b). In flood years, the southwesterly monsoon flow is stronger over the Arabian Sea, Bay of Bengal, and continental India (Fig. 10c).

The maximum decrease (increase) in the wind speed in the flood (drought) composite is about 1.2 m s−1 only (Fig. 10b and c). These analyses suggest that the changes in the seasonal mean wind speed in the anomalous monsoon years are about 6–7% of the mean wind speed in normal monsoon years. Also, the weakening of the low-level southwesterly flow in the drought years is confined to the southern part of the Arabian Sea and the eastern part of the SIO close to the equator. The eastward component of the flow over the Bay of Bengal appears to be strengthened in the drought year composite. In the flood composite, the low-level monsoon circulation is strengthened over the Arabian Sea and the Bay of Bengal. However, the flow weakened over a section of the IO over 65°E–80°E and 5°S−5°N. The analysis reveals that wind speed variability is much smaller than precipitation variability. Between 1980 and 2018, the seasonal mean precipitation anomalies over central India and the 850 hPa wind speed anomalies over the Arabian Sea exhibit a correlation of 0.47. This moderate correlation suggests that the majority of precipitation variability arises from factors other than atmospheric circulation. The relatively weak interannual variability in the low-level monsoon flow aligns with the similarly subdued variability in OHT. In addition, the opposing changes in the low-level wind pattern in different parts of the IO can induce contradictory changes in the OHT, which can be the reason for a muted change in the OHT between drought and flood monsoons. The drought composite shows an enhanced low-level flow from the northern Arabian Sea and West Asia towards continental India. This is particularly noteworthy, as this pattern resembles the dry air advection suggested by42. The decreased moisture transport coupled with the dry air intrusion can lead to large deficits in ISM rainfall over continental India. The dry air intrusion over continental India during the summer monsoon is also suggested to be a contributor to the interannual variability of rainfall over Northwest India43.

Furthermore, we also examined the flood and drought composites of the Hadley circulation, along with their differences (Fig. 11). The analysis revealed no substantial differences between the flood and drought composites of the Hadley cell. The interannual variability in precipitation does not reflect in the Hadley cell or zonal mean meridional stream function. Similar to OHT, these results reveal that Hadley cell circulation does not play a significant role in the interannual variability of ISM.

Fig. 8
figure 8

JJAS mean rainfall departure (in percentage) over central India (75°E–85°E and 19°N–26°N).

Full size image
Fig. 9
figure 9

(a) JJAS mean OHT composite of flood, drought and normal monsoon years and (b) difference of flood and drought OHT composite from normal.

Full size image
Fig. 10
figure 10

The JJAS mean composite of 850 hPa wind speed composite, for (a) normal years, and (b) difference between drought and normal years, and (c) difference between flood and normal years. The stippling indicates statistically significant regression coefficients at 95% confidence level as revealed by a two-tailed Student’s t-test.

Full size image
Fig. 11
figure 11

(a–b) Drought and flood composites of zonal mean meridional streamfunction, (b) Difference between drought and flood composites.

Full size image

Relationship between ENSO and ISM of the following summer

Given the subdued influence of cross-equatorial OHT on ISM variability, we proceed to examine ENSO, which has long been recognized as a key factor modulating interannual variations in the monsoon. The effect of the ENSO on ISM rainfall variability has been widely studied44,45,46. Most of the literature on the influence of ENSO on ISM examined the ENSO-monsoon relationship in the same calendar year. In line with the OHT analysis, we also performed a regression using the Niño3.4 index following a similar approach to explore its relationship with the monsoon rainfall in the following year. To understand the influence of ENSO on the SST over the IO, we performed a linear regression of DJF mean Niño3.4 index on the following MAM and JJA mean SSTs (year+1) (Fig. 12 a and b). The positive regression patterns over the NIO indicate that a warm (cold) ENSO can lead to warm (cold) SST anomalies over the NIO in the following spring and summer seasons. The regression patterns over the SIO are rather weak, with patches of both positive and negative regression coefficients. This SST response is favourable for a stronger (weaker) ISM following a warm (cold) ENSO event.

The JJA mean wind vectors show a weak relationship with the Niño3.4 SST anomalies of the previous winter, while the precipitation shows a positive correlation with it over most of continental India (Fig. 12c and d). Nevertheless, the regression patterns of precipitation are statistically insignificant over continental India. This indicates that the effect of ENSO in the year 0 on the summer monsoon of year+1 is rather weak. This is consistent with the previous studies that showed weak positive correlations between ENSO and ISM at lead/lag of one year and a strong negative correlation in the same calendar year47. This suggests that even the effect of Niño3.4 SST anomalies in the winter of year 0 on the ISM of year+1 diminishes substantially. Thus, a simple explanation of the interannual fluctuations of ISM rainfall in the range of ± 10% through the ocean-atmosphere coupling is difficult. We did not examine the complicated mechanisms, such as the tropospheric biennial oscillation, which might be better suited to explain the interannual variability of ISM28.

Fig. 12
figure 12

(a–b)The linear regression of DJF mean Niño3.4 index (year 0) on SST anomalies averaged for (a) MAM (+1), (b) JJA (+1), and c–d) same for precipitation and 850 hPa wind from 1998-2018. The stippling indicates statistically significant regression coefficients at 95% confidence level as revealed by a two-tailed Student’s t-test.

Full size image

Discussion

The interannual variability in ISM rainfall shows a quasi-biennial nature, with more than two consecutive droughts or floods a rare occurrence. This led to a claim that the ISM is a self-regulated coupled ocean-atmosphere system, in which the cross-equatorial OHT over the IO by the ISM winds is assumed to have a crucial role15,21. A long ocean “memory” is needed to support this mechanism of self-regulated monsoon interannual variability, as the warm/cold oceanic anomalies must persist for one year. The ability of OHT to modulate the heat balance of IO is important for achieving this self regulation. However, a direct analysis of the relationship between the summer OHT, oceanic temperature anomalies in the following seasons, and the subsequent ISM was lacking in the literature. Hence the main focus of this paper is confined to testing the hypothesis proposed by Webster et al.15 and Chirokova and Webster21 using reanalysis data.

Here, we computed the OHT for all summer monsoon seasons in the 1980–2018 period. Furthermore, we examined the influence of this summer OHT on the following seasons’ SST, low-level wind vectors, and precipitation. The focus of this study is understanding how OHT in year 0 affects the ISM in year +1. These analyses revealed that the effect of OHT on SST and other meteorological fields does not last beyond the following winter season. Moreover, the low transfer entropy values between OHT and SST further confirm the absence of a strong or sustained causal influence of cross-equatorial OHT on SST evolution over the SIO. Thus, the OHT variability cannot contribute to the interannual variability of ISM. The difference in seasonal mean precipitation over central India between drought and flood monsoons is approximately 20%, whereas the corresponding difference in OHT is less than 2%. Further analysis shows that the difference in low-level monsoon wind speed between flood and drought monsoons is also much smaller than the precipitation difference. This aligns with the weaker interannual variability in OHT, which is primarily governed by Ekman transport. Therefore, we conclude that interannual variability in ISM rainfall is not a self-regulated mechanism through the ocean-atmosphere coupling over the IO. Nevertheless, we acknowledge, as a caveat, that while the weak regression coefficients generally indicate a weak relationship between OHT and ISM precipitation on the interannual scale, it does not fully capture the complex non-linear nature of interactions between climate variables. The OHT may still have an indirect effect on ISM rainfall through the modification of other processes that in turn may affect ISM.

The roles of ENSO in the interannual variability of ISM have been investigated. Our analyses revealed that the ENSO significantly influences the evolution of SST anomalies over the NIO in the following spring and summer. However, this strong correlation does not imply causality, especially in the backdrop of a weak relation between the Niño3.4 SST anomalies of the year 0 winter and the ISM circulation and precipitation in year+1. This is particularly relevant in light of the Indo-Western Pacific Ocean Capacitor (IPOC) mechanism proposed by Xie et al.48, which highlights a delayed ENSO influence on the southwest Indian Ocean (SWIO) through oceanic Rossby wave propagation and associated SST anomalies. While our regression patterns show a warming signal over the SWIO consistent with this framework, the associated atmospheric response over the Indian region, especially in terms of rainfall and low-level winds is weak, or inconsistent. Similarly, it is worth noting that Izumo et al.49 showed that ENSO-related anomalies during the preceding winter can influence monsoon rainfall along the west coast of India, particularly through mechanisms involving western Arabian Sea upwelling. This suggests that although ENSO-related SST memory in the IO is present, its capacity to modulate the subsequent summer monsoon may be limited or masked by competing internal variability. It also highlights that the ENSO–monsoon teleconnection may exhibit spatial heterogeneity, with stronger lagged effects over specific regions. Therefore, it remains difficult to attribute interannual ISM rainfall fluctuations in the ±10% range solely to ocean–atmosphere coupling, underscoring the likely importance of internal atmospheric processes such as low-pressure system activity. Webster et al.15 explicitly studied the role of the IOD in the self-regulation of the ISM. However, in our regression coefficients of SST during SON and DJF (Fig. 4), there is no statistically significant IOD-like pattern evident, which is notable given that IOD typically peaks in the post-monsoon season. We did not explore the IOD-monsoon relationship in the same way as the ENSO-monsoon, because ENSO tends to influence the monsoon starting from June, while the IOD’s impact is generally confined to the second half of the monsoon. The influence of the IOD, in conjunction with OHT, can be addressed in a separate future study.

Apart from external drivers like ENSO, internal monsoonal processes, such as low-pressure system activity, also play a role in the interannual variability of the ISM. Krishnan et al.50 highlighted that monsoon LPS events, when coinciding with a positive IOD, enhance monsoon activity over the Indian subcontinent. As LPS are a component of sub-seasonal variability, understanding how sub-seasonal variability impacts the interannual variability of ISM remains an important area for future research. This study specifically examines the impact of OHT on the interannual variability of ISM, while the influence of sub-seasonal variability, such as LPS activity, warrants further exploration in future research. Our analysis of low-level monsoon circulation suggests an increased advection from the northern Arabian Sea and the West Asian desert region towards continental India. This might be another contributor to the ISM rainfall variability as suggested by earlier studies42,43. Thus, both internal and external factors contribute to the interannual variability of ISM rainfall, while the cross-equatorial OHT, as argued earlier, is not one of those factors.

Methods

Data

This study uses European Centre for Medium range Weather Forecasting (ECMWF) Ocean Reanalysis System 5 (ORAS5)51 to analyze the OHT. Compared to the predecessor ORAS4, significant improvements have been made to the ocean model resolution of ORAS5. Specifically, the horizontal resolution has been increased to an eddy-permitting resolution of 0.25 degrees, allowing for a more detailed representation of oceanic eddies and smaller-scale features. This enhancement enables a more accurate depiction of the ocean’s dynamics and improves the model’s ability to capture fine-scale processes. Furthermore, ORAS5 incorporates a near-surface resolution of 1 meter in the vertical dimension. This refinement enables a more precise representation of the upper layers of the ocean, which are crucial for understanding air-sea interactions and capturing the intricate processes occurring near the ocean surface. Additionally, ORAS5 features a prognostic thermodynamic-dynamic sea-ice model. Potential temperature and meridional velocity are used for OHT calculation from ORAS5. The study period spans from 1980 to 2018.

The atmospheric variables such as zonal and meridional wind components from the fifth generation ECMWF reanalysis (ERA5) are used52. NCEP-NCAR reanalysis data used for the calculation of the zonal mean meridional stream function53.

Two different datasets are used for precipitation. Tropical Rainfall Measuring Mission (TRMM) (0.25x0.25, 1998-2018) daily data was used to analyze the rainfall over the IO54 and Indian Meteorological Department(IMD) New High Spatial Resolution (0.25x0.25 degree) daily gridded rainfall data (1980-2018) for the central India for the calculation drought-flood index55.

Methodology

In our study, the method used to compute OHT over the IO (40°E–120°E and 40°S–30°N) at different latitudes involves integrating the product of meridional velocity and potential temperature over the depth up to 2000 m across the basin. This integration provides an estimate of the heat transport in the north-south direction (meridional) at each latitude. It is important to note that this calculation does not uphold mass conservation due to the influence of volume transport through the Indonesian Throughflow, and it is referenced to a baseline temperature of 0 °C56. Nevertheless, our analysis primarily focuses on anomalies relative to the monthly climatology. This approach reduces the dependence on the arbitrarily chosen reference temperature, as highlighted in the work of Trenberth and Zhang57 and Zhang et al.58.

$$\begin{aligned} OHT = \rho Cp \int \int (\theta - \theta _t) * v dx dz \end{aligned}$$
(1)

where \(\rho\) is density, \(\theta\) is potential temperature, \(\theta _t\) is reference temperature, and v is meridional velocity. The reference temperature is considered zero.

The statistical significance of the analysis was determined by applying a two-tailed Student’s t-test. The TE, a metric used to quantify the information transfer between two variables59,60 is computed between the OHT time series (averaged over 5°S–5°N) and the SST across the SIO (40°S–EQ, 40°E–120°E) at varying time lags.

The zonal mean meridional streamfunction for the Hadley Cell can be expressed as:

$$\begin{aligned} \psi (\phi , p) = 2 \pi a \cos (\phi ) \int _p^{p_s} \frac{v(\phi , p')}{g} \, dp', \end{aligned}$$
(2)

where:

$$\begin{aligned} \psi (\phi , p:\text {Zonal mean meridional streamfunction (kg/s)} \\ a:\text {Earths radius } (6.371 \times 10^6 \, \text {m}) \\ \phi :\text {Latitude} \\ p:\text {Pressure level (Pa)} \\ p_s:\text {Surface pressure (Pa)} \\ v(\phi , p'):\text {Meridional wind (m/s)} \\ g:\text {Gravitational acceleration } (9.81 \, \text {m/s}^2). \end{aligned}$$