Contrasting marine and terrestrial responses of South Asian summer monsoon system caused by hemispheric insolation

Significance

On orbital timescales, terrestrial rainfall and Arabian Sea upwelling records of South Asian summer monsoon (SASM) evolve largely out-of-phase, with the former and latter in-phase with Northern Hemisphere (NH) and Southern Hemisphere (SH) summer insolation, respectively. This leaves a long-standing debate on whether SASM is driven by NH or SH insolation. Given that insolation varies globally, resolving this debate has remained particularly challenging. Here, we isolate the NH and SH insolation in transient climate simulations, showing that, while SASM rainfall is driven by NH insolation, upwelling is controlled by SH insolation through its influence on South Africa. Our finding reconciles the debate by revealing that hemispheric insolation forcing drives the marine and terrestrial responses of SASM through distinct mechanisms.

Abstract

South Asian summer monsoon (SASM) delivers substantial rains to the Indian subcontinent and drives strong upwelling in the Arabian Sea, making marine upwelling records and terrestrial rainfall records two primary proxies for reconstructing past SASM variability. However, on orbital timescales, these two sets of records vary largely in opposite directions: the upwelling records are in-phase with Southern Hemisphere (SH) summer insolation, whereas the rainfall records are in-phase with Northern Hemisphere (NH) summer insolation. This leaves a long-standing debate on whether SASM is driven by NH or SH insolation. Here, combining paleoclimate records with transient climate simulations that explicitly separate the effects of the NH and SH insolation forcing, we show that the SASM rainfall is dominated by the NH insolation, whereas the Arabian Sea upwelling is forced predominantly by the SH insolation. When boreal summer occurs at perihelion, insolation is strongly enhanced not only in the NH but also in the tropical-subtropical SH. The former enhances the SASM rainfall through Eurasian warming, while the latter weakens the Arabian Sea upwelling by inducing South African warming and subsequent atmospheric teleconnections over the Indian Ocean. Our study reconciles the long-standing debate, and more broadly, reveals that warming in South Africa could exert a significant and previously overlooked remote forcing on the SASM system in past and future climate changes.

The South Asian summer monsoon (SASM), a major component of the global monsoon system, is the prime source of freshwater for the Indian subcontinent (1). In addition, monsoon dynamics profoundly affect marine ecosystems across the northern Indian Ocean where fisheries are closely tied to primary production (2). During boreal summer, increased insolation initiates SASM wind and rainfall, characterized by heavy rainfall over the continent and robust southwesterly winds. The southwesterly monsoon winds drive large-scale upwelling and phytoplankton blooms along the western Arabian Sea (3), making the Arabian Sea one of the most productive marine ecosystems among the tropical oceans (4). This upwelling-related productivity has been used as a marine proxy to infer SASM variability in past climate changes (5, 6). Interestingly, however, while terrestrial proxies for monsoon rainfall, notably cave speleothem oxygen isotope ratios ($\delta$¹⁸O_c), appears to respond directly to the Northern Hemisphere summer insolation (NHSI; June–July–August) on precession timescales (7–9), marine upwelling proxies appear largely out-of-phase with the NHSI, and responding directly to the Southern Hemisphere summer insolation (SHSI; December–January–February) (10–15). This observational contrasting behavior between marine and terrestrial proxies has remained a great puzzle to the monsoon community, leading to a fundamental question on the response mechanism of the SASM: Is the SASM variability driven primarily by the insolation from the Northern Hemisphere (NH) or Southern Hemisphere (SH) (10, 16–18)?

Traditional monsoon studies have suggested that the SASM rainfall is enhanced by increased local NHSI via enhanced land-sea thermal contrast on precession timescales (19–21). This NH forcing has been further hypothesized to also explain the contrasting marine upwelling signal through a wind-shift mechanism: Increased NHSI shifts the southwesterly winds northward, contracting the region of positive Ekman pumping in the western Arabian Sea and thereby weakening upwelling (22–24). However, this mechanism remains speculative, as previous simulations applied insolation forcing globally and therefore cannot isolate the specific contribution from the NH. Alternatively, the direct relationship of marine upwelling proxies to SHSI has prompted the hypothesis that SH insolation can drive the SASM through a remote latent heat mechanism (10): increased SHSI warms the subtropical South Indian Ocean; the resultant latent heat is then exported across the equator onto the South Asia by the monsoon wind in the subsequent boreal summer, strengthening SASM monsoon rainfall, wind, and coastal upwelling. However, this mechanism predicts in-phase variability of terrestrial and marine proxies, failing to account for their largely out-of-phase relation in observations. Thus, neither existing hypothesis fully explains the fundamental discrepancy in the proxy records. As such, the mechanism for the contrasting responses of terrestrial and marine proxies still remains outstanding.

Here, combining paleoclimate records with transient climate simulations that explicitly separate the effects of the NH and SH insolation forcing, we reveal a mechanism that SH insolation controls the Arabian Sea marine upwelling through the South African warming and atmospheric teleconnection, leading to the upwelling varying opposite to the monsoon rainfall forced by NH insolation.

Results

Model-data Comparison of SASM Terrestrial Rainfall and Marine Upwelling.

To understand the orbital response of the SASM system, we simulate the climate evolution during the past 150,000 y using the isotope-enabled Community Earth System Model (iCESM1.3) (detailed in Materials and Methods). The model reproduces the major features of the SASM climatology during boreal summer, including the substantial terrestrial rainfall, lower precipitation oxygen isotope ratios ($\delta$¹⁸O_p), strong southwesterly winds, and strong Arabian Sea upwelling (see discussion in Materials and Methods and SI Appendix, Fig. S1). The apparent correspondence between model upwelling and observational productivity (23, 25) in the western Arabian Sea (outlined by red box in SI Appendix, Fig. S1D) suggests that the simulated upwelling in this region can be treated as the marine productivity proxies in observations. This is further confirmed in simulated water ideal age, which measures the water ventilation time from the ocean surface (SI Appendix, Fig. S1E; see Materials and Methods). As expected, water ideal age is old near the surface of western Arabian Sea due to strong upwelling, indicating supply of old and nutrient subsurface waters. The monsoon climatology suggests that over South Asia, the terrestrial rainfall, and Arabian Sea upwelling should vary in-phase, which has established both terrestrial rainfall and marine upwelling records as fundamental archives for reconstructing past SASM variability.

The contrasting terrestrial rainfall and marine upwelling responses indicated by proxy records over the past 150,000 y are well simulated by the orbital forcing experiment (ORB, detailed in Materials and Methods, and Fig. 1A). In particular, NHSI maxima (boreal summer occurs at perihelion) is associated with increased rainfall and depleted $\delta$¹⁸O_c over the South Asia continent (Fig. 1 A3 and A4), consistent with terrestrial cave speleothem records (8, 9, 26, 27) (Fig. 1 A2). The simulated $\delta$¹⁸O_c and rainfall is in the sense of “amount effect”, with the regional averages highly correlated to NHSI. In contrast, the Arabian Sea upwelling weakens along with younger water age during NHSI maxima, with a negative correlation with NHSI (Fig. 1 A6 and A7), also consistent with marine productivity records in the western Arabian Sea (Fig. 1 A5). This Arabian Sea upwelling variability, nevertheless, appears in phase with the SHSI, because the SHSI is out-of-phase with NHSI (red and blue lines in Fig. 1 A1). Notably, the observed inverse relationship between terrestrial $\delta$¹⁸O_c records and Arabian Sea upwelling records is most clearly expressed during intervals of large precession (i.e., 150 to 60 ka BP). During weak precession intervals (i.e., 60 to 0 ka BP), the relation appears unstable, which could be contributed partly by the obliquity signals and partly by the uncertainties of chronology and time resolution of the proxies.

Fig. 1.

Simulated and observed SASM climate on precession timescales. (A1) Precession (gray; Y-reverse), 40°N June insolation anomaly relative to the 150,000-y mean (or NHSI; red; W/m²), and 40°S December insolation anomaly relative to the 150,000-y mean (or SHSI; blue; W/m²). The insolation is calculated from Berger (48). Higher -precession index (or lower precession index) corresponds to higher precession-band filtered NHSI and lower SHSI. (A2) Time series of observed cave speleothem δ¹⁸O_c records composite from Bittoo (7), Tianmen (71), and Xiaobailong caves (8). (A3) Time series of simulated δ¹⁸O_c in SASM region as denoted by the red box in Fig. 1B. (A4) Time series of simulated terrestrial SASM rainfall. (A5) Time series of upwelling-induced productivity records (stacked records following principal components analysis, hereafter referred to as the summer-monsoon factor) in the western Arabian Sea (10). (A6) Time series of simulated Arabian Sea upwelling averaged over the red box in Fig. 1C. (A7) Time series of simulated water ideal age averaged over the red box in Fig. 1C. The r values denote the correlation coefficient between SASM variables and -precession index. As more negative δ¹⁸O_c value corresponds to intensified monsoon, we calculate correlation coefficient between -δ¹⁸O_c value and –precession index. (B) Regressed coefficients of boreal summer rainfall (shading), precipitation-weighted δ¹⁸O (δ¹⁸O_p; contours; dashed for negative), and 850 hPa wind (vectors) from ORB experiment onto the normalized –precession index. (C) As in (B), but for boreal summer ocean upwelling averaged over 0-100 m depth from ORB experiment. In (B) and (C), the circles denote the observation sites, and all variables are conducted by a precession-band filter before regression.

The contrasting evolution between terrestrial $\delta$¹⁸O_c records and Arabian Sea upwelling records are dominated by precessional signal, as is more clearly shown in the precessional band-passed response, with much stronger correlations with NHSI relative to their original time series (SI Appendix, Fig. S2). Moreover, on precession timescales, the observed inverse relationship between terrestrial $\delta$¹⁸O_c records and Arabian Sea upwelling records becomes substantially more stable across the past 150,000 y. This confirms that the seemingly unstable relationship in their original evolution is influenced by obliquity signals, as discussed later. The spatial pattern of contrasting responses between terrestrial rainfall and marine upwelling is evident in Fig. 1 B and C, where increased terrestrial rainfall is accompanied by weakened Arabian Sea upwelling. This weakened upwelling is further shown in the cross-section of ocean response, which show overall weakened upwelling and younger water age in the western Arabian Sea, extending to greater depths (SI Appendix, Fig. S3 A and B).

Overall, our ORB simulation captures the largely contrasting features of the terrestrial rainfall and marine upwelling during the past 150,000 y. These features are also simulated when orbital parameters, greenhouse gases, and ice sheet forcings are all changed in the all-forcing experiment (ORB + GHG + ICE) (detailed in Materials and Methods and SI Appendix, Fig. S4 A and B), indicating the dominant role of insolation in forcing the responses of the terrestrial and marine records.

Previous modeling studies have attributed the contrasting terrestrial and marine responses to NHSI (22, 23). They propose that increased NHSI strengthens SASM rainfall by amplifying the land-sea thermal contrast over Asia, while concurrently weakens Arabian Sea upwelling by shifting the monsoon circulation northward, which contracts the region of Ekman pumping upwelling. However, the SH insolation is also known to modulate the SASM (10, 28). Therefore, in standard orbital forcing simulations in previous studies (e.g., ORB), insolation varies concurrently in both hemispheres, making it impossible to isolate the unique contribution of NH forcing unambiguously. Thus, whether the contrasting response of terrestrial rainfall and marine upwelling is dominated by NH insolation, SH insolation, or both, is far from clear. To unravel the fundamental cause, novel sensitive experiments that explicitly separate the NH from SH insolation are required.

Separating SASM Response to NH and SH Insolation Forcing.

Here, we employ an approach to separate the NH and SH insolation explicitly in two transient simulations (Materials and Methods). These experiments are the same as the ORB experiment except that the varying insolation (or orbital parameters) is applied separately in the NH (ORB-NH experiment) and SH (ORB-SH experiment). The climate responses to the two hemispheres’ insolation forcing are largely additive (SI Appendix, Fig. S5), allowing for a linear decomposition of the SASM response (21).

Under NH insolation forcing (ORB-NH experiment), both terrestrial rainfall and marine upwelling exhibit an in-phase response to NHSI on precession timescales (solid lines in Fig. 2A). A high NHSI forces a rainfall increase and $\delta$¹⁸O_c depletion, with the rainfall and $\delta$¹⁸O_c strongly correlated with NHSI (solid lines in Fig. 2 A2 and A3). This is also indicated in the regressed pattern over the SASM region (Fig. 2B). These results are largely consistent with those from the global forcing experiment of ORB (Fig. 1 A3, A4, and B), supporting the classic NHSI forcing theory of SASM rainfall (21, 29). Meanwhile, the high NHSI leads to an enhanced upwelling in the western Arabian Sea and older water age (solid lines in Fig. 2 A4, A5, and C) that are opposite to the weakened upwelling response in the ORB simulation (Fig. 1 A6, A7, and C). This is also evident in the deeper ocean response, where strong upwelling and older water ages in the western Arabian Sea extend to depths of more than 100 m (SI Appendix, Fig. S3 C and D). Therefore, the NH insolation alone cannot explain the contrasting terrestrial rainfall and marine upwelling response shown in proxy records, contrary to the previous hypothesis that attributed it to NH forcing via wind-shift mechanism (23).

Fig. 2.

Simulated SASM climate on precession timescales under hemispheric insolation forcing. (A) Precession-band filtered time series. (A1) is for precession (gray; Y-reverse), 40°N June insolation anomaly relative to the 150,000-y mean (or NHSI; red), and 40°S December insolation anomaly relative to the 150,000-y mean (or SHSI; blue). The insolation is calculated from Berger (48). (A2) is for simulated δ¹⁸O_p. (A3) is for simulated terrestrial SASM rainfall. (A4) is for simulated Arabian Sea upwelling. (A5) is for simulated water ideal age. The solid line is for ORB-NH experiment, while the dashed line is for ORB-SH experiment. The r values denote the correlation coefficient between precession-band filtered SASM variables and -precession index. As more negative δ¹⁸O_p value corresponds to intensified monsoon, we calculate correlation coefficient between -δ¹⁸O_p value and –precession index. (B) Regressed coefficients of boreal summer rainfall (shading), δ¹⁸O_p (contours; dashed for negative), and 850 hPa wind (vectors) from ORB-NH experiment onto the normalized –precession index. (C) As in (B), but for boreal summer ocean upwelling averaged over 0-100 m depth from ORB-NH experiment. (D and E) As in (B and C), but the results are from the ORB-SH experiment. In (B–E), all variables are conducted by a precession-band filter before regression.

In contrast, under SH insolation forcing when the varying insolation is applied only in the SH (ORB-SH experiment), both terrestrial rainfall and marine upwelling vary opposite to NHSI. When boreal summer occurs at perihelion, it corresponds to reduced terrestrial rainfall, along with weakened upwelling in the western Arabian Sea (dashed lines in Fig. 2 A, D, and E). It is this strongly weakened upwelling that overwhelms the enhanced upwelling under the NH forcing (Fig. 2 A4), finally resulting in the weakened upwelling as shown in the ORB simulation (Fig. 1C). These results are further illustrated by the cross-section of the ocean response, where the western Arabian Sea exhibits strongly weakened upwelling and younger water ages (SI Appendix, Fig. S3 E and F). In comparison, SH insolation alone generates only a weak rainfall reduction over the SASM region (dashed line in Fig. 2 A3 and D), in sharp contrast to the strong increase of rainfall forced by the NH insolation alone (solid line in Fig. 2 A3 and B). This rainfall change leads to small response of $\delta$¹⁸O_p (dashed line in Fig. 2 A2). Therefore, the contrasting marine upwelling and terrestrial rainfall stem from the combined effects of both hemispheric insolation forcings. SH insolation dominates the Arabian Sea upwelling response, while NH insolation dominates the rainfall response, leading to their contrasting variability.

Mechanisms for SASM Rainfall and Upwelling Response to NH Insolation Forcing.

To elucidate the dynamic process through which NH insolation influences the SASM, we employ statistical analysis and sensitive experiments. As noted earlier, NH insolation enhances both SASM rainfall and upwelling during NHSI maxima on precession timescales. Physically, when boreal summer occurs at perihelion, increased NHSI causes continental warming over Eurasia, which increases the land-sea thermal contrast (Fig. 3A), and enhances the southwesterly monsoon winds in the Arabian Sea (Fig. 3B). This enhances moisture convergence into, and therefore rainfall over, the South Asian subcontinent (SI Appendix, Fig. S6A), as proposed in classical monsoon theory (29). Moreover, the pronounced meridional temperature gradient over North Africa drives anomalous westerlies that extend from West Africa to the Arabian Sea. These westerlies likely reinforce the southwesterly monsoon winds over the Arabian Sea, and facilitate moisture transport from North Africa toward South Asia, contributing to enhanced rainfall over the Indian subcontinent (30).

Fig. 3.

Regressed coefficients of climate variables onto the normalized –precession index. (A) Boreal summer surface air temperature (SAT; shading) and sea level pressure (SLP; contours; solid for positive and dashed for negative) from ORB-NH experiment. (B) As in (A), but for boreal summer rainfall (shading) and 850 hPa wind (vectors) from ORB-NH experiment. (C) As in (A), but for boreal summer Ekman pumping (shading) and surface wind (vectors) from ORB-NH experiment. (D–F) As in (A to C), but the results are from the ORB-SH experiment. All variables are conducted by a precession-band filter before regression. (G) LBM results: Prescribed heating anomaly at surface (shading) and 600 hPa (red contours) to mimic the South African diabatic heating induced by surface warming and rainfall, respectively. The maximum heating rate is 5 K/day. The vectors denote the changes in boreal summer 850 hPa wind (m/s) in response to heating anomalies. (H) As in (G), but for changes in boreal summer Ekman pumping (shading; cm/day) and surface wind (vector; m/s).

Meanwhile, the continental warming also migrates the low-level jet northward (SI Appendix, Fig. S7), leading to a contraction of positive wind curl region in the western Arabian Sea (SI Appendix, Fig. S7 G and I), which is consistent with previous hypothesis (22, 23). However, the near-coastal region of the western Arabian Sea, where most upwelling records are located, experiences enhanced upwelling (Fig. 2C), as the intensified southwesterlies over the western Arabian Sea generate positive wind curl and consequently Ekman pumping upwelling in this region (Fig. 3C, SI Appendix, Fig. S7 G and I, and Materials and Methods). The dynamic linkage between meridional land-sea thermal contrast and Ekman pumping over the western Arabian Sea can be seen in the scatter diagram that a stronger meridional warming contrast favors stronger monsoon circulation and Ekman pumping upwelling in the region (SI Appendix, Fig. S8). This dynamic linkage is further confirmed in a sensitivity experiment using atmospheric general circulation model (GCM) (Materials and Methods). An idealized land warming and, in turn, enhanced meridional land-sea warming contrast produces enhanced southwesterly winds over the western Arabian Sea, leading to increased Ekman pumping in the region as well as increased terrestrial rainfall (SI Appendix, Fig. S9), consistent with the coupled model results. Furthermore, the strong meridional thermal gradient over North Africa drives westerly anomalies extending from West Africa to the Arabian Sea and South Asia, which may contribute to both increased monsoon rainfall and enhanced upwelling. Together, these findings confirm that such meridional warming contrast induced by increased NHSI can drive the increased monsoon rainfall and enhanced upwelling. Although the northward shift of the low-level jet has been proposed to suppress Arabian Sea upwelling (22, 23), this shift alone cannot explain the weakened upwelling that is recorded by the proxies, given that these records are located in the near-coastal region of the western Arabian Sea. This indicates that the observed weakening of Arabian Sea upwelling must be additionally controlled by SH insolation.

Mechanisms for SASM Rainfall and Upwelling Response to SH Insolation Forcing.

To further assess how SH insolation weakens Arabian Sea upwelling and dampens terrestrial rainfall, we examine large-scale atmospheric circulation change under SH insolation forcing. Notably, on precession timescales when boreal summer occurs at perihelion, insolation increases not only in the NH but also in the tropical-subtropical SH (SI Appendix, Fig. S10A). The increased SH insolation causes continental warming over South Africa, which forces lower sea level pressure and increased local rainfall (Fig. 3 D and E). The combined surface warming and rainfall-induced latent heating release create a diabatic heating anomaly over South Africa. This heating generates anomalous easterlies at lower troposphere over the northern Indian Ocean (Fig. 3E), which attenuate the climatological southwesterly monsoon winds. This wind attenuation weakens moisture convergence into the South Asian continent, reducing terrestrial rainfall (SI Appendix, Fig. S6B). Furthermore, the wind attenuation, rather than wind shift over the western Arabian Sea, induces a negative wind curl (SI Appendix, Fig. S11), driving Ekman downwelling that suppresses the upwelling in the region (Fig. 3F).

To further substantiate this dynamical linkage, we conduct sensitivity experiments with prescribed heating in South Africa using a Linear baroclinic model (LBM) (Fig. 3G and Materials and Methods). The heating anomaly triggers a baroclinic Kelvin wave response over the tropical Indian Ocean east of the forcing region, with anomalous easterlies in the lower troposphere. This anomalous surface easterlies produces negative wind curl and, in turn, downwelling Ekman pumping north of the jet axis in the western Arabian Sea (Fig. 3H). Therefore, on precession timescales when boreal summer occurs at perihelion, increased SH insolation warms the South Africa, which induces anomalous easterlies in the northern Indian Ocean, drives Ekman downwelling and therefore the weakened upwelling in the western Arabian Sea.

Discussion

For over 30 y, it has remained a great puzzle in paleoclimatology that why terrestrial rainfall and marine upwelling proxies of the SASM appear evolving in the opposite direction on precession timescales. This land-sea paradox has further led to a longstanding debate whether SASM is driven primarily by NH or SH insolation. Here, employing transient experiments that explicitly separate the NH and SH insolation forcing, we show that this contrasting variability is the result of competing hemispheric influences. On precession timescales, when boreal summer occurs at perihelion, insolation is increased in both hemispheres. The increased NH insolation strengthens marine upwelling and terrestrial rainfall through the Eurasian warming, whereas concurrently increased SH insolation weakens marine upwelling and dampens the terrestrial rainfall through the South African warming. The observed contrasting feature is the outcome of this competition, featured by SASM rainfall governed predominantly by NH forcing, whereas Arabian Sea upwelling dominated by SH forcing. These processes are summarized schematically in Fig. 4. The contrasting responses of terrestrial rainfall and marine upwelling records on precession timescales are dynamically consistent with their in-phase responses on obliquity timescale (10) (see discussion in Materials and Methods). This obliquity modulation disrupts the pure precession-driven inverse relationship between terrestrial rainfall and Arabian Sea upwelling, contributing to their unstable relationship over the past 150,000 y.

Fig. 4.

Schematic diagram for the SASM response to insolation forcing when boreal summer occurs at perihelion. (A) ORB forcing: Increased NHSI warms the Eurasian continent, which intensifies the southwesterly SASM winds and increases terrestrial rainfall over South Asia. Moreover, the negative wind curl in the western Arabian Sea weakens marine upwelling in the region. (B) ORB-NH forcing: When insolation forcing is only applied to NH, the SASM rainfall is increased, and the positive wind curl enhances upwelling in the western Arabian Sea. (C) ORB-SH forcing: when insolation forcing is only applied to SH, the SASM rainfall is decreased, and the negative wind curl weakens upwelling in the western Arabian Sea. Consequently, the observed contrasting feature represents the outcome of this competition: monsoon rainfall is governed predominantly by NH forcing, whereas Arabian Sea upwelling is dominated by SH forcing.

The primary contribution of our study is the quantification of the hemispheric insolation effect on SASM rainfall and upwelling change. Unlike previous studies that primarily attribute SASM evolution to the NH insolation (26, 29), or SH insolation (10, 31) based on qualitative in-phase relationship between proxies and hemispheric insolation, our results provide robust evidence that the contrasting behavior of terrestrial rainfall and marine upwelling results from the distinct influence of insolation from each hemisphere. This finding advances the traditional view that the SASM system should vary coherently in response to a single mechanism. Furthermore, our results suggest that future interpretations of upwelling proxies in the western Arabian Sea should be approached with caution, as the relationship between upwelling and the monsoon varies across different timescales and is also determined by the locations of the observations.

Our results underscore the critical, yet previously overlooked, role of South African warming in influencing the SASM systems by cross-equatorial atmospheric teleconnections. Given that the variability of Arabian Sea upwelling records is consistent with latter marine records across the Indian Ocean on precession timescales (31–33), we propose that these widespread marine proxies, that vary out-of-phase to terrestrial rainfall proxies, are influenced by SH insolation via this remote pathway.

Beyond paleoclimatology, our findings have broad implications. For future global warming scenarios, climate models consistently simulate the future strengthening of SASM rainfall, along with reduction in net primary productivity (34) and nutrient inventory (35) in the Arabian Sea, as well as warming over South Africa (SI Appendix, Fig. S12). Therefore, South African warming may play a role in influencing the SASM systems, triggering the reduction of Arabian Sea productivity and threatening the fisheries in the region. Given that marine ecosystems serve as primary protein sources for the planet’s densest and most vulnerable populations of the subcontinent (2), we propose that South Africa warming in the future should be considered as a key factor in the assessments of the regional hydroclimate and marine food security across South Asia.

Materials and Methods

Transient Simulations.

The Community Earth System Model enabled with a water isotope module (iCESM1.3) is used in this study. The iCESM is developed by the National Center for Atmospheric Research (NCAR), which is a fully coupled model including atmosphere, ocean, land, sea ice, and river runoff components (36). Our model resolution is f19_g16, which is substantially higher than other models used in long-time transient simulations (37, 38). The atmosphere model (CAM5) resolutions are ~1.9° × 2.5° with 30 vertical levels (39). The land model (CLM4) shares the same horizontal grid with the atmosphere model (40). The ocean model (POP2) and sea ice model (CICE4) share the same horizontal grid of a nominal 1° horizontal resolution (41, 42). The ocean model has 60 vertical levels. The iCESM1.3 reliably reproduces key features of both modern and past climate (23, 43, 44), as well as water isotope distributions (44–46).

The 150 ky transient simulation is performed following a similar strategy as in the previous transient simulation of 300 ky by NCAR-CCSM3 (38, 47). Three transient simulations are included here. The first simulation (ORB) was forced by transient insolation change calculated from the Berger78 astronomical solution (48) starting from 170 ka BP and ending in the present day. The first 20 ky is for spin-up, and the last 150 ky is for analysis. The second (ORB + GHG) and third (ORB + GHG + ICE) simulations are branched off from 150 ky ago, with the orbital parameters starting at 150 ka BP. The ORB + GHG simulation was further forced by GHG that is retrieved from the Antarctica ice core reconstruction (49). The ORB + GHG + ICE simulation is additionally forced by the changing continental ice-sheet variations as the boundary condition of the atmosphere. The last 21 ka ice sheet conditions are based on the ICE-6G reconstruction (50) and were changed every 1,000 y in terms of ice-sheet extent and thickness. Before 21 ka BP, ice sheets were interpolated using the global sea-level reconstruction (51) [more details in Thomas et al. (38)]. All these three simulations employed a 100-fold accelerated scheme. That is, at the end of each year of the simulation, we advance the forcings by 100 y. Following Lu et al. (52), all monthly model output was converted to the “fixed-angular” calendar from the “fixed-day” calendar to remove the artificial phase shifts on the climatic responses (53). The contrasting terrestrial rainfall and marine upwelling responses of SASM are primarily forced by the orbital change, as shown by the similar results from ORB and ORB + GHG + ICE simulations.

The ORB simulation, in which insolation varies globally, cannot estimate the fundamental cause of the contrasting land-sea responses, as these may be independently influenced by insolation from each hemisphere. To address this, we conducted two sensitivity experiments (ORB-NH and ORB-SH) that separate the hemispheric insolation forcing.

In ORB-NH experiment, orbital parameters are varied only in the NH, while SH parameters are fixed at their 150 ka BP state. Comparing this with the ORB simulation isolates the pure influence of NH insolation. Conversely, in the ORB-SH experiment, orbital parameters are varied only in the SH, with NH parameters fixed at 150 ka BP, thereby isolating the SH insolation signal. The climate responses to the two hemispheric insolation forcings are largely additive, indicating a linear response at the leading order (SI Appendix, Fig. S5). This validates the decomposition of the total signal into distinct hemispheric contributions.

We note some differences in the detailed phase relationships among proxy records between our model and observations, which could be attributed to model deficiency and proxy uncertainty associated with their interpretation, chronology, and temporal resolutions (11, 54, 55). Nevertheless, the success in reproducing the gross phase relationships in the observed SASM proxies justifies the utility of our model for understanding the mechanism of these responses.

Justification of the 100-Factor Acceleration Scheme.

The 100-factor acceleration scheme is justified given that the response time of the atmosphere-upper ocean system is of the order of a decade, or about 1,000 times faster than the shortest orbital cycle. Meanwhile, the change in insolation forcing from year to year with the acceleration factor of 100 is generally less than 0.5 W/m² at all latitudes. Previous work (56) suggest that deviation of accelerated simulation compared with nonaccelerated simulation is expected to occur in deep-water formation regions and below the thermocline. The near-surface quantities such as temperature and precipitation are expected to be largely unaffected by the acceleration technique. The 100-factor acceleration scheme is further validated given that the simulated tropical Pacific sea surface temperature aligns with reconstructions (57, 58), and the simulated monsoon rainfall and water isotopes are consistent with independent proxy records (23, 44).

To assess the impact of the acceleration scheme on the SASM simulation, we compared two experiments from Liu et al. (59) conducted with the CESM1.2 model: one employing the 100-fold acceleration and one without (SI Appendix, Fig. S13). The CESM1.2 model is structurally identical to iCESM1.3 except for the exclusion of water-isotope processes. In each simulation, only the precession was changed with time, all the other forcings were held steady. Precession is arbitrarily initialized at a value of 0.034. Obliquity is fixed at 23.0°, while the eccentricity is held at a high value of 0.05 to consider the climate impact of a relatively strong precessional variation. The atmospheric CO₂ is set to 240 ppmv. The atmospheric concentrations of CH₄, and N₂O are held steady at 480 ppbv and 250 ppbv, respectively. Vegetation and continental ice sheets are prescribed to their present-day distributions. The land ice sheets and vegetation were not allowed to change.

It is clearly shown that the phase of SASM rainfall under the forcing with and without orbital acceleration scheme is identical, with the correlation coefficient of 0.9, exceeding the 95% confidence level (SI Appendix, Fig. S13). Thus, we conclude that our acceleration scheme does not affect the main results of this work.

Model Evaluation.

The 150 ka transient simulation can largely capture the SASM climatology. The rainfall is maximized in the Western Ghats and northeastern India (SI Appendix, Fig. S1 A and B), accompanied by the depletion of $\delta$¹⁸O_p along the moisture transport pathway that follows the southwesterly monsoon winds (SI Appendix, Fig. S1C). The strong southwesterly winds drive strong upwelling in the western Arabian Sea (SI Appendix, Fig. S1F), which upwells subsurface nutrients to the surface along with deeper water of older ages (SI Appendix, Fig. S1E), consistent with the observed high chlorophyll concentration (SI Appendix, Fig. S1D). The capability of iCESM in reproducing the SASM climatology makes it reasonable to study the past monsoon variability.

Model Water Ideal Age.

Ideal age is an idealized age tracer that measures the ventilation time of a water parcel since it was last at the surface. In the model, the ideal age is a passive tracer that is set to zero at the ocean surface and ages by 1 year/year in the ocean interior (60). In its climatology, the ideal age usually ages with depth, reflecting a longer ventilation time at a deeper depth away from the surface. The ideal age also contains the signal of coastal upwelling. Above about 100 m, the ideal age becomes older toward the coast in the Arabian Sea (SI Appendix, Fig. S1E), which is consistent with the strong upwelling of older water from depth (SI Appendix, Fig. S1F).

Contrasting Terrestrial and Marine Responses of SASM on Precession Timescales Is Dynamically Consistent with Their In-phase Responses on Obliquity Timescales.

Different from their contrasting response on precession timescales, terrestrial rainfall and marine upwelling records vary in-phase on obliquity timescales (10). Proxy records and our ORB simulation consistently show that high obliquity corresponds to both increased monsoon rainfall (depleted $\delta$¹⁸O_p) and intensified Arabian Sea upwelling (SI Appendix, Fig. S14). Previous work attributed this covariability to SH insolation via a latent-heat mechanism (10). However, as NHSI and SHSI are positively correlated on obliquity timescales, the globally insolation change in ORB simulation makes us unable to identify whether the in-phase response between terrestrial rainfall and marine upwelling is driven by NH or SH insolation.

Our sensitive experiments further reveal that the in-phase response between terrestrial rainfall and marine upwelling on obliquity timescales is primarily caused by NH insolation change, with SH insolation playing negligible role (SI Appendix, Fig. S15). Physically, increased NH summer insolation during high obliquity periods enhances meridional thermal contrast. This intensifies southwesterlies in the Arabian Sea, that converge moisture into the South Asian continent, resulting in terrestrial rainfall increase (SI Appendix, Fig. S16 A–C). Meanwhile, the positive wind curl in the western Arabian Sea intensifies Ekman pumping upwelling (SI Appendix, Fig. S16D). This mechanism is consistent as it is found on precession timescales. The minimal influence of SH insolation on obliquity timescales is also consistent with our precessional findings. On obliquity timescales, we demonstrate that the limited influence of SH insolation results from its relatively small change, which is largely in contrast to the substantial SH insolation increase during boreal summer on precession timescales (SI Appendix, Fig. S10). Due to its weak forcing, SH insolation induces minimal change in surface temperature over South Africa, and thereby the minimal response in Ekman pumping, and upwelling in the western Arabian Sea (SI Appendix, Figs. S16 and S17).

Therefore, the in-phase responses of terrestrial rainfall and marine upwelling on obliquity timescales are dynamically consistent and, in turn, concurrent, with their contrasting responses on precession timescales.

Atmospheric GCM Experiments.

To investigate the dynamic link between meridional thermal contrast and Ekman pumping in the western Arabian Sea, we conduct atmospheric GCM experiments using Community Atmosphere Model version 4 (CAM4), with a resolution of 3.75° $\times$ 3.75°. This atmospheric GCM allows us to isolate atmospheric processes from ocean feedbacks while maintaining computational efficiency. One is a control experiment CAM_ctrl, which is prescribed with the observed SST climatology as the default setting in CAM4. The other is albedo reduction experiments CAM_ABlow, in which we artificially lower the surface albedo by 80% over the region of 15° to 50°N, 15°W to 100°E according to previous studies and to simulate the Eurasian continent warming (61). We use CAM_ABlow minus CAM_ctrl to estimate the effect of meridional temperature contrast on Ekman pumping change. Each experiment runs for 40 y, and the last 20 y are used for equilibrium analysis.

LBM Experiments.

To examine the atmospheric process directly forced by diabatic heating response, we employ the LBM (62). LBM is a powerful tool for understanding linear atmospheric dynamics to a prescribed diabatic heating forcing as has been used extensively in previous works (62, 63, 64). It can well capture the observed teleconnection patterns while excluding nonlinear processes and oceanic feedback that complicate the interpretation in a fully coupled GCM.

In our setup, the LBM has a horizontal resolution of approximately 5.6° and 20 vertical σ-levels. According to the heating profile in South Africa (SI Appendix, Fig. S18), we impose an idealized surface heating of 5 K/day over the South African region (10°S to 30°S, 10°E to 40°E) to mimic surface warming, and a heating at 600 hPa over the tropical South African region (10°S to 0, 10°E to 30°E) to mimic the precipitation-induced latent heat release at mid-low troposphere (Fig. 3G). The wind and Ekman pumping response to these combined forcings is shown in Fig. 3 G and H. The simulations run for 30 d, and the equilibrium response was obtained by averaging the final 10 d of integration. It is worth noting that diabatic heating over other SH regions (e.g., Australia) has a negligible impact on the wind and Ekman pumping responses over the Indian Ocean and western Arabian Sea (SI Appendix, Fig. S18).

Proxies-Model Comparison.

Upwelling-based proxy.

Clemens and Prell (10) use five summer-monsoon proxies from the western Arabian Sea to create multiproxy records of summer-monsoon variability. Each proxy independently reflects monsoon-driven upwelling or wind intensity. Detailly, $\delta$¹⁵N in core RC27-61 is linked to wind-driven upwelling. Enhanced upwelling intensifies the oxygen minimum zone, increasing denitrification and consequently $\delta$¹⁵N values (10). Planktonic foraminifer Globigerina bulloides is typically reflecting the upwelling strength. High abundance of G. bulloides in the core RC27-61 indicates an intensified SASM wind that drives upwelling and therefore the enhanced productivity (11). Biogenic opal of diatoms and radiolaria in core 722B are productivity records that are also linked to monsoon-driven upwelling (65). Lithogenic grain size in core RC27-61 is linked to monsoon wind strength. Coarser particles in core RC27-61 reflect enhanced transport capacity of the southwesterly monsoon winds, which carry larger dust grains from the Somali and Arabian deserts (6). Excess Ba accumulation record in core 722B is tied to monsoon wind-driven upwelling and productivity (10). All five proxies exhibit synchronous variability, confirming their shared response to monsoon strength (10). The composite summer monsoon factor reveals a negative correlation to NHSI on precession timescales (SI Appendix, Fig. S2E), but positive correlation on obliquity timescales (SI Appendix, Fig. S14A5).

Additionally, the relative abundance of G. bulloides in core GeoB 3004 from the western Arabian Sea exhibits a good correlation to the summer monsoon factor of Clemens and Prell on precession timescales (10). Notably, the monsoon-driven upwelling intensity recorded in this core is also characterized by reduced upwelling during periods of high NHSI (14). A similar pattern is observed in core 70KL, where the Ba/Al ratio serves as a proxy for productivity and monsoonal wind strength (15). Our simulations broadly capture these features.

Rainfall-Based Proxy.

The $\delta$¹⁸O_c record presented here is the composite value from Bittoo (7), Tianmen (66), and Xiaobailong (8) caves, which shows a decreased value, or enhanced monsoon intensity, when the NHSI is high (Fig. 1 A2 and SI Appendix, Fig. S2B). This result is consistent with the speleothem $\delta$¹⁸O_c records from Hoti (67) and Qunf (68) caves spanning the Holocene, showing enhanced monsoon during the early Holocene when NHSI is high.

Model $\mathbf{\delta }$¹⁸O_p and Cave $\mathbf{\delta }$¹⁸O_c.

Annual precipitation-weighted $\delta$¹⁸O ($\delta$¹⁸O_p) is calculated as

$${\delta }^{18}{\mathrm{O}}_{\mathrm{p}}={\displaystyle \sum _{m=1}^{12}}\frac{P_m}{P}\bullet {\delta }^{18}{\mathrm{O}}_{\mathrm{m}},$$

where $m$ denotes the calendar month, $P_m$ is precipitation from month $m$, ${\delta }^{18}{\mathrm{O}}_{\mathrm{m}}$ is the $\delta$¹⁸O from month $m$, and $P$ is annual total precipitation. Calcite $\delta$¹⁸O_c in speleothems differs from $\delta$¹⁸O_p owing to the temperature dependency of fractionation between calcite and water. For a direct comparison with the observation, we derive the model $\delta$¹⁸O_c from $\delta$¹⁸O_p. First, the unit of $\delta$¹⁸O_p-SMOW is converted to water $\delta$¹⁸O_W-PDB:

$$\delta {{}_{}{}^{18}\mathrm{O}}_{\mathrm{W}-\mathrm{PDB}} = 0.97002 \times \delta {{}_{}{}^{18}\mathrm{O}}_{\mathrm{p}-\mathrm{SMOW}} - 29.98.$$

Second, ${\delta }^{18}\mathrm{O}$ _W-PDB is substituted into the temperature-dependent equilibrium equation of inorganic carbonate oxygen isotope to derive $\delta$¹⁸O_c:

$$\delta {{}_{}{}^{18}\mathrm{O}}_{\mathrm{c}} = \mathrm{\delta }{{}_{}{}^{18}\mathrm{O}}_{\mathrm{W}-\mathrm{PDB}} + 2.70 \times {10}^6/T^2 - 3.29.$$

Here, T is the absolute cave temperature (in K). In general, the cave temperature equals the annual mean surface air temperature at the cave site, and we thus use the model annual mean surface air temperature in our calculation. These calculations follow ref. 46.

Ekman Dynamics.

The monsoon wind over the Arabian Sea can drive the ocean upwelling by Ekman pumping over the open sea and coast upwelling near the coastal region. The Ekman pumping velocity is calculated as:

$$W_{\mathit{Ekman}}=\frac{1}{{\rho }_w}\nabla \times (\frac{\tau }{f})\approx \frac{1}{{\rho }_{w}f}\nabla \times \tau ,$$

where $\tau$ is surface wind stress, $\mathrm{\nabla }\times \tau$ is the wind-stress curl, ${\rho }_w$ is seawater density of 1,024 kg/m³, $f$ is the Coriolis parameter.

Coastal upwelling can be estimated qualitatively as:

$$W_{\mathit{coast}}=\frac{T}{R_d},$$

where $T=\frac{{\tau }_a}{{\rho }_{w}f}$ is the offshore Ekman transport, ${\tau }_a$ is the alongshore wind stress, and $R_d$ is the Rossby deformation radius. These calculations follow ref. 23. The Rossby deformation radius is approximately 100 km in the tropical Arabian Sea (69, 70), which corresponds to a single grid cell in the iCESM. Consequently, the coastal upwelling signal is confined to the first grid point off the Oman coast, while the broader western Arabian Sea region outlined in this work is dominated by Ekman pumping dynamics.

Supplementary Material

Appendix 01 (PDF)