Introduction

Precession exerts a strong influence on Earth’s glacial–interglacial cycles and monsoons1,2. However, past studies have largely focused only on the orbital phase when perihelion occurs close to summer or winter solstice3,4. The extent to which the precessional change during the equinox season shapes the seasonal timing of monsoons5 remains unclear. A lack of understanding of paleoclimate monsoon seasonality and regionalization6,7,8,9 may bias the attribution of paleo-monsoon evolution and cause reconstruction errors10. A related concern is how the variation in the seasonal timing of perihelion and aphelion is related to the process of deglaciation, including some of the leads and lags between insolation, monsoon proxies, ice volume, and greenhouse gases10,11,12.

Equatorial and tropical climate changes could offer insights into climate changes during spring and autumn equinoxes and possibly the semiprecessional signal (with periods of approximately 11.5 kyr)13. In the eastern equatorial Pacific, sea surface proxies during the last glacial cycle indicated a stronger upwelling along the margin region (Costa Rica Thermal Dome), related to the precession-driven insolation maxima in spring and autumn14. A semiprecessional temperature change was also recorded in stalagmite data over the northwestern side of Borneo Island (Mulu)15. Recent water isotope–enabled modeling studies suggest that autumn precipitation substantially contributed to the wetting of South China during the last deglaciation16. Another question is whether and how monsoon seasonal migration is influenced by precession during the seasonal transitions.

In spring and early summer, the Asian monsoon advancement is closely related to the heating distribution over the Asian–Australian land bridge17 and to the jet position across the Tibetan Plateau18,19,20. Paleoclimate studies have noted a critical role of the changes in the Tibetan Plateau snow cover in the Indian monsoon in spring21. During the last deglaciation, warming driven by perihelion occurring in boreal spring caused early snow melting over the Tibetan Plateau, favoring early monsoon onset22. The Sunda Shelf was exposed during the Last Glacial Maximum (LGM)23, which could amplify the contrasting land–sea responses to the orbitally induced insolation change24,25. Thus, the role of precession-induced spring warming might be more important than previously acknowledged. Climate simulations with glacial boundary conditions suggest the enhancement of premonsoon convection over the Arabian Sea and the Bay of Bengal26. This study aims to explore the effect of perihelion and aphelion over the Asian monsoon region, focusing on transition seasons. We focus on the premonsoon advancement and the conditional stationary of the oceanic monsoon and relate those changes to the tropical and midlatitude changes in response to precessional seasonality.

Results

Afro–Asian monsoon changes caused by seasonal orbital precession

Climate modeling may provide insights into the seasonal sensitivity of precession. We utilized the climate simulations produced by Chiang, et al.27, which span the space of eccentricity and longitude of perihelion (LOP). We first examined the change in insolation with the varying LOP by using the difference between eccentricity = 0.04 and eccentricity = 0.01. The insolation difference at 30°N exhibits a maximum in March–April (premonsoon) and a minimum in August–September (late–summer monsoon) when the LOP is around 180° (corresponding to perihelion at vernal equinox), whereas the opposite occurs when the LOP is around 0° (perihelion at autumnal equinox; Fig. 1a). We further explored the extent to which the monsoon pattern might be modulated by the contrasting seasonal insolation. We examined seasonal increases in the upper tropospheric geopotential height in the region (10°N–40°N) over West Asia (50°–70°E), South Asia (80°E–100°E), and East Asia (110°E–130°E). With a relatively small eccentricity (0.01), the monsoon-related geopotential heights reached a peak value in July without a considerable difference in the seasonal migration at LOP = 0° compared with LOP = 180° (blue-dashed contours, Fig. 1b–g). With a larger eccentricity (0.04), a precessional effect on that shifts the seasonal geopotential height pattern can be observed. At LOP = 180°, which corresponds to a warmer spring and a cooler autumn, the monsoon seasonality is observed to shift earlier. By contrast, delayed monsoon seasonality occurs because of a cooler spring and a warmer autumn at LOP = 0° (red contours, Fig. 1b–g).

Fig. 1: Precession-driven monsoon geopotential height changes.
figure 1

a Seasonal insolation (unit: W m−2, blue solid line for March–April and green dashed line for August–September) in precessional cycle at eccentricity of 0.04 compared with 0.01 along 30°N. b, c Annual variation in geopotential height at 200 hPa in West Asia (50°E–70°E) when precession = 180° and 0°; red-solid and blue-dashed contours denote heights of 12080 and 12280 gpm (or higher) when eccentricity is 0.04 and 0.01, respectively, and shadings denote difference in between (gray bars). d, e Same as b, c except for South Asia (80°E–100°E). f, g Same as b, c except for East Asia (110°E–130°E).

Full size image

Using the Transient Climate Evolution (TraCE) simulation, we explored the modeling of atmospheric fields over the most recent precessional cycle since 21 ka. Boreal summer insolation peaked in the early Holocene, given the larger orbital tilt and the occurrence of perihelion in boreal summer (Fig. 2). The precession-induced seasonal insolation changes can be clearly identified in the positive summer insolation and negative winter insolation anomaly at 30°N (Fig. 2). Compared with the insolation at 21 ka (similar to present-day), the Asian monsoon region experienced a warmer spring at around 17 ka and autumn at ~6 ka.

Fig. 2: Seasonal insolation change at 21–0 ka.
figure 2

a Annual variation in global insolation along 30°N at 21–0 ka (shaded, unit: W m−2); contours denote that insolation at 21 ka is subtracted (unit is % compared with conditions at 21 ka). b Same as a except for seasonal mean condition averaged in March–April (blue line), May–July (orange line), and August–September (dashed green line).

Full size image

Changes in the midlatitude westerly jet stream (since 21 ka)

The northward migration of the westerly jet stream has been previously hypothesized to control Asian seasonality28. Following this, we examined the seasonal variation in strength and latitudinal position of the westerlies from 21 ka to the present. We divided this period into two intervals centered on the early Holocene to distinguish between conditions with and without a substantial ice volume influence (Fig. 3). In winter, the jet stream is strong and located near 30°N in the Asia–Pacific region (50°E–150°E), whereas it is relatively weak and shifts poleward by ~10° in summer. Solar heating affects the westerlies through mechanisms associated with the land–sea configuration, topographic effects, and monsoon precipitation. Thus, we consider orbital forcing on the westerly jet as a midlatitude metric for paleoclimate seasonality changes28,29.

Fig. 3: Leading EOF results of annual upper-level zonal winds.
figure 3

a EOF1 and (b) EOF2 of annual zonal winds averaged at 200–500 hPa and 50°E–150°E (unit: m s−1) at 21–11 ka, and (c) corresponding PC1 (green dotted line) and PC2 (dashed blue line); contours denote wind field at 19 ka, and dashed black line approximately denotes position of westerly jet core. d–f Same as a–c except for data at 11–0 ka and contour wind field at 9 ka. S and W denote strengthening and weakening of westerly jet in seasons during period marked by same-colored vectors. Percentage of total variance explained by EOF is also provided. g Seasonal regional average of zonal winds at 21–0 ka [blue curve for March–April in 50°E–150°E, 15°N–30°N, as a box marked in b; green curve for August–September in 50°E–150°E, 25°N–40°N, as a box marked in e].

Full size image

The first empirical orthogonal function (EOF) mode of the upper-level westerlies (Fig. 3) shows that the summer jet stream weakened before the Holocene, with a rapid decrease in magnitude at 17–15 ka, after which it gradually strengthened during the Holocene. As expected, the EOF1 and its principal component (PC1) in the westerly jet reflects the summer insolation change. Given that the second EOF is orthogonal to the first, it should provide information for the transition seasons (Fig. 2b and Fig. 3c, f). EOF2 indicates that the jet stream between 17 and 15 ka is anomalously poleward before summer and equatorward after summer; this indicates an earlier south-to-north transition in response to a warmer spring and vice versa in response to a cooler autumn (Fig. 3b, c). Compared with the considerable change at 17–15 ka, the PC1 is relatively unchanged at 19–17 ka, indicating that the change to the westerlies during this period is dominated by EOF2. For the mid-to-late Holocene, the second EOF mode suggests a weakening of the spring jet and a strengthening and equatorward shifting of the jet in late summer and autumn (Fig. 3e, f).

The explained variance of EOF1 from the LGM to the early Holocene (65%) is smaller than that during the Holocene (76%). Thus, the precessional contribution to the transition seasons could be more important before the Holocene (28% explained by EOF2). The seasonal change in the jet evolution supports this speculation, including the shift in the spring jet (weakening in 15°N–30°N) in the early part of the last deglaciation and the evolution turning across the mid-Holocene (Fig. 3g). The EOF2 of global spring precipitation and the upper tropospheric divergent circulation support this contention (Supplementary Fig. 1). From 19 to 17 ka (or later), the upper tropospheric divergent circulation strengthened with an increase in precipitation over Southeast Asia and the Maritime Continent. This suggests a northward advancement of convective activity and implies early monsoon activity. By contrast, tropical convection is suppressed over the equatorial central Pacific, the Indian Ocean, and Africa, suggesting an overall weakness of the tropical Hadley circulation (Supplementary Fig. 1).

Monsoon transitions

In spring, the upper tropospheric westerly jet stream is located over the southern edge of the Tibetan Plateau and exerts a major dynamic influence on the downstream East Asian spring rain28,30. Moreover, strong solar heating excites convective activity over the Maritime Continent and increases the divergent circulation in the upper troposphere (Fig. 4a)31. With the seasonal progression of insolation, the westerly jet stream shifts northward over the Tibetan Plateau32. Correspondingly, the upper tropospheric divergent circulation migrates northwestward from the Maritime Continent to South Asia (Fig. 4b) along the so-called Asian–Australian land bridge17,33. These premonsoon characteristics are dynamically and thermodynamically related, and the forcing on these premonsoon changes could exert substantial effects on the monsoon condition, including the heating over the Tibetan Plateau34,35.

Fig. 4: Seasonal evolution of monsoon-related circulation.
figure 4

Seasonal evolution of upper tropospheric circulation in (a) March–April and (b) May and of lower tropospheric circulation in (c) July and (d) August–September. In a, b, shadings with contours denote strong (200 hPa) zonal flow (with speeds >30 m s−1); contours denote negative values of velocity potential at 200 hPa (<−8 × 106 m2 s−1); zonal winds across Tibetan Plateau (black shadings) in 80°E–100°E are also shown. In c, d, 850-hPa streamfunction (shaded, 106 m2 s−1) and wind vectors (black for westerly and blue for easterly when speeds >2 m s−1) are shown; thick-dashed lines denote the oceanic monsoon represented by the zero line of the streamfunction.

Full size image

Co-incident with monsoon activity over South Asia and East Asia (Fig. 4c), the broad-scale monsoon over the Afro–Asian region enters another monsoon phase in mid-to-late July. Monsoon activity can be clearly identified in August and September over the subtropical region (10°N–20°N), including the oceanic monsoon over the western North Pacific (WNP; Fig. 4d). In the Holocene, the co-evolution of the subtropical monsoons could be closely related to the LOP). Unlike the weakening of the continental monsoons since the early Holocene resulting from a decrease in summer insolation, insolation did not decrease in August until the mid-Holocene. As we will show, this feature of the insolation combined with the continued weakening of the continental monsoons is responsible for the rapid enhancement of the oceanic monsoon over the WNP36.

Given the question of how the influence of precession on the transition seasons might be observed (or simulated) over the period since the LGM, in the next two sections we discuss the premonsoon advancement in the early part of the last deglaciation, and the evolution of the late-summer oceanic monsoon over the most recent precessional cycle.

Spring warming impact on pre-monsoon South Asia in the early part of the last deglaciation

Mechanisms for early deglacial warming in the tropics have been proposed and explored, including changes in the concentration of atmospheric carbon dioxide, teleconnected impacts of the Atlantic meridional overturning circulation, and shift in the location of the Intertropical Convergence Zone12,37. These mechanisms improve our understanding of the observed constraints of glacial–interglacial and monsoon proxy records on a precessional scale2. One factor to consider is the low sea level during Heinrich Stadial 1 (HS1); the presence of the larger tropical land area during this time might amplify the response to increased insolation during boreal spring resulting from precession. The tropical and premonsoon responses to precessional spring warming might serve as a dynamical precursor of the early part of the last deglacial warming.

Over South Asia (80°E–100°E), the spring changes in atmospheric circulation are closely related to the evolution of spring insolation between the LGM and the Holocene (21–11 ka), as indicated by the EOF1 of winds and circulation over the Indian Ocean and Tibetan Plateau region (Fig. 5). Corresponding to perihelion during boreal spring from 19 to 17 ka, the westerly jet stream strengthened over the Tibetan Plateau, suggesting an early poleward shift from the southern flank of the plateau. The northward shift of the upper-level jet is attributed to the strengthening of the cyclonic circulation to the south of the Tibetan Plateau, consistent with the strengthened southerly flow in the lower troposphere and northerly flow in the upper troposphere over North India (Fig. 5 and Supplementary Fig. 2). Over the southern Tibetan Plateau and North India, the strengthened meridional circulation indicates early monsoon advancement. Notably, the tropical Hadley circulation weakened corresponding to the advanced premonsoon season, characterized as the weakening of an anticyclonic (cyclonic) streamfunction circulation in the tropics north (south) of the equator (Supplementary Fig. 2c). This observation again suggests an early seasonal migration in South Asia and the northern Indian Ocean.

Fig. 5: South Asian circulation changes in March–April at 21–11 ka.
figure 5

EOF1 (shaded) of (a) zonal winds (m s−1), (b) meridional winds (m s−1), (c) mass streamfunction (1010 kg s−1) in 80°E–100°E in March–April and (d) corresponding principal components (orange: zonal winds; dashed green: meridional winds; blue with dots: mass streamfunction). Contours denote the corresponding fields at 19 ka. E’ and W’ in (a) denote anomalous easterly and westerly, N’ and S’ in (b) denote anomalous northerly and southerly, and C’ in (c) denotes anomalous cyclonic streamfunction at 17 ka, respectively.

Full size image

Evolution of the oceanic monsoon over the last precessional cycle

Fundamentally, monsoon circulations are driven by the land–sea thermal contrast. Orbitally induced insolation changes thus alter the monsoon distribution. In the early Holocene, high summer insolation shifted the Afro–Asian monsoon further northward relative to its present-day distribution. The Pacific anticyclonic gyre became stronger and expanded westward, suppressing the monsoon trough over the subtropical WNP and thus weakening the oceanic WNP monsoon. Thus, the dynamically-related evolution of these regional monsoons could have changed in the early Holocene from the present38.

Presently, the summer monsoons occur in West Africa and the WNP in mid-to-late July, which is several weeks later than the monsoon onset in South Asia and East Asia (Fig. 3d, e). We measured the strength of the oceanic WNP monsoon using the low-level cyclonic streamfunction over the South China Sea and Philippine Sea (Fig. 6), which corresponds to the deepening of the WNP monsoon trough and the retreat of the WNP high. Consistent with previous modeling results38,39, the TraCE simulated oceanic monsoon in late summer is only observed before the Holocene and in the late Holocene. This suggests that the timing of perihelion during late summer (August–September) weakens the oceanic WNP monsoon.

Fig. 6: Evolution of oceanic monsoon at 21–0 ka.
figure 6

a Temporal evolution at 21–0 ka averaged at 110°E–130°E, 10°N–20°N, b spatial patterns at 17, 13, 9, 6, 3, and 0 ka of streamfunction (106 m2 s−1) at 850 hPa in August–September.

Full size image

In the mid-to-late Holocene, the timing of aphelion in late summer and thus weakened continental monsoons may have been responsible for the strengthening of the oceanic monsoon trough over the subtropical WNP36, and the strengthening of the westerly jet acts as a bridge for transmitting monsoon changes globally. The deglacial world before the Holocene provides another scenario where this connection can be tested, and in a mean state climate that is different from the mid-late Holocene. Thus, we shift our focus to an extremely active oceanic monsoon around 17 ka (Fig. 6); like the late Holocene, aphelion occurred in late summer during this stage (Fig. 2b). Of note is the strength (Fig. 3g) and southward position (Fig. 3b) of the westerly jet stream. The monsoon onset mechanisms on the basis of the jet stream perturbation, as documented by Wu and Chou40, might work more efficiently in the early deglacial period toward 17 ka.

Hypothesis

The precessional modulation of the midlatitude westerly jet may exert a major influence on the seasonal dynamics of the Asian monsoon. Based on the leading EOFs on westerly jet changes, the perihelion/aphelion effect on the transition seasons could be significant compared with perihelion-induced summer season changes centered on the early Holocene. With a weak summer insolation and aphelion timed to autumn, the autumn jet stream is shifted equatorward during the early part of the last deglaciation and the late Holocene, indicating an early seasonal retreat. Another notable finding is the influence of perihelion during boreal spring to advance seasonal jet migration in the early part of the last deglaciation. Unlike the dominance of perihelion during summer on the weakening of westerly jet, the occurrence of perihelion/aphelion in transition seasons might be more impactful in the early part of the last deglaciation and the late Holocene (as summarized in the schematic in Fig. 7a).

Fig. 7: Schematic of precessional impact on transition seasons.
figure 7

Schematic of (a) advanced pre-monsoon and nonstationary oceanic monsoon driven by precession and (b, c) proposed mechanisms. With perihelion occurring during boreal spring, the Hadley circulation weakens in spring over the northern Indian Ocean, coinciding with enhanced Tibetan Plateau heating and an early migration of the midlatitude westerly jet stream, leading to an advancement of the premonsoon. For the oceanic monsoon, a weak continental monsoon and aphelion occurring during autumn leads to a weaker North Pacific subtropical high and hence a stronger western North Pacific monsoon.

Full size image

We further hypothesize a possible precessional contribution when perihelion and aphelion occur over the transition seasons. With perihelion during boreal spring (Fig. 7b) driving a spring convection enhancement along the Asian–Australian land bridge, the Hadley circulation weakens over the northern Indian Ocean. Increased insolation during boreal spring and sensible heating over the Tibetan Plateau induces early candle-like heating19 and an early migration of the westerly jet stream. Thus, monsoon onset may occur earlier than in present-day climate. Then, with a weak continental summer monsoon and weak Pacific high over the east of coastal East Asia, a vigorous oceanic WNP monsoon may be observed in the late summer and autumn (Fig. 7c). The late-summer Pacific–Japan pattern41,42 might be intensified.

Discussion

Several outstanding research questions remain unanswered. To what extent obliquity contributes to the behavior driven by precession remains unknown. Our previous modeling experiments revealed that changes caused by obliquity are comparable to those caused by precession39,43. Even with perihelion in boreal summer, the oceanic monsoon could be relatively unchanged if obliquity is smaller39. The combined effects of obliquity and precession on altering seasonality could be considerable. The origin of the monsoon-like phase might be tested in a conceptual model involving thresholds in seasonal radiation44. Controlled by the present-day insolation level, gradual tropospheric moistening could provide a necessary boundary condition, eventually triggering a threshold monsoon seasonal phase transition45.

To relate the precessional effect during transition seasons to the observed climate constraints in a glacial–interglacial cycle, we examine the individual contributions of perihelion during boreal spring and aphelion during boreal autumn to the early part of the last deglaciation. Prior to and around 17 ka, increased spring insolation can be characterized by the premonsoon advancement in climate, which might correspond to increased ice melt consistent with the deglacial process. In contrast, corresponding to decreased autumn insolation, the oceanic monsoon (Fig. 6b) and jet stream (Fig. 3g) are coincidentally strong. Although a weak subtropical high might be expected in this season, the seasonal contribution to the deglacial process remains unclear. A further diagnosis of teleconnection of the increased oceanic convection to the extratropics could be helpful for identifying the contribution. A recent study16 also considers a remote control of the enhanced convective activity and precipitation over South China by the slowdown of the Atlantic thermohaline circulation in the last deglacial autumn.

Recent high-resolution oxygen isotope records from stalagmites in central China suggest a considerable contribution of dry-wet pattern changes from tropical sources, including changes in rainout from the tropical Pacific Ocean and Indian Ocean and changes in oxygen isotopes during the HS146. Proxy data could further provide insights into the semiprecessional signal in the change between the LGM and the Holocene. Isotope changes in the Afro–Asian summer monsoon region show a distinctly opposite trend from 19 ka to 17 ka compared to 16 ka to 14 ka (Supplementary Fig. 3a, b), which is consistent with the simulated peak timing of monsoon circulation around 17 ka. Compared with the change based on benthic oxygen isotope records, the identified change based on water oxygen isotopes exhibit a larger amplitude in the early part of the last deglacial warming; thus, the semiprecessional signal can be observed more clearly (Supplementary Fig. 3c–h). These observations support our speculation of the sea level and premonsoon changes, with perihelion occurring in spring rather than summer. However, how the monsoon response during transitional seasons is related to deglacial warming remains unanswered and worthy of further investigation.

If one focuses on only perihelion during summer or winter, we may miss out on the significant changes to the monsoon during transition seasons. Reconstruction errors, partly due to incorrectly recorded seasonal precipitation changes, could further bias the attribution of monsoon evolution.

Methods

Proxy and reanalysis data

We analyzed 17 proxy records including cave and marine core oxygen isotopes spanning 21 ka to the present, as listed in Supplementary Table 1. We determined the difference between two paleoclimate states (i.e., 17 ka vs. 19 ka and 14 ka vs. 16 ka; as shown in Supplementary Fig. 3a, b), giving planktonic oxygen isotopic records preference over benthic records if both were available; for example, the change from 19 ka to 17 ka was calculated as the average of 17–18 ka subtracted by that of 18–19 ka. For current climatology (1981–2010), we analyzed the dynamical fields obtained from the Japanese 55-year Reanalysis project (JRA55)47, which has a temporal resolution of 6 h and a latitude–longitude spatial resolution of 1.25°. We also used surface pressure data obtained from ECMWF Reanalysis version 5 (ERA5)48, with a temporal resolution of 1 h and a latitude–longitude spatial resolution of 0.25°.

Climate simulations

We used output from the TraCE simulation for a coupled atmosphere–ocean–sea ice–land surface climate model from 21 ka to the present49. The full TraCE simulation with monthly outputs was conducted using the Community Climate System Model version 3 (CCSM3) at a horizontal resolution of approximately 3.75° in the atmosphere. The simulation was forced by transient forcing changes in greenhouse gas concentrations, orbital insolation variations, ice sheets, and meltwater fluxes (further information is detailed in the original paper and on the project website). We averaged 30 years of data around the target time slice to form a climatological condition. A time series comprising climatological conditions with a 500-year interval was further analyzed for the last glacial–interglacial cycle.

Another series of climate simulation50 varying the longitude of precession (at intervals of 30° from 0° to 330°) and eccentricity (0.01 and 0.04) were also explored. The simulations were conducted using the Community Earth System Model (CESM) version 1.2 at approximately a 2° finite-volume grid in the atmosphere and at a 1° rotated pole grid in the ocean and sea ice. Obliquity was set to 23.439°, boundary conditions were set to preindustrial, and seasonal climatology was averaged for the last 20 years of a 25-year run [refer to the original paper27 for further details of the experimental design].

EOF analysis

We used EOF analysis to model output to examine changes in climate fields including precipitation and wind fields. Before performing the EOF analysis, we applied a 3-month running mean to account for annual variation. Global and regional domains were both focused on the millennial scale variation in focused seasons. Generally, the first two modes of the investigated EOF patterns can explain more than 90% of the variance. We also looked at the difference between two climatological periods, including the changes from 19 to 17 ka, to compare with the EOF-derived changes. We can also calculate confidence intervals (90%, Student’s t test) for comparisons.