Introduction

The Asian/Indian summer monsoon (ASM/ISM) has a profound impact on the regional ecosystem it traverses the livelihoods of billions of people in South and Southeast Asia through the amount of precipitation it conveys. Researches have been done to understand the fundamentals and dynamics of the past ASM/ISM system on orbital1,2 and millennium timescales3,4. The monsoon is clearly influenced by the boreal summer insolation on orbital timescales, and tracks ocean circulation changes in the North Atlantic on millennial scales. However, the forced monsoon variability on centennial to multidecadal timescales is less understood for the reason that it lacks of well-defined climate forcings. Radiatively, the sun is the ultimate source of energy for climatic variations on Earth, and it potentially affects climate via energy balance in all time scales5,6. Sunspot observations represent a direct measure of solar energy output for centennial to decadal timescales, but the instrumental record is too short to provide sufficient evidence for the climate response.

Terrestrial paleoclimate proxy data is a valuable tool to extend instrumental records on centennial and decadal timescales, because the higher temporal resolution can filter out noise and reveal underlying patterns. A growing number of studies7,8,9,10,11,12 have found periodicities in proxy data that match the known frequencies of solar irradiance fluctuations, suggesting that solar activity is a pivotal driving force of climate change. On the contrary, other studies have found that the link between solar activity and climate is not always statistically significant, and may be confounded by other factors including internal mechanisms that cause minor climate variability13,14,15. This is likely due to the relatively small radiative effects of solar activity against the climate forcings. For this reason, the Holocene period, without significant known forcing and anthropogenic influence, should be an ideal and promising testbed for the solar climate. Indeed, there is evidence that climate has responded to solar variations in Asia10,16,17, the tropics18,19,20,21, and northwestern North America22,23,24. However, the correlation between solar activity and climate is less clear during the glacial period.

In this work, we show that the solar–climate linkage sustains even in glacial period given high enough resolution and chronology. We examined two published sub-decadal resolved speleothem records from northwestern Vietnam25,26 (Fig. 1). These records represent one of the highest resolutions and among the best age models based on U–Th dating during the marine isotope stage III. When combined with other nearby records from the ASM/ISM region (Fig. 1) and conducted spectral analysis, an evident feature of solar-related periodicities can be observed. We then used a single-forcing transient numerical model experiment to identify a plausible mechanism for how solar irradiance can influence the ASM/ISM region. This study suggests that the monsoon system is able to positively respond to solar activity, even as little as 1–2 Wm-2, regardless of background climate states. Therefore, we argue that changes in solar irradiance should be considered in ASM/ISM regional climate projections.

Fig. 1: Map showing compiled speleothem records and climatological wind fields.
figure 1

TT cave is marked as the red star. Circles and rectangles indicate speleothem δ18O records covering Holocene and last glacial periods, respectively. Color codes are the same as in Fig. 4. Arrows represent average early monsoon season (May–August) wind fields at the 850-hPa level from 1980 to 2020 based on the NCEP/NCAR reanalysis.

Full size image

Results

Correlation between TT δ18O record and solar activity

The composite high-resolution δ18O record from Vietnamese speleothems (TT-3 & 5), covering the period from 32.5 to 27.5 kyr BP, has been established as a proxy for regional precipitation levels and is connected to the variations in the strength of ASM/ISM25,26. We first check whether the coherence between Total Solar Irradiance (TSI) and ASM/ISM regional precipitation matches the conventional monsoon theory. According to this theory, an increase in solar energy reinforces land-sea contrast through heterogeneous heating, subsequently enhancing monsoon circulation and intensifying monsoonal rainfall. A comparative analysis was conducted on the TT record, atmospheric Δ14C and ice core 10Be records27,28, which are cosmogenic nuclides and commonly employed as proxies for solar irradiance. The TT δ18O values decreased in the period of 30.2–29.3 and 28.3–27.9 kyr BP, which coincided with the change in Δ14C and 10Be concentration. It indicates the presence of both wetter regional conditions and strengthened solar activity (see Fig. 2). On the other hand, increases in TT δ18O also align with Δ14C and 10Be trend at 31.0–30.3 and 29.2–28.3 kyr BP. This correlation implies a positive climate response between, tallying with the expectation; while the mechanism-in-detail remains incompletely understood (simulation result hereafter).

Fig. 2: Comparison of the TT δ18O and cosmogenic nuclides record.
figure 2

Comparison between TT δ18O record (blue)25,26, cosmogenic nuclide 10Be from GISP227, atmospheric Δ14C from Hulu Cave28, and oceanic Δ14C record from Cariaco51. The 30 U–Th dates and calculated growth rate for TT are indicated by the blue error bars and dark yellow line, respectively.

Full size image

In exploring the correlation between regional monsoon strength and solar activity, we are not arguing that TSI is the primary forcing, nor are we directly comparing the two proxy records. The TT record’s resemblance to millennial-scale features observed in the ice core δ18O record suggests that perturbations originating in the North Atlantic, such as Dansgaard–Oeschger and Heinrich events, played a dominant role in influencing the ASM/ISM climate during this period and at the millennial timescale. Our study instead aims to investigate other factors that may influence the monsoonal precipitation, providing a broader understanding of the hidden drivers.

Based on above association, we therefore expect the known solar cycles at centennial to decadal scales can be identified with the exceptional high resolution of TT record. This domain has been less explored to date but holds significance as the temporal scale and magnitude of subtle energy changes induced by TSI variations are pertinent to future climate projections. In the compiled TT record, five centennial-scale events, each lasting 140–230 years, were visually identified during the interval of 30.0–28.2 kyr BP, marked by δ18O minima at 30.0, 29.9, 29.7, 28.5, and 28.2 kyr BP, respectively (Fig. 2). Power spectral analysis of the composite TT δ18O over the span of 32.8–27.5 kyr BP also revealed the presence of an approximately 180-year periodicity (see Fig. 3a) with a significance level exceeding 90%. Individual spectral analyses of TT-3 and TT-5 records also displayed periodicities close to 200 years, as well as the spectral results of Δ14C and 10Be concentration (Fig. 3b). Acknowledging the potential influence of chronology on spectral analysis outcomes29, an alternative chronology was constructed using the StalAge algorithm30 to verify the robustness of the spectrum results (see Supplementary).

Fig. 3: Spectral analysis of the TT δ18O and cosmogenic nuclides record.
figure 3

Results of power spectral analysis for (a) the compiled TT δ18O time series and (b) Cariaco Δ14C51 and GISP2 10Be27 records. Black (gray) line was calculated from 10Be data in 3000–8000 year BP (11,000–18,000 year BP). The cyclicities above the 95% (dashed) and 90% (dotted) confidence line are shown in red and blue.

Full size image

This sub-millennial periodicity coincides with the de Vries cycle (~ 205 years; hereafter DVC) of solar activity, a phenomenon observed in many Holocene speleothem δ18O records10,11,21,31,32,33,34. Intriguingly, reported records with the DVC cycle seemed much less during glacial periods than in the Holocene. Here we argue this is due primarily to low sampling resolution and insufficiently dense age control points (refer to subsequent discussions).

Key for the TSI fingerprinting in the ASM/ISM regime

As described above the TT record offers distinct advantages in documenting sub-millennial hydroclimate changes during a glacial period owing to its exceptional age model. Firstly, the thirty U–Th dates covering approximately 5000 years in the TT record are not only in high density but also evenly distributed (Fig. 2), a feature contrasting with the sparse age control points found in other cave records (typically less than 2 dates per thousand years). Secondly, the TT record achieves a temporal resolution averaging around 6 years due to its rapid growth rate, which is equal if not superior to that of many published cave records from the last glacial period. Such an observation hints the importance of both high sampling resolution and sufficient age controls in identifying solar-related periodicities at centennial timescales.

To assess the extent of solar influence in the ASM/ISM domain particularly during the glacial period, the geographical range of 20°–35°N and 75°–120°E is selected after carefully comparing the regional precipitation distribution associated with δ18O value and wind field in monsoon seasons (indicated by monsoon circulation in Fig. 1). We then examined another thirty-five published speleothem δ18O records originating from twenty-eight locations within the area (Table 1). These records span both interglacial (Holocene) and glacial periods. Among them, twenty out of the thirty-five spectrum results exhibited a periodicity similar to the DVC. Meanwhile, ten records—KM-A, TM18, XL2, XL26, WY12, GZXND21-1, HZ11, YX175, YX51, and Chy-1—revealed solar cycles of varying durations, suggesting climate associations with solar activity as well. This finding provides support for the solar modulation of ASM/ISM intensity on sub-millennial timescales. The well-known 11-year solar cycle, marked by fluctuations in the number of sunspots on the sun’s surface, is not resolvable in these speleothem records due to their interannual to decadal time resolution. Another ~60-year cycle is present but with lower confidence. Therefore, the absence of the strongest 11-year and weak 60-year frequency among these records should result from the inherent temporal limitations of speleothem δ18O records.

Table 1 List of compared caves
Full size table

We then propose that the pivotal factor in capturing solar fingerprinting in proxy records lies in the combination of sampling resolution and accurate chronological sequencing. Figure 4 elucidates the impact of sampling resolution and the quantity of dates on the solar imprint within collected speleothem δ18O records. Among the five records exhibiting no DVC, Y1, SB46, MWS 2, Sw5, and Wu3 (hollow squares in Fig. 4), the latter four have fewer than two dates per thousand years, even though two of them, Sw5 and Wu3, boast a relatively high sampling resolution. Further, two paired cases underscore the significance of date numbers. Firstly, in Wulu cave, DVC was identified in Wu88 stalagmite, characterized by a relatively lower sampling resolution of approximately 16 years but a denser distribution of U–Th dates, averaging around two dates per thousand years. Conversely, Wu3 stalagmite, with an ~8-year sampling resolution, features an average of less than 1 date per thousand years. The second case pertains to Mawmluh cave. The presence of more dates in MAW-6 stalagmite compared to MWS 2 resulted in the subsequent observation of a distinct DVC in MAW-6 (Table 1). Yamen Cave’s record, spanning from deglaciation to the early Holocene, stands as an exception with both high temporal resolution and dense age control points. However, it does not exhibit any solar-related periodicities. While the exact reason remains unclear, other local factors may be prevailed and overprinted.

Fig. 4: Chronological control points and sampling resolution for the ASM/ISM records.
figure 4

The blue symbols are records with the de Vries cycle, while the greens show the ones with other solar-related cycles. Hollow symbols indicate records without solar-related cycles. Circles and rectangles indicate speleothem δ18O records covering Holocene and last glacial periods, respectively. Compiled TT, TT-3 and TT-5 records are all marked in red.

Full size image

Two key findings can be concluded: (1) the detection of DVC in the ASM/ISM-region records suggest solar activity significantly serves as a robust forcing (potential mechanism follows) in influencing short-term sub-millennium monsoon intensity, despite minor radiative variability ( < 1 W/m2), and (2) sufficient age control points are essential for conducting sub-millennial climate frequency studies on fluctuations, probably more crucial than sampling resolution. We in addition advocate for a 500-year age control, encapsulated in the “2 dates per 1kyr” criterion, as an empirical guideline for capturing the DVC solar footprint in speleothem δ18O records, with the benefits of readily available U–Th dating and enhanced stable isotope analysis efficiency.

Discussion

Kodera (2004) suggested a direct solar influence on the Intertropical Convergence Zone (ITCZ). Numerical simulations indicate that during solar maximum, temperatures in the lower stratosphere of the tropics to subtropics regions rise, enhancing convective activity of the equatorial region35,36. Consequently, increased convection during high solar activity in boreal summer leads to greater precipitation over South Asia, as evidenced in proxy records. Another mechanism suggests that solar activity minima induce tropospheric cooling and lower North Atlantic sea surface temperatures at high latitudes11. This cooling, in turn, reduces surface water evaporation. Furthermore, changes in solar activity may directly influence regional moisture variations through the direct heating effect on air temperature37,38,39. Although various mechanisms have been proposed to explain the solar–monsoon link, here we show a dynamical process particularly for the ASM/ISM region aided by climate model simulation.

In contrast to previous low verse high TSI sensitivity approaches investigating the solar-monsoon link, we employ LME (Last Millennium Ensemble40) experiments to examine the Asian Monsoon’s response to solar irradiance changes over the last millennium. These experiments encompass transient decadal and centennial solar cycles, enabling the exploration of the temporal and spatial complexity of the climate system. By conducting a single forcing run using only solar irradiance changes, we can identify solar influence on Asian monsoonal precipitation, aligning with observations from the TT record. To isolate the impact of solar activity from other forcing factors, we employ Multi-Linear Regression analysis, a method frequently used to disentangle the effects of various sources of variability on climate variables41,42,43,44,45.

The multi-linear regression analysis reveals the regression coefficients of precipitation against solar activity, as depicted in Fig. 5a. Employing a bimonthly time scale to illustrate monsoon stages, we found this relatively small irradiance forcing being more effective on precipitation during the monsoon onset stage. The solar signals within the bimonthly mean precipitation (May–June) indicate a correlation between increased total solar irradiance and elevated rainfall across South and East Asia (Fig. 5a). Aligned with the typical summer monsoon circulation, intensified northward cross-equatorial wind near the surface, the precipitation surge primarily manifests on the windward side of India and Indochina. A similar pattern, albeit of smaller magnitude, is observed in the precipitation response during peak summer (July–August).

Fig. 5: Regression map of climate variables on solar irradiance.
figure 5

a Precipitation (shaded) and 850-hPa wind field (vector) coefficients to solar activity obtained by multi-linear regression analysis using LME TSI single forcing experiment. b zonal averaged (60°–90°E) temperature (shaded), omega (contour: blue lines denote anomalous upward motion), and meridional winds (vector). The solid black dots denote the grids where the coefficients are statistically significant at 95% confidence level.

Full size image

To probe into the alterations in atmospheric circulations concurrent with anomalous early-summer precipitation in Asia, Fig. 5b displays the vertical structure of anomalous temperature and meridional circulations within the 60°E to 90°E section. Consistent with the increased precipitation over South Asia (Fig. 5a), an augmented updraft (Fig. 5b, contour) and landward flow (Fig. 5b, arrow), propelled by anomalous warming over the Tibetan Plateau (Fig. 5b, shading), characterize the region. This pattern mimics the candle heating process, playing a pivotal role in fostering anomalous convection and enhancing precipitation in the present-day ISM development. Drawing on the Last Millennium Ensemble (LME) analysis, our findings suggest that the solar signal observed in the TT record stems from circulation responses driven by cyclic solar irradiance heating. Importantly, this signal underscores the significance of relatively minor radiative forcing, capable of exerting hydrological impacts on the decadal to centennial timescale of monsoonal Asia’s regional climate.

The assumption that speleothem deposition was uniform between age control points may not hold true in light of various factors. Conventionally, this assumption was widely accepted, especially when dating capacity was limited and the subject of interest spanned much longer time scales. However, with widely-applied dating techniques and our understanding of geological processes, it has become evident that deposition can vary significantly over shorter time intervals. The rise of high-frequency and abrupt paleoclimate events, such as rapid changes in centennial timescale, poses a challenge due to discontinuities or anomalies of non-uniform depositions. Essential as it is to carefully evaluate the age model when interpreting geological records, there is no definitive guidance for the baseline requirement of age control points.

The existence of known solar cycle such as DVC can serve as a verification target in this regard. By decomposing the record resolution into sampling and dating resolution (x and y axes respectively in Fig. 4), we confirm that dating resolution is critical in resolving DVC in the ASM/ISM region. Cave records that resolve the ~200-yr DVC periodicity (blue in Fig. 4) have sampling intervals ranging from 60 years to sub-decadal timescales (x-axis), while maintaining at least ~500 years of dating resolution. This feature appears to be independent of sample growth rate; while the glacial sites (squares in Fig. 4), where the sampling resolution is often limited by slower growth rates, still preserve the cyclic DVC with a high enough dating resolution.

Further extending our findings in a general context, we suggest that a ratio of 2–3 between age control interval to event frequency serves as a reference for detecting a particular periodicity, independent of other sample growth and preservation conditions. This feature implies that when studying millennium to centennial periodic climate events, a corresponding dating resolution promises a more confident result, and interpolation between age control points should be constrained for a reliable chronology.

Methods

Location of Thuong Thien Cave and climatology

Thuong Thien cave (103°56’E, 21°20’N) is located at ∼700 m above sea level in northwestern Vietnam25,26. The entrance is ~80–90 m below top of the mountain within a formation of Middle Triassic limestone46. The mean annual temperature and precipitation at the study site is 21 °C and 1328 mm, respectively, according to a local meteorological station. The relative humidity in the cave is ~85%. The summer rainfall (May–September) accounts for ∼80% of the annual precipitation, which is governed primarily by the Indian summer monsoon (ISM). The description and climatology of study site were detailed in Nguyen et al. 25.

Age model

The age model is constructed by a linear interpolation between the thirty U–Th dates covering approximately 5000 years in the TT record25,26. An alternative chronology was compared using the StalAge algorithm to verify the robustness of the spectrum results (see Supplementary).

Spectrum analysis

The power spectrum analyses of δ18O data was carried out by use of the PAST data analysis software47. The periodicity is calculated with the REDFIT algorithm48, with a Rectangle window (oversample size 5, number of segments 3) at 90% and 95% confidence level.

Community earth system model-last millennium ensemble (CESM–LME)

The model simulation dataset utilized in this study is from the Community Earth System Model-Last Millennium Ensemble (CESM–LME) modeling project40. The CESM–LME prescribed CMIP5 climate forcing reconstructions to conduct ensembles of last millennium simulations, containing “full-forcing” experiments with all forcings and “single-forcing” experiments with each forcing individually. All ensemble members, perturbed with slightly different initial conditions of air temperature, were branched from year CE 850 of the 850 control experiment and ended in CE 2005. In order to single out the influence of solar activity, we used the data obtained from the solar-only forcing (changes in total solar irradiance only) experiment during CE 850–1850 and analyzed the ensemble mean of four ensembles. Regarding the solar forcing, the changes in total solar irradiance are prescribed by the reconstruction49, which has decadal and long-term solar cycle over the last millennium. Upon the total solar irradiance, an estimated 11-yr solar cycle has been imposed and its linear regression at each spectral interval is used to derive spectral solar irradiance50. We utilized the solar forcing to represent the changes in solar irradiance at the top of atmosphere in this study.