The Holocene Asian Monsoon: Links to Solar Changes and North Atlantic Climate

See allHide authors and affiliations

Science  06 May 2005:
Vol. 308, Issue 5723, pp. 854-857
DOI: 10.1126/science.1106296


A 5-year-resolution absolute-dated oxygen isotope record from Dongge Cave, southern China, provides a continuous history of the Asian monsoon over the past 9000 years. Although the record broadly follows summer insolation, it is punctuated by eight weak monsoon events lasting ∼1 to 5 centuries. One correlates with the “8200-year” event, another with the collapse of the Chinese Neolithic culture, and most with North Atlantic ice-rafting events. Cross-correlation of the decadal- to centennial-scale monsoon record with the atmospheric carbon-14 record shows that some, but not all, of the monsoon variability at these frequencies results from changes in solar output.

The impacts of decadal- to centennial-scale solar variability on the climate system during the Holocene have been reported from mid to high northern latitudes (13) to low-latitude regimes (46), including the Asian monsoon (AM) (4,5). To test the degree to which the Holocene AM may be linked to solar variability, a high-resolution, precisely dated, continuous record of the monsoon is needed. Such a record could also be used to test the degree to which changes in the interglacial AM are related to climate change elsewhere. For example, a number of studies have demonstrated close ties between the glacial AM and the climate in the North Atlantic region (79). The degree to which such links extend into interglacial periods is an open question. We have previously reported on a Chinese Holocene record of the AM that addresses some of these issues (10). Here, we build on that work with a higher resolution absolute-dated Holocene AM record from Dongge Cave, southern China, which we compare in detail with the atmospheric 14C record (as a proxy for solar output) and climate records from the North Atlantic region (2, 1113).

Stalagmite DA was collected from Dongge Cave (25°17′N, 108°5′E, elevation 680 m) in southern China. Today, the cave site has two distinct seasons: a cool, dry season during the boreal winter when the Siberian high establishes a strong anticyclone on the Tibetan Plateau and a warm, wet season during the summer months when the intertropical convergence zone (ITCZ) shifts northward and monsoonal convective rainfall reaches its maximum. Our previous studies have shown that shifts in the oxygen isotope ratio (δ18O) of the stalagmite from the cave largely reflect changes in δ18O values of meteoric precipitation at the site, which in turn relate to changes in the amount of precipitation and thus characterize the AM strength (10, 14).

Chronology of the 962.5-mm-long stalagmite DA is established by 45 230Th dates (table S1), all in stratigraphic order, with a typical age uncertainty of 50 years. Sample DA grew continuously from approximately 9000 years before the present (ky B.P., where the “present” is defined as the year 1950 A.D.) until 2002 (when the stalagmite was collected), with a nearly constant growth rate of ∼100 μm per year. A total of 2124 δ18O measurements were obtained from along the growth axis, with an average temporal resolution of 4.5 years (table S2). We present two time scales, one of which is based on linear interpolation between the 230Th dates and is independent from all of the chronologies to which we compare our data (Fig. 1). Another is tuned within dating error to INTCAL98 (15) (Fig. 1) and is used to determine the extent to which fine-scale dating errors may affect our correlations with other records.

Fig. 1.

(A) δ18O time series of the Dongge Cave stalagmite DA (green line). Six vertical yellow bars denote the timing of Bond events 0 to 5 in the North Atlantic (2). The Chinese events that correlate with Bond events 3 and 5 coincide within error with the collapse of the Neolithic Culture of China (NCC) (23) and the timing of an abrupt outflow event from a Laurentide ice-margin lake (22), respectively. Two vertical gray bars indicate two other notable weak AM events that can be correlated to ice-rafted debris events (2). (B) DA age-depth (mm, relative to the top) relations. Black error bars show 230Th dates with 2σ errors (table S1). We use two different age-depth curves, one employing linear interpolation between dated depths and the second slightly modified by tuning to INTCAL98 (15) within the 230Th dating error (26).

The δ18O values vary between –9.2 and –6.5 ‰, with a typical amplitude of somewhat less than 1 ‰ over time scales of decades to centuries. As verified by our previous studies (10, 14), Dongge Cave δ18O becomes lower as Asian summer monsoon intensifies, and vice versa. Such anticorrelation is also observed in the modern precipitation records near the cave site (16).

DA δ18O data shows a strong AM interval from 9 to 7 ky B.P., followed by a gradual weakening. This overall temporal pattern resembles high-resolution Holocene precipitation records from a southern Oman stalagmite (5), titanium concentration data from the Cariaco Basin, tropical Atlantic (17), and our earlier work (10), which suggests that shifts in mean position of the ITCZ may control temporal variability of precipitation throughout the entire low-latitude region (18). This general weakening of the Asian monsoon during the Holocene corresponds with orbitally induced lowering of Northern Hemisphere summer solar insolation during this interval, which indicates that the broad trend is caused by insolation change. The general trend is punctuated by eight weak AM events, each lasting ∼1 to 5 centuries, centered at 0.5, 1.6, 2.7, 4.4, 5.5, 6.3, 7.2, and 8.3 ky B.P., with a temporal spacing averaging ∼1.2 ky (Fig. 1). These events are, within dating error, in phase with weak southwest AM events recorded in marine cores from the Arabian Sea (19) and are possibly linked to Holocene ice-rafting events in the North Atlantic (2) (Fig. 1). Among those events, the two at 8.4 to 8.1 and 4.5 to 4.0 ky B.P. are longer in duration and larger in magnitude. They are similar in terms of the abrupt transitions and magnitude (0.8 to 1.0 ‰), and they have one or two brief reversions within the events. The event at 8.4 to 8.1 ky B.P. correlates with the strongest Holocene cooling/drying event recorded at high northern latitudes (1, 20) and subtropical temperate regions (21) and in tropical ocean and terrestrial records (4, 5, 19); it also coincides within error with the 8.2-ky event recorded in the Greenland ice cores (11, 12), possibly related to abrupt outflow from a Laurentide ice-margin lake (22). Another major event at ∼4.4 to 3.9 ky B.P., although not clear in Greenland records, has been reported in various localities in China (23). Among the most abrupt events in the Holocene Dongge record is the abrupt lowering of AM intensity at ∼4.4 ky B.P. over several decades (Fig. 1 and table S2), which supports the idea that this sharp hydrological change might be responsible for the collapse of the Neolithic culture around Central China about 4.0 ky ago (23). Strongly enhanced aridity at this time is also a main feature of the Indian monsoon as recorded in western China (24) and is in phase with the Mesopotamian dry event in western Asia (25).

To assess the link between solar activity and AM intensity, we compared the detrended DA δ18O (Δ18O) record to the detrended atmospheric 14C record (Δ14C), a proxy for solar activity (15) (Fig. 2), using the tuned time scale for DA (26). As the time scale is tuned, we consider the resulting correlation to be a “best case” scenario. Visually, (Fig. 2) the larger amplitude fluctuations in the AM (>±0.2 ‰ in the Δδ18O record) broadly agree with Δ14C events on centennial time scales, similar to the relation observed in the record from a southern Oman stalagmite (5). The correlation coefficient for the full record is r = 0.30, which indicates that some of the variability in the AM can be attributed to solar changes. The main discrepancy between the two records comes at decadal time scales, plausibly reflecting fine-scale errors in chronology or, alternately, indicating that at these frequencies other factors may be more important in controlling AM variability, such as changes in atmospheric and oceanic circulation.

Fig. 2.

Time series of the DA Δ18O record (five-point running average, green line) and the atmospheric Δ14C record (red line) (15). All data have been detrended using singular spectrum analysis. Higher solar irradiance (smaller Δ14C) corresponds to a stronger AM (smaller Δδ18O value). The correlation coefficient is 0.30 for the entire profile and 0.39 between 9 and 6 ky B.P.

Power spectral analysis of the tuned DA δ18O record shows statistically significant centennial periodicities centered on 558, 206, and 159 years (fig. S1A). These periodicities are close to significant periods of the Δ14C record (512, 206, and 148 years) (27) and to previously reported findings from spectral analysis of another Chinese speleothem (10). Cross-spectral analysis of the DA record and the 14C record further shows some common periodicities (232, 129, 116, 104, 89, 57, and 54 years) (fig. S1B). Our data, together with the other Chinese work (10) and two Oman stalagmite δ18O records (4, 5), support the idea that solar changes are partly responsible for changes in Holocene AM intensity (28).

We have previously demonstrated a close correlation between last glacial period AM variability and the temperature change over Greenland on millennial time scales (9, 14). The present high-resolution DA δ18O record enables a more precise correlation between the AM and Greenland climate on centennial time scales and under interglacial conditions. The smoothed, detrended DA Δ18O record shows a broad similarity to the δ18O records of Greenland ice—Greenland Ice Core Project (GRIP) (11) and Greenland Ice Sheet Project 2 (GISP2) (12)—in terms of frequent decadal-scale and centennial-scale fluctuations (Fig. 3). Similar to Greenland ice core records, the centennial- and multidecadal-scale AM variations during the Holocene are considerable (∼0.2 to 0.7 ‰ in δ18O) but not as large as glacial millennial-scale variability (∼1 to 2 ‰ in δ18O) (9). Because of fine-scale uncertainties in dating of records from both sites, it is not possible to determine whether decadal-scale variations correlate. However, the general correlations between DA and Greenland records are apparent on the multicentennial scale (Fig. 3). This broad correlation is also noticeable between DA and the new δ18O record of Greenland ice [NGRIP (13)], which also has a long-term trend similar to DA between 0 and 3.8 ky B.P. (fig. S2). Over this time interval, the Pearson correlation coefficient of the records reaches its highest value of 0.57 when setting a 150-year-phase lead of the tuned DA δ18O record over the NGRIP δ18O time series (fig. S2). A lead of this magnitude in this time interval would be larger than the combined uncertainty in the DA 230Th dating and the Greenland layer-counting chronology, but not by a large amount, because both records have errors of up to several decades. If the lead is real, given that we can attribute at least some of the variability in the AM to solar changes, it is plausible that the AM responds almost immediately to solar changes by rapid atmospheric response to solar forcing. Because Greenland's climate is closely tied to the rate of production of North Atlantic Deep Water, it is plausible that Greenland temperature lags solar forcing because of the time constants involved in changing ocean circulation. Alternatively, it is plausible that the apparent lead is not a “true” lead and that the high Pearson correlation coefficient is simply, by chance, higher with a 150-year offset. We note that the DA record has significant power at both 159-year and 206-year periods (fig. S1). Thus, the lead could plausibly represent an offset of one period at one of these frequencies.

Fig. 3.

Comparison of the smoothed (5-point running average) detrended DA Δ18O record (green) with the smoothed 20-year averaged GRIP δ18O record (5-point running average, red) (11)(A)and the GISP2 δ18O record (20-point running average, red) (12)(B) over the past 9 ky. The broad correlations between DA and Greenland records are apparent at the multicentennial scale.

In summary, the broad decline in AM intensity through the latter part of the Holocene correlates well with other northern low-latitude records and results directly from the orbitally induced lowering of summer insolation affecting ITCZ position and low-latitude precipitation patterns. The centennial- and multidecadal-scale events that characterize the AM record throughout can, in part, be ascribed to responses to changes in solar output. There are similarities and correlations between the Holocene AM and the North Atlantic climate, including both the ice-rafted debris record and the Greenland ice core records. Some of these correlations result from solar forcing affecting climate in both regions. It is also possible, with the 8200-year event as the main example, that oceanic circulation changes in the North Atlantic triggered changes in the AM. Thus, changes in the Holocene AM result from a number of factors, including orbitally induced insolation changes, changes in solar output, and changes in oceanic and atmospheric circulation.

Supporting Online Material

Figs. S1 and S2

Tables S1 and S2


References and Notes

View Abstract

Navigate This Article