Central Indo-Pacific Rainfall Variations and ENSO

The most notable feature of the TK δ¹⁸O record is the long-term increasing trend, which signifies decreasing rainfall in the NCIP over the past 2,700 y. This suborbital-scale rainfall decline is recognizable in paleohydrological records from numerous locations in the northern tropics, such as Southeast Asia (25), Mesoamerica (35, 36), and the Caribbean (15). An opposite increasing rainfall trend is documented in the southern tropics, including East Africa (37), the Western Pacific Warm Pool (38, 39), the eastern tropical Pacific (ETP) (40), and South America (41) (SI Appendix, Fig. S12). This tropical interhemispheric precipitation see-saw pattern on suborbital timescales is similar to observations during the last glacial period (42, 43) and is most likely driven, on a longer timescale, by precessional changes in summertime insolation in each respective hemisphere.

The centennial- to decadal-scale rainfall variation in the NCIP is highlighted in the detrended TK δ¹⁸O record (Fig. 3B). Comparison with other hydroclimate sequences in the southern central Indo-Pacific (SCIP) region during the last 2,000 y, including the hydrogen isotopic ratios of terrestrial higher plant leaf waxes (δD_wax) in marine sediments from southwest Sulawesi in central Indonesia (39), and a synthesized stalagmite-based rainfall record from Liang Luar Cave in eastern Indonesia (3), are given in Fig. 3 B–D. All records feature deceased rainfall during the 20th century [Current Warm Period (CWP)] and at periods of 950–1150 AD and 1200–1300 AD during the Medieval Warm Period (MWP). In contrast, both NCIP and SCIP experienced increased rainfall during the LIA and the Dark Ages Cold Period (DACP) (9). A dry MWP is also confirmed by the low accumulation rate of peat in Lake Pa Kho (<0.1 mm/y) in northeastern Thailand (44). On the contrary, much higher accumulation rates were observed during the LIA (∼1 mm/y) and the DACP (∼1.3 mm/y) (44). The extremely wet late 14th century in our TK record coincides with the timing of extensive floods and destruction of the water management systems in Angkor, Cambodia, confirming the role of climate on the decline of the Khmer Empire (45).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Comparisons of the detrended TK stalagmite δ¹⁸O record with high-resolution rainfall records from SCIP region, and ENSO and volcanic activities during the last 2,000 y. (A) Northern Hemisphere temperature (NHT) (65). (B) TK δ¹⁸O record (gray line) from southern Thailand (this study). The blue line is the 20-y moving average. (C) Leaf waxes δD in marine sediments from southwest Sulawesi, central Indonesia (39). (D) Multiproxies synthesized stalagmite record (LLPC1) from Liang Luar Cave in eastern Indonesia (3). Records in B−D are detrended to emphasize their centennial to decadal timescale variations. (E) Reconstructed SOI record (14). (F) El Niño activities recorded by percent of sand in lake sediments of El Junco, Galápagos; an increase in sand abundance represents more intense rainfall events associated with El Niño events (40). (G) Model simulated Niño3.4 sea surface temperatures variability (48). (H) Volcanic sulfate recorded in the Greenland Ice Sheet Project 2 (GISP2) ice core (66). Yellow bars denote dry conditions during the MWP and CWP, corresponding to enhanced El Niño activities. Gray bars mark the wet LIA and DACP, corresponding to enhanced La Niña activities.

The spatial patterns of centennial- to decadal-scale dry conditions in the CIP region during the MWP and CWP are similar to that during El Niño events today (Fig. 1C). Indeed, a 2,000-y reconstruction of Southern Oscillation Index (SOI) shows intensified El Niño activity during the MWP and CWP (14). A significant negative correlation (r = −0.28, P < 0.01) was observed between TK δ¹⁸O and the SOI record, with more positive TK δ¹⁸O values (less rainfall) correlated to more negative SOI values (El Niño-dominated conditions), and vice versa (Fig. 3E). In addition, increased sand abundance and red color intensity of lake sediments from ETP during the MWP (40, 46) reveal enhanced rainfall in this region, also suggesting El Niño-like conditions during the MWP (Fig. 3F). On the other hand, relatively wet conditions during the LIA and DACP in the CIP correspond to La Niña-like conditions, as suggested by increased SOI values and decreased ETP rainfall (Fig. 3). Modern observations and model results indicate that, since the mid-19th century, the warming of the tropical Pacific caused a weakening of the Walker circulation (47). Some model simulations also suggest La Niña-dominated conditions and enhanced Walker circulation during the LIA and El Niño-dominated conditions during the MWP and CWP (48) (Fig. 3G). All lines of evidence suggest ENSO influence on CIP rainfall variability. ENSO can impact tropical rainfall through the east−west displacement of the ascending and descending branches of the Walker circulation (49). During El Niño (La Niña) conditions, the ascending branch of the Walker circulation moves eastward (westward), and the increased descent (ascent) is distributed over CIP, suppressing (enhancing) monsoon rainfall in this region.

Our results contrast with hydroclimate records from western United States (12) and sea surface temperature reconstructions from the tropical Pacific (10, 11) which suggest La Niña-like conditions during the MWP and El Niño-like conditions during the LIA. Conroy et al. (11) suggested the disparity of ENSO reconstructions might result from 1) the age model errors inherent in ¹⁴C-dated records, 2) North Atlantic temperature’s impact on the precipitation over the western United States, and 3) different spatial expressions of El Niño and La Niña events. Our rainfall reconstruction, showing dry MWP and wet LIA in the CIP, is robust, with superior chronological control. If a strengthened zonal sea surface temperature gradient across the tropical Pacific during the MWP, as suggested by marine proxies, is confirmed, the physical mechanism of ENSO, Walker circulation, and rainfall in the tropical Indo-Pacific needs to be reconsidered. Additional high-resolution, radiometrically dated climate records covering the past 1,000 to 2,000 y from different regions will contribute to a comprehensive understanding of past ENSO and Walker circulation variability, impacts, and their responses to global climate change.

The drought lasting many decades during 1220–1300 AD coincides with the largest volcanic eruption, in 1257 AD (50), over the past 2,000 y (Fig. 3H). Decreased rainfall in the TK record is also observed during the 2 large eruptions of 1600 AD (51) and 1815 AD (52), suggesting a consistent response of tropical rainfall to volcanic eruptions (53). However, the annual-scale volcanic aerosol effect (53) could not have been the primary driving force of droughts in the tropics lasting decades. As shown in Fig. 3H, the intensities of volcanic eruptions in the 20th century and during the MWP are smaller than those during the 13th century, but the droughts during the 2 periods apparently were much more severe. The timing and magnitude of the rainfall changes in the CIP imply that forcings in addition to volcanic aerosols, as well as natural climate variability, affect tropical rainfall.

ITCZ Shifts over the Past 2,000 y

In addition to zonal Walker circulation variability induced by ENSO, tropical rainfall is also influenced by meridional ITCZ shifts (1, 3, 15). Considering the synchronous, direct impact of ENSO on rainfall variations in the CIP of both hemispheres (Fig. 1C), the history of ITCZ shifts is revealed from antiphased rainfall differences between the NCIP and SCIP. Indeed, the observed annual mean cross-equatorial rainfall gradient between the 2 sectors (SI Appendix, Fig. S13) for the years 1983–2012 AD supports this argument (SI Appendix, Fig. S12). The time series of this gradient shows a significant positive correlation (r = 0.57, P < 0.01) with the observed location of the ITCZ in boreal summer over this region (1) (SI Appendix, Fig. S14). Higher (lower) rainfall in the NCIP relative to the SCIP corresponds to a northward (southward) movement of the ITCZ.

The TK δ¹⁸O record from Klang Cave in southern Thailand and the stalagmite δ¹⁸O record (LL) from Liang Luar Cave (8°32′S, 120° 26′E) in eastern Indonesia (3) (Fig. 1) are negatively correlated with the regional rainfall amount in the NCIP and SCIP, respectively. Although other factors, including cave air temperature (related to regional sea surface temperatures), surface seawater δ¹⁸O, and ENSO (3), might exert a relatively minor influence on the stalagmite δ¹⁸O in both caves, their impacts are similar due to homogeneous environmental conditions in both regions. We thus construct a new ITCZ shift index (SI) sequence for the past 2,000 y (1−2004 AD) by subtracting the standardized LL record from the standardized TK record (Fig. 4 and Materials and Methods). Negative (positive) ITCZ SI values represent higher (lower) rainfall in the NCIP as relative to the SCIP, and a northward (southward) shift of the CIP ITCZ.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Reconstruction of the ITCZ SI for CIP during the past 2,000 y (1−2004 AD). (A) Composite TK δ¹⁸O record. (B) Normalized annual TK δ¹⁸O record (TK_n). (C) Original LR06-B1 δ¹⁸O record (LL) from eastern Indonesia (3). (D) Normalized annual LL δ¹⁸O record (LL_n). Red lines in B and D are the 20-y moving averages of the normalized TK and LL records, respectively. (E) The ITCZ SI series (green line) results from subtracting the smoothed LL_n dataset from the TK_n record, and its comparison with the temperature gradient between the Northern Hemisphere (NH) and Southern Hemisphere (SH) extratropics (red line, ref. 56).

The ITCZ SI sequence (Dataset S2) shows a general trend of southward movement over the past 2 millennia (Fig. 4E). Superimposed on the long-term trend, it reveals a relatively northern mean ITCZ position during the first century, followed by a southward shift during the second century. From the third century AD onward, the CIP ITCZ began to move northward, and maintained its most northern position between the fourth and early seventh centuries. Thereafter, an abrupt 40-y-long southward movement was followed by multidecadal oscillations for the next 350 y, until another rapid southward shift occurred in the early 11th century. Since the late 11th century, a northward−southward−northward pattern of ITCZ movements is observed, with each phase lasting ∼200 y. The CIP ITCZ retracted southward since the 18th century, and reached its most southerly position in the 1980s (Fig. 4E).

The reconstruction reveals a southward mean ITCZ position during the early MWP and the CWP. Spectral analysis indicates significant ∼1,000-, ∼400-, 130- to 140-, and ∼60-y cycles in the ITCZ SI record (SI Appendix, Fig. S15). Further ensemble empirical mode decomposition analysis suggests that these cycles contribute 6%, 54%, 28%, and 8%, respectively, of ITCZ variance during the last 2,000 y (SI Appendix, Fig. S16). The 130- to 140- and ∼60-y cycles are significant throughout the entire ITCZ SI reconstruction (SI Appendix, Fig. S17).

Our ITCZ SI record is consistent with a discontinuous reconstruction of ITCZ rainfall changes in the ETP during some events (54) (SI Appendix, Fig. S18). The only discrepancy is observed in the seventh century, when the reconstructed ITCZ rainfall in the ETP shows an unusual southward shift (54). While our ITCZ SI also suggests a southward movement during this period, the amplitude was smaller (SI Appendix, Fig. S18). We also find a weak but significant negative correlation between our new ITCZ SI and a marine sediment Ti record (r = −0.17, P < 0.01) from the Cariaco Basin (15), with relatively low ITCZ SI values (interpreted as northward ITCZ movement) corresponding to rainfall-induced high Ti content (SI Appendix, Fig. S18). Modern observations of rainfall over the CIP (SI Appendix, Fig. S14) and North Africa confirm a northward ITCZ shift since the 1990s (55), which is also recorded in our SI record (Fig. 4E). The correspondence between proxy records provides additional evidence for large-scale movements of the ITCZ over the past 2,000 y. Discrepancies between the records could be attributed to the complex relationship between the biomarker index/grain size variations in lake sediments and rainfall over the ETP, on one hand (54), and regional differences between ITCZ movements in the CIP and the ETP, on the other hand. The differences between our ITCZ SI time series and the Cariaco Basin Ti record most likely results from ENSO’s influence on rainfall in the tropical Atlantic, as well as a regionally different behavior of the ITCZ in the Atlantic versus the Pacific. Our record suggests that ITCZ migrations on millennial to multicentennial timescales are coherent with the temperature contrast between the northern and southern extratropics (56) (Fig. 4E). According to the atmospheric energy balance ITCZ mechanism (1), a warming of the northern relative to the southern extratropics differentially reduces meridional temperature gradients in the northern extratropics. This can reduce the energy export out of the tropics into the Northern Hemisphere, resulting in a northward ITCZ repositioning. In contrast, warming of the southern relative to the northern extratropics can lead to a southward ITCZ shift (1).

The southward mean position of the ITCZ is concurrent with enhanced El Niño conditions during the early MWP and the CWP. This is coherent with observations of a southward shift of the ITCZ in the Indo-Pacific during strong El Niño events (such as in 1983 AD and 1998 AD) (57). Our highly resolved and accurately dated stalagmite record suggests that the drying trend in the northern tropics since the 20th century is similar to that during the historical warm period (950–1150 AD), which was caused by intensified El Niño activities and a southward ITCZ shift. Anthropogenically forced regional rainfall changes may, therefore, remain difficult to detect against the relatively large background of natural hydroclimate variability.