The Indonesian throughflow (ITF) is an important component

Transcription

1 ARTICLES PUBLISHED ONLINE: 25 JULY 21 DOI: 1.138/NGEO92 Holocene evolution of the Indonesian throughflow and the western Pacific warm pool Braddock K. Linsley 1 *, Yair Rosenthal 2 and Delia W. Oppo 3 High sea surface temperatures in the western Pacific warm pool fuel atmospheric convection and influence tropical climate. This region also hosts the Indonesian throughflow, the network of currents through which surface and thermocline waters are transported from the western equatorial Pacific Ocean into the Indian Ocean. Here we show, using records of the δ 18 O and Mg/Ca of planktonic foraminifera from eight sediment cores, that from about 1, to 7, years ago, sea surface temperatures in the western sector of the western Pacific warm pool were about.5 C higher than during pre-industrial times. We also find that about 9,5 years ago, when the South China and Indonesian seas were connected by rising sea level, surface waters in the Makassar Strait became relatively fresher. We suggest that the permanent reduction of surface salinity initiated the enhanced flow at lower, thermocline depths seen in the modern Indonesian throughflow. However, the uniformly warm sea surface temperatures found upstream and downstream of the Indonesian throughflow indicate that the early Holocene warmth in this region was not directly related to reduced heat transport by the throughflow that may have resulted from surface freshening of the Makassar Strait. Instead, we propose that the elevated temperatures were the result of a westward shift or expansion of the boundaries of the western Pacific warm pool. The Indonesian throughflow (ITF) is an important component of the global thermohaline circulation 1,2. As the only lowlatitude interocean conduit, it annually transports a large amount of water (1 15 Sv, where one Sverdrup equals one million cubic metres per second) and heat (.5 PW, where 1 PW = 1 15 W) from the western Pacific warm pool (WPWP) north of the Equator to 12 S in the Indian Ocean. The WPWP is part of the Indo-Pacific warm pool (IPWP) and is known to be influential to the dynamics of the El Niño/Southern Oscillation and the East Asian monsoon (Fig. 1). Throughout the IPWP, warm water fuels atmospheric convection and heavy rainfall with a net 1 2 m yr 1 of freshwater input to the ocean 3,4. The warmest region of the IPWP is the WPWP region east of Indonesia where the surface ocean temperature is >29 C(Fig. 1a). The relative warmth of this region leads in part to the 1 cm average difference in sea level between the WPWP and eastern Indian Ocean that drives the ITF (ref. 5). Despite the importance of this region, we have a very limited understanding of long-term variability of the WPWP and ITF, and of the relationship between ITF transport and WPWP surface properties. Here we examine sea surface temperature (SST) and surface salinity variations in the WPWP and ITF region over the Holocene epoch using a network of sediment cores from the region. At present, the ITF exports relatively cool and fresh water to the Indian Ocean thermocline 6. Approximately 8% of the total 1 15 Sv of the ITF moves through the Makassar Strait with maximum annual average volume transport between 15 and 4 m (refs 6 9). During the boreal winter monsoon (January April/May) eastward zonal winds drive low-salinity South China Sea (SCS) water southeast across the Sunda shelf (Java Sea) into the southern Makassar Strait, creating a northward pressure gradient in the surface layer (see Fig. 1b, Supplementary Fig. S1 and Supplementary Information). During these months salinity throughout the upper 5 m of the water column in the entire southern Makassar Strait drops to pss (practical salinity scale), 2 3 pss lower than during the boreal summer monsoon (July September) (see Supplementary Fig. S1). Concurrently, there is a small increase in surface salinity at the Sulu Sea, the open western Pacific and to a lesser extent in the Banda Sea (see Fig. 1). The upper 5 m plug of relatively fresh water in the southern Makassar region inhibits the flow of warm surface water from flowing southward into the Indian Ocean. The summer wind reversal eliminates the obstructing pressure gradient by transferring more saline Banda Sea water into the southern Makassar Strait (see Fig. 1c and Supplementary Fig. S1). Mooring observations during the boreal winter show that the meridional velocity profile in the Makassar Strait has a pronounced vertical shear, with surface flow in the opposite direction of thermocline flow 6. The net result is a transport-weighted ITF temperature of slightly less than 15 C (ref. 1) and the overall flux of relatively cool and fresh water into the upper thermocline of the eastern Indian Ocean. Modelling results with and without this low-salinity SCS inflow into the Makassar Strait suggest that the ITF flow is centred at 11 m depth with SCS inflow, and at the surface without SCS inflow 1. Transport-weighted ITF temperatures are 2 C higher without SCS inflow 1. This results in a.18 PW difference (32%) in southward heat transport by the ITF with or without SCS inflow (.38 PW with SCS inflow and.56 PW without SCS inflow). Thus, both modelling and mooring results suggest that SCS inflow into the southern Makassar Strait has an important impact on mean climate throughout the Indo-Pacific region. To evaluate if any link exists among regional WPWP temperatures, southern Makassar surface salinity and ITF flow, we generated new surface-water oxygen isotopic (δ 18 O) and Mg/Ca records using sediment gravity cores from the southern Makassar Strait (Table 1) and synthesize these with existing palaeo-reconstructions from the region. We analysed the tests of the mixed-layer dwelling planktonic foraminifera Globigerinoides ruber s.s. for comparison to other previously published G. ruber data in the region. Our southern 1 Department of Atmospheric and Environmental Sciences, University at Albany-State University of New York, Albany, New York 12222, USA, 2 Institute of Marine and Coastal Sciences, and Department of Earth and Planetary Sciences, Rutgers, The State University, New Brunswick, New Jersey 891, USA, 3 Department of Geology and Geophysics, Woods Hole Oceanographic Institution, Woods Hole, Massachusetts 2543, USA. * blinsley@albany.edu. NATURE GEOSCIENCE ADVANCE ONLINE PUBLICATION 1

2 ARTICLES NATURE GEOSCIENCE DOI: 1.138/NGEO92 a 2 N 2 N Indo-Pacific warm pool Latitude 1 N MD9 South China Sea 13 GGC MD62 MD41 Makassar Strait 7GGC MD81 Banda Sea MD76 2 S Western Pacific warm pool (WPWP) Sea surface temperature 1 S MD65 MD S 1 E 11 E 12 E 13 E 14 E 15 E 16 E 17 E 18 Longitude b 8 N 4 N MD41 34 ITF path MD81 c 8 N 4 N 33 MD ITF path MD81 Latitude 4 S 8 S 12 S 34 7GGC 32 MD62 13GGC MD65 January March; sea surface salinity MD MD76 Latitude 4 S 8 S 12 S 34 7GGC 33.5 MD62 13GGC MD65 July September; sea surface salinity MD78 MD E 18 E 116 E 124 E 132 E 14 E Longitude 1 E 18 E 116 E 124 E 132 E 14 E Longitude Figure 1 Sea surface temperature and salinity in the WPWP and Makassar Strait. a, Annual average SST in the WPWP. The locations of the sediment cores discussed in this study are indicated. SST data from ref. 46. b,c, Sea surface salinity during the boreal winter (northwest monsoon) (January March) (b) and boreal summer (southwest monsoon) (July September) (c). Note the location of core sites in the southern Makassar Strait (13GGC, 7GGC and MD62). The grey arrows show generalized flow of the ITF. The dark green arrows show the flow of relatively fresh water from the South China and Java seas into the Makassar Strait during the boreal winter monsoon. Salinity data from ref. 47. Table 1 Sediment cores used in this study. Core ID, location Core ID in text Lat./Long. Water depth (m) Sample interval (yr) Reference 1GGC*, Bali Basin, Makassar St. 1GGC 7 22 S, E This study 13GGC, Bali Basin, Makassar St. 13GGC 7 24 S, E This study 7GGC, SW Sulawesi, Makassar St. 7GGC 3 34 S, E This study MD , S. Makassar St. MD S, E 1, Ref. 18 MD , Sumba, Indonesia MD S, E 2,1 29 Ref. 49 MD123-78, Timor Sea MD S, E 1, Ref. 21 MD , Sulu Sea MD N, E 3,633 9 Ref. 14 MD , W. Pacific, Mindanao MD N, E 2, Ref. 15 MD , Banda Sea MD S, E 2, Ref. 15 MD1-239*, South China Sea MD N, E 1, Ref. 5 *Not used in SSTa composite reconstructions (MD9 and 1GGC). Only G. ruber δ 18 O measured on this core. SSTa calculated relative to average of SST from 4,3 4,9 yr BP, as sediments younger than 4,3 yr BP are missing from this core. Not used in SST composite reconstruction owing to geographic location. 2 NATURE GEOSCIENCE ADVANCE ONLINE PUBLICATION

3 NATURE GEOSCIENCE DOI: 1.138/NGEO92 a b SSTa ( C) SSTa ( C) Insolation; N JJA MD78 SSTa (Timor Sea) 3 MD81 SSTa (Mindanao) MD41 SSTa (Sulu Sea) MD76 SSTa (Banda Sea) MD62 SSTa (Makassar St.) 4 MD65 SSTa (Sumba Indonesia) 13GGC SSTa (Bali basin) 7GGC SSTa (Makassar St.) 5 5, 1, 15, 2, Medieval Warm Period Age (year BP) Holocene climatic optimum Age (year BP) Flooding of SSTa (all 8 Indonesian Sunda shelf WPWP cores) SSTa (just Makasar St. cores: GGC, 7GGC, MD62, MD65) SSTa from ref. 12, 29 Barbados RSL (m) Tahiti RSL (m) (for Makassar, SW Sulawesi) 2.5 3, 6, 9, 12, 15, Figure 2 Planktonic foraminifera Mg/Ca records of mixed-layer temperatures in the WPWP. a, Comparison of Indonesian and WPWP Globigerinoides ruber Mg/Ca-based surface temperature anomaly records. Anomalies are calculated as departures relative to the average of the past 2, years (except for MD41, see Table 1). The bold grey curve is June August (JJA) solar insolation at N for reference. b, 2 yr non-overlapping binned averages of all eight cores shown in a (black) and average of just the four southern Makassar region cores (green). The light green and dashed bounding lines show the standard error of all measurements in each 2 yr bin. The dark grey curves are sea level reconstructions from Tahiti 24 and Barbados 48. Makassar Strait cores extend back to 1,5 14,2 yr bp; (see the Methods section). We converted Mg/Ca to SST using a widely used calibration 11 after analysis of core-top G. ruber Mg/Ca values indicated that this relationship is appropriate for our study area 12 (see Supplementary Information). Following previous studies, we also reconstructed the δ 18 O of sea water (δ 18 O sw ) (which has a positive linear relationship with salinity) from the δ 18 O of G. ruber and our Mg/Ca-based SST estimates A uniform Holocene cooling trend in the WPWP Comparison of the eight G. ruber Mg/Ca-based SST reconstructions from the region (see Figs 1 and 2, Table 1 and Supplementary Fig. S2) indicate that there is a remarkably consistent pattern of SST variability across the entire western sector of the WPWP. Owing to the different mean annual temperatures at each site, and to avoid possible interlaboratory analytical offsets (see the Methods section), we show SST departures (anomalies, SSTa) from average SST of Insolation JJA O N (w m 2 ) Relative sea level (m) ARTICLES the past 2, years to facilitate comparison (except for MD41, see Table 1). This highlights the very similar pattern and amplitude of SST change across the region. Given the similarity of these Mg/Ca-derived SSTa reconstructions, we calculate an eight-core average SSTa record with standard error for the entire region and also a separate four-core average for just the southern Makassar Strait cores (see Fig. 2b and the Methods section). Beginning in the Last Glacial Maximum, WPWP SSTa followed the boreal summer tropical insolation trend and shows a consistent 3 C warming to the early Holocene (Fig. 2a). This warming trend preceded sea level rise by several thousand years as previously described 17. Peak temperatures.5 C above present were reached in the early Holocene throughout this entire sector of the WPWP. The early Holocene peak temperatures lasted from 1, yr bp to 7, yr bp, roughly concurrent with the timing of the Holocene Thermal Maximum in the high-latitude Northern Hemisphere 18,19. Peak WPWP temperatures remained until 7, yr bp despite declining tropical boreal summer insolation. The Holocene cooling SST trend also did not follow the Asian monsoon weakening trend, a system known to directly respond to insolation forcing 2. Combined, this indicates that early Holocene elevated WPWP SSTs were apparently not a simple response to direct insolation forcing. After 7, yr bp there was a gradual cooling to the modern with a noted regional warming during the Medieval Warm Period and subsequent.5 C cooling during the Little Ice Age as previously reported for just the Makassar Strait 12,16. Previous studies have already noted that the WPWP has cooled since the early Holocene 15,21. However, our composite reconstruction documents, for the first time, the regional consistency of the Holocene WPWP SST variability and provides more details of the temporal evolution of WPWP SST. Our SST reconstruction shows that SSTs during the early Holocene were within error or perhaps greater than SST in the Medieval Warm Period. The similarity of all eight G. ruber Mg/Ca-based SSTa records from throughout Indonesia and the far western tropical Pacific indicates that this is a robust regional signal and supports the interpretation of G. ruber Mg/Ca in this region as a proxy for SST. Given that mean annual salinity in the southern Makassar Strait is >1.5 pss fresher than the eastern Indian or western Pacific oceans (see Supplementary Fig. S1) this also argues for minimal salinity influence on G. ruber Mg/Ca in this region 22,23, and is also consistent with a >35 pss threshold for possible salinity effects on G. ruber Mg/Ca 22 (see Supplementary Information). Freshening in the Makassar Strait beginning 9,5 yr BP Armed with these regionally coherent SST reconstructions, we compare the southern Makassar Strait surface-water δ 18 O and δ 18 O sw results to comparable data from the same cores used in the composite SST reconstruction (Fig. 3 and Table 1). This comparison reveals that the southern Makassar Strait and Sumba Indonesia (Indian Ocean) G. ruber δ 18 O and δ 18 O sw records show a.3 δ 18 O and.4 δ 18 O sw decrease 9,5 yr bp relative to the cores from the Sulu Sea, Mindanao and northeastern Banda Sea. In contrast, the cores located farther east and north in the WPWP (Sulu Sea 14 ; western Pacific near Mindanao 15 ; and the Banda Sea 15 ) show a.2 increase in G. ruber δ 18 O and δ 18 O sw 9,5 yr bp. The Timor Sea δ 18 O and δ 18 O sw records show an intermediate response and the data from the SCS indicate that this basin was fresher than the other core sites until 9, yr bp. The δ 18 O sw divergence observed in the southern Makassar Strait beginning 9,5 yr bp occurred when rising post-glacial sea level reached 3 m and the flooding of the Sunda shelf reached a level where the shallowest point, the Karimata Strait, flooded, reconnecting the SCS to the Makassar Strait through the Java Sea (see Supplementary Fig. S3). This evidence indicates that low-salinity waters, which accumulated in the SCS because of its semi-enclosed NATURE GEOSCIENCE ADVANCE ONLINE PUBLICATION 3

4 ARTICLES nature before 9,5 yr bp, could now reach the southern Makassar Strait, particularly during the boreal winter (Fig. 1), and explains both the δ 18 O sw increase in the SCS and the δ 18 O sw decrease in the Makassar Strait. The freshening of the southern Makassar Strait beginning 9,5 yr bp would have resulted in a more restricted southward flowing warm surface current during the boreal winter monsoon as occurs today 6. Likewise, the salinity increase observed in cores MD41, MD81 and MD76 resembles the seasonal response during the boreal winter monsoon. Other studies have postulated that the relationship between sea level and freshening of the southern Makassar Strait should exist 1,21. However, our results are the first direct evidence that the low-salinity plug in the southern Makassar Strait developed 9,5 yr bp when the post-glacial sea level rise reached 3 m below the present sea level and flooded the Sunda shelf (Fig. 3). To more precisely evaluate the temporal evolution of the δ 18 O sw gradient between the southern Makassar cores and the cores in the western Pacific, we stacked the δ 18 O sw reconstructions for each region and binned the data into 4 yr, non-overlapping bins (Fig. 3b). We then calculated the δ 18 O sw difference between the two regions for each 4 yr bin (see Fig. 3b top and the Methods section). The difference between the southern Makassar and WPWP δ 18 O sw records effectively normalizes for global ice volume effects on these δ 18 O sw records. The relative δ 18 O sw difference between the southern Makassar region and the western Pacific increased from.3 before, to.7 after 9,5 yr bp, consistent with the re-routing of low-salinity water from the SCS to the southern Makassar Strait causing a freshening of the region. On the basis of the present mean annual mixed-layer spatial salinity pattern, if the lower salinity water influx from the Java and SCS was cut off, as would have been the case before 9,5 yr bp, mean annual salinity in the South Makassar Strait would have been approximately 1. pss higher 6 (see Supplementary Fig. S1). The.3.4 lower δ 18 O sw values in the South Makassar Strait during the 2, yr interval immediately after 9,5 yr bp (see Fig. 2b) are in line with this predicted salinity change, assuming a δ 18 O sw salinity relationship of.3/1 pss (ref. 29). This agreement between the expected and measured δ 18 O sw change after 9,5 yr bp also suggests negligible effects of possible changes in G. ruber production rate (seasonality) in the surface ocean that could have biased the results towards one season (see Supplementary Information). Although reconnection of the SCS with the Makassar Strait played a primary role in setting the δ 18 O sw difference between the Makassar Strait cores and the cores farther north and east, subsequent changes in the gradient may have resulted from changes of regional hydrography. We note that the largest offset of.7.8 is observed from 9, to 5, yr bp, possibly reflecting high regional rainfall as the continental shelf flooded 3 (see Supplementary Fig. S4 and Supplementary Information). The relative freshening of the southern Makassar Strait at 9,5 yr bp is nearly coincident with the initiation of abrupt cooling in the upper thermocline of the Timor Sea 21 (MD78; see Figs 1 and 3c). Downcore changes in Mg/Ca of the upper thermocline foraminifera Pulleniatina obliquiloculata in the Timor Sea show a 1.5 C drop just after 9,5 yr bp followed by continued cooling into the late Holocene. Combined, these data are consistent with the hypothesis that freshening of the surface ocean in the southern Makassar Strait 9,5 yr bp increased the northward pressure gradient and inhibited the flow of warmer surface-layer water into the Indian Ocean, cooling the transport-weighted mean ITF by the same amount predicted by models 1. Thus, ITF heat transport was probably enhanced in the earliest Holocene (pre- 9,5 yr bp) relative to today, and 9,5 yr bp may have marked the initiation of the thermocline-enhanced cool ITF transport that is observed today 6. a b c ) sw Upper thermocline temperature anomaly ( C) ) PDB ) δ 18 O ) 18 O δ NATURE GEOSCIENCE DOI: 1.138/NGEO92 Core key 13 GGC Bali basin MD65 Sumba Indonesia 1GGC Bali basin (18O only) 7 GGC S. Makassar St. MD62 S Makassar St. δ 18 O G. ruber Bali basin, S. Makassar Sulu Sea, Mindanao, MD76, MD65 Bali basin, S. Makassar δ 18 O sw Sulu Sea, Mindanao, MD76 MD41 Sulu Sea MD81 Mindanao MD76 Banda Sea MD78 Timor Sea MD9 South China Sea South China Sea Sunda shelf flooded 2 2, 4, 6, 8, 1, 12, 14, Age (yr BP) Difference Sunda shelf flooded Southern Makassar St. 2, 4, 6, 8, 1, 12, 14, Age (yr BP) Sunda shelf flooded 1. 2, 4, 6, 8, 1, 12, 14, Age (yr BP) Surface δ O ) δ O difference ( sw sw Figure 3 Planktonic foraminifera δ 18 O and δ 18 O sw in the WPWP and Makassar Strait. a, Downcore Globigerinoides ruber δ 18 O (top) and δ 18 O sw (bottom) records in Indonesia and the WPWP region. b, 4 yr non-overlapping binned averages of the southern Makassar cores (MD7, MD62, 13GGC, MD65) (green curve) and western Pacific cores (MD41, MD81, MD76) (orange curve). The black curve is the difference between the two composite binned curves. The light green and orange bounding lines indicate the standard errors of data in each 4 yr bin. c, MD78 (near Sumba Indonesia) Pulleniatina obliquiloculata Mg/Ca upper thermocline temperatures (8 12 m). Early Holocene WPWP expansion and/or repositioning Despite this change in heat export through the ITF 9,5 yr bp, regional WPWP SST and SST gradients seem unaffected by the freshening event in the southern Makassar. As the broad region of early Holocene surface ocean warming in the WPWP includes areas that would both cool and warm under conditions of restricted surface ITF flow through the Makassar Strait (from areas both downstream and upstream of the Makassar Strait, respectively), the elevated early Holocene Indonesian temperatures do not seem to be related to reduced ITF heat transport after the 9,5 yr bp freshening event. To explain the early Holocene warmth in the WPWP, any mechanism must account for the regional consistency of the ) ) 4 NATURE GEOSCIENCE ADVANCE ONLINE PUBLICATION

5 NATURE GEOSCIENCE DOI: 1.138/NGEO92 temperature response across this broad region. This requirement would seem to rule out upwelling effects in the southern sector of the region, heat trapping by an enhanced low-salinity barrier layer 31,32 and East Asian monsoon 2 or Australian Indonesian summer monsoon 3 changes, all of which would affect SST in a non-uniform way across the study area. Our preferred explanation for the early Holocene elevated SSTs in the far western equatorial Pacific is a westward shift of the region of maximum temperatures >29 C in the WPWP and/or an expansion of the region of most elevated WPWP temperatures to the west. Today the region of temperatures >29 C is located between 13 E and 17 E (see Fig. 1). A shift of this warm water to the west would explain our observations of elevated SSTs in the early Holocene. This inferred change in the WPWP may have been independent of, but is also consistent with, the hypothesis that El Niño activity gradually increased over the Holocene Under this scenario, as the mean position of the WPWP gradually shifted to the east in the middle to late Holocene, El Niño events and warming in the eastern Pacific became more frequent and the trade winds may have weakened. Another related possibility is that elevated early Holocene WPWP temperatures were associated with an overall warming and/or expansion of the WPWP and not just a re-positioning of the region of maximum SST. An increase in the area-averaged heat content in the WPWP mixed layer would have had global consequences by triggering feedbacks on ocean atmosphere interactions in the WPWP, tropical atmospheric water vapour and heat, and, in the absence of other influences, increase Hadley cell strength. Either WPWP temperature scenario for the early Holocene supports a century-scale ocean dynamical control of WPWP SST as has been proposed on shorter timescales 37,38. Present climate models do not capture the early Holocene warmth in the tropics indicated by the WPWP Mg/Ca-based SST estimates. Rather, in atmosphere ocean general circulation models, changes in mean annual insolation forced by obliquity changes dominate the tropical response, and result in an early Holocene cooling and mid Holocene warming throughout most of the tropics However, a warmer state for the early Holocene WPWP is consistent with evidence for a greater mid-holocene zonal Pacific SST gradient 36 and more generally with the clear evidence for early Holocene Northern Hemisphere warmth as also recorded in Greenland ice cores 19. Our Indonesian WPWP results lead to two important conclusions: (1) there was a major change in the nature of the ITF following the Sunda shelf flooding 9,5 yr bp that probably resulted in a decrease in ITF heat transport; (2) early Holocene WPWP SSTs seem to be insensitive to this change, suggesting that other mechanisms may compensate for the changes in the ITF heat transport, keeping SSTs in the WPWP relatively stable. One possibility is that air sea heat and moisture exchange in the WPWP surface layer quickly minimizes the small regional SST differences that might have occurred because of reduced ITF heat transport following Sunda shelf flooding. In accord with modelling studies 1,42,43 it is likely, however, that the development of the early Holocene low-salinity plug in the southern Makassar Strait and consequent reduction in ITF heat transport 21 had a greater effect on sea surface and upper thermocline temperatures in the central and western Indian Ocean as observed for other periods in the geological record. Analytical methods Age models. Accelerator mass spectrometry (WHOI AMS facility) 14 C dates were obtained on mixed planktonic species (primarily G. ruber and G. sacculifer) and corrected and converted to calendar age using a reservoir-age correction of 5 years for core 7GGC and 4 years for cores 1 and 13GGC (see Supplementary Table S1; ref. 44). A smaller reservoir correction was applied to the dates from cores 13 and 1GGC after considering the depth of the Tambora ash (ad 1815) in each core. The sampling interval in core 7GC was every 4 cm, yielding approximately centennial time resolution. Core 13GGC was sampled every 2 cm and core 1GGC ARTICLES every 8 cm, yielding 5 yr and 2 yr resolution respectively. In core 13GGC we identified the Mount Tambora volcanic ash layer at 3 cm depth and therefore assigned it an age of ad Mg/Ca measurements. We made Mg/Ca measurements on the mixed-layer planktonic species, G. ruber s.s. (white variety µm size) in our shallow water cores. Approximately 4 planktonic tests were picked and weighed with a Mettler M3 microbalance (nominal precision ±1 mg). After weighing, each sample was gently crushed to open the chambers and then samples were prepared using a modified reductive, oxidative cleaning and analysed at Rutgers Inorganic Analytical Laboratory using a sector-field inductively coupled plasma mass spectrometer (Thermo Element XR) following the methods outlined in ref. 45. Fe/Ca and Al/Ca were used to assess for evidence of sedimentary contamination and for evidence of other environmental changes. On the basis of past analyses, the external precision of the Mg/Ca ratio is better than 1% as determined by repeated measurements of three consistency standards. We converted Mg/Ca to SST using a calibration, Mg/Ca =.38exp(.9SST), based on seasonal Mg/Ca variations in multiple species of planktonic foraminifera from Sargasso Sea sediment trap samples 11.A summary of the cleaning protocols and Mg/Ca SST calibrations used for all of the cores discussed in this study are listed in Supplementary Table S2. Possible biases among different Mg/Ca-derived SST records might be attributable to the use of different cleaning methods. However, these offsets although affecting the absolute temperature derived from Mg/Ca records do not significantly alter the relative changes. To overcome this problem and also to normalize for different mean annual SSTs in each core site we present the data as SST anomalies (SSTa), which represent the downcore change relative to the average SST over the past 2, years at each site (see the Methods section, Fig. 2 and Supplementary Fig. S2). Oxygen isotopes. For cores 13GGC and 1GGC 2 individual G. ruber tests in the µm fraction were analysed for δ 18 O and δ 13 C by reacting them with 1% H 3 PO 4 at 9 C in a MultiPrep carbonate preparation device. The resulting CO 2 gas was analysed with a Micromass Optima dual-inlet mass spectrometer at the University at Albany, State University of New York. The standard deviation of the National Institute of Science and Technology international reference standard (NBS-19) analysed over the time of these analyses was.4 for δ 18 O. For core 7GGC, δ 18 O on G. ruber in the µm was measured at WHOI on a Finnigan MAT253 stable isotope mass spectrometer with the Kiel III carbonate device. Long-term precision of δ 18 O measurements of standards is.7. Bin lengths for data in Figs 2,3 and Supplementary Fig. S2. Given that all of the sediment cores except one (MD62) had a sample resolution <21 years (see Table 1) we have elected to bin the Mg/Ca SST estimates into 2 yr non-overlapping bins (see Fig. 2b and Supplementary Fig. S1). As the composite reconstructions shown in Fig. 2b are based on binning of raw data and not interpolation, we feel this choice of bin length accurately reflects variability in the data. For the δ 18 O sw composites shown in Fig. 3b we used a 4 yr bin size because we were interested in calculating the difference between the southern Makassar records and wanted to minimize errors in the δ 18 O sw reconstruction. Received 1 February 21; accepted 23 June 21; published online 25 July 21 References 1. Vranes, K., Gordon, Al. L. & Ffield, A. The heat transport of the Indonesian Throughflow and implications for the Indian Ocean heat budget. Deep Sea Res. (II) 49, (22). 2. Gordon, A. L. Interocean exchange of thermocline water. J. Geophys. Res. 91, (1986). 3. Chen, G., Fang, C. Y., Zhang, C. Y. & Chen, Y. Observing the coupling effect between warm pool and rain pool in the Pacific Ocean. Remote Sensing Environ. 91, (24). 4. Huang, B. & Mehta, V. M. The response of the Indo-Pacific Warm Pool to interannual variations in net atmospheric freshwater. J. Geophys. Res. 19, C622 (24). 5. Bray, N. A., Hautala, S., Chong, J. & Pariwono, J. Large-scale sea level, thermocline, and wind variations in the Indonesian throughflow region. J. Geophys. Res. 11, (1996). 6. Gordon, A. L., Susanto, R. D. & Vranes, K. Cool Indonesian throughflow as a consequence of restricted surface layer flow. Nature 425, (23). 7. Fieux, M. et al. Measurements within the Pacific-Indian oceans throughflow region. Deep Sea Res. 41, (1994). 8. Ffield, A., Vranes, K., Gordon, A. L. & Susanto, R. D. Temperature variability within the Makassar Strait. Geophys. Res. Lett. 27, (2). 9. Susanto, R.D. & Gordon, A. L. Velocity and transport of the Makassar Strait throughflow. J. Geophys. Res. 11, C15 (25). 1. Tozuka, T., Qu, T. & Yamagata, T. Dramatic impact of the South China Sea on the Indonesian Throughflow. Geophys. Res. Lett. 34, L12612 (27). 11. Anand, P., Elderfield, H. & Conte, M.H. Calibration of Mg/Ca thermometry in planktonic foraminifera from a sediment trap time-series. Paleoceanography 18, 15 (23). NATURE GEOSCIENCE ADVANCE ONLINE PUBLICATION 5

6 ARTICLES 12. Oppo, D. W., Rosenthal, Y. & Linsley, B. K. 2-year-long temperature and hydrology reconstructions from the Indo-Pacific Warm Pool. Nature 46, (29). 13. Lea, D. W., Pak, D. K. & Spero, H. J. Climate impact of late Quaternary equatorial Pacific sea surface temperature variations. Science 289, (2). 14. Rosenthal, Y., Oppo, D. W. & Linsley, B. K. The amplitude and phasing of climate change during the last deglaciation in the Sulu Sea, western equatorial Pacific. Geophys. Res. Lett. 3, 1428 (23). 15. Stott, L. et al. Decline of sea surface temperature and salinity in the western tropical Pacific Ocean during the Holocene Epoch. Nature 431, (24). 16. Newton, A., Thunell, R. & Stott, L. Climate and hydrologic variability in the Indo-Pacific Warm Pool during the last Millennium. Geophys. Res. Lett. 33, L1971 (26). 17. Visser, K., Thunell, R. C. & Stott, L. Magnitude and timing of temperature change in the Indo-Pacific warm pool during deglaciation. Nature 421, (23). 18. Renssen, H. et al. The spatial and temporal complexity of the Holocene thermal maximum. Nature Geosci. 2, (29). 19. Vinther, B. M. et al. Holocene thinning of the Greenland Ice Sheet. Nature 461, (29). 2. Wang, Y. et al. The Holocene Asian monsoon: Links to solar changes and North Atlantic Climate. Science 38, (25). 21. Xu, J., Holbourn, A., Kuhnt, W., Jian, Z. & Kawamura, H. Changes in the thermocline structure of the Indonesian outflow during Terminations I and II. Earth Planet. Sci. Lett. 273, (28). 22. Kisakurek, B., Eisenhaur, A., Bohm, F., Garbe-Schonberg, D. & Erez, J. Controls on shell Mg/Ca and Sr/Ca in cultured planktonic foraminferan, Globigerinoids ruber (white). Earth Planet. Sci. Lett. 273, (28). 23. Mathien-Blard, E. & Bassinot, F. Salinity bias on the foraminifera Mg/Ca thermometry: Correction procedure and implications for past ocean hydrographic reconstructions. Geochem. Geophys. Geosyst. 12, Q1211 (29). 24. Bard, E. et al. Deglacial sea-level record from Tahiti corals and the timing of global meltwater discharge. Nature 382, (1996). 25. Hanebuth, T., Stattegger, K. & Grootes, P. M. Rapid flooding of the Sunda Shelf: A late-glacial sea level record. Science 288, (2). 26. Peltier, W. R. & Fairbanks, R. G. Global glacial ice volume and Last Glacial Maximum duration from an extended Barbados sea level record. Quat. Sci. Rev. 25, (26). 27. Sathiamurthy, E & Voris, H.K. Maps of Holocene sea level transgression and submerged lakes on the Sunda Shelf. Nat. Hist. J. Chulalongkorn Univ. (suppl. 2), 1 44 (26). 28. Siddall, M. et al. Sea-level fluctuations during the last glacial cycle. Nature 423, (23). 29. LaGrande, A.N. & Schmidt, G.A. Global gridded data set of the oxygen isotopic composition in seawater. Geophys. Res. Lett. 33, L1264 (26). 3. Griffiths, M.L. et al. Increasing Australian Indonesian monsoon rainfall linked to early Holocene sea-level rise. Nature Geosci. 2, (29). 31. Lukas, R. & Lindstrom, E. The mixed layer of the western equatorial Pacific Ocean. J. Geophys. Res. 96, (1991). 32. Vialard, J. R. M. & Delecluse, P. An OGCM Study for the TOGA Decade. Part I: Role of salinity in the physics of the Western Pacific fresh pool. J. Phys. Ocean 28, (1998). 33. Rodbell, D. T. et al. A 15, year record of El Niño-driven alluviation in southwestern Ecuador. Science 283, (1999). 34. Moy, C.M., Seltzer, G. O., Rodbell, D. T. & Anderson, D. M. Variability of El Niño/Southern Oscillation activity at millennial time-scales during the Holocene epoch. Nature 42, (22). NATURE GEOSCIENCE DOI: 1.138/NGEO Conroy, J. L., Overpeck, J. T., Cole, J. E., Shanahan, T. M. & Steinitz-Kannan, M. Holocene changes in eastern tropical Pacific climate inferred from a Galápagos lake sediment record. Quat. Sci. Rev. 27, (28). 36. Koutavas, A., demenocal, P. B., Olive, G. C. & Lynch-Stieglitz, J. Mid-Holocene El Niño-Southern Oscillation (ENSO) attenuation revealed by individual foraminifera in eastern tropical Pacific sediments. Geology 34, (26). 37. Clement, A. C., Seager, R., Cane, M. A. & Zebiak, S. E. An ocean dynamical thermostat. J. Clim. 9, (1996). 38. DiNezio, P. N. et al. Climate response of the equatorial Pacific to global warming. J. Clim. 22, (29). 39. Liu, Z., Bradly, E. & Lynch-Stieglitz, J. Global ocean response to orbital forcing in the Holocene. Paleoceanography 18, 141 (23). 4. Lorenz, S. J., Kim, J-H., Schneider, R. R. & Lohmann, G. Orbitally driven insolation forcing on Holocene climate trends: Evidence from alkenone data and climate modeling. Paleoceanography 21, PA12 (26). 41. Jansen, E. et al. in IPCC Climate Change 27; The Physical Science Basis. 4th Assessment Report IPCC (eds Solomon, S. et al.) (Cambridge Univ. Press, 27). 42. Wajsowicz, R. C. Air sea interaction over the Indian Ocean due to variations in the Indonesian throughflow. Clim. Dyn. 18, (22). 43. Karas, C. et al. Mid-Pliocene climate change amplified by a switch in Indonesian subsurface throughflow. Nature Geosci. 2, (29). 44. Fairbanks, R.G. et al. Marine radiocarbon calibration curve spanning to 5, years B.P. based on paired 23 Th/ 234 U/ 238 U and 14 C dates on pristine corals. Quat. Sci. Rev. 24, (25). 45. Rosenthal, Y., Field, F. & Sherrell, R. M. Precise determination of element/calcium ratios in calcareous samples using sector field inductively coupled plasma mass spectrometry. Anal. Chem. 71, (1999). 46. Locarnini, R. A., Mishonov, A. V., Antonov, J. I., Boyer, T. P. & Garcia, H. E. in World Ocean Atlas 25, Temperature Vol. 1 (ed. Levitus, S.) 182 (NOAA Atlas NESDIS, Vol. 61, US Government Printing Office, 26). 47. Conkright, M. E. et al. World Ocean Database 1998 Documentation and Quality Control (National Oceanographic Data Center, Silver Spring, 1998). 48. Fairbanks, R. G. A 17, year glacio-eustatic sea level record: Influence of glacial melting rates on the Younger Dryas event and deep ocean circulation. Nature 342, (1989). 49. Levi, C. et al. Low-latitude hydrological cycle and rapid climate changes during the last deglaciation. Geochem. Geophys. Geosyst. 8, Q5N12 (27). 5. Steinke, S. et al. Proxy dependence of the temporal pattern of deglacial warming in the tropical South China Sea: Toward resolving seasonality. Quat. Sci. Rev. 27, (28). Acknowledgements We are indebted to Y. S. Djajadihardja, F. Syamsudin, the captain and crew of our 23 R/V Baruna Jaya VIII cruise, the Indonesian Agency for Assessment and Application of Technology (BPPT) and the Center of Research and Development for Oceanography (LIPI) of Indonesia for their support of this project. This research was supported by the US NSF. We thank S. Howe, S. Langton, L. Zou, D. Ostermann, K. Rose, S. Pike and M. Chong for technical assistance, A. Gordon for helpful discussions and the NOSAMS facility at WHOI. Author contributions All authors contributed extensively to this work including project planning, field and analytical work and data interpretation. Additional information The authors declare no competing financial interests. Supplementary information accompanies this paper on Reprints and permissions information is available online at Correspondence and requests for materials should be addressed to B.K.L. 6 NATURE GEOSCIENCE ADVANCE ONLINE PUBLICATION

7 SUPPLEMENTARY INFORMATION Holocene Evolution of the Indonesian Throughflow and the Western Pacific Warm Pool by Braddock K. Linsley, Yair Rosenthal and Delia W. Oppo Supplementary Discussion: Salinity Variability in the southern Makassar Strait Supplementary Figure 1 shows salinity at 2m water depth throughout Indonesian and the western Pacific (data from Conkright et al., ). The relatively low salinity in the southern Makassar Strait is evident. This is due to the seasonal influx of low salinity water from the South China Sea (SCS) and Java Sea during the boreal winter monsoon. The low salinity of the SCS and Java Sea results from heavy precipitation and river runoff from Southeast Asia and Borneo. Seasonal changes in the vertical salinity profiles in the regions of the two gray boxes are indicated in the lower half of figure. Note that the during the boreal winter monsoon, salinity throughout the entire southern Makassar Strait drops to pss (practical salinity scale), 2-3 pss lower than during the boreal summer monsoon (July-September). Supplement Page1

8 Supplementary Figure 1: Mixed layer salinity in the WPWP and seasonal vertical salinity profiles in the southern Makassar Strait. (top) Annual mean salinity at 2m depth in the study area (Salinity data from Conkright et al. (1998) 47 ). The core locations of the SST and 18 O sw reconstructions shown in Figures 2 and 3 are indicated. The gray squares in the southern Makassar Strait indicate the location of the vertical salinity climatology profiles shown in the bottom two panels (Bali Basin on left; Makassar Strait on right). These profiles show a large 2 to 3 pss salinity reduction in the upper 5-75m of the water column during the boreal winter (NW monsoon)(january-june). The magnitude of the salinity reduction is the same at both of our study sites in the southern Makassar Strait. Supplement Page2

9 Seasonality of Globigerinoides ruber flux to the sediment We considered the possible effects of changes in the seasonal flux of G. ruber to the sediments (G. ruber seasonality) in the southern Makassar Strait on our interpretations. Sediment trap studies indicate that in the tropics the seasonal preference of G. ruber varies with location Detailed sediment records over the last ~15 years in the southern Makassar Strait suggest that G. ruber seasonality varied through time with greater flux in JAS during the cooler periods 12. Over long time periods, if G. ruber is preferentially produced" during the JAS boreal summer monsoon upwelling season in the Makassar Strait, then the 18 O difference between the southern Makassar cores and the cores further east in the WPWP (MD81, MD76, and MD41) would be even greater after 9.5Kyr. This is because salinity is 2-3 pss higher in the southern Makassar Strait during the SE Monsoon but relatively unchanged over the course of the year in the cores further to the east (see Figures 1 and Supplementary Figure 1). Thus a JAS G. ruber seasonality preference in the southern Makassar Strait would tend to dampen the 18 O and 18 O sw difference between the Makassar Strait cores and the cores farther to the north and east outside of the strait. 2 year Binned Averages of G. ruber Temperatures Supplementary Figure 2 shows 2-year non-overlapping binned averages of the eight individual WPWP G. ruber Mg/Ca-based temperature records. Anomalies calculated as departures relative to average of last 2, years for each record (except for MD41, see table 1). This figure is meant to complement Figure 2A. Supplement Page3

10 Supplementary Figure 2: Globigerinoides ruber Mg/Ca SST in the WPWP. 2 year nonoverlapping binned averages of the eight individual WPWP G. ruber Mg/Ca-based temperature records. Anomalies calculated as departures relative to average of last 2, years for each record (except for MD41, see table 1). This figure is another way of depicting the data shown in Figure 2A. Possible Salinity Influence on Foraminifera Mg/Ca It has been argued that seawater salinity has an effect on the Mg/Ca in planktonic foraminfera 22,23. G. ruber Mg/Ca results for core top samples and plankton tows compared to modern mean annual salinity indicate that the effect is such that a 1 pss decrease in salinity will result in an apparent warming of 1.6 C in foraminifera test Mg/Ca,23. However, culture experiments with G. ruber suggest that a 1 pss decrease Supplement Page4

11 should result in an apparent warming of.5 C 22. It seems, however, that there is also a salinity threshold where below 35 pss there is no measureable effect on foraminifera Mg/Ca (see data in references 22,23 ). The G. ruber Mg/Ca SST estimates from the WPWP are from regions with different salinity regimes (i.e.; > 1 pss difference between mean annual salinity in the Makassar Strait in eastern Indian Ocean and western Pacific or Banda Sea)(see Supplementary Figure 1). The fact that all eight Mg/Ca-SST records show the same basic pattern from the Last Glacial Maximum to the present argues for minimal salinity influence on G. ruber Mg/Ca in this region where mean annual salinity is <34.5 pss. In addition, our calculation of SST departures (anomalies) from the last 2, year average at each site, would have the effect of minimizing any potential salinity influence on G. ruber Mg/Ca SST. Sea Level Flooding of the Sunda Shelf Using current bathymetry and reconstructed regional sea level since the Last Glacial Maximim 25,26, Sathiamurthy and Voris (26) 27 determined that the Karimata Strait in the Java Sea is the critical connection point to the SCS (see Supplementary Figure 3). The flooding of the Karimata Strait would have started around 9,5 yr BP and continued over a 2, to 3, year period until ~6, yr BP. Thus the timing of the observed lowering of surface salinity beginning ~9,5 yr BP agrees with the timing of the reconnection of the South China Sea with the Makassar Strait through the Java Sea. Supplement Page5

12 Supplementary Figure 3: Flooding of Sunda Shelf at 1,2 and 9,5 year BP. Maps depicting the distribution of land and sea in Indonesia at 1.21 Kyr and 9.53 Kyr (modified from Sathiamurthy and Voris 26)27. Green and yellow areas depict exposed land areas. The location the Karimata Stait in the Java Sea is indicated. Flooding of this sill at ~9,5-1, yr BP would have reconnected the South China Sea with the Makassar Strait. The locations of cores discussed in this study are also indicated (with the exception of MD76 which is located further east in the Banda Sea). Supplement Page6

13 Comparison to Holocene Speleothem and Coral Records in the Region In Supplementary Figure 4 we compare our SSTa and 18 O sw paleo-records to: (1.) speleothem-based 18 O data from Flores 3,55 (eastern Indonesia), (2.) to Borneo speleothem 18 O ref 56, and (3.) to uplifted fossil coral Sr/Ca estimates of SST from the southern edge of the WPWP off southern Sumatra and off northeastern Papua New Guinea (PNG) 57. The Borneo and Flores 18 O records may show evidence of a more northward position of the ITCZ in the mid-holocene near 5, yr BP. The mid- Holocene is also a relatively cooler time in the fossil coral record. This could partly be explained by the location of the coral sites on the southern edge of the WPWP in Sumatra and PNG. A more northerly position of the ITCZ in the mid-holocene could have been associated with a northward contraction of the IPWP. Our composite WPWP SSTa and 18 O sw reconstructions contain no clear evidence of concordant changes in the mid- Holocene. We note, however, that the coral SSTa records from the southern edge of the WPWP show larger amplitude Holocene variability (~-2-3 C) than the <1 C range observed in any of the WPWP foraminifera-based SSTa reconstructions. Supplement Page7

14 Supplementary Figure 4: Regional Palaeoclimatic Data in the WPWP region. Comparison of our composite SSTa and 18 O sw reconstructions to other published paleodata in the region. (a) Coral Sr/Ca SST reconstructions for corals on the southern edge of the WPWP in Sumatra and Papua New Guinea (Abram et al., 29) 57. (b) Composite SSTa reconstruction from all 8 WPWP cores shown in Figure 2B. (c) 4 year binned 18 O sw from the same cores depicted in Figure 3B along with difference between S. Makassar and openocean WPWP cores (as in Figure 3b), (d) Borneo speleothem 18 O (Partin et al., 27) 56, (e); Flores speleothem 18 O (Griffiths et al., 29; 21) 3,55 Supplement Page8

15 Supplementary Table 1. Radiocarbon measurements made on mixed planktonic foraminifera at the National Ocean Sciences Accelerator Mass Spectrometry Facility (NOSAMS) and converted to calendar age using a reservoir age of 5 years for 7GGC and 4 years for 1 and 13GGC 48,51. CORE NOSAMS ID Depth (cm) 14 C Age (yr BP) S.D. (yr) Calendar age (yr BP) S.D. (yr) BJ8 7GGC OS BJ8 7GGC OS BJ8 7GGC OS BJ8 7GGC OS BJ8 7GGC OS BJ8 7GGC OS BJ8 7GGC OS BJ8 7GGC OS BJ8 7GGC OS BJ8 7GGC OS BJ8 1GGC OS BJ8 1GGC OS BJ8 1GGC OS BJ8 1GGC OS BJ8 13GGC OS BJ8 13GGC OS BJ8 13GGC OS BJ8 13GGC OS BJ8 13GGC OS Supplement Page9

16 Supplementary Table 2: Mg/Ca cleaning protocols and Mg/Ca-SST calibrations used. Core ID; Location 13GGC, Bali Basin- Makassar St. 7GGC, SW Sulawesi-Makassar St. MD , S. Makassar St. MD , Sumba, Indonesia Core ID in text 13GGC 7GGC MD62 MD65 Lat./Long. Reference Mg/Ca cleaning method: 7 24 S, E 3 34 S, E 4 41 S, E 9 39 S, E This study This study Red/Ox: = full reductive & oxidative trace metal cleaning** Red/Ox: Rosenthal et al. (1999) 64 Red/Ox: Rosenthal et al. (1999) 64 Visser et al., Red/Ox: Boyle et al. (1995) 59 Levi et al., Red/Ox: Barker et al. (23) 58 Mg/Ca to SST calibration used Dekens et al. (22) 6, Anand et al. (23) 11 Dekens et al. (22) 6, Anand et al. (23) 11 Dekens et al. (22) 6, Hastings et al. (21) 65 Dekens et al. (22) 6 MD123-78, Timor Sea MD S, E Xu et al., Red/Ox: Martin and Lea (22) 61 Dekens et al. (22) 6, Anand et al., MD , Sulu Sea MD N, E Rosenthal et al., Red/Ox: Rosenthal et al. (1999) 64 Rosenthal and Lohmann, MD , W. Pacific-Mindano MD N, E Stott et al., Red/Ox: Martin and Lea (22) 61 Nurnberg et al. (1996) 62 MD , Banda Sea MD S, E Stott et al., Red/Ox: Martin and Lea (22) 61 Nurnberg et al. (1996) 62 MD1-239 ", South China Sea MD 'N, ' E Steinke et al., Red/Ox: Barker et al. (23) 58 Dekens et al. (22) 6, Anand et al., " = not used in SSTa composite reconstructions due to location **: all these cleaning methods relate back to the original developmental work of Boyle and Keigwin (1985) 66 and Rosenthal et al. (1997) 67 Supplement Page1