To explore the seasonal to decadal variability in surface water masses that contribute to the Indonesian throughflow, we have generated a 115-year bimonthly coral-based radiocarbon time series from a coral in the Makassar Straits. In the pre-bomb (pre-1955) era from 1890 to 1954, the radiocarbon time series occasionally displays a small seasonal signal (10–15‰). After 1954 the radiocarbon record increases rapidly, in response to the increased atmospheric 14C content caused by nuclear weapons testing. From 1957 to 1986 the record displays clear seasonal variability from 15 to 60‰ and the post-bomb peak (163 per mil) occurred in 1974. The seasonal cycle of radiocarbon can be attributed to variations of surface waters passing through the South Makassar Strait. Southern Makassar is under the influence of the Northwest Monsoon, which is responsible for the high austral summer radiocarbon (North Pacific waters) and the Southeast Monsoon that flushes back a mixture of low (South Pacific and upwelling altered) radiocarbon water from the Banda Sea. The coral record also shows a significant 14C peak in 1955 due to the bomb-14C water advected into this region from nuclear weapons tests in the Marshall Islands in 1954.
 Throughflow variability and flow volume are correlated with the state of ENSO [Gordon et al., 1999]. During La Niña events an increase in the sea surface height in the Western Pacific Warm Pool region results in an increase in the dynamic height gradient with the Indian Ocean and a corresponding increase in the throughflow volume, with the converse occurring during El Niño events [Gordon et al., 1999]. Vertical mixing influences the composition of the ITF. Heat and freshwater are transported down toward the thermocline and cool water mixes upward potentially influencing the atmosphere-ocean heat flux [Ffield and Gordon, 1992, 1996]. The ITF results in a significant export of heat and freshwater from the tropical Pacific into the Indian Ocean and may influence atmosphere-ocean coupling, tropical SST patterns and the Asian Monsoon [Schneider, 1998]. One way to examine the sources and variability of the throughflow is to use a water mass tracer. The radiocarbon (Δ14C) content of waters can be used to track ocean currents, vertical mixing, and air-sea CO2 exchange [Druffel, 1985, 1987; Guilderson et al., 1998, 2000a, 2000b; Moore et al., 1997].
 Measurements of subannual samples from coral skeletal material provide a proxy time series record of the Δ14C of the dissolved inorganic carbon (DIC) of the surrounding seawater [Druffel, 1981; Druffel and Suess, 1983; Guilderson et al., 1998; Moore et al., 1997]. In this paper, we present the radiocarbon content in roughly bimonthly samples from a coral in the Makassar Strait to study the processes governing water mass evolution in the Indonesian Seas over time scales long enough to study the linkage between processes in the Indonesian Seas and decadal modulation of ENSO, Asian Monsoon and global climate.
 A 2.3m core was drilled from a Porites lutea coral off Langkai Island (5°02′ S, 119°04′E; Figure 1) [Moore et al., 1997]. The core was cut to a thickness of 9 mm [Moore et al., 1997] and sonicated in distilled water. This study builds on the original Langkai Δ14C data set that spanned 1970–1985 published by Moore et al. . The samples reported here were remilled from the same coral core as Moore et al. along a main growth axis. Samples were milled sequentially in 1.8 mm increments using a manual low-speed mill with a 2 mm diameter mill bit, corresponding to ~6–8 samples per year. 14C sample splits (8–10 mg) were reacted in individual chambers (BD Vacutainer), evacuated, heated and acidified with orthophosphoric acid at 90°C [Guilderson et al., 1998, 2000b]. The CO2 was purified, trapped, and converted to graphite using an iron catalyst following a method similar to that described by Vogel et al. . The graphite targets were analyzed at the Center for Accelerator Mass Spectrometry, Lawrence Livermore National Laboratory. The 14C results are reported as age-corrected Δ14C (‰) as defined by Stuiver and Polach  and include δ13C correction for isotope fractionation, and a blank subtraction based on 14C-free calcite. Radiocarbon accuracy and precision is ±3.5‰ (1σ) based on long-term statistics of secondary standards. A coral tertiary standard was analyzed with the unknown coral samples with a resulting total precision of ±0.75‰ (1σ) (n = 88) (Figure 2). This result is consistent with the analytical uncertainty of the measurements. Stable isotopes (δ13C and δ18O) were measured on sample splits (∼100 μg of material) and analyzed on a Finnigan MAT 252 with a Kiel carbonate device using orthophosphoric acid at 70°C. Stable isotope data are reported relative to the Vienna Pee Dee Belemnite (V-PDB) with an average analytical precision of ±0.03‰.
 An initial age-model and time series was constructed using the seasonal variability in coral δ13C. In some tropical reef locations the coral δ13C is modulated by seasonal light cycles. Higher light levels are associated with more positive δ13C whereas lower light levels are associated with more negative δ13C [Fairbanks and Dodge, 1979; McConnaughey, 1989]. For this project we are assuming that the coral δ13C primarily tracks variability in solar insolation and that δ13C inflection points are interpreted as markings of seasonal change. In this part of Indonesia, the maximum cloudiness and maximum rainfall occur in February/March. To generate the primary time series we pinned the minimum δ13C to February of each year, and linearly interpolated the data between marker points. We then used the seasonal cycle of δ18O, which is influenced by water temperature and fresh water to fine-tune the time series [Charles et al., 2003]. Coral δ18O is negatively correlated with water temperature, the summer δ18O coral signal is heavily influenced by rainfall and runoff, so we matched the coral δ18O maxima to the winter SST minimum from the GOSTA SST atlas [Bottomley et al., 1990]. With our sampling resolution this results in a time series that is precise to ±2 months.
3. Results and Discussion
 The coral Δ14C time series spans 1870 to 1990. The pre-bomb interval (1870–1950) has an average Δ14C value of −56.5‰ (Figure 3a) and does not exhibit a consistent seasonal cycle. During this period there is a very weak secular decrease of ~8‰ (based on a least squares fit). Δ14C begins to slowly increase in January 1954 with a more rapid increase in February 1955 (Figure 3). The first large, sustained peak above the pre-bomb average value occurs in July 1955 (Figure 3), where values jump from −50‰ in January 1955 to −7.2‰ in July 1955. After this early peak, Δ14C values decrease until October 1957 after which time they begin to slowly rise until after 1962 when they rise rapidly to a post-bomb max of 163.3‰ in January 1974 (Figure 3). The peak occurs at approximately the same time as that in the northern hemisphere subtropics [Guilderson et al., 2000b] which implies a lateral mixing process dominated by N. Pacific source waters. The pre to post bomb Δ14C amplitude is 220‰. A strong Δ14C seasonal signal occurs after 1957 with varying amplitude ranging from 15 to 65‰ (Figure 3b).
 The seasonal cycle of radiocarbon can be attributed to variations in the 14C of surface waters passing the coral in Makassar Strait. During the NW Monsoon (January–March) Gordon et al.  showed that low salinity surface waters from the Java Sea (due to high rainfall/runoff) can lower the temperature and salinity of water entering the Indian Ocean (Figure 4a). The NW monsoon wind pushes the low-salinity surface water from the Java Sea into South Makassar and possibly even northward into the strait itself. This low-salinity surface water also flows into the Banda Sea and ultimately the Indian Ocean. During the opposite season (Southeast Monsoon) the winds push the Banda Sea surface waters westward back toward South Makassar and perhaps even slightly northward due to the northward flowing meridional winds [Gordon et al., 2003]. Sea surface temperature and salinity records indicate the Banda Sea is an area of intense vertical mixing (upwelling), resulting in lower surface seawater 14C content [Cresswell and Luick, 2001; Ffield and Gordon, 1996]. We attribute the coral seasonal radiocarbon signal to this seesaw effect of the monsoon.
 The strong seasonal Δ14C cycle observed in this coral displays high radiocarbon values during the austral summer (February). The seasonal maximum Δ14C values are associated with the maximum eastward component of the Java Sea Zonal wind and the fresh/warm water signal preserved by the coral δ18O record [Moore et al., 1997; Charles et al., 2003]. At first this seems to fit nicely with the Gordon et al.  idea of fresh warm water from the Java Sea during the NW monsoon being advected into the southern end of Makassar Strait. However, the high coral radiocarbon values are more indicative of unmodified North Pacific source waters entering Makassar Strait and flowing southward past the Langkai coral site and not waters arriving via the Java Sea. We suggest that there may be northward flowing low-salinity water from the Java Sea into Makassar Strait (as per Gordon et al. ) but it is tightly constrained along the western boundary of the Makassar Strait. (Figure 4a).
 Seasonally low radiocarbon values occur during the austral winter when the zonal wind component is at its westward maximum, the meridional component is northward, and the coral δ18O indicates cool/dry conditions (Figure 5) [Moore et al., 1997]. We infer that the low radiocarbon waters are sourced from the Banda Sea (East Indonesian Seas) (Figure 4b), which has inputs from the South Pacific (lower in Δ14C than N. Pacific water) modified by mixing with upwelled, subsurface, lower-14C water. During the NW monsoon the Banda Sea acts as a reservoir, filling up and deepening the thermocline by ∼55 m [Cresswell and Luick, 2001; Ffield and Gordon, 1996]. The shoaling of the thermocline during the austral winter SE monsoon and the extensive upwelling in the Banda Sea can account for the low 14C surface waters [Wyrtki, 1961]. These data imply that surface water flows from the Banda Sea back into the southern confines of Makassar Strait during the SE monsoon. (Figure 4b). This supports the ideas of the “flush back” of Banda Sea waters of Gordon et al. . Therefore our coral record observes the movement of high radiocarbon surface waters through Makassar Strait during the warm/wet NW monsoon and the “flush back” of low-14C “upwelling altered” surface waters from the Banda Sea during the cooler SE monsoon. Our interpretation is slightly different to the one proposed by Moore et al. They argue that the seasonal cycle is due to seasonal radiocarbon variability in the open waters of the Pacific as inferred from a 4-year 14C-coral series at Guam. The “seasonal” cycle at Guam is not very clear or consistent and does not fit the timing of warm water temperature and high Δ14C [see Moore et al., 1997, Figure 3].
4. Early Bomb Peak
 This coral Δ14C record exhibits an interesting set of radiocarbon peaks commencing in February 1955. The coral/seawater Δ14C DIC increases by 40‰ from −50‰ to −10‰ in 3 months with high values continuing until November 1955 (Figure 6). This is a tremendous change and a significant source of 14C is needed to alter the seawater Δ14C DIC. During the first part of 1954 the United States of America conducted a series of thermonuclear weapons tests at Bikini Atoll (11° 35N, 165° 23E). This started with the 15 MT “Bravo” test on 1 March 1954, followed by tests on 27 March (11 MT), 26 April (6.9MT) and 5 May (13.5 MT) [Yang et al., 2000]. A significant labeling of waters around the Marshall Islands and to the west must have take place either by gas phase CO2 exchange or particulate fallout (adsorbed onto CaCO3 particles). The large rise observed in this coral suggests that fallout and not air/sea CO2 exchange must be the source of the 14C signature in the surface waters observed at Langkai. General open ocean air-sea CO2 exchange is too slow to account for the rapid rise in coral/seawater Δ14C.
 The general westward flow of the North Equatorial Current (NEC) must have brought this labeled water toward the Philippines and into the ITF via the Mindanao Current (Figure 1). Seawater 90Sr measurements in the Philippines, show high levels (850 dpm/100 L of seawater) by the spring of 1955 suggesting that weapons test labeled water was there in high concentration [Miyake and Saruhashi, 1958]. Since the waters of the Philippines and the Mindanao Current are sourced from the NEC it follows that this labeled water entered the ITF and flowed past the coral site at Langkai. In an elegant study, Toggweiler and Trumbore  reported elevated 90Sr measurements in coral skeletal material from Cocos Island (12.5 S, 97E Indian Ocean) band counted to the year 1955. Konishi et al.  observed a 40‰ increase in Δ14C in a coral from Okinawa from 1955 to 1956. All of these observations suggest a 1 yr. transit time for waters originating in the Marshall Islands to reach the bottom of Makassar Strait.
 Another interesting feature of this part of the record is that the high Δ14C values occur during May–November 1955 (the SE monsoon) and not during the NW monsoon as the post 1958 record (Figures 2 and 5). The subsequent year, 1956, has 2 cycles peaking in April and November (Figure 6). The time period 1954 to 1957 was a moderately strong La Niña, suggesting that the surface wind and sea level pressure field during the La Niña event (more North Pacific water) and not the Indonesian monsoon was the dominant surface flow during this time period. Since there was no testing in 1955 the two peaks in 1956 indicate that some type of “storage” or new input of labeled (high Δ14C) water occurred. The peaks in 1957 could be due to further testing in 1956 (Figure 6).
4.1. Pre-Bomb Era
 The Langkai pre-bomb (1870–1950) Δ14C record exhibits an overall secular decrease of ∼8‰. Although consistent with the dilution of atmospheric 14C by the combustion of 14C-free fossil fuel carbon (the so called Suess effect) and its subsequent transfer into the surface ocean, we do not believe that the time series is only influenced by the Suess Effect. Δ14C values between 1870–1890 are ∼5‰ higher than the 1891–1905 average (Figure 7a), thus most of the decrease occurs prior to a significant decrease in atmospheric Δ14CO2, particularly when considering the decadal time constant of air-sea isotopic equilibrium. We hypothesize that the reason for the decrease is a change in either the proportion of Banda Sea water ‘back flushing’ the Southern Makassar Strait, or an increase in the upwelling in the Banda Sea that are coincident with an overall decrease of higher Δ14C waters sourced from the Northern hemisphere subtropics.
 Superimposed on this shift are interannual Δ14C oscillations that may be related to the longer-term variations in the monsoon system or the redistribution of Pacific surface waters driven by ENSO. The coral δ18O record (not shown) suggests that 1876–1878 was a significant drought on the same order of magnitude as the 1982/1983 El Niño drought. This time period corresponds to the low Δ14C seen in Figure 7a (arrow). This low period is surrounded by higher Δ14C in 1874/75 and 1879/80, which may be due to stronger La Niña conditions bringing more N. Pacific water through the ITF and less dilution by Banda Sea water during the southern hemisphere winter. In the instrumental record, stronger than average El Niño events often are terminated by stronger than average La Niña events. In the context of the weaker pre-bomb spatial surface ocean Δ14C gradients, only strong events are likely to provide a consistent signal in our surface water reconstruction. The pre-bomb interval is dominated by an inconsistent seasonal Δ14C cycle, although some years do appear to follow the δ18O cyclicity similar to the post-bomb era (data not shown). The lack of a distinct seasonal cycle is due to the lower spatial Δ14C gradient between the water masses prior to nuclear weapons testing.
 To examine the entire coral record in relation to ENSO frequency fluctuations of ITF water we applied a (2.6–8 yr.) band-pass filter to the coral Δ14C data and to the Southern Oscillation Index (Figures 7b and 7c). The comparison implies a gross coupling between the two records (Figure 7b). Higher Δ14C occurs during positive SOI events, which is analogous to the SE Monsoon and greater throughflow (more North Pacific water) (Figure 7b). Also seen are low Δ14C and negative SOI (El Niño events), analogous to the NW Monsoon or less throughflow (less North Pacific water) (Figure 7b). We remind the reader that the amplitude of surface Δ14C is not a direct proxy for the strength and intensity of the ENSO event, particularly during the post-bomb interval [Guilderson et al., 1998]. The Makassar record is a reflection of changes in the Δ14C signature of waters that are sourced into the ITF. These dynamic processes are imprinted upon an evolving Δ14C background where, in the post-bomb period, the bomb-14C transient is being mixed into and with waters with less 14C.
 The two records are coherent during the 1889 and 1905 ENSO events; the strong events of 1878 also appear in the Δ14C record (Figure 7b). Throughout the record there are mismatches, especially in the post-bomb (1950–1990) period that may be due to the variable nature of ENSO and a temporally varying Δ14C pattern. We also cannot discount the effect of the monsoon winds and their interactions, both positive and negative with ENSO, causing different surface circulation in southern Makassar resulting in misfits between the two data sets. Thus the Δ14C time series reinforces that the Indonesian Seas’ surface water is a complex interaction between seasonally varying local winds, and externally forced variability that incorporates both ENSO and the Asian Monsoon. This complex interaction excludes the common and simplistic ENSO-centric interpretation of surface ocean 14C time series. Additional and strategically located coral-based records in concert with data-model comparisons [Rodgers et al., 2004] will provide a strong tool to explore the seasonal to decadal controls on the mixing and dynamics of the Indonesian Seas.
4.2. Implications for ITF Surface Circulation
 The surface coral radiocarbon record presented here documents a complicated surface water circulation pattern of North Pacific waters arriving during the SH summer NW Monsoon and a flush back of mixed (S. Pacific and upwelling altered) waters from the Banda Sea during the SE Monsoon. Recent studies have shown that the main flow of the ITF is at depths of 200–400m and that surface flow can act independently of the subsurface flow [Gordon et al., 2003]. Although not a panacea to place all variability in a dynamical context, El Niño and La Niña events do manifest themselves as surface water mass variations that are often times discernible in the Δ14C time series. A potential measurement of surface water transit speeds in the West Pacific is provided by the arrival of a 14C signature resulting from early weapons tests (in 1955–57) that documents a transit time from the Marshall Islands through the ITF of approximately one-year.
 We thank Chris Charles and Michael Moore for access to the Langkai coral samples and, at the time, unpublished data. Paula Zermeño, Dot Kurdyla, and Sam Heller assisted in the graphite lab pressing graphite targets. Special thanks to Dave Mucciarone and Rob Dunbar for allowing us to analyze stable isotopes at Stanford University. This work was performed under the auspices of the U.S. Department of Energy by the University of California, Lawrence Livermore National Laboratory under contract W-7405-Eng-48. Funding for this project was supplied by UC/LLNL LDRD 01-ERI-009 to T.G. Data will be archived at WDC-A, Boulder CO.