Analytical methods, age model and calibrations
For δ18Ocoral and Sr/Ca analyses, the slabs were microsampled at 0.4-mm intervals following a method developed at Australian National University (Gagan et al., 1994). The analytical procedure and the δ18Ocoral record for the years 1970–2002 have been published previously (Liu et al., 2013). We report here for the first time continuous δ18Ocoral and Sr/Ca records from the entire core. The δ18Ocoral values of samples covering the period from 1890 to 2002 was analyzed using an online system comprising an IsoPrime Isotope-ratio mass spectrometer (GV Instruments Ltd.) coupled to a Multicarb automatic sample treatment system at the Geological Survey of Japan (GSJ), whereas the δ18Ocoral values of samples predating 1890 were determined by the Center for Advanced Marine Core Research at Kochi University (KCC). All δ18Ocoraldata were normalized to Vienna Pee Dee Belemnite (V-PDB) using the NBS-19 standard (δ18O = +2.2‰) from the U.S. National Institute of Standards and Technology. The International Atomic Energy Agency standard IAEA-603 (δ18O = –2.37 ± 0.04‰) was also used for some measurements conducted at KCC. The standard deviations of replicate δ18O measurements of the standards during the mass spectrometer runs were 0.04‰ and <0.10 ‰ at GSJ and KCC, respectively.
Concentrations of Ca and Sr were measured by inductively coupled plasma (ICP) optical emission spectrometry using an IRIS Advantage system (Thermo Electron Co., Ltd) at GSJ and an Agilent 720 system (Agilent Technologies) at Okayama University. We used an analytical method that involves the insertion of a reference solution after every third sample and makes rapid determination of high-precision Sr/Ca ratios feasible (Schrag, 1999). The reference solution was made from JCp-1, a coral material with a known chemical composition provided by GSJ (Okai et al., 2002). Each sample was first weighed (100 ± 10 µg) and then dissolved in 2% HNO3. Sr/Ca analyses of samples covering the period after 1883 were conducted at GSJ, and samples predating 1890 were analyzed at Okayama University, respectively, with a 6-year overlap to confirm measurement consistency. The analytical precision of the measurements made at both facilities in terms of the inferred SST was 0.55 °C (1σ).
An age model for the period before 1982 was constructed using the Sr/Ca ratios, which reflect SST, with Sr/Ca maxima (indicating relatively cool SSTs) corresponding to growth in January and February, the coolest months in this region. For the period 1982–2002, the precise timing of the coolest SSTs was determined using IGOSS-nmc SSTs. Other Sr/Ca ratios in this time interval were then converted to the time domain by simple linear interpolation. This age model was confirmed by counting annual density bands on the X-radiographs of the coral core. Age–depth models of the remaining core were developed using the annual density bands and then fine-tuned using the seasonal Sr/Ca cycles. At around 1883, when the Krakatau eruption occurred, geochemical data were disturbed for 1–2 years and the density bands were blurred for about 3 years; thus, the 1982–2002 age model may include an age error of 1–3 years before around 1885. The temporal resolution of the Sr/Ca analyses for the most recent 46 years of the record was approximately monthly, whereas for earlier periods, the resolution was mostly bimonthly. The temporal resolution of the δ18O analyses was bimonthly throughout most of the core.
The SST calibration obtained by comparing Sr/Ca minima and maxima to SST maxima and minima, respectively, is (Figure S2a):
Sr/Ca = –0.084SST + 11.137.
Seawater δ18O values (δ18Osw) were calculated from the paired proxy records as (Cahyarini et al., 2008):
δ18Osw = (δ18Ocoral – δ18Ocoral, m) – γ11 (Sr/Ca – Sr/Cam),
where δ18Ocoral is the measured coral δ18O value; δ18Ocoral, m is the mean δ18Ocoral value; Sr/Ca is the measured coral Sr/Ca ratio and Sr/Cam is its mean; and γ1 and β1 are the slopes of the linear regressions of δ18O (–0.143; Figure S2b) and Sr/Ca (–0.084; Figure S2a), respectively, against IGOSS SSTs. Then the minima (maxima) of δ18Osw and SSS were used to calibrate SSS as (Figure S2c):
δ18Osw = 0.790SSS – 27.103.