Past summer upwelling events in the Gulf of Oman derived from a coral geochemical record

We used a high-resolution oxygen isotope (δ18Ocoral), carbon isotope (δ13Ccoral) and Sr/Ca ratios measured in the skeleton of a reef-building coral, Porites sp., to reveal seasonal-scale upwelling events and their interannual variability in the Gulf of Oman. Our δ13Ccoral record shows sharp negative excursions in the summer, which correlate with known upwelling events. Using δ13Ccoral anomalies as a proxy for upwelling, we found 17 summer upwelling events occurred in the last 26 years. These anomalous negative excursions of δ13Ccoral result from upwelled water depleted in 13C (dissolved inorganic carbon) and decreased water-column transparency. We reconstructed biweekly SSTs from coral Sr/Ca ratios and the oxygen isotopic composition of seawater (δ18OSW) by subtracting the reconstructed Sr/Ca-SST from δ18Ocoral. Significant δ18OSW anomalies occur during major upwelling events. Our results suggest δ13Ccoral anomalies can be used as a proxy for seasonal upwelling intensity in the Gulf of Oman, which, driven by the Indian/Arabian Summer Monsoon, is subject to interannual variability.

(SSS) is derived from δ 18 O SW , which is generated by subtracting the temperature component (obtained from coral Sr/Ca) from δ 18 O coral 12 . Stable isotopes of carbon in coral skeletons (δ 13 C coral ) are influenced by kinetic isotopic fractionation, vital effects (photosynthesis and respiration) and the carbon isotopic composition of dissolved inorganic carbon in seawater (δ 13 C DIC-SW ) [13][14][15][16] . Coral skeletons are precipitated in isotopic disequilibrium with ambient seawater as a result of kinetic and vital effects. The kinetic effect selectively depletes 12 C and 16 O in coral skeletons and is particularly important when coral growth rates are very low (<4 mm per year) 13,14 . Photosynthetic activities of zooxanthellae affect δ 13 C coral by changing the carbon isotopes in the internal dissolved inorganic carbon pool of the coral 17 . A 50% weakening of solar radiation induces a decrease of approximately 0.5‰ VPDB in δ 13 C coral 18 . The amount of solar radiation received by the coral varies depending on incoming solar radiation, cloud cover and water transparency 17,19,20 . Upwelling can reduce water transparency and change the sea-surface δ 13 C DIC-SW . Therefore, upwelling events should be registered by the coral via a decrease in δ 13 C coral . We used a coral record from the Gulf of Oman to reconstruct the timing and frequency of upwelling events using high-resolution records of Sr/Ca ratios, δ 18 O SW and δ 13 C coral based on a 26-year-old coral core.

Results and Discussion
We determined δ 18 O coral , δ 13 C coral and Sr/Ca ratios from 664 samples. Each powdered sample was split for paired stable isotope and Sr/Ca analysis. Sr/Ca ratios and δ 18 O coral showed 26 distinct annual cycles ( Fig. 1a and b). The average of the Sr/Ca ratios was 9.28 (mmol × mol −1 ), with values ranging from 8.98 to 9.56 (mmol × mol −1 ). The δ 18 O coral averaged −4.33 (‰ VPDB ) and ranged from −4.92 to −3.41 (‰ VPDB ). We calculated the regression line between satellite SST and Oman coral Sr/Ca ratios using seasonal maxima and minima to avoid potential biases due to intra-seasonal age model uncertainties, as follows: Sr/Ca ratios (mmol mol ) 0044 0 003 SST 10 46 0 18 (r 095: P 0 01) 1 We established a regression line between satellite SST and δ 18 O coral , assuming that δ 18 O coral reflect only SST variations, with the same δ 18 O coral samples with Sr/Ca ratios, as follows: The correlation coefficient between δ 18 O coral and Sr/Ca ratios was 0.77 (P < 0.01). δ 18 O sw were calculated by subtracting the temperature component (estimated from coral Sr/Ca ratios) from δ 18 O coral , following the method proposed by Nurhati et al. 21 . The slope of the δ 18 O coral -SST regression is −0.104 ± 0.005‰ VPDB /°C, which is too high to be consistent with published estimates 12,13,22 . This suggests a significant contribution of δ 18 O SW to δ 18 O coral . We therefore used the published regression slope of −0.18 ± 0.03 (‰/°C) 12 to convert δ 18 O coral to SST, and our slope of −0.044 mmol × mol −1 /°C for SST estimation. The δ 18 O SW anomalies were calculated by applying a band-pass filter to remove the periodicity components longer than 2 years and subtracting the seasonal cycle. Relative changes of δ 18 O SW are on the order of ± 0.424‰ VSMOW (2σ). Anomalies above or below this threshold were marked as significant δ 18 O SW anomalies (Fig. 1c). The uncertainty of calculated δ 18 O SW is ± 0.113‰ VSMOW (following Nurhati et al. 21 ).
The average δ 13 C coral was −1.62 (‰ VPDB ) and ranged from −3.28 to +0.29 (‰ VPDB ). The δ 13 C coral also showed clear seasonal variation (Fig. 1d) and distinct short-term negative anomalies (Fig. 1d). The δ 13 C coral analysis was performed to avoid contamination from organic matter. We measured each CO 2 gas sample 6 times using a dual inlet system loaded on a MAT251. Analytical precision of the δ 13 C coral (standard deviations) were below 0.05‰. Growth rate disturbances and anomalous-colored annual band were not observed on X-ray photographs and coral cores. Therefore, the variations of δ 13 C coral were assumed to reflect environmental changes rather than the coral growth disturbances 23 .
Kinetic effects have been recognized as simultaneous 18 O and 13 C enrichment in coral skeletons with low extension rates 13 . Strong kinetic effects mask vital effects 13 . In our core, δ 13 C coral values showed a weak negative correlation with the δ 18 O coral record (r = −0.317, n = 634, P < 0.001: Fig. S2a). Summer δ 13 C coral did not correlate significantly with δ 18 O coral (r = 0.140, n = 181, P > 0.05: Fig. S2a). Winter δ 13 C coral had no significant correlation with winter δ 18 O coral (r = 0.04, P > 0.05, n = 159: Fig. S2b). The extension rates show that the Oman coral grew very quickly, on average 25.1 mm/year with a range between 19 to 31.5 mm. These values were considerably higher than the critical value estimated for kinetic isotopic fractionation effects (4 mm/year) (Fig. 1f) 14 . Therefore, the coral growth history and the lack of correlation between δ 13 C coral and δ 18 O coral suggest that the kinetic isotopic effect did not significantly affect this coral record.
Previous studies reported δ 13 C coral on seasonal and inter-annual variations are attributable to solar radiation 17,26 . To investigate the processes driving these δ 13 C coral fluctuations, we compared δ 13 C coral with satellite-based outgoing longwave radiation (OLR) (Fig. S3a) which reflect cloud cover. For a comparison of δ 13 C coral with monthly-resolved OLR data, biweekly resolved δ 13 C coral data were resampled at a monthly resolution using the software AnalySeries (version 2.0.8) 27 . The δ 13 C coral were compared with OLR, and we calculated the correlation coefficients between these time series. δ 13 C coral without anomalous δ 13 C coral excursions positively correlated with OLR at a significant level (r = 0.411, P < 0.01, n = 302: Fig. S3a and S3b). A significant correlation appeared between the mean seasonal cycle of δ 13 C coral and OLR averaged over the past 26 years (r = 0.702, P = 0.01, n = 12: Fig. S3c and d). The positive correlations between δ 13 C coral and OLR ( Fig. S2b and S2d) suggest that δ 13 C coral captured the variation of photosynthetic activity caused by the seasonal solar radiation cycle. At inter-annual resolution, the 15 month-moving average profile of δ 13 C coral positively correlate with that of OLR (r = 0.347, P < 0.01, n = 303: Fig. S4a and S4b). The duration of low OLR and coeval δ 13 C coral decreased from 1992 to 1993. We propose that insolation and OLR had decreased in globally as a result of up-stirred volcanic aerosol from the eruption of Mount Pinatubo, the Philippines in June 1991 28 . Low δ 13 C coral from 1992 to 1993 would be influenced by decreasing insolation which resulted from the volcanic eruption of Mount Pinatubo.
We calculated the δ 13 C coral anomaly (δ 13 C anomaly ) by removing the 15 month-moving average (31 bi-weekly data point) after subtracting the averaged seasonal cycle of δ 13 C coral . The threshold for δ 13 C coral anomalous excursions was determined as a standard deviation of 1σ: ± 0.343‰ VPDB . In summer, the anomalous negative excursions of the δ 13 C anomaly occurred 17 times in summer, while 1 anomalous negative excursion occurred in the spring of 1993 (Fig. 1f). Anomalous positive δ 13 C anomaly excursions were also observed prior to summer negative δ 13 C anomaly excursions. The δ 13 C anomaly had no significant correlation with OLR anomaly calculated by same procedure (r = 0.05, P > 0.3 Fig. S4c and S4d), suggesting that anomalous negative excursions of δ 13 C anomaly in the summer (AN-δ 13 C) would not be generated from OLR variations.
We examined the timing of the AN-δ 13 C with the compiled evidence of each past upwelling event documented from in situ and satellite observations ( Fig. 1f and g). Abrupt SST decreasing events in summer were revealed in 1987-1989 29 and 2000 30 from satellite SST data, in 1992 6 , 1994 4 , 2001 30 , 2002 30 (Fig. S5) and 1990 31 based on in situ SST data, and 2010 based on our vertical seawater temperature profile (Fig. S6). The vertical profile of seawater temperature deduced by temperature sensors attached to the diving gear of local volunteer divers in 2010, also suggest that the thermocline was closer to the surface during summer upwelling events (Fig. S6). In addition, Al-Azri et al. 1 (Fig. 1g). The AN-δ 13 C corresponds with these past upwelling events.
The possible controlling factors of the AN-δ 13 C with upwelling events are: (1) decreasing water-column transparency 33 , (2) variations of δ 13 C DIC-SW 16, 34 , and (3) change to heterotroph feeding 10 . It is known that increasing chlorophyll-a concentrations correspond with upwelling events 1, 7 inducing phytoplankton blooms, thereby decreasing water-column transparency and depleting 13 C coral with low photosynthetic activities of zooxanthellae 33 . Moreover, lower δ 13 C DIC-SW supply from greater depths decreases δ 13 C DIC-SW at the sea surface 24,25,34 . Upwelling events may produce an AN-δ 13 C due to sudden decreases in water-column transparency and δ 13 C DIC-SW . Heterotrophic feeding would also be the controlling factor of negative δ 13 C coral with upwelling events. A study 35 reported that corals feeding 13 C-depleted zooplankton decreased their δ 13 C coral . The coral records from the Gulf of Aqaba, Red Sea suggested that increasing heterotrophy with upwelling decreased δ 13 C coral for an approximately half a year 10 . Afterwards, δ 13 C coral could be increased by the preferential uptake of 12 C by phytoplankton at the sea surface 16 . In the western Indonesian coast, it was reported that δ 13 C coral increased by approximately 2.2‰ VPDB after large phytoplankton blooms due to upwelling 16 .
We propose the following mechanism to explain the short-term negative peaks in the δ 13 C coral : 1. Upwelling events bring deep, cold and nutrient-rich water with low δ 13 C DIC-SW to the surface in summer. Upwelling events cause unusually high nutrient conditions in the Gulf of Oman. Photosynthesis activities in zooxanthella would be emphasized in eutrophic conditions and temporarily increased δ 13 C coral . 2. Lower δ 13 C DIC-SW from the deep sea decreases δ 13 C coral . 3. Phytoplankton blooms arise from a nutrient supply to the sea surface. 4. Phytoplankton primarily depletes 12 CO 2-SW . Active phytoplankton photosynthesis increases 13 CO 2-SW . 5. δ 13 C coral increases with the restoration of δ 13 C DIC-SW .
We compared the AN-δ 13 C minima with the upwelling periods (number of the days) in summer (Figs 2a, S5 and S6). In situ daily to weekly SST data in 1992SST data in , 1994SST data in , 2001SST data in , 2002 and 2010 revealed that SST during upwelling events was as same as winter SST (23.5 °C), and daily fluctuations of SST in upwelling periods ranged within 3 °C. Therefore, the numbers of the days for upwelling periods were defined as the duration of SST lower than 26.5 °C in summer. δ 13 C anomaly values of no upwelling years (0 days) was estimated from in situ δ 13 C DIC-SW in Arabian Sea (+1.325‰ VPDB at 0-10 m depth in non-upwelling seasons 36 ) and the value of δ 13 C in isotopic equilibrium between coral carbonate and seawater 37 . The AN-δ 13 C minima were correlated to the upwelling periods as below.
Then, past upwelling periods in the year with no in situ SST data were reconstructed from each AN-δ 13 C using this equation (Fig. 2b). The estimated uncertainty for reconstructed upwelling-periods was 12.66 days (1σ) including the analytical precisions of δ 13 C coral , the intercept and the slope of this equation. In 1987In , 2006In , 2008In , 2009, each upwelling period was extremely long, over 120 days (Fig. 2b). In those years, coral extension rates decreased to 23 mm/year (Fig. 1e). The long upwelling events would therefore have a negative effect on coral extension rate due to eutrophic conditions and decreased water-column transparency.
We compared the reconstructed upwelling events from AN-δ 13 C (Fig. 2b) with Sr/Ca ratios and δ 18 O SW-anomaly (Fig. 1b and c). Sr/Ca ratios showed 1-month increasing (cooling) in summer except in 1994, 2001, 2002, 2006, and 2009, however, these did not correspond to reconstructed upwelling events. In non-AN-δ 13 C (upwelling) years (1989, 1991, 1997-1998, 2003, 2007, 2011-2012), the δ 18 O SW-anomaly was low in summer. Upwelling events in the Gulf of Oman are driven by the SW Monsoon, which causes strong seasonal winds parallel to the coast of Southern Oman in the Arabian Sea, while the associated Ekman transport creates strong upwelling along the coastal margins, bringing cold, nutrient-rich water to the surface 1, 4 . This upwelled water has indirect impacts on corals and reef areas farther north through gyres and eddy systems that sweep into the Oman Sea 1, 4 . In addition, upwelling may be influenced by vertical seawater density, depending on SST and SSS 38 . The δ 18 O SW-anomaly record suggested that deep seawater did not reach the sea surface as low-density water masses might form a cap on the sea surface in the Gulf of Oman.
Observations suggest that the primary productivity of the Gulf of Oman is subject to inter-annual variability 1 , but long-term observational records are lacking. Our new δ 13 C coral record captured past upwelling events and their periods in the Gulf of Oman for 26 years. Thus, coral skeletal archives fill an important gap in the observational record and have great potential for increasing our understanding of the upwelling mechanisms in the Gulf of Oman. Moreover, it is possible to reconstruct past SST, SSS and upwelling frequency/intensity during the Holocene (from 0 to 10 ka) by applying the same methods to fossil corals from the Arabian Peninsula.

Methods
Coral sampling. On February 23, 2013, we drilled a Porites sp. coral colony in the Gulf of Oman (23°30′ N, 58°45′ E: Fig. 3a and b). This Porites colony was living at a 2 m water depth in a small bay (Bandar Khayran) south of Muscat. There was no dry-riverbed (locally name: wadi) nearby; thus, we excluded the influence of occasional plumes of freshwater from coastal runoff at the site. In total, the coral core was 71 cm long. On the sampling date, we measured in situ SST and SSS at 24.3 °C and 38.2 PSU (practical salinity unit). Meteorological records from the weather station at Seeb Airport (23.60°N, 58.30°E) showed low precipitation rates, with less than 14.0 mm/month (the monthly average precipitation climatology for the past 23 years was 0.28-14.0 mm/month; GHCN-Monthly ver. 2). For coral proxy calibration, we used Advanced Very High Resolution Radiometer (AVHRR) satellite SST data, SODA satellite SSS data (http://iri.columbia.edu: Fig. 3c) and OLR data (https://climexp.knmi.nl/: Fig. 3c) [39][40][41] . Salinity records decrease in summer suggesting a possible occurrence of upwelling events.
Subsampling. The coral core was sliced into 5-mm-thick slabs. We took X-radiographs of the coral slabs to identify the coral growth axis (Fig. 4). We prepared ledges of 1.5 mm in thickness along the maximum growth axis and obtained coral powder at a resolution of 0.5 mm for geochemical analysis.
Oxygen and carbon isotope measurements. The coral powder was weighed, and 100 μg (±20 μg) were taken for oxygen and carbon stable isotope analysis. The sample powder was reacted with 100% H 3 PO 4 at 70 °C in an automated carbonate preparation device (Kiel II). The δ 13 C coral and δ 18 O coral were analyzed with a Finnigan MAT251 stable isotope ratio mass spectrometer system installed at Hokkaido University. Analytical errors for δ 13 C coral and δ 18 O coral were determined to be 0.08 and 0.07‰, respectively, based on replicate measurements of the NBS-19 standard (1σ, n = 40). Trace element measurements. We measured Sr/Ca ratios with a SPECTRO CIROS CCD SOP inductively coupled plasma optical emission spectrophotometer installed at Kiel University following a combination of methods described by Schrag 42 and de Villiers et al. 43 . Approximately 250 μg of coral powder was dissolved in 4 mL of HNO 3 . The sample solution for the measurement of trace elements was prepared via serial dilution with 2% HNO 3 for a Ca concentration of ca. 8 ppm. Analytical precision of the Sr/Ca determinations was 0.07% RSD or 0.01 mmol × mol −1 (1σ).

Data analysis.
We used the coral Sr/Ca ratios to develop an age model for all proxies. Minima and maxima of the coral Sr/Ca ratios were chosen as anchor points and tied to the maxima and minima of SST, respectively. To obtain a time series with equidistant time steps, we resampled the proxy data at a biweekly resolution using the AnalySeries software, version 2.0.8 27 . Annual extension rates were estimated from the distance (in mm) between the winter anchor points in each sclerochronological year.