Abrupt upwelling and CO2 outgassing episodes in the north-eastern Arabian Sea since mid-Holocene

Identifying the causes and consequences of natural variations in ocean acidification and atmospheric CO2 due to complex earth processes has been a major challenge for climate scientists in the past few decades. Recent developments in the boron isotope (δ11B) based seawater pH and pCO2 (or pCO2sw) proxy have been pivotal in understanding the various oceanic processes involved in air-sea CO2 exchange. Here we present the first foraminifera-based δ11B record from the north-eastern Arabian Sea (NEAS) covering the mid-late Holocene (~ 8–1 ka). Our record suggests that the region was overall a moderate to strong CO2 sink during the last 7.7 kyr. The region behaved as a significant CO2 source during two short intervals around 5.5–4 ka and 2.8–2.5 ka. The decreased pH and increased CO2 outgassing during those abrupt episodes are associated with the increased upwelling in the area. The upwelled waters may have increased the nutrient content of the surface water through either increased supply or weaker export production. This new dataset from the coastal NEAS suggests that, as a potential result of changes in the strength of the El-Nino Southern Oscillation, the region experienced short episodes of high CO2 outgassing and pre-industrial ocean acidification comparable to or even greater than that experienced during the last ~ 200 years.

The recent rapid increase in anthropogenic CO 2 has led to ocean acidification negatively impacting marine carbonate ecosystems 1 . The level of carbon dioxide (CO 2 ) in the atmosphere has a direct role in regulating the chemistry of the ocean. The dissolution of anthropogenic CO 2 in seawater has lowered ocean pH and carbonate ion concentrations and potentially reduced the ability of marine organisms to calcify 2 . Pre-industrial variations in atmospheric CO 2 are believed to be strongly influenced by changes in circulation and carbon cycling in the ocean [3][4][5] where the interaction between large deep ocean carbon reserves and dynamic ocean circulation plays an important role in regulating the ocean carbonate system and air-sea gas exchange of CO 2 3,4 . The mechanism involved in this exchange of CO 2 consists of several complex processes which alter the magnitude of oceanic sinks (mainly cold, high latitude regions) and sources (mainly warm, low latitude and tropical regions) 3,6,7 . The ocean carbonate system consists of six co-varying parameters (pH, [CO 2 ], [HCO 3 − ], [CO 3 2− ], total alkalinity (TA) and dissolved inorganic carbon (DIC)), with the full system resolvable provided any two of these six components are known. Moreover, surface water pH and pCO 2 are very tightly coupled in natural systems and thus knowledge of one is enough to infer the other with high confidence 8 . On a glacial-interglacial timescale, the processes influencing the concentration of CO 2 in the atmosphere are thought to leave a fingerprint on the surface water pH 9 . Also, the pH of the surface water is the primary control of the boron isotope composition in planktonic foraminifera [10][11][12][13][14] . Therefore, the boron isotope signatures (δ 11 B) of planktonic foraminifera provides a good opportunity to better understand the changes in surface ocean chemistry (pH) and the mechanisms involved in regulating the ocean carbonate system before human influence and industrialization, as well as providing a way to produce reconstructions of past pH and atmospheric CO 2 15 . The Arabian Sea is one of the most productive basins of the world, having the capability to release around 90 TgC yr −1 to the atmosphere 16,17 . The monsoon-influenced lateral and vertical circulation of nutrients in the

Results and discussion
The surface water above core SK-240/485, is strongly influenced by the SWM with very little contribution from the NEM 41,42 , despite being located in the NE Arabian Sea. The SWM is considered as the major regulator of SST and SSS, particularly in coastal regions. This can also be observed in the monthly salinity record of nearby location (21.5°N; 68.5°E) which shows an extreme SSS decrease around October (post-SWM) ( Supplementary  Fig. 1). The surface-dwelling foraminifera G. ruber inhabits the upper water column (down to ~ 25 m) when  26  www.nature.com/scientificreports/ the surface water is well stratified due to increased fresh water discharge in the basin during the SWM season. Therefore, the reconstructed signals from the G. ruber proxy may be biased towards the SWM season (JJAS) 43 (Fig. 2b). The δ 18 O sw has been extensively used as a proxy for SWM-influenced evaporation-precipitation budget in the northern Indian Ocean 50,53 . The δ 18 O sw at our site varies between +0.93 and −0.70 ‰ (VSMOW), with most negative value observed during the late Holocene (after ~ 4 ka; Fig. 2b and Supplementary Fig. 2b). Such a lowering of δ 18 O sw suggests a significant decrease in the evaporation-precipitation budget of the area, perhaps due to an overall intensification of the SMW during 8-1 ka. A Mann-Kendall trend test reveals a significant increasing trend in δ 18 O sw over our study interval [Ʈ = 0.686; p value (two tailed) < 0.0001] ( Supplementary Fig. 2b). These observations thus point towards the increased intensification of SWM between 8 and 1 ka, as the major contribution of fresh water to the core site comes from the Indus River and small seasonal tributaries during the SWM. Previously, a δ 18 O G.ruber based statistical and spectral study from the same www.nature.com/scientificreports/ site (SK-240/485) also suggested an intensifying SWM trend during the Holocene 41,42 . In addition, it is in good agreement with other studies from eastern Arabian Sea during this interval 54,55 .
The boron isotopic composition of the foraminifera (δ 11 B) shows an average value of 21.0‰ and ranges between 21.7‰ (at around 1.44 ka) and 19.8‰ (at around 4.44 ka; Fig. 2c). The reconstructed pH in the core SK-240/485 has an average value of 8.19 pH units for last ~ 7.7 kyr (Fig. 2d), which shows a good agreement with the existing pH records of the Holocene from the Pacific (~ 8.15 pH units), Atlantic (~ 8.16 pH units), and Indian (~ 8.22 pH units) Oceans 9,27,28,56-58 . More specifically, the variation of pH in the core SK-240/485 ranges between 8.02 (around 4.44 ka) and 8.28 (around 1.44 ka; Fig. 2d). The Mann Kendall trend test suggests a significant increasing trend [Ʈ = 0.386; p-value (two tailed) < 0.028] in surface water pH since 7.7 ka ( Supplementary  Fig. 2a). The calculated average pCO 2 sw at the study site for last 7.7 kyr is 247 µatm which is ~ 22 µatm (or ppm) lower than the average atmospheric CO 2 value (around 269 ppm) obtained from EPICA Dome-C ice core record (pCO 2 atm ) over the same period 59 . The pCO 2 sw attains a maximum of 415 µatm at around 4.4 ka and minimum of 170 µatm at around 1.4 ka (Fig. 2e). While the average pCO 2 sw at the site SK-240/485 is considerably lower than the average Holocene pCO 2 sw values recorded in Atlantic and Pacific Ocean 9,58,60 , it is significantly higher (~ 30 µatm) than that recorded at site AAS9/21 in the south-eastern Arabian Sea 28 since ~ 8 ka (247 µatm at SK-240/485 vs ~ 214 µatm at AAS9/21). A modelling-based study of the evolution of pCO 2 sw from the Arabian Sea suggests the existence of a strong north-south gradient in the Arabian Sea due to the seasonal sea surface circulation dynamics 17 . The south-eastern Arabian Sea shows comparatively lower pCO 2 sw due to the fresh water input from the Bay of Bengal through the West Indian Coastal Current (WICC) during the NEM 41 . On the contrary, the NE Arabian Sea (including the present site) does not receive fresh water input from the WICC and therefore shows comparatively higher pCO 2 sw during modern times 41,55 . For a better understanding of the source/sink nature and quantification of air-sea CO 2 exchange in the studied region, we calculated the ΔpCO 2 (the difference of pCO 2 sw and pCO 2 atm ) at each time interval (Fig. 2f). The crossplot between ΔpCO 2 and salinity at the site SK-240/485 shows a significant positive correlation (r = 0.65; p < 0.01; R squared = 0.41 with exclusion of two points of high ΔpCO 2 (which are probably related to intense upwelling, see below, Supplementary Fig. 3). This suggests that the SWM-influenced freshwater input plays a significant role in the ΔpCO 2 variation of the area in the absence of strong upwelling. We found that the area was an overall CO 2 sink during the last 7.7 kyrs with average ΔpCO 2 sw around − 25 µatm (ranging between 170 and 415 µatm) which is a value characteristic of shelf regions 30 . Between 7.7 and 6.1 ka, the region was in quasi-equilibrium with the atmosphere with respect to CO 2 (ΔpCO 2 < 10 µatm).
The record shows two abrupt periods since 7.7 ka when the region behaved as a significant CO 2 source, ~ 5.4-4.1 ka and ~ 2.8-2.6 ka. During these periods, pCO 2 sw and ΔpCO 2 were as high as 415 µatm and 145 µatm, respectively and surface water shows a notable 0.25 units decline in pH. During the same intervals, the δ 13 C G.ruber (δ 13 C GR ) from the same site (SK-240/485) shows a decrease of 0.5‰ centred around 4.5 kyr 41 (Supplementary Fig. 4). Hence, the excess CO 2 in the surface water we see at site SK-240/485 around ~ 4.4 ka may be a result of weaker export production and/or enhanced nutrient-rich water masses reaching to the sea surface (e.g. upwelling). A negative excursion in δ 13 C G.ruber associated with the increased pCO 2 sw due to increased upwelling of nutrient-rich water has also been observed in the south-eastern Arabian Sea during the Bølling-Ållerød (warm period) 28 . A similar pace of pH decrease along with high amounts of CO 2 outgassing was also observed around site MD01-2416 in sub-polar North Pacific during the Bølling-Ållerød which shows the significant response of local upwelling on the pCO 2 outgassing 61 . Based on these previous findings, we suggest that the outgassing at ~ 4.4 ka at SK-240/485 may be due to the supply of respired photosynthetic products to the surface along with the upwelled water, which increased the DIC/alkalinity ratio and thereby lowered pH and increased surface water CO 2 61 . However, the increase in Mg/Ca SST at the same time may either be related to the near-surface habitat of G. ruber and upper water column stratification, rather than upwelling to the surface; the replacement of surface water currents from a lower pH and warmer region; or the influence of non-thermal effects on Mg/Ca 62 . This latter possibility is discussed in more detail in the next section.

Non-thermal influences on foraminiferal Mg/Ca. Several studies suggest that the foraminifera bound
Mg/Ca may be impacted by other environmental variables than temperature known as non-thermal effects 62 , including salinity 63 or carbonate system parameters 61,64,65 . These studies highlight that the local changes in seawater pH may exert a significant bias on the planktonic foraminifera Mg/Ca which lessens its correspondence to sea surface temperature (SST) 61,65 . Based on such observations, a correction protocol has been suggested for the non-thermal influence on planktonic foraminifera Mg/Ca using MgCarb package 62 . We compared the original data with MgCarb-adjusted results to check the effect of such influences in our Arabian Sea record (Fig. 3). As expected, for much of the record, the MgCarb-adjustment causes little variation. However, a difference in calculated SST of ~ 2°C is observed at 4.4 and 2.7 ka (Fig. 3) where the lowest pH is observed. Because SST has a more minor control on the calculated pH, the adjusted pH record is largely within error of the original record but with a difference of ~ 0.06 units during 4.4 and 2.7 ka, corresponding to a maximum difference of ~ 50 µatm in ΔpCO 2 (with the adjusted pCO2 values always being lower). Nevertheless, this adjustment represents only a minor change in the overall pH and ΔpCO 2 trends and hence the main inferences of our results remain unchanged (Fig. 3). Additionally, the adjusted SST during the periods of abrupt drop in pH and high CO 2 outgassing (marked by dark grey bands in Fig. 3) further supports the increased upwelling in the area at 2.7 ka.

Teleconnection of Arabian Sea upwelling with the El-Nino Southern Oscillation. The upwelling
of sub-surface waters plays an important role in controlling the CO 2 saturation of the ocean surface. During intense upwelling, the surface waters are higher in CO 2 saturation relative to the atmospheric CO 2 level, decreasing surface water pH and causing CO 2 to outgas. The modern Arabian Sea experiences intense upwelling due www.nature.com/scientificreports/ to the action of southwest monsoon winds over the sea surface through Ekman pumping 67 . Moreover, recent studies suggest a link between SWM and related upwelling in the Arabian Sea with the El Niño Southern Oscillation (ENSO) 68 . In this context, we compared our record of pH and ΔpCO 2 from site SK-240/485 with the proxy record of long-term ENSO activity during the Holocene (Figs. 3, 4): the variation of sedimentation around Laguna Palcacocha, South Ecuador that is thought to reflect the variation of ENSO activity in the area 66 . The comparison reveals that the abrupt spikes in SST, ΔpCO 2 and pH at 4.4 and 2.7 ka coincide with periods with high ENSO activity (Figs. 3, 4). The ENSO is an important phenomenon of the tropical climate system. During positive ENSO phases basinwide warming occurs throughout the tropical Indian Ocean 68 , with several studies suggesting that SST variations in the Indian Ocean have a strong correlation with the ENSO variability [71][72][73][74] . The Indian Ocean Basin Mode (IOBM) is the primary mode of variation of SST which has a strong impact on ENSO-induced heat-flux anomalies around Equatorial and northern Indian Ocean (including Arabian Sea) 68 . During frequent ENSO events, the IOBM shows warming causing an increase in precipitation over tropical Indian Ocean and further strengthening the South Asian High 68 . These climatic features influence the rising southwest monsoon winds over the Arabian Sea and result in the intensification of the Findlater Jet (FLJ) 75 . In turn, the FLJ transports the moisture from the southern Indian Ocean to the Indian landmass through the Arabian Sea and inter-annual variation of FLJ has been linked to the strength of El Niño and La Niña episodes 76 . The northward advancement of FLJ is reported to induce upwelling in the Arabian Sea 76-78 and during an El Niño event the intensified FLJ advances towards the northern Arabian Sea where it induces strong coastal upwelling in the Arabian Sea via coastal Kelvin waves 68,77,[79][80][81] . These existing teleconnections are yet to be explored in records of past climate and are difficult to test, however, our observations suggest a strong influence of ENSO events on the upwelling around the NE Arabian Sea during mid-late Holocene, possibly similar to the modern-day relationship via coastal Kelvin waves. We therefore propose such a mechanism may have induced strong upwelling around the study area during the periods of high ENSO activity, increasing ΔpCO 2 and surface water pH.

Comparison of global pre-industrial records with the modern analogue. For a better insight of
our results in a global perspective, we have compared our pCO 2 sw record with the existing records from Arabian Sea and other parts of the world ocean over the last 8-1 ka (Fig. 5 and Supplementary Table 2). We have also compared our results with global pCO 2 /pH composite curves prepared by considering several δ 11 B based pH and pCO 2 sw records to understand the behaviour of north-eastern Arabian Sea in context of global source/sink patterns 69 (See Supplementary Table 2 for more details). The pCO 2 records from equatorial and southwest Pacific  ) and the south-western Pacific has changed to a modest sink of atmospheric CO 2 (Fig. 1) 83 . The pCO 2 sw record from the ODP Site 999 in Caribbean Sea (Atlantic) indicates that the region was in near equilibrium (pCO 2 sw ≈ pCO 2 atm ) over the last 8 kyr and continues to remain this way to the present 9 . On the other hand, the site PS2498 in the south of Atlantic (~ Southern Ocean) appears as a larger CO 2 source ~ 8 ka and is currently a modest CO 2 sink during modern times 58 . A recent study from site TAN1106/28 around offshore New Zealand in the sub-Antarctic Pacific shows high pCO 2 sw leading to the CO 2 outgassing from the region around 4 ka 84 and a similar increase in CO 2 outgassing potentially caused by enhanced upwelling of CO 2 -rich deep waters around 4.5 ka, this has also been reported from site MD972106 in Atlantic sector of Sub Antarctic Zone 85 . The change in the Southern Ocean from a significant CO 2 source (during the deglacial and into the Holocene) to a significant sink in modern times has been attributed to the resumption of Antarctic upwelling and northward advection of CO 2 -rich waters via Ekman pumping 58,86 .
The currently available δ 11 B based records from Indian Ocean are restricted to the Arabian Sea 27,28 . The ΔpCO 2 record around site AAS9/21, albeit at low resolution, indicates that the south-eastern Arabian Sea was a significant sink during last ~ 8 kyr 28 . The present study provides new information about the changes in carbonate system around the coastal NE Arabian Sea during 7.7-1.4 ka. The data shows an overall sink in the area along with the high CO 2 outgassing episodes in the area (as discussed in the previous sections). At the present time, the average ΔpCO 2 and pH nearby site SK-240/485 is around 37 µatm and 8.04 units respectively 26,70 . It is worth mentioning here that the recent (2010 AD) pH nearby site SK-240/485 (8.04) is 0.03 units lower than the average pH of global ocean in 2010 AD 70 (Fig. 4 and Supplementary Fig. 5). We have compared the pre-industrial (1770 AD) and modern (2010 AD) pH at our core site based on a recent study of global pre-industrial, modern and future pH estimates 70 . The pre-industrial pH value (during 1770 AD or ~ 0.18 BP) nearby the study area is estimated to be 8.15 which was also 0.03 units lower than the average pH of the global ocean (~ 8.18) at that time 70 . During industrialisation (between 1770 AD-present) the surface water pH around the study area varied at a similar pace to the global surface water pH (i.e. dropped ~ 0.15 pH units) due to the significant increase in global atmospheric CO 2 levels 70 . Furthermore, with the current rate of pH change, it is estimated that the Arabian Sea, similar to the global ocean, may experience around 0.2-0.3 pH units drop in pH by 2100 if no attempt is made to mitigate emissions 70 .

Conclusion
The present δ 11 B based pH and pCO 2 sw results from the site SK-240/485 in NE Arabian Sea suggest that the region overall behaved as a net CO 2 sink during Mid-Late Holocene, which is a typical characteristic of a shelf region. In addition, the region was nearly in equilibrium with the atmosphere during 7.7-6.1 ka. Thereafter, the area behaved as an abrupt CO 2 source during two short periods of strong ENSO activity during 5.5-4 ka (very strong CO 2 source) and 2.8-2.5 ka (moderate CO 2 source). The periods of abrupt increase in ΔpCO 2 are likely associated with the enhanced upwelling of nutrient-rich waters to the surface. It is estimated that the two intense upwelling episodes (centring around 4.4 and 2.7 ka) triggered CO 2 outgassing and preindustrial ocean acidification similar to the modern levels. Subsequently, during 2.5-1.4 ka, the region behaved as a strong CO 2 sink probably due to the absence of upwelling. We suggest therefore that there is evidence of intense natural ocean www.nature.com/scientificreports/ acidification episodes in the region over the past at times of sudden increase in carbon and nutrient supply to the surface due to increased coastal upwelling. We found that the reconstructed pH from site SK-240/2485, as well as the pre-industrial and modern value, appears to be in the range of global pH variation during the Holocene (within uncertainty) except during these two abrupt periods of intense upwelling. A comparison of our data with available records of past ENSO activity suggests a possible teleconnection of NE Arabian Sea upwelling with the timing and intensity of ENSO events during the Holocene. However, this aspect in particular needs to be confirmed in future studies from other regions of the northern Indian Ocean.

Methods
Study area. The analysed marine sediment core (SK-240/485) was recovered from offshore Saurashtra (Lat. 21° 16′; Long. 68° 55.99′ E; 88 m water depth), NE Arabian Sea (Fig. 1). The core covers period between the Pleistocene-Holocene boundary and the Holocene 42 (last 12.4 kyr). However, the present study is restricted to cover the period between ~ 7.7 and ~ 1.4 ka due to insufficient foraminiferal abundance towards the bottom of the core. The Arabian Sea is the north-western extension of Indian Ocean. It spreads around the coasts of India, Pakistan, Yemen, Oman, Somalia and the Addu Atoll (Maldives). The Indus River is the primary source of freshwater and sediment deposition in the NE Arabian Sea which receives fluxes of meltwater discharge from the Himalaya as well as runoff from monsoon precipitation [87][88][89] . The Indus Fan is the most extensive physiographic feature of the Arabian Sea and considered to be the second-largest submarine fan in the world 89 . Apart from the primary source, minor freshwater input also comes from the Bhadar River (a seasonal river) during the SWM season. The average annual salinity around the study area today is 36.2 psu, and the average annual sea surface temperature (SST) and ΔpCO 2 (the difference between pCO 2 sw and atmosphere) are 27.1 °C and + 36.7 µatm respectively 26,90 (Supplementary Fig. 1) with the latter having an annual cycle of ~ 18 µatm.
Geochemical proxies for studying ocean carbonate system. We analysed Mg/Ca ratios and δ 11 B in planktonic foraminifera (G. ruber). The ratio of Mg to Ca (or Mg/Ca ratio) in calcite (particularly foraminifera) serves as a proxy of past seawater temperature and has been extensively used in the world oceans 62,[91][92][93][94][95][96] .The boron isotopic composition, expressed as δ 11 B (Eq. 1), of foraminifera is strongly influenced by the pH of the www.nature.com/scientificreports/ surface water [9][10][11]14 . Hence, the δ 11 B of planktonic foraminiferal shells is used as a proxy to reconstruct the pH of surface seawater and further to decipher other parameters of the ocean carbonate system, most notably pCO 2 sw9 .
The geochemical analyses of foraminifera (Mg/Ca ratio and δ 11 B) were carried out at the University of Southampton, UK. For determination of Mg/Ca as well as δ 11 B for each sample 170 individuals of Globigerinoides ruber (white variety; sensu-stricto type) with a size range of 300 to 350 µm were hand separated. The hand-separated specimens were then cleaned following Barker et al. 97 , without applying the reductive cleaning step. Briefly, a multi-step cleaning protocol was applied consisting of crushing the foraminifera shells, followed by the removal of clay, silicate and organic matter, and a weak acid leaching with final dissolution of the carbonates using 0.5 M HNO 3 as detailed in Foster et al. 98 . A small aliquot (~ 7%) of the cleaned and dissolved sample was analyzed for Mg and Ca (and other elemental ratios) by inductively coupled plasma mass spectrometry 9 (ICP-MS) following Henehan et al. 99 while the remaining sample was analysed for δ 11 B using Multi collector inductive coupled plasma mass spectrometry (MC-ICP-MS) following Foster 9 . External reproducibility of the MC-ICPMS δ 11 B method (at 95% confidence, 2SD) is based upon repeat measurements of an in house carbonate standard and is a function of sample concentration this ranges from 0.4‰ to 0.2‰ for the samples presented here 57,98,100 . Analytical precision for elemental ratios is determined at 95% confidence by the reproducibility of several in house standards and is < 5% for all elements. Strict criteria are used to rule out clay contaminated results (samples with Al/Ca > 100 μmol/mol removed). All samples are matrix matched to a gravimetrically determined standard, to eliminate Ca concentration effects on the ratios produced 99 .
The measured Mg/Ca (mmol/mol) ratio were converted to SST (°C) by using the calibration equation 101 Further, δ 18 O of seawater (δ 18 O SW ) was calculated by using the SST derived from Mg/Ca to deconvolve ice volume effect values 95 from δ 18 O of G. ruber and following equation of Bemis et al. 102 The Sea Surface Salinity (SSS) was then calculated by using δ 18 O SW -salinity relation for Arabian Sea suggested in equation 103 .
The pH was calculated from the analysed δ 11 B values using the species-specific calibration for G. ruber where pK * B is the dissociation constant for boric acid dissociation in the in-situ (temperature, salinity and pressure) conditions and calculated as per Dickson 104 , δ 11 B sw is the isotopic composition of seawater having value 39.61‰ 105 . 11-10 KB is the equilibrium constant for the two existing forms of boron in seawater i.e. borate and boric acid 106 having a value of 1.0272 ± 0.0006 107 . Since the borate ion is predominantly incorporated into the foraminifera calcite 9 , the δ 11B CaCO 3 can be used to obtain the pH with the species and size specific (G. ruber) coefficients as estimated by Henehan et al. 57 (See Eq. 5) Further, the pCO 2 sw was calculated from pH (derived from δ 11 B), SST and SSS (derived from Mg/Ca ratio) and TA (calculated using modern day relationship between SSS and TA around the area using the Global Data Analysis Project data set) 70 .
All the original data calculations are done with the R computer programming platform 108 (R Core Team 2014) using "seacarb" package, version 6.8 109 . Foraminifera bound Mg/Ca may be impacted by non-thermal effects 62 (e.g. Gray and Evans 2019) including salinity 63 or carbonate parameters 64 . We explored the influence of carbonate parameters using the parametrisation using MgCarb package as mentioned in Ref 62 . The uncertainties were calculated using Monte Carlo approach (n = 10,000) with consideration of the uncertainty in all relevant input parameters (with 95% confidence level) δ 11 B ± 0.25‰; Mg/Ca derived SST ± 1 °C; SSS ± 1psu; TA ± 50 µmol/kg.