Observations and analytical methods.
We collected data during two Chinese National Arctic Research Expedition (CHINARE) cruises (01 August–08 September 2008, and 20 July–30 August, 2010) in the western Arctic Ocean. During CHINARE 2008 and 2010, hydrographic casts were performed using a 24-bottle rosette sampler equipped with a Sea-Bird SBE 911plus CTD. Seawater samples were analysed for dissolved inorganic carbon (DIC), total alkalinity (TA), nutrients (NO3−, PO43−) and oxygen isotope ratio H218O/H216O (δ18O).
DIC and TA samples were stored in borosilicate glass bottles and preserved with HgCl2. The samples were shipped back to Wei-Jun Cai’s lab at the University of Georgia in USA for analysis (within 8 weeks). DIC and TA were measured by a DIC Analyzer and an Alkalinity Titrator (Apollo Scitech) following the analytical techniques and procedures documented in refs 31,32. A 0.75 ml DIC sample was acidified using 10% phosphoric acid and the evolved CO2 gas was extracted and carried by pure N2 gas to an non-dispersive infrared CO2 analyser (Li-Cor 7000) for quantification. TA was determined using a 25 ml sample by potentiometric Gran titration with 0.1 M hydrochloric acid and an open-cell titration system. Both analyses had a precision of better than 0.1% (±2 μmol kg−1) and were calibrated using the Certified Reference Materials (CRM) from A. G. Dickson, Scripps Institution of Oceanography. An internal data consistency assessment suggests that the accuracy of the DIC and TA data is also within ±2 μmol kg−1 (ref. 31). δ18O samples were measured using a mass spectrometer referenced to Vienna-Standard Mean Ocean Water (V-SMOW). The precision of δ18O measurements is ±0.1‰. The nutrients (NO3−, PO43−) were analysed on board, using a San++ automated continuous flow analyser (SKALAR Inc.) according to the WOCE protocol33, with a precision of ±0.1 μM for NO3− and ±0.03 μM for PO43−. Oxygen was determined using the automatic Winkler titration system, with a precision of ±1 μmol kg−1.
Comparison of TA and DIC data among various cruises.
In addition to CHINARE 2008 and 2010, we included Arctic Ocean Section 1994 (AOS 1994), Surface Heat Budget of the Arctic Ocean 1998 (SHEBA1998), and Beringia 2005 (ODEN 2005) historical data (Supplementary Fig. 3). Certified Reference Materials (CRM) were used in all analyses with the exception of the AOS1994 cruise, during which CRM was not yet available. AOS 1994 data were quality controlled and adjusted by the community (reduced by ~24 μmol kg−1).
The measurements from all the cruises were expected to be systematically consistent. A consistency check of the data collected from the Canadian Basin show that DIC and TA data below ~400 m have small systematic differences (Supplementary Fig. 1). Deep Arctic Ocean water (>2,000 m) is extremely uniform in almost all of the properties due to the long residence time—hundreds of years34—therefore biogeochemical processes, such as organic matter remineralization and the temporal variability of calcium carbonate dissolution, are insignificant and can be neglected35. Thus, the deep water (>2,000 m) in the Canadian Basin is an ideal region to conduct a basin-wide consistency analysis.
To correct the differences between the five trans-western Arctic Ocean cruises we determined ‘relatively reasonable’ DIC and TA values of the deep waters and then used them to correct the small systematic differences among the five cruises. This method is consistent with that of ref. 36. This correction and consistency check allows us to use all existing data from a total of 14 cruises (Supplementary Table 2; note most of them are limited to a subregion and some of them have data quality issues that we chose not to include in Fig. 1 in this paper). First, DIC and TA values of the deep water were normalized to a salinity of 35 (NDIC and NTA, NDIC = DIC/Salinity ×35 and NTA = TA/Salinity ×35, respectively). Second, we divided the Canadian Basin into three distinct subregions (Supplementary Table 2)—the southern Canada Basin (SCB), northern Canada Basin (NCB), and Makarov Basin (MB)—to evaluate spatial variation due to inconsistent sampling density. Then we averaged the NDIC and NTA values in each subregion for every cruise to decrease the sensitivity of different sample size among cruises (Supplementary Table 2). The next step was to further average NDIC and NTA for all of the cruises for every subregion to evaluate if there were significant differences among them. Finally, we averaged the values of the three subregions to create ‘relatively reasonable’ values of NDIC (2,157 μmol kg−1) and NTA (2,307 μmol kg−1) for the whole Canadian Basin. Subsequently, for all the cruises, the water column DIC and TA data were adjusted to the above NDIC and NTA using equations (1) and (2):
After the correction described above, TA and DIC data from the five trans-western Arctic Ocean cruises below 400 m agreed well (Supplementary Fig. 1). Any slight errors associated with the correction of these deep-water values do not affect our conclusion.
Ωarag was calculated using the CO2SYS program37, with the carbonic acid dissociation constants K1 and K2 from Roy et al. 38, KSP from Mucci39 and KSO4 from Dickson40. CO2SYS was aso used to simulate Ωarag dynamics caused by an increase in anthropogenic carbon dioxide emission in the AW layer of the Canadian Basin. The dissociation constants (K1 and K2) of Roy et al. 38 were recommended by previous works41 and were found16, 31 to yield only slightly higher Ω values (~0.01) than that using Mehrbach et al. 42 as refit by Dickson and Millero43.
Qualitative tracer data—TA/Sal ratio.
Various tracers are used to identify freshwater accumulation and Pacific Winter Water (PWW) intrusion, including salinity (Sal), potential temperature (°C), potential density anomaly (kg m−3), and TA/Sal ratio (Supplementary Figs 5 and 6). TA in subsurface and deep water, unlike DIC and nutrients, is only slightly affected by organic matter cycling. Small increases in TA can be driven by aerobic respiration, whereas larger increases are primarily a result of CaCO3 dissolution, and denitrification in oxygen-limited regions, as global ocean water circulates from the Atlantic to the Pacific Ocean44, 45. Thus, PWW has a much higher ratio of TA to salinity (TA/Sal) than that of AW (Supplementary Fig. 5 and Tables 3 and 4). Therefore, the TA/Sal ratio can be used as a quasi-conservative tracer of water mixing between the PWW and AW in the Arctic Ocean. The TA/Sal ratio (Supplementary Fig. 5o–u) exhibited spatial and temporal distribution tendencies similar to those of Ωarag (Fig. 1 in the main article).
Quantitative tracer data—Fraction of Pacific water (PW).
There are several water masses in the western Arctic Ocean, including Atlantic water (AW), Pacific water (PW), sea-ice meltwater (SIM) and meteoric water (MW, runoff and precipitation/evaporation). In this study we focused on the fraction of PW (fPW(%)), which was determined using the nitrate–phosphate relationship as a tracer27, 28. This method fits best-fit regression lines for nitrate and phosphate for PW and AW. The intercepts are significantly different for Atlantic and Pacific lines due to denitrification occurring on the sea floor of western Arctic Ocean shelves46, 47, 48. The relative distance between data points and the best-fit line can be used to calculate the mixing ratio of AW and PW. As a result, the N/P method can be used to distinguish the Pacific water (PWW and PSW) and AW (Supplementary Fig. 2). The fractions of all water sources in any seawater sample were calculated using the following mass balance model (equations (3)–(5)), with details shown in ref. 28. Atlantic and Pacific contributions were first distinguished using the N/P relationship (Supplementary Fig. 2), then PW fractions were estimated using δ18O and salinity values:
where f is fraction, S is salinity, and the subscripts, PW, AW, SIM, MW, and m, denote Pacific water, Atlantic water, sea-ice meltwater, meteoric water, and measured values, respectively. The S and δ18O values for the four endmembers used in the calculations are given in ref. 28.
Model simulation of the impact of anthropogenic CO2 uptake on Ωarag in the AW endmember.
Oceanic anthropogenic CO2 uptake from the atmosphere since 1765 has caused the Ω values of surface waters to decrease by approximately 0.4 units globally1. North Atlantic surface water, which is the source of the subsurface Atlantic water layer in the western Arctic Ocean, is in equilibrium with the atmosphere when it sinks in transit to the western Arctic Ocean49. The transit time for North Atlantic surface water to reach the Canada Basin is 16 ± 2 yr (ref. 50). Thus, the water parcels that formed the Atlantic layer water in the Arctic Ocean in 1994 and 2010 are estimated to have left the Atlantic surface layer in 1978 and 1994, respectively. The atmospheric CO2 concentration over the North Atlantic in 1994 (359 ppm) was 24 ppm higher than in 1978 (335 ppm). Therefore, from 1994 to 2010, anthropogenic CO2 uptake could have increased surface pCO2 (ΔpCO2) by approximately 24 μatm. Thus, the Ωarag change in AW from 1994 to 2010 can be estimated from the known atmospheric pCO2 increase and measured TA data: S = 34.75, T = 0.6 °C, TA = 2,291 μmol kg−1, ΔpCO2 = 24 μatm. CO2SYS calculated carbonate variables and Ωarag results show that anthropogenic CO2 uptake would reduce Ωarag by ~ 0.09 in the Atlantic water from 1994 to 2010. As a result of the pCO2 increase, DIC would also increase from 2,145 μmol kg−1 (1994) to 2,155 μmol kg−1 (2010). In addition, [NO3 + NO2] and AOU have increased slightly by 0.7 and 10 μmol kg−1, respectively (Supplementary Table 4), with corresponding DIC increases, on top of those due to pCO2, of 4.6 and 7.7 μmol kg−1, respectively; estimated using the O/C/N ratio of 138/106/16. The respiration-induced DIC increase would lead to a decrease in Ωarag by 0.03–0.06. Thus, we conclude that the combined effects of anthropogenic CO2 uptake and organic matter respiration would lead to a Ωarag reduction of 0.12–0.15, which is consistent with the observed Ωarag reduction (from measured TA and DIC data) of 0.17 ± 0.03 (from 1.55 to 1.38 units, Supplementary Table 4).
Percentage area calculation for subsurface acidified water along the transect.
The percentage area occupied by subsurface acidified water along a transect is calculated by dividing the subsurface area of acidified water by the total area. Given that the Ωarag < 1 seawater extended northwards only to 77° N during the AOS 1994 cruise, this calculation assumes that there was no acidified area to the north of 77° N in 1994. In addition, SHEBA1998 data show that the Ωarag < 1 seawater extended northwards only to 80° N; however, there is no data south of 75° N. According to the observed maximum vertical expansion of Ωarag < 1 seawater, usually located between 75°–77° N from the data sets reported here, we assume the acidified area south of 75° N has the same expansion as that at 75° N to extrapolate the area of subsurface acidified water in 1998. It should be noted that the assumption described above might slightly underestimate and overestimate the percentage area in 1994 and 1998, respectively. The uncertainty, however, will not affect our conclusion of an increase in the percentage area of acidified seawater from 1994 to 2010.
We calculated two estimates for Ωarag < 1 and Ωarag < 1.25. Ωarag = 1 is typically used to indicate when calcium carbonate dissolution could begin to occur; however, Ωarag = 1.5 is also adopted as the critical index to evaluate if acidification affects calcifying organisms in most oceans51, 52, 53. The Arctic Ocean has the lowest aragonite saturation values noted in the world’s oceans, with most of the observed Ωarag values less than 1.5 in the top 250 m (Fig. 1). Therefore, here we use Ωarag = 1.25, the mean of 1.0 and 1.5, as the critical index to evaluate the percentage area change of acidified seawater.
CO2 system data are deposited in the Chinese National Arctic and Antarctic Data Center (http://www.chinare.org.cn/pages/index.jsp) and CDIAC Ocean CO2 Data (http://cdiac.ornl.gov/oceans). The data that support the findings of this study are available from the corresponding authors upon reasonable request.