Long-term trend of CO 2 and ocean acidification in the surface water of the Ulleung Basin , the East / Japan Sea inferred from the underway observational data

Anthropogenic carbon is responsible for both global warming and ocean acidification. Efforts are underway to understand the role of ocean in a high CO 2 world on a global context. However, marginal seas received little attention despite their significant contribution to biogeochemical cycles. Here we report the CO 2 increase and ocean acidification in the surface waters of the Ulleung Basin (UB) of the East/Japan Sea, and possible causes are discussed. Fourteen observations of surface f CO2 were made in the period from 1995 to 2009. The contribution of temperature variation to the seasonality of f CO2 was almost equivalent to the non-thermal effect in the UB. However, the difference of relative contribution with the season makes two seasonal peaks of f CO2 in the surface water of the UB. Non-thermal effect contributed to the surface f CO2 drawdown in summer, whereas the surface f CO2 elevation in winter. The decadal trend of f CO2 increment was estimated by harmonic analysis. The estimated rates of increase of f CO2 were 1.8± 0.4 μatm yr −1 for the atmosphere and 2.7± 1.1 μatm yr −1 for the surface water. The ocean acidification trend, calculated from total alkalinity and f CO2, was estimated to be −0.03± 0.02 pH units decade −1. These rates seem to be higher than observations at most other ocean time-series sites during the same period of time. Sustained observations are required to understand more accurate trend in this area.


Introduction
The IPCC (2007) announced that the anthropogenic CO 2 that has accumulated in the atmosphere since the Industrial Revolution is responsible for the enhanced greenhouse effect.The ocean is the ultimate mobile carbon storage reservoir in the Earth system.The surface ocean absorbs atmospheric CO 2 by means of both physical and biogeochemical processes.In the oceanic carbonate system, the dissolved carbon is transported to the deep ocean by the oceanic carbon pumps and eventually sequestrated in deep-sea sediments.Sabine et al. (2004) reported that about one-third of the CO 2 from the total anthropogenic CO 2 emissions is stored in the ocean.Currently, results from established oceanic time-series stations show that the trend of increasing CO 2 in surface seawater exceeds that of the atmosphere (Bates, 2001;Keeling et al., 2004) and is accelerating ocean acidification (Caldeira and Wickett, 2003;Key et al., 2004;Orr et al., 2005).
Ocean carbon uptake is commonly described as a variety of carbon pumps, which operate by thermodynamics (e.g., temperature effects on the solubility of CO 2 ), physical transport (e.g., mixing and advection of water masses carrying various forms of carbon), sinking of soft tissue (e.g., carbon uptake/release and export by phytoplankton during photosynthesis/respiration), and sinking of carbonate shells (e.g., formation and dissolution of CaCO 3 ) (Volk and Hoffert, 1985).These processes are easily affected, directly or indirectly, by temperature.The world ocean's temperature from the surface to 3000 m increased by 0.06 • C between the mid-1950s and mid-1990s because of an enhanced greenhouse effect (Levitus et al., 2005), and previous works have shown that the trend of increasing global sea surface temperature (SST) is about 0.08-0.14• C decade −1 (Nicholls et al., 1996;Molinari et al., 1997;Casey and Cornillon, 2001).Soaring SST creates stronger stratification between the surface sea (upper mixed layer) and the deep ocean.It might affect the great ocean conveyor system by weakening deep water formation.The warming will also weaken the solubility pump and the biological carbon pump through fortified stratification.Overall, a positive feedback exists between global warming and ocean carbon uptake, thus making the situation worse.
Nearly all the ocean's dynamics can be observed and studied in the East/Japan Sea (EJS) despite its relatively small dimensions (Kim and Kim, 1996;Kim et al., 2001;Kang et al., 2003).The Ulleung Basin (UB), located on the western side of the southern EJS, has complex hydrography.The bifurcation of warm current entering through the Korea Strait leaves numerous spin-off mesoscale eddies at the surface, and beneath the surface, a number of cold currents originate from the winter overturning along the Siberian coast of the EJS.Winter-mode water formation is occasionally observed (Kim et al., 1991;Seung and Kim, 1993;Talley et al., 2003).The UB supports a high productivity (> 220 gC m −2 yr −1 ) owing to its surface dynamics and coastal upwelling (Yamada et al., 2004;Yoo and Park, 2009).Because the UB has strong solubility pumps and biological pumps, it is an ideal laboratory for ocean carbon uptake studies.For this reason some studies were carried out in the UB (Oh et al., 1999;Choi et al., 2011Choi et al., , 2012)).In particular, it has been noted that the role of the coastal ocean in the global carbon budget has been underestimated (Wollast, 1998;Borges et al., 2005) and under debate (Cai and Dai, 2004;Thomas et al., 2004).Therefore, the role of the carbon pump in marginal seas should be reconsidered.
To reduce the scientific uncertainty in the prediction of future climate, a better understanding of the carbon flux among various reservoirs, especially the fluxes across the air-sea interface, is required.The Takahashi climatology (Takahashi et al., 2009) utilized all of the available data for modeling but still neglected marginal seas.Here, we like to emphasize the role of the marginal sea, as exemplified by the UB, and compare it to that for the entire ocean.There is no time-series station comparable to BATS or HOT in the EJS.However, we attempted to fill the gap of oceanic CO 2 time series in the UB by interpolation using a harmonic function analysis of the data from repeated measurements.Although simplified and crude, an attempt to discern the long-term trend of CO 2 uptake and acidification in a marginal sea that operates its own conveyor belt at a much faster timescale than the global ocean is here made for the first time.
2 Experimental methods and data analysis

Experimental methods
Measurements of the partial pressure of CO 2 (pCO 2 ) in the surface water and overlying atmosphere, sea surface temperature (SST), and sea surface salinity (SSS) were carried out 14 times in the EJS from 1995 to 2009 (Table 1).The data that were collected in the UB, the southwestern part of the EJS, were used in this study.The study area was defined as the region from 36-38 • N latitude and 130-133 • E longitude (Fig. 1).
The pCO 2 of surface seawater and the overlying atmosphere was measured by LiCor model 6252 non-dispersive infrared (NDIR) gas analyzer with a two-stage Weiss-type equilibrator.The CO 2 concentration from the NDIR was acquired every 2 s and, after statistical treatment for 1 min or 2 min, averaged data were obtained and used as the final data.Three different standard gases were used to calibrate the NDIR.Every 12 h, a series of standard gases was analyzed for calibration; one of the standard gases was analyzed every 6 h to check for drift of the machine.Atmospheric and surface seawater pCO 2 were analyzed every 15 min and 45 min on the hour, respectively.Surface seawater pCO 2 was analyzed by equilibrating air with surface seawater fed to the equilibrator.In situ SST and SSS were measured using thermosalinograph (SeaBird Electronics, Model SBE-21 or SBE-45).
f CO 2 is defined as the difference between f CO 2 in the surface water (f CO sea 2 ) and overlying atmosphere (f CO atm 2 ) ( f CO 2 ≡ f CO sea 2 − f CO atm 2 ).When we calculated f CO 2 , we applied an average value of prior and posterior f CO atm 2 data to f CO sea 2 data, since we do not have both data at the same time and the CO 2 variation in the atmosphere is much smaller than in the seawater.
Total alkalinity (TA) was determined on board by a potentiometric titration method using a closed cell (Millero et al., 1993).Total alkalinity was calibrated by certified reference materials provided by A. Dickson of Scripps Institute of Oceanography, which are measured at every cruise.The TA data were obtained during cruises 9906, 0306, 0405, and 0410 (Table 1).Surface seawater pH values were obtained immediately after sampling by a spectrophotometric method using the indicator dye m-cresol purple (Clayton and Byrne, 1993).The extinction coefficients for m-cresol purple from Clayton and Byrne (1993) were used.Determination of surface seawater pH values was carried out during cruises 9906, 0306, 0406, and 0410 (Table 1).pH values were calculated at in situ sea surface temperature.

Air-sea flux estimation
The CO 2 flux (mmol C m −2 d −1 ) was estimated from following equation:  where k is the gas transfer velocity (cm h −1 ), and s is the solubility of CO 2 gas in seawater (mol kg −1 atm −1 ; Weiss, 1974).We choose the long-term scale formulas for k and the wind speed relationships used by Wanninkhof (1992).
The NCEP wind speed data (http://www.cdc.noaa.gov/cdc/reanalysis) averaged on monthly scale in 36-38 • N latitude and 130-133 • E longitude were used to estimate for flux.

General characteristics of f CO 2 and CO 2 flux in the UB
The secular variation of mean f CO sea 2 and f CO atm 2 of the UB from 1995 to 2009 is shown in Fig. 2. The f CO sea 2 ranged from 125 to 499 µatm with a mean value of 350.9 µatm (SD = 41.8), and the f CO atm 2 varied from 344 to 397 µatm with a mean value of 370.5 µatm (SD = 8.7).The f CO sea 2 was generally lower than the f CO atm 2 indicating that the surface seawater was undersaturated to atmospheric CO 2 in this area.
The secular trend of f CO 2 includes the long-term linear trend as well as the cyclic seasonal variation (Keeling et al., 2004).The harmonic function, which is the sum of two components defined as the 1-year and half-year component, has been used primarily to reconstruct time-series data.Since this analysis method was first introduced by Nojiri et al. (1999), harmonic functions have been used extensively for f CO 2 variability studies (Zeng et al., 2002;Lüger et al., 2004;Chierici et al., 2006).We applied the harmonic function analysis to f CO 2 data from the UB to estimate the decadal trend of CO 2 and to separate thermal and non-thermal effects on the f CO 2 .Because harmonic function analysis is a technique for annual data, we added the long-term linear trend to the harmonic function as follows: where t is year.
The set of six constant coefficientsc 0 , c 1 , c 2 , c 3 , c 4 , and c 5 -ensured a satisfactory fit of the harmonic function of Eq. (1) to the observed data (RMSD = 17.22 µatm, R 2 = 0.78 for seawater, RMSD = 2.02 µatm, R 2 = 0.96 for air; Fig. S1 in the Supplement), when the constant coefficients were equal to these values (Fig. 3 Uncertainties for the annual trend were estimated by the equation , where σ 2 = [( is the variance around the fitted Eq. ( 1), and Y is f CO 2 , X is year, and N is number of data (Takahashi et al., 2009).Among the results, the c 5 value, which means annual trend for f CO sea 2 , was estimated to be 2.7 ± 1.1 µatm yr −1 and that for the f CO atm 2 was 1.8 ± 0.4 µatm yr −1 (see Fig. S1 in the Supplement).
The secular variation of f CO 2 is shown in Fig. 4a.Although the f CO sea 2 and f CO atm 2 vary with time, the secular variation of f CO 2 can be insignificant on the variation of f CO 2 , since f CO 2 was calculated with f CO 2 values in the seawater and in the atmosphere at the same time.We can hardly find the long-term trend of f CO 2 .However, the seasonal variation of f CO 2 shows that seawater is generally undersaturated with respect to the atmosphere with exception in summer (Fig. 4b) as other studies reported (Oh et al., 1999;Choi et al., 2012).Therefore, the UB serves as a sink of atmospheric CO 2 , in general.
The sea-air CO 2 fluxes in 1995 and 2004 were estimated in order to evaluate the effect of the last decadal increasing trend in f CO 2 .The sea-air CO 2 fluxes were evaluated to be −0.95 ± 0.53 mol m −2 yr −1 for 1995 and −0.81 ± 0.49 mol m −2 yr −1 for 2004.This result shows that the UB acts as a carbon sink and its carbon sink efficiency in unit area is almost 1.7( ± 0.3)-fold higher than global ocean (−0.51 mol m −2 yr −1 , Takahashi et al., 2002).Choi et al. (2012) and Oh et al. (1999) reported that the annual integrated CO 2 flux in this area was −2.47 ± 1.26 mol m −2 yr −1 and −2.2 mol m −2 yr −1 , which were larger than our results.The transiency of their observations could lead to overesti- mation in CO 2 flux because of fewer observations (less than four times) despite of the complexity of the monthly variability and significant seasonal amplitude in f CO sea 2 in the UB.

Seasonal variability and controlling factors of f CO 2 in the UB
The parameters that can control the f CO sea 2 such as sea surface temperature (SST), mixed layer depth (MLD) based on the climatology data (de Boyer-Montégut et al., 2004), and chlorophyll a concentration inferred from the SeaWiFS data between 1998 and 2007 are represented in Fig. 5.The SST reached a minimum in winter (February) and a maximum in summer (August and September) and had a mean value of 18.5 • C. The monthly variation in f CO atm 2 shows opposite in phase to that of SST.The monthly variation pattern of f CO atm 2 followed the general seasonal cycle of f CO 2 (or pCO 2 ) monitored in the North Hemisphere; that is, the f CO atm 2 reached its maximum value in spring and minimum value in summer.The f CO sea 2 , meanwhile, showed more complex variation than did the f CO atm 2 .  of the f CO sea 2 in February-June and September-November was lower than the f CO air 2 , but that in July and August was above the f CO air 2 (Fig. 4b).The f CO sea 2 in December data appeared close to the f CO air 2 .During the warm season (June to August), relatively high f CO sea 2 was associated with elevated SST.During the fall and winter (October to February), monthly f CO sea 2 values were higher than those in spring (April and May).This difference was a result of vertical entrainment of CO 2 -rich subsurface water despite the decreasing SST.Deepening of MLD (> 50 m) supported high f CO sea 2 phenomenon in winter (Fig. 5b).Comparison of the normalized total CO 2 (NTCO 2 ) in the EJS showed the difference of ∼ 100 µmol kg −1 in surface layer between summer and winter (Park, 1997).It also could be a result of the biological drawdown of CO 2 in March and April, when the f CO sea 2 attained its lowest value.During March and April, satellite-based chlorophyll a concentration reached a peak value (> 1 mg m −3 ) suggesting a possibility of biological CO 2 drawdown (Fig. 5c). ) in the UB.The MLD was based on the climatology by de Boyer-Montégut et al. ( 2004).The chlorophyll a was based on the SeaWiFS data between 1998 and 2007.Takahashi et al. (2002) proposed a method for estimating the relative importance of the effects of biological activity and seasonal temperature change on the pCO 2 of surface seawater.In order to estimate the relative magnitude of these effects, the ratio of the thermal effect (effect of temperature change) to the non-thermal effect (effect of biological activity and vertical mixing) was adapted after the method proposed by Takahashi et al. (2002) as the following equations: where SST is the surface seawater temperature in • C, and the f CO 2 and SST refer to the annual average of observed f CO 2 and SST values, respectively.The dependence of f CO 2 variability in the UB on the thermal (f CO therm 2) and non-thermal (f CO non−therm 2 ) effects was separated by the equations above (Eqs.2 and 3).The results Fig. 6.A plot of the rate of increase of surface seawater CO 2 versus latitude.Dashed lines represent the range of long-term trend at major ocean time-series stations (Rhein et al., 2013).
are shown in Fig. 5f and g.The f CO therm 2 followed the same pattern as the seasonal SST variation (Fig. 5a and f).The peak-to-peak amplitudes of the seasonal cycles of SST and f CO therm 2 ( f CO therm 2 , Eq. 4) were ∼ 17 • C and 275 µatm, respectively, which means that an 1 • C temperature change makes a 16 µatm change in f CO 2 .However, the f CO therm 2 was larger than that observed at Station "P" (100 µatm; Wong and Chan, 1991) and at BATS (150 µatm, Bates, 2001), which are located at the comparable latitude as the UB.Larger seasonal variation of SST (> 17 • C) may have caused such a difference.
On the other hand, f CO non−therm 2 reached its highest value during the cold season and its lowest value during the warm season.The peak-to-peak amplitude was 190 µatm, which was smaller than the value from the Ross Sea (260 µatm, Sweeney, 2000) but much larger than the value from BATS (115 µatm, Bates, 2001).This high f CO non−therm 2 value in winter was due to (1) relatively low biological carbon uptake during the cold season and (2) stronger vertical mixing that pumps up high-CO 2 subsurface water.The decrease in the f CO non−therm 2 value of more than 100 µatm in spring was attributed to carbon fixation by the spring bloom (Fig. 5e and  c).In summer, the value decreased because of weakening of vertical mixing caused by stratification (Fig. 5b and e).
The seasonal variation in f CO 2 can be explained by the sum of the thermal and non-thermal effects.The ratio between the thermal effect and the non-thermal effect was estimated to be almost equal to one ( f CO therm  / f CO non−therm 2 = 1.4) in the UB, which suggested that the thermal effect was almost balanced with the non-thermal effect.A similar result (0.9) was obtained at Station "P" in the eastern subarctic Pacific Ocean (50 • N, 145 • W) during the period 1973-1978(Wong and Chan 1991).On the other hand, the ratio differed significantly from that found for the BATS (2.7) (Bates 2001) or the Ross Sea (0.02) (Sweeney, 2000) values (Table 2).
In summary, the contribution of temperature variation to the seasonality of f CO 2 was almost equivalent to the nonthermal effect in the UB.However, the relative contribution varied with the season (Fig. 5h).Non-thermal effect contributed to the surface f CO 2 drawdown in summer, while the surface f CO 2 elevation in winter.According to Sarmiento and Gruber (2006), pCO 2 in the North Pacific is not affected by one dominant factor among SST, biological activity, and vertical mixing but is affected by their combined effect.Their explanation about the controlling factors of pCO 2 in the North Pacific could be applied to similar features of the f CO 2 in the UB.

Long-term trend of f CO 2 and ocean acidification in the UB
Long-term trend of the f CO 2 in the UB is shown in Fig. 3.The long-term atmospheric pCO 2 variation observed at Gosan station (33 • 17.4 N, 126 • 9.9 E) on Jeju Island, WMO Global Atmosphere Watch Station (Cho et al., 2005) Wong and Chan (1991), Takahashi et al. (2002) shown in the figure for comparison with f CO atm 2 measured in the UB (Fig. 3b).
The f CO atm 2 in the UB coincided with pCO 2 at Gosan owing to the shorter turnover time of atmospheric CO 2 than that in the ocean.The trends of increasing pCO 2 at Gosan and f CO atm 2 in the UB were 1.9 ppm yr −1 and 1.8 µatm yr −1 , respectively.These values are comparable to the several ocean time-series data (between 1.88 ± 0.02 µatm yr −1 and 1.92 ± 0.01 µatm yr −1 ) during same period (Rhein et al., 2013).
Since the work of Inoue et al. (1995), numerous studies have focused on the rate of CO 2 increase based on monitoring and survey data.For the sake of comparing the long-term trend of the f CO sea 2 in the UB with other regions, previously published data covering more than 30 different oceanic areas are listed in Table 3.The f CO sea 2 in the UB had been increasing gradually at a rate of 2.7 ± 1.1 µatm yr −1 , which is in the middle of the observed range of increasing rates in the mid-latitude of the Northern Hemisphere (Fig. 6).Although it has been reported for some regions of the oceans that the rate of f CO 2 change is dependent on the period of observations (Ishii et al., 2009;McKinley et al., 2011;Midorikawa et al., 2012), the increasing rate in most areas is close to the long-term trend at major ocean time-series stations (between 1.58 ± 0.07 µatm yr −1 and 2.22 ± 0.11 µatm yr −1 ; Rhein et al., 2013).However the rate in the mid-latitudes of the Northern Hemisphere is higher.Compared to observations from the same period of time, the increasing rate of f CO 2 in the UB that we found appears to be higher than most (Table 3) and warrants further investigation.
As surface water CO 2 has been increasing, the pH of the surface seawater has been decreasing, an effect called ocean acidification.The long-term trend of pH in the UB was estimated based on total alkalinity (TA) data from this study and f CO sea 2 .The pH values were calculated from TA and simulated f CO sea 2 by CO2SYS (Lewis and Wallace, 1998).The carbonate dissociation constants (K 1 and K 2 ) used in these calculations were those of Mehrbach et al. (1973) as refit by Dickson and Millero (1987).We assumed the TA of surface seawater to be constant at 2266 ± 17 µeq kg −1 , the average value of 60 measurements during the study period, because the seasonality, secular trends, and the relationships between total alkalinity with salinity and temper-ature (Lee et al., 2006) were not significant (see Fig. S2 in the Supplement).These calculated pH values were in good agreement with the measured ones (Fig. 7b).As the surface f CO 2 increased, the pH value in the surface seawater of the UB, while fluctuating seasonally, has changed at the rate of −0.03 ± 0.02 pH units decade −1 since 1995 (Fig. 7).The pH in the surface ocean decreased by 0.1 pH units between 1750 and 1994, which was noted as an unprecedented decline by Sabine et al. (2004).
To compare f CO 2 increasing trend and pH decreasing trend in the UB with other ocean time-series stations, we also plotted time series of f CO 2 and pH from Station ALOHA data during the same period.Since 1995, the pCO 2 at Station ALOHA has increased at a rate of ∼ 2.16 µatm yr −1 , and mixed layer pH has declined by ∼ 0.02 pH units per decade (Fig. 7).Rhein et al. (2013) reported that the increasing rates of f CO 2 in the seawater were in range of 1.58 ± 0.07 µatm yr −1 and 2.22 ± 0.11 µatm yr −1 , and the rates of decrease of pH were estimated between 0.015 ± 0.001 pH units decade −1 and 0.022 ± 0.004 pH units decade −1 based on long-term ocean time-series station data in the North Atlantic and North Pacific oceans.Compared with those data, the trend of CO 2 and ocean acidification in the surface seawater of the UB seems to be slightly faster as the faster warming trend than the global one in this area (Kim et al, 2007;Yeh et al., 2010).
Under the assumption of constant TA and sea surface temperature, 10 % increase of f CO 2 in the UB (i.e., 26.6 µatm increase in a decade) may reduce pH by 0.027 pH units, which contributed about 87 % of pH decrease in the UB during the last decade.A seawater temperature rise of 1 • C may also decrease pH by 0.01 pH units at a pressure of 1 atm (Gieskes, 1969).The sea surface temperature in the EJS increased by 0.2-0.7 • C during the last decade (Kim et al., 2007;Yeh et al., 2010), which is a faster warming than the global average (0.13 • C decade −1 ; Rayner et al., 2006).This warming contributed to a pH decrease of 0.002-0.007pH units, which was equivalent to 7-23 % of the pH decline in the UB during the last decade.Therefore, the ocean acidification in the UB was mainly driven by the f CO 2 increase, while the warming effect was relatively small.

Fig. 1 .
Fig. 1.Maps showing the tracks of the cruises conducted between 1995 and 2009 for this study.

Fig. 2 .
Fig. 2. The secular variations in f CO sea 2 (filled circles) and f CO atm 2 (open diamonds) in the Ulleung Basin, East/Japan Sea, from 1995 to 2009.Error bars represent one standard deviation from the mean value.

Fig. 3 .
Fig. 3. Long-term trend of f CO 2 at the Ulleung Basin from 1995 to 2009.(a) f CO sea 2 (filled circles) of surface seawater of the Ulleung Basin, (b) f CO atm 2 (filled diamonds) of overlying atmosphere of the Ulleung Basin.Atmospheric pCO 2 data, measured at Gosan, Jeju Island, South Korea (gray dots), are also shown for comparison with f CO atm 2 .The gray curves represent f CO 2 fitted from a harmonic function analysis by Eq. (1).

Fig. 4 .
Fig. 4. (a) The secular variations in f CO 2 and (b) monthly variation of f CO 2 in the Ulleung Basin, East/Japan Sea, from 1995 to 2009.Error bars represent one standard deviation from the mean value.

Fig. 7 .
Fig. 7. Secular variation of (a) f CO 2 (µatm) (black filled circles and fitting line), and (b) pH (black diamonds and fitting line) of the surface water in the UB.The time-series pCO 2 (µatm) (gray open circles) and pH (gray open squares) at Station ALOHA in the subtropical North Pacific Ocean are also shown for comparison.The pCO 2 and pH data at Station ALOHA are from D. Karl, University of Hawaii (http://hahana.soest.hawaii.edu)and are updated from Doney et al. (2009) and Dore et al. (2003).

Table 1 .
Information for cruises in this study: SST, SSS, f CO atm 2 , f CO sea 2 , f CO 2 , total alkalinity (TA), and pH are reported as averages and standard deviations.The average values are arithmetic mean of 1 or 2 min averaged values in the cruise.The number of data values IS listed in parentheses.

Table 2 .
Comparison of the contribution of thermal and non-thermal effects on the f CO 2 of surface seawater in various regions.

Table 3 .
Rate of increase of surface seawater CO 2 at various regions of the world ocean.