Decreasing pH trend estimated from 25-yr time series of carbonate parameters in the western North Pacific

We estimated long-term trends of ocean acidification in surface waters in latitudinal zones from 3°N to 33°N along the repeat hydrographic line at 137°E in the western North Pacific Ocean. Estimates were based on the observational records of oceanic CO2 partial pressure and related surface properties over the last two decades. The computed pH time series both for 25 yr in winter (late January.early February) and for 21 yr in summer (June.July) exhibited significant decreasing trends in the extensive subtropical to equatorial zones, with interannual variations that were larger in summer. The calculated rates of pH decrease ranged from 0.0015 to 0.0021 yr-1 (average, 0.0018 ± 0.0002 yr-1) in winter and from 0.0008 to 0.0019 yr-1 (average, 0.0013 ) 0.0005 yr-1) in summer. The thermodynamic effects of rising sea surface temperature (SST) accounted for up to 44% (average, 15%) of the trend of pH decrease in the subtropical region in winter, whereas a trend of decreasing SST slowed the pH decrease in the northern subtropical region (around 25°N) in summer. We used the results from recent trends to evaluate future possible thermodynamic changes in the upper ocean carbonate system.


Introduction
The recent uptake of anthropogenic carbon by the ocean has given rise to changes in chemical equilibrium in the surface ocean CO 2 system (Intergovernmental Panel on Climate Change [IPCC], 2007). In particular, the ocean acidification due to the increase in the surface concentrations of CO 2 has recently attracted increasing interest because of possible future positive feedback to global warming resulting from a reduction in the capacity of the ocean to take up additional CO 2 (i.e. to buffer increasing atmospheric CO 2 ) and because of the potential effects on marine ecosystems.
A decrease in surface pH by 0.1 over the global ocean has been calculated from the estimated uptake of anthropogenic carbon since pre-industrial times (Royal Society, 2005), and a further decrease in pH by 0.3-0.4 over the next century is predicted from a model experiment based on estimates of future CO 2 emissions (Caldeira and Wickett, 2003;Orr et al., 2005). To detect such a gradual, long-term ocean acidification requires pH measurements of high precision and traceability. It is only recently that a spectrophotometric method with high precision has been developed (e.g. Clayton and Byrne, 1993), so modernday changes in surface-water pH have not yet been precisely determined.
The recent time series data for pH at 25 • C directly measured at the European Station for Time Series in the Ocean at the Canary Islands (ESTOC) in the North Atlantic exhibited a long-term decrease in pH of 0.0017 ± 0.0005 yr −1 averaged for 1995-2003(González-Dávila et al., 2007. Similar trends of decreasing pH at ambient sea surface temperature (SST) were calculated from dissolved inorganic carbon (DIC) and total alkalinity (TA) data at the Hawaii Ocean Time series (HOT; Dore et al., 2009; 0.0019 ± 0.0002 yr −1 ) and the Bermuda Atlantic Time series Study (BATS; Bates, 2007; 0.0017 ± 0.0001 yr −1 ) sites. However, differences in pH trends between regions or basins over the world oceans are poorly known because of the paucity of longterm time series observations. The repeat hydrographic line at 137 • E in the western North Pacific, which is a longitudinal section located near the centre of the highest sea-surface dynamic height in the subtropical gyre ( Fig. 1), is one of the few areas where oceanic CO 2 measurements have been routinely made since the early 1980s. We previously described the latitudinal and seasonal variations of the partial pressure of CO 2 in surface seawater (pCO 2 sea ) and overlying air (pCO 2 air ) along 137 • E, and the significant increasing trend and interannual variability of winter pCO 2 sea (Inoue et al., 1987(Inoue et al., , 1995Midorikawa et al., 2005Midorikawa et al., , 2006. The pCO 2 sea increased in response to the increase in pCO 2 air in all regions studied, from equatorial to subtropical. The increase in pCO 2 sea was caused primarily by that in surface DIC due to the oceanic uptake of anthropogenic CO 2 , and the relatively small regional differences were attributed mostly to the effects of SST. However, the results indicating unchanged pCO 2 and CO 2 influx over the subtropical region for two decades implied that the northern subtropical region could continue to act as a strong CO 2 sink in the near future. The interannual variations of pCO 2 sea from 1983 to 2003 were derived from the different degrees of compensation between the SST and DIC effects on pCO 2 sea , which varied with latitude.
In the present study, we specifically focus on the changes in the surface carbonate system producing ocean acidification, on the basis of the long-term pCO 2 sea record along 137 • E in the western North Pacific, and discuss the influence of ocean warming on the progress of acidification.

Methods
The time series of pH values over the last two decades was estimated based on the observational records of pCO 2 sea and related surface properties in the western North Pacific.

Observation and data
The Japan Meteorological Agency (JMA) and Meteorological Research Institute have been periodically conducting observations of pCO 2 sea , pCO 2 air and related hydrographic properties [SST, sea surface salinity (SSS), and others] from south of Japan to the equatorial region along 137 • E (Fig. 1) in the western North Pacific during nearly the same periods: from late January to early February since the early 1980s, and in June and July since 1990 (JMA, 2009). The details of the methods have been described by Inoue et al. (1995) and Midorikawa et al. (2006). The data for 1983-2007 used in this study are available from the World Meteorological Organization's World Data Centre for Greenhouse Gases (WDCGG; http://gaw.kishou.go.jp/wdcgg.html) operated by JMA.
We have also been conducting precise measurements of DIC, using a coulometric technique and Certified Reference Material provided by Dr. A. G. Dickson at Scripps Institution of Oceanography, at irregular intervals since June 1994 (during the WOCE P9 one-time observations) (Ishii et al., 2001), and pH using a spectrophotometric technique and m-cresol purple as an indicator dye since April 2003 (Saito et al., 2008).

Estimation of total alkalinity
TA was not measured, but instead calculated by the methods of Dickson et al. (2007) from SST, SSS, pCO 2 sea and DIC for the period after 1994, when these four surface parameters were measured. The calculations used a solubility constant for CO 2 in seawater given by Weiss (1974) and dissociation constants of carbonic acid given by Lueker et al. (2000). The uncertainty of TA estimates, evaluated from uncertainties of the respective measurements, was less than 4 μmol kg −1 . The calculated TA normalized to a salinity of 35 (n-TA) at the respective latitudes ranged from 2294 ± 4 to 2298 ± 3 μmol kg −1 in winter and from 2289 ± 7 to 2296 ± 8 μmol kg −1 in summer for the past 14 yr, suggesting that n-TA was almost constant (average n-TA of 2295 μmol kg −1 at SSS = 35) over the study region and remained unchanged for the study periods, that is, over the extensive regions from 3 to 33 • N, the calculated n-TA showed little variability in either space or time over the past 14 yr (Fig. 2).

Estimation of pH
The time series of pH values were calculated for the period of 1983-1993 by the methods of Dickson et al. (2007) using the time series data for pCO 2 sea , SST and SSS, and assuming TA of 2295/35 * SSS μmol kg −1 . For TA, a constant n-TA of 2295 μmol kg −1 at S = 35, the average value over the most recent 14 yr, was converted to the TA value at ambient SSS, proportional to the salinity. For this study, pH values were expressed on the pH scale of total hydrogen ion concentration (Dickson et al., 2007).  For 2003-2008, when pH was measured along with SST, SSS and surface pCO 2 sea and DIC, surface pH values calculated from these four parameters were compared with measured ones (Fig. 3). The calculated and measured pH values for 2003-2008 were in good agreement; the differences averaged 0.0026 ± 0.0050 (mean ± 1σ ) in winter and 0.0003 ± 0.0052 in summer, supporting the validity of the calculation procedures.

Normalization of pH and pCO 2 sea
Normalization of pH to a constant SST of 25 • C was also based on the computation routine of carbonate equilibrium using the methods of Dickson et al. (2007). In this study, because TA throughout the entire period was assumed, time series of DIC were also able to be calculated. Then, pH at a constant SST of 25 • C (pH T25 ) was calculated from time series of observed pCO 2 sea and SSS as well as calculated DIC. Normalization of pCO 2 sea to constant values of SST and SSS was usually calculated by using the empirical equation (e.g. Weiss et al., 1982). In this study, however, because time series of DIC were estimated, as mentioned above, pCO 2 sea at the mean SST and SSS (n-pCO 2 sea ) was calculated from time series of observed pCO 2 sea and calculated DIC, using the methods of Dickson et al. (2007). The calculated n-pCO 2 sea values were in good agreement with those calculated empirically according to the equations by Midorikawa et al. (2006), derived from the integration of the equation by Weiss et al. (1982); the differences between two calculation were 0.02 ± 0.33 μatm, which was smaller than the overall precision of pCO 2 sea measurements (<3 μatm) (Midorikawa et al., 2006).

Long-term changes in pCO 2 air and pCO 2 sea
The time series data for pCO 2 air showed similar, linearly increasing trends with little interannual variability at all latitudes along 137 • E in both winter and summer, whereas pCO 2 sea showed substantial increasing trends with a larger interannual variability over the entire area, from the subtropical to equatorial regions, in both winter and summer (Figs 4a and c). The values for pCO 2 sea were lower than pCO 2 air in winter because of lower SST at higher latitudes, whereas, in summer, pCO 2 sea values were higher than pCO 2 air because of the effects of high SST exceeding those of lowering DIC. The magnitude of pCO 2 sea interannual variability was relatively large at latitudes south of 20 • N in winter and north of 24 • N in summer. Furthermore, the interannual variability of pCO 2 sea in summer was larger than that in winter, especially owing to larger SST variability from 25 • N to 31 • N.

Long-term changes in pH at ambient SST
The estimated pH (Fig. 5a) was higher at high latitudes than at low latitudes in winter because of a large latitudinal gradient of SST (Fig. 4b). By contrast, pH normalized to a constant SST of 25 • C, pH T25 , was lower at high latitudes than at low latitudes because of higher DIC concentrations at higher latitudes (Ishii et al., 2001). In summer, the latitudinal differences in pH disappeared because of a much smaller North-South SST gradient. The time series pH data in both winter and summer at all latitudes exhibited distinct long-term decreasing trends with notable interannual variation but without readily recognizable periodicity. The interannual variability of pH was also large in summer.
We calculated the average rate of pH decrease for each 1 • of latitude along 137 • E for the entire study period using the linear least-squares method. The rates of decrease ranged from (mean ± 1 standard deviation [1σ ]) 0.0015 ± 0.0002 to 0.0021 ± 0.0002 yr −1 (average, 0.0018 ± 0.0002 yr −1 ) in winter (Fig. 6e) and from 0.0008 ± 0.0004 to 0.0019 ± 0.0005 yr −1 (average, 0.0013 ± 0.0005 yr −1 ) in summer (Fig. 7e). These trends were significant at most latitudes south of 33 • N: p < 0.001 at all    Bates, 2007). These similar trends in different regions of the ocean could reflect the rapid response of pCO 2 sea to changes in pCO 2 air , and furthermore, suggest no intrinsic difference in the changes in the surface carbonate system between these subtropical regions.

Contribution of changes in SST to those in pH and DIC
The rates of decreasing pH T25 were similar in winter (average, 0.0015 ± 0.0003 yr −1 ) and summer (average, 0.0014 ± 0.0004 yr −1 ). The steeper decreasing trends of pH at ambient SST in winter were attributed to the effects of a relatively greater rate of  Fig. 7d). The rate of increase in pCO 2 sea (average, 1.58 ± 0.12 μatm yr −1 in winter and 1.37 ± 0.33 μatm yr −1 in summer; Figs 6c and 7c) was not significantly different from that of pCO 2 air (average, 1.65 ± 0.05 μatm yr −1 in winter and 1.54 ± 0.08 μatm yr −1 in summer). However, the time series pCO 2 sea data in each latitude range normalized to the mean temperature and salinity for the entire period, n-pCO 2 sea , exhibited a pattern of lower rates of n-pCO 2 sea increase in latitudes showing increasing trends in SST, and higher rates in latitudes with decreasing SST trends, yielding a latitudinal distribution with relatively lower rates of increase in the southern region of the study area (especially, Fig. 7c).
These characteristics were the same in the trends of n-DIC and pH T25 (Figs 6 and 7) calculated on the assumption of a constant n-TA. These results suggest that the long-term response of the surface carbonate system to the anthropogenic CO 2 increase in the atmosphere could be predominantly through increases in pCO 2 sea , in response to the changes in pCO 2 air . Therefore, in latitudes with increasing SST trends, larger temperature effects could allow pCO 2 sea to keep up with the pCO 2 air level even in areas with a smaller rate of n-DIC increase, and consequently, further depress the n-DIC increase rate; decreasing SST trends would conversely induce higher rates of DIC increase.
These temperature effects could also extend to the magnitude of pH rate of decrease; in latitudes with increasing SST trends, a smaller rate of n-DIC increase would produce a more gradual decreasing pH T25 trend. For pH at ambient SST, larger effects of increasing SST trends, acting to decrease pH, could make up for those of the lower rates of n-DIC increase, leading to significant decreasing pH trends.
For the individual rate at each latitude, the effects of changing SST accounted for −25 to 44% (average, 15%) in winter and −85 to 37% (average, −10%) in summer of the decreasing pH trends (Figs 6e and 7e). Compared with the rates of pH T25 , the trend of increasing SST elevated the decreasing trend in pH at ambient SST in most latitudes (27 latitudes) in winter and the decreasing SST reduced the rates of decreasing pH at ambient SST in the northern subtropical region (a maximum of 46% against pH T25 at 25 • N) in summer.
Conspicuously low rates of change in pH T25 were found around 23-24 • N in winter, together those of n-pCO 2 sea and n-DIC, corresponding to high rates of SST increase (Figs 6 and  7). Midorikawa et al. (2006) indicated that the close coupling of interannual variations between the surface DIC concentrations and SST in these latitudes could be attributed to the influence of the interannual North-South movement of the Subtropical Front, which forms the boundary between the northern cold, DIC-rich water mass and the southern warm, DIC-poor water mass (Ishii et al., 2001), in addition to inverse DIC-SST anomaly relationships derived from winter vertical mixing. The position of the Subtropical Front, denoted by a surface potential density σ θ of 24 (or SST of 22 • C) in winter, exhibited a long-term tendency to shift to the north: 0.5 ± 0.3 • of latitude per decade during 1983-2007 (p = 0.10) and 0.29 ± 0.15 • per decade during 1967-2007 (p = 0.07), based on hydrographic observations (JMA, 2001(JMA, , 2008, and 0.22 ± 0.09 • per decade during 1951-2007 (p = 0.01) based on SST reanalysis . The long-term northward shift of the Subtropical Front could lead to the trends described earlier for the carbonate system properties and SST in these latitudes.

Long-term changes in pH since 1969
Paired data for SST, SSS, TA, and pCO 2 sea was acquired at five latitudes from 9 • N to 30 • N around 158 • E in February 1970 (Inoue et al., 1999). The n-TA values ranged in 2297 ± 3 μmol kg −1 , which is similar to those estimated for 1994-2007 (Fig. 2). Considering the effects of SST variations, the pH T25 values from these 1970 data were lower by 0.002 ± 0.006 to 0.030 ± 0.009 than those estimated for 1970 from the extrapolation of the pH T25 trends in the same latitudes for the past 25 yr in the present study. The rate of pH T25 decrease for 1970-1983 was calculated as 0.0009 ± 0.0006 yr −1 , which is smaller than those for the more recent trends. The difference between SST observed in 1970 and the average at the same latitude for 1983-2007 was 0.12 ± 0.82 • C, which was smaller than 1σ value (0.66 • C, the average in the entire area) for the interannual variability of SST during the past 25 yr, indicating that 1970 was not an anomalous year. Moreover, data set of SST, SSS, and pCO 2 sea at latitudes from 3 • N to 30 • N around 138 • E in February 1969 (the average anomaly, −0.17 ± 0.60 • C for 1969; Inoue et al., 1999) also gave a low average rate of pH T25 decrease (0.0008 ± 0.0006 yr −1 ,) for 1969-1983, assuming TA of 2295/35 * SSS μmol kg −1 . Although a determination of acidification based on snapshot observations from two specific years is difficult because interannual variations must also be considered, these data suggest an acceleration of acidification after the 1980s.

Estimation of changes in pH during the next 50 yr
The observational records of pCO 2 sea and the time series of pH estimated on the basis of these data in the western North Pacific for the past 25 yr were in agreement with long-term trends calculated from the atmospheric CO 2 increase. We therefore estimated the future trends of ocean acidification in the western subtropical North Pacific Ocean on the assumption that only the changes derived from the atmospheric CO 2 increase could continue to dominate changes in the surface carbonate system in these regions without altering the ecosystems and other factors related to the carbon cycle, excluding SST. Thus, the magnitude of future ocean acidification attributable to thermodynamic changes in the surface carbonate system of the western subtropical North Pacific over the next 50 yr was tentatively evaluated for the following two cases. Case 1 used the same rate of pCO 2 sea increase as that observed for the past 25 yr, that is, the extrapolation of the recent trend of pCO 2 sea increase. Case 2 assumed a rapid response of pCO 2 sea to the pCO 2 air changes predicted in accordance with future CO 2 emission scenario IS92a in IPCC (2007). Calculations were performed using the results of winter observations averaged in two representative realms: the northern (25-28 • N) and southern (11-14 • N) subtropical regions. The northern region is located near the centre of the highest sea surface dynamic height in the western subtropical gyre, and the southern region is located in the southern half of the North Equatorial Current (Fig. 1). For the past 25 yr, SST showed an insignificant trend (0.009 ± 0.018 • C yr −1 , p = 0.6) in the northern region, whereas a trend of increasing SST (0.020 ± 0.013 • C yr −1 , p = 0.12) was observed in the southern region, although this trend was not statistically significant. For Case 1, two sets of calculations, with and without SST changes, were compared (Table 1).
In Case 1, the magnitudes of n-DIC and pH changes from calculations ignoring SST increases were 45-50 μmol kg −1 for n-DIC and −0.073 to −0.085 for pH, respectively, over the next 50 yr (Table 1, Fig. 8). Under a constant SST, the rates of n-DIC increase in the two regions for 2040-2060 were 14-24% lower, and the rates of pH decrease were 21-22% lower, than the rates for 1983-2007. We attribute this decrease in the rates of change to the increase in the buffer factor (Revelle and Suess, 1957) from the increase in DIC.
When the same trend of increasing SST for the past 25 yr is assumed for the future, the rates of n-DIC increase are reduced an additional 5-18%, whereas there is almost no change to the projected rates of pH decrease in either region for 2040-2060 (a reduction of <1 to 1%). Under these conditions, the effects of increasing SST compensated for the slowing of the rate of pH decrease resulting from the reduction in the rate of DIC increase from the increase in the buffer factor.
In Case 2, under the future CO 2 emission scenario IS92a, larger amounts of oceanic CO 2 uptake were predicted considering the past SST trend, leading to an additional n-DIC increase of 24 μmol kg −1 and subsequent additional pH decrease of 0.04 (total pH decrease of 0.11-0.13 over the next 50 yr; Table 1, Fig. 9), compared with Case 1. The rates of both n-DIC increase and pH decrease for 2040-2060 were notably higher in the two regions, compared with those for the past 25 yr and those in Case 1. The predicted rates of pH decrease for 2040-2060 were about 40% greater than those from recent years. It is therefore possible that future ocean acidification could be accelerated, depending on increases in CO 2 emissions. The magnitudes of the total pH decreases estimated in the present study are less than that predicted for the average over the global ocean in previous reports (a decrease of approximately 0.2 by the year 2050) (Caldeira and Wickett, 2003;Orr et al., 2005).
Our calculations in the present study are confined within the equilibria among the dissolved and ionic carbonate species, as mentioned in the section of Methods. Recent studies have indicated possible future changes in marine ecosystems resulting from global warming, ocean acidification, or both. Recent  reports on the effects of ocean acidification on the structure and function of marine microbial communities under high CO 2 conditions (e.g. Doney et al., 2009) suggest significant feedbacks to the upper ocean carbon cycle through related biogeochemical processes. It is important to monitor the surface carbonate system through time series observations, including high-precision pH measurements as well as the information on the formation of CaCO 3 and organic matter, to determine the future trends of acidification in the upper water column, and to detect changes in the surface carbonate system. Future studies using time series data over extensive regions are required to determine the effects of changes in marine ecosystems on the upper-ocean carbon cycle.

Conclusions
We firmly evaluated the trend of ocean acidification in surface layer of the western North Pacific using the data of partial pressure of CO 2 acquired for the period 1983-2007 when no precise data of pH was available. Time series of pH estimated at ambient SST exhibited distinct long-term trend of acidification (∼0.01 to 0.02 pH unit per decade) over the regions from the tropical to the northern subtropical zone in both winter and summer. The magnitude of pH decrease in these regions was close to those reported for other subtropical regions (HOT, BATS and ESTOC sites). These similar trends in different regions of the ocean suggest that the long-term response of the surface carbonate system to the anthropogenic CO 2 increase in the atmosphere could be predominantly through increases in only pCO 2 sw (but not DIC), in response to the changes in pCO 2 air . The comparisons of the pH decreasing rate for the past 25 yr with those for a decade since 1969 and for the next 50 yr suggest an acceleration of the acidification in recent years, as well as in the future depending the scenario. Our analysis exhibited that, in latitudes with increasing SST trends, its larger thermodynamic effects to decrease pH at ambient SST exceeded those to depress n-DIC increase, leading to larger decreasing pH trends. These results suggest that the future ocean warming could also be a factor accelerating the acidification through its thermodynamic effects. Continual monitoring of the qualitative and quantitative changes in the surface carbonate system through time series observations, including high-precision pH measurements, is required in future study.