Decadal change of the surface water pCO2 in the North Pacific: A synthesis of 35 years of observations

Authors


Abstract

[1] Surface water pCO2 data observed over the 3 decades between 1970 and 2004 are analyzed for space and time (mean decadal) variability in thirty-two 10° × 10° box areas over the North Pacific Ocean north of 10°N. During this period, the pCO2 values at SST increased at a mean decadal rate of 12.0 ± 4.8 μatm decade−1 in all but four areas located in the vicinity of the Bering and Okhotsk Seas, where they decreased at a mean rate of −11.1 ± 5.7 μatm decade−1. The mean rate of increase for the open ocean areas is indistinguishable from the mean atmospheric CO2 increase rate of 15 μatm decade−1 (or 1.5 ppm yr−1) suggesting that the North Pacific surface waters as a whole have been following the atmospheric CO2 increase. However, the rate of increase varies geographically, reflecting differences in local oceanographic processes including lateral mixing of waters from marginal seas, upwelling of subsurface waters and biological activities. The decrease observed in the southern Bering Sea and the peripheries of the Okhotsk Sea may be accounted for by the combined effects of intensified biological production and changes in lateral and vertical mixing in these areas. The natural logarithm of wintertime pCO2 values normalized to a constant temperature and salinity of 14.3°C and 34.0 (the basin mean values, respectively) is correlated with winter SST. Using this relationship, the wintertime TCO2 in mixed layer can be expressed as a function of winter SST with a standard error of ±5 μmol kg−1.

1. Introduction

[2] The global oceans are one of the major dynamic reservoirs for CO2 containing about 50 times as much CO2 as the atmosphere, and continuing to absorb a large portion of the excess CO2 emitted by anthropogenic activities [Battle et al., 2000; Keeling and Garcia, 2002; Quay et al., 2003; Sabine et al., 2004a; Sarmiento et al., 2000; Takahashi et al., 2002]. The mean annual rate of CO2 uptake by the oceans for the past several decades has been estimated to be about 2 Pg-C yr−1. It is therefore important to know how the oceans have responded to the increased loading of CO2 into the atmosphere. Because of the availability of research and commercial ships of opportunity, the North Pacific Ocean has become one of the most frequently sampled regions of the world oceans for the investigation of seasonal and interannual variability of CO2 and nutrient chemistry [Landrum et al., 1996; Sabine et al., 2004b; Takahashi et al., 1993; Wong and Chan, 1991; Wong et al., 2002; Zeng et al., 2002].

[3] The purpose of this paper is to estimate decadal mean rates of change in the partial pressure of CO2 in surface mixed layer waters, (pCO2)sw, over the North Pacific Ocean and their spatial variability using the measurements made by many investigators over the past 35 years. Since the time-space variability of (pCO2)SW over the equatorial Pacific has been investigated recently [Feely et al., 2006; Takahashi et al., 2003], this paper deals with the areas north of 10°N. In addition, we will show that the wintertime (pCO2)sw is related to the wintertime mixed layer temperature (SST), and that it is also correlated tightly with the winter time total CO2 concentrations (TCO2) in the mixed layer. These relationships may be used for interpolating observations, as well as for testing ocean-atmosphere carbon cycle models.

2. CO2 Partial Pressure in Seawater

[4] The pCO2 in seawater is a vapor pressure of CO2, and the difference between pCO2 in seawater and that in the overlying air is one of the important factors governing the CO2 transfer flux across the sea-air interface. It is a sensitive function of temperature, doubling with every 16°C [Takahashi et al., 1993]. It is also a sensitive function of the total concentration of CO2 (TCO2), a sum of CO2 species dissolved in seawater: TCO2 = [CO2]aq + [HCO3] + [CO3=]. [CO2]aq represents the concentration of free CO2 molecules in aqueous media and is proportional to pCO2. TCO2 in seawater depends on the net biological community production, the rate of upwelling of subsurface waters rich in CO2, and the air-sea CO2 flux. The sensitivity of pCO2 to changes in TCO2 may be expressed in terms of the Revelle factor (= (∂ ln pCO2/∂ ln TCO2) T, S, Alk), which varies from 8 in tropical waters with lower TCO2 concentrations to 15 in polar waters with higher TCO2 concentrations [Takahashi et al., 1993].

[5] In the surface mixed layer, the effect of seasonal warming on pCO2 is counteracted by lower TCO2 caused by photosynthetic fixation of CO2, as often seen during spring bloom periods; the effect of winter cooling is counteracted by increasing TCO2 caused by upwelling of subsurface waters rich in respired CO2. Consider the following example: If a parcel of polar ocean water at −1.9°C is warmed to an equatorial temperature of 30°C without change in TCO2 and other chemicals, its pCO2 is increased by a factor of 4. On the other hand, deep waters upwelled in polar regions contain high nutrient concentrations, typically ∼35 μmol kg−1 nitrate. If the nitrate in this water is completely utilized by biological growth with the Redfield N/C ratio of 16/106, then the TCO2 in the same water would decrease from 2150 μmol kg−1 to 1920 μmol kg−1. As a result, pCO2 is decreased by a factor of 3 ((1920/2150)10 ≈ 0.33, using a Revelle factor of 10). The reduction is partially counteracted by the growth of CaCO3 shell forming organisms, which is on the average about 20% of the organic carbon production. This lowers the alkalinity from 2300 to 2200 μeq kg−1 and results in an increase of pCO2 by about 40%. Therefore, over the global oceans, the effect of change in temperature is roughly compensated by changes in TCO2 and alkalinity, and the time-space variation in surface seawater pCO2 is dictated by competing effects of temperature, net biological production and the deep water upwelling.

[6] Figure 1 shows the climatological mean distribution of (pCO2)sw for non–El Niño years over the North Pacific in February and August, normalized to a reference year of 1995. This is an updated version of the earlier distribution maps by Takahashi et al. [2002], and is based upon an improved database consisting of about 450,000 pCO2 measurements made over the tropical and North Pacific since the late 1960s. Since the atmospheric pCO2 was about 360 μatm in 1995, the blue-magenta areas in Figure 1 are a sink for atmospheric CO2, the green areas are nearly neutral and the yellow-orange areas are a CO2 source. The central and eastern equatorial Pacific between 5°N and 15°S is a strong CO2 source throughout a year due to upwelling of deep waters. The areas of the Kuroshio and its extension are a strong CO2 sink during winter due primarily to cooling, and are a weak source during the summer due to warming. The western subarctic areas along the Kuril and western Aleutian arcs are a strong CO2 source during winter due to convective mixing of deep waters rich in respired CO2 and nutrients, and these areas become a strong sink in spring and early summer due to intense photosynthesis fueled by the nutrients that were supplied by the upwelling during the previous winter. In subtropical gyres, the primary cause for seasonal changes in pCO2 is seasonal temperature changes, whereas those in subpolar and polar waters are due primarily to TCO2 changes caused by winter upwelling of deep waters and spring time plankton blooms [Takahashi et al., 2002]. Using the NCEP-42 year mean monthly wind speeds (at 10 m above the sea surface) [National Centers for Environmental Protection, 2004] and the wind speed squared dependence of sea-air gas transfer coefficient of Wanninkhof [1992], the pCO2 distributions shown in Figure 1 yield a mean annual net sea-to-air CO2 flux of about 0.01 Pg-C yr−1 over the area north of 50°N and an air-to-sea net flux of 0.5 Pg-C yr−1 over the area between 14°N and 50°N. These flux estimates are uncertain by about ±35%, about one half of which is due to variability in (pCO2)SW, and the remainder is due to wind speed variability (1σ) of ±2 m s−1.

Figure 1.

Distribution of surface water pCO2 for February and August in a reference year 1995. This has been constructed using the method described by Takahashi et al. [2002], but is based on an updated database containing about 450,000 pCO2 measurements over the tropical and North Pacific (15°S–75°N), which has been extracted from a global database of 1.7 million pCO2 measurements. The yellow-orange areas are a strong source for atmospheric CO2, and the blue-magenta areas are a strong CO2 sink.

3. Data Sources and Computational Procedures for the Decadal Change of pCO2 and SST

[7] For an investigation of time-space variability of CO2 chemistry, surface water pCO2 is chosen for the following reasons. First, a large number of measurements are available over the North Pacific for all seasons over the past 3.5 decades since 1968. Second, the measurements which have been made by various investigators using the air-seawater equilibration method are compatible as a result of the use of common CO2 gas mixture standards for calibrations at sea. Third, (pCO2)SW is a major property that governs sea-air CO2 flux. Fourth, pCO2 in seawater is a sensitive function of the total CO2 concentration dissolved in seawater because of the chemical amplification represented by the Revelle factor, so that small changes in TCO2 can be detected.

3.1. Sources of the Data

[8] The surface water pCO2 data and associated measurements (e.g., SST, salinity) used for this study have been obtained by the following research groups during the period from 1968 to 2004; Lamont-Doherty Earth Observatory (LDEO), Pacific Marine Environmental Laboratory (PMEL) of NOAA, Atlantic Oceanographic and Meteorological Laboratory (AOML) of NOAA; Institute of Ocean Sciences (IOS), Canada; and Japan Meteorological Center (JMC) and National Institute for Environmental Studies (NIES), Japan. The original data files are available at the Web sites of the respective groups. The total number of pCO2 measurements (north of 10°N) used in this study is about 327,000. Since (pCO2)SW computed using TCO2, alkalinity and/or pH are not always consistent with the directly measured values, only those measured by equilibration methods are used in this study. The assembled data files and the time trend plots (annual and seasonal) are available at 〈 http://www.ldeo.columbia.edu/CO2〉. The distribution of these data is shown in Figure 2.

Figure 2.

Numbers of the observations for surface water pCO2 in 10° × 10° box areas used in this study over the North Pacific Ocean. “K” after numbers indicates a multiplier of 1000. There are about 327,000 pCO2 measurements in the areas north of 10°N.

3.2. Spatial Resolution

[9] The 10° × 10° spatial resolution used in this study is selected primarily on the basis of the time-space density of the observations. While smaller box areas would permit the resolution of narrow oceanographic features such as the Kuroshio and equatorial currents, they would reduce the number of observations made in a box during different seasons and hence fail to demonstrate seasonal variation clearly. Accordingly, we have chosen to bin the data into 10° × 10° area boxes. Each box is defined according to its center point and contains ±5° latitude and longitude inclusive of 5.0° borders, with the exception of the six irregular size boxes which are located near the Aleutian and Kuril arcs along the northern edge of the Pacific. Four of these six boxes have rectangular shapes with a 10°E-W width but with varying N-S dimensions: the 55°N–165°W box has a N-S dimension of 54°N–60°N; the 45°N–165°W box has a N-S dimension of 40°N–54°N; the 55°N–175°E box has a N-S dimension of 52°N–60°N; and the 45°N–175°E box has a N-S dimension of 40°N–52°N. Two adjacent boxes that straddle the Aleutian arc have a slanted border, so that the 55°N–175°W box (with its southern border slanted at an angle of about 40°) contains only the data from the Bering Sea; and the 45°N–175°W box (with its northern border slanted at an angle of about 40° to fit the box located north of it) contains only the North Pacific data south of the Aleutian arc. Since we have no measurements inside the Okhotsk Sea, the boxes located along the Kuril chain contain only the Pacific data, although some of them reflect the outflow from the Okhotsk Sea.

3.3. Seasonal Correction and Annual Mean

[10] In many of the 10° × 10° box areas, measurements are available only in some months in a year, and fewer are available especially for pre-1997 periods. In order to estimate the decadal mean rate of change in (pCO2)SW based upon measurements made during various seasons in different years, it is necessary to eliminate the seasonal bias from the observations. The following method is used for correcting the (pCO2)SW data obtained during a certain month to yield an annual mean. Of a total of 55 boxes in the North Pacific north of 10°N, none of them has measurements in every month in a single year. However, 43 of them have observations for 6 months or more for the most recent 7-year period 1997 through 2004. Twelve boxes with less than 6 months data are not used in this study. We estimate a mean seasonal variability in each box using 1997–2004 7-year composite data. Assuming that the seasonal variability thus estimated remained unchanged over time, we correct monthly mean values to deseasonalize them.

[11] Figure 3 shows an example for our procedure. In each 10° × 10° box, mean monthly values for (pCO2)SW and SST values are computed (open circles) using the 7-year composite data. If measurements are available for a given month in different years, an average is obtained (open squares). Values for the months with no observations are estimated by linearly interpolating two adjacent monthly mean values (solid squares). A single year is wrapped around from December to January. The annual mean (pCO2)SW value is computed using all 12 monthly values including the interpolated values (the thick horizontal line), and the difference between the annual mean and each monthly value (vertical arrows) represents a seasonal adjustment to be applied to mean monthly observations. In the bottom panel of Figure 3, the monthly mean SST values, that are computed from the 1° × 1° data of Reynolds et al. [2003], are shown in gray dots for the same box area. The variability of their values reflects mostly the SST trend over 10° latitude. Our SST values measured concurrently with (pCO2) SW are consistent with their values. In each of the 43 box areas, mean seasonal variation for (pCO2)SW and SST are similarly established. We assume that the seasonal variability is invariant with time, and apply the seasonal adjustment values thus derived to mean values observed for the corresponding months during different years (including the pre-1997 years) in order to deseasonalize the observations.

Figure 3.

The 1997–2004 7-year composite data for (pCO2)SW and SST in Box 25°N–165°W, showing the seasonal variability. The open circles are mean monthly values, and the open squares are a mean of the mean monthly values if measurements were made during the same month in different years. The solid squares are values interpolated linearly using two adjacent mean monthly values. The annual mean is shown with a thick horizontal line, and the differences between the monthly values and the annual mean (vertical arrows) are seasonal adjustments applied for mean monthly values to deseasonalize the observations made in different years. The gray lines in the bottom panel indicate the range of climatological monthly mean SST values on 1° × 1° grid reported by Reynolds et al. [2003] for this box area.

3.4. Mean Decadal Trends

[12] The mean multidecadal trend of (pCO2)SW is estimated using observations made as far back as 1968. Since measurements were not always available for every month in each box, and since seasonal variability of (pCO2)SW is large, each mean monthly value is deseasonalized using the method described above. The deseasonalized mean monthly (pCO2)SW values are regressed linearly against time to obtain the mean decadal rate of change for (pCO2)SW. The SST values observed concurrently with (pCO2)SW are deseasonalized similarly, and the mean rate of change is computed by means of a linear regression. Uncertainties for the rates of change are computed using ±[σ2/(Σ(Xi2) − N(Xmean)2)]1/2, where σ2 = [(Σ(Yi − aXi − b)2)/(N − 2)] is the variance around the fitted equation Y = a X + b, and Y is (pCO2)SW or SST and X is year.

4. Decadal Trends of Surface Water pCO2 and SST

[13] The mean decadal rates of change of (pCO2)SW and SST determined in 43 box areas are discussed in this section. The data and their analysis for the three areas which show contrasting trends are presented first in some detail. Overall features observed in the North Pacific will follow.

4.1. Area (55 ± 5°N, 145 ± 5°W) and Weather Station “P”

[14] Figure 4 shows the surface ocean pCO2 and SST data obtained in Box 55°N–145°W (50°N–60°N, 140°W–150°W) between 1970 and 2002. This box includes the Weather Station “P” (50°N and 145°W), where measurements were made in 1973–1979 by Wong and Chan [1991], in 1984–1989 by Takahashi et al. [1991], and in 1993–2002 by NIES, PMEL and others. This is one of the best documented areas in the North Pacific. One of the unique features of this site is that the seasonal amplitude of pCO2 is small (on the average about 50 μatm) in spite of the fact that the seasonal amplitude of SST is as large as 13°C. The warming alone should increase the (pCO2)SW by about 70% from a typical winter value of 300 μatm to 500 μatm during summer. The effect of warming, however, is largely canceled by the biological drawdown of CO2 in spring-summer months. The top panel shows all the pCO2 data at in situ temperature (solid dots) and the mean monthly values (open circles) that are deseasonalized using the method described in section 3.3. The solid line represents a linear regression line computed using the deseasonalized monthly mean values yielding a mean rate of increase of (pCO2)SW at SST of 19.9 ± 1.7 μatm decade−1 (with 87 mean monthly values). Because of the irregular data distribution and the large amplitude of seasonal changes, we are unable to identify the effects of the North Pacific Decadal Oscillation and the El Niño events in terms of a slope change and displacement of the time trends. During the same period 1970–2003, the atmospheric pCO2 has been increasing with decadal rate ranging from 12 to 19 μatm decade−1 (or 1.5 ppm yr−1 ranging from 1.2 to 1.9 ppm yr−1 over the 30-year period) as a result of anthropogenic emissions. Although the surface water pCO2 at this box area appears to be increasing at a rate consistent with the atmospheric CO2 increase, other contributing factors will be evaluated below.

Figure 4.

Surface water pCO2 and SST observations in Box 55°N–145°W. This area includes the Weather Station “P.” The dots indicate individual observations, and the solid lines indicate the linear regression lines computed using the deseasonalized mean monthly values (open circles). (top) The pCO2 at SST, (middle) the SST measured concurrently with pCO2, and (bottom) the pCO2 values normalized to a constant temperature of 7.91°C, the mean of the monthly mean values. The mean decadal rate of change in each property is shown along the bottom of each panel, and N indicates the number of mean monthly values used.

[15] Since seawater pCO2 is a sensitive function of temperature, we need to examine the contribution of SST changes to (pCO2)SW. The middle panel in Figure 4 shows the SST data (solid dots) obtained concurrently with pCO2, and the open circles indicate the deseasonalized monthly mean SST. The straight line indicates a linear regression line computed using these monthly mean values, with a mean temporal rate of −0.1 ± 0.1°C decade−1, that includes zero change. The validity of this estimate is tested by comparing the time trend obtained using the monthly mean SST data set from Reynolds et al. [2003], that has complete monthly coverage for 1981–2003. Their data, when processed in the identical way used for this study, yield a mean temporal rate of −0.11 ± 0.07°C decade−1 for this box, which is consistent with that based upon our own SST data. Thus, although the time-space distribution of our data is irregular and incomplete, it appears to yield a credible estimate for the SST time trend for this box. Over the past 3 decades, the mean SST appears to have stayed constant within 0.1°C. This means that the observed increase rate in (pCO2)SW is caused primarily by change in seawater chemistry, and is not affected significantly by SST changes. Below, the nature of chemical changes will be further explored.

[16] Freeland et al. [1997] reported the following changes in upper ocean layers at and near the Station “P” from 1970 to 1994; (1) a slight warming of mixed layer at a mean rate of +0.2 ± 0.1°C decade−1 (which differs from our estimate of −0.1 ± 0.1°C decade−1 for 1970–2002 representing a broader sampling area and a longer time span), (2) a decrease in salinity at a mean rate of −0.04 ± 0.03 decade−1, (3) a decrease in the winter mixed layer thickness at a mean rate of −6.3 ± 2.8 m decade−1, and (4) a decrease in the winter-average mixed layer nitrate concentration by about 30% the from 16.2 μmol kg−1 in 1970 to 12.3 μmol kg−1 in 1994 at a mean rate of −1.6 (±1.1) μmol kg−1 decade−1. These observations are consistent mutually: As the winter mixed layer becomes shallower, subsurface waters with lesser nutrient concentrations should be mixed into the mixed layer. Their one-dimensional vertical mixing model (without biological effects) yields a rate of decrease in nitrate of −0.43 μmol kg−1 decade−1 that is nearly consistent with the field observations with the lower limit of the estimated uncertainty. The low model value suggests that other processes such as change in lateral transport may be involved. We test below whether their observations are consistent with our increasing pCO2 trend observed in this area.

[17] Changes in the carbon and nutrient concentrations that are associated with the observed decrease in nitrate concentration may be estimated using the properties of sub-mixed layer waters. We assume that the chemical properties of the sub-mixed layer water that is added to the winter mixed layer is represented by the measurements made near a depth of 200 m at a WOCE Station (48.2°N and 146.6°W): TCO2 (2200 μmol kg−1)/NO3 (34.0 μmol kg−1) = 64.70, TALK (2262 μeq kg−1)/NO3 = 66.53, SiO3 (56.0 μmol kg−1)/NO3 = 1.65 and PO4 (2.15 μmol kg−1)/NO3 = 0.063. The reduction of these properties corresponding to the nitrate reduction may be estimated by multiplying the observed rate of nitrate reduction of −1.60 μmol kg−1 decade−1 with each property/nitrate ratio: −106 μeq kg−1 decade−1 for the total alkalinity (TALK), −103 μmol kg−1 decade−1 for TCO2, −2.6 μmol kg−1 decade−1 for silica and −0.1 μmol kg−1 decade−1 for phosphate. Using an inorganic chemical equilibrium model with a typical winter mixed layer chemistry, we obtain a pCO2 increase of about +13 μatm decade−1 at a mean winter SST of 4.0°C and salinity of 32.6. The increase rate in pCO2 as subsurface water addition is reduced may be accounted for by the fact that the alkalinity is reduced faster than TCO2 (−106 μeq kg−1 versus −103 μmol kg−1), and that an increasing effect on pCO2 of the lower alkalinity wins out the reducing effect of lower TCO2. This indicates that the decrease in nitrate observed in the winter mixed layer water is consistent with the observed increase in the mixed layer pCO2.

[18] Wong et al. [2002] observed that during the 1995–1997 period, the production of CaCO3 was low or negligibly small in the western subarctic Pacific, whereas it was high in the eastern sector ranging from 6% to 75% of the organic carbon production. This indicates that during the post-winter growing season, the surface water pCO2 was controlled by a competition of the decrease in mixed layer pCO2 by biological CO2 fixation against the increasing effects of warming SST and alkalinity reduction by CaCO3 production. This may contribute to the small seasonal pCO2 amplitudes observed in this region.

[19] C. S. Wong et al. (unpublished manuscript, 2006) reported cooling and significant increases in surface water salinity and concentrations of nutrients and TCO2 in the Station “P” area in coincidence with the1976 and 1989 La Niña periods. As shown in Figure 4, the pCO2 data failed to show the 1976 event (and no pCO2 data for the 1989 event). During this event, SST decreased by about 1.5°C, while TCO2 and salinity increased by about 20 μmol kg−1 and 0.1, respectively. Thus the increase in TCO2 should have increased pCO2 by 11% (∼40 μatm), while the cooling should have decreased pCO2 by about 6.3% (∼22 μatm). In addition, an increase of about 7 μeq kg−1 in the alkalinity that is estimated assuming its proportionality with the salinity would have decreased pCO2 by about 3% (∼11 μatm). Therefore the effect on pCO2 of the TCO2 increase by the La Niña event is compensated largely by the cooling and alkalinity increase. This suggests that we must measure as many carbon-nutrient parameters as possible in order to understand the regulatory mechanisms for surface water pCO2 and hence the sea-air CO2 exchange.

[20] In summary, the observed pCO2 increase rate of +19.9 μatm decade−1 appears to be regulated not only by sea-air gas transfer and net community production of organic matter and CaCO3, but also by changes in the upper layer dynamics including a reduction in mixed layer depth and changes in lateral transport with time.

4.2. Area (25 ± 5°N and 155 ± 5°W) and Station ALOHA

[21] Next, we examine the 25°N–155°W box, which includes the Station ALOHA (22.7°N and 158°W) of the Hawaii Ocean Time-series (HOT) project. The top panel of Figure 5 shows individual observations (solid dots) and deseasonalized mean monthly pCO2 values (open circles) in surface waters. A linear regression for the monthly mean values yields a mean slope of 13.0 ± 2.2 μatm decade−1 for the 1970–2003 period. As observed for Box 55°N–145°W, the rate of increase in surface water pCO2 is consistent with the mean rate of increase for atmospheric pCO2. The SST in this area is virtually unchanged over the 30-year period (the middle panel, Figure 5): 0.1 ± 0.2°C decade−1 on the basis of our SST data obtained concurrently with the pCO2 measurements and 0.04 ± 0.04°C decade−1 based on the monthly data of Reynolds et al. [2003]. Hence the observed pCO2 increase rate is due mainly to changes in the water chemistry.

Figure 5.

Surface water pCO2 and SST observations in Box 25°N–155°W. This area includes the Station ALOHA of the HOT program. The dots indicate individual observations, and the solid lines indicate the linear regression lines computed using the deseasonalized mean monthly values (open circles). The crosses indicate the (pCO2)SW and SST values at the Station ALOHA reported by Dore et al. [2003]. Since pCO2 was not measured during the HOT program, these pCO2 values have been computed using the measured alkalinity and TCO2 values, and hence are not used in our analysis. (top) The pCO2 at SST, (middle) the SST measured concurrently with pCO2, and (bottom) the pCO2 values normalized to a constant temperature of 24.76°C, the mean of the monthly mean values, that are used to estimate changes in TCO2.

[22] On the basis of time series measurements of the alkalinity, TCO2, temperature and salinity in surface waters at the Station ALOHA, Dore et al. [2003] computed (pCO2)SW and reported a mean increase rate of 24.6 ± 2.8 μatm decade−1 for the 13-year period, 1989–2002. Keeling et al. [2004] used also the alkalinity and TCO2 data, that were determined by them independently for the Station ALOHA samples, and reported that (pCO2)SW increased at a mean rate of 14 ± 2 μatm decade−1 for the first 8-year period, 1988–1996, and at a much faster rate of 32 ± 4 μatm decade−1 for the following 5-year period, 1997–2002. On the basis of a thorough analysis of the data using a diagnostic box transport model, they concluded that the post-1997 increase in (pCO2)SW was a result of local decrease in precipitation as well as to a regional change in water mass distributions, that is perhaps related to the Pacific Decadal Oscillation (PDO). Thus their mean rate of increase of 25 ± 1 μatm decade−1 for their 14-year study period of 1988–2002 is in agreement with that reported by Dore et al. [2003].

[23] The 14-year mean rate of (pCO2)SW increase determined by these investigators are twice as large as our 35-year mean rate 13.0 ± 2.2 μatm decade−1estimated using the directly measured pCO2 values, whereas the pre-1997 8-year mean rate of 14 ± 2 μatm decade−1 obtained by Keeling et al. [2004] is consistent with ours. Two reasons may be considered to account for the difference: The first is our undersampling problem, and the second is incompatibility of the computed with the measured pCO2 values.

[24] First, since we have no pCO2 observations during 1992–1996 and a limited number of observations during 1996–2004 as shown in Figure 5, our observations are not sufficient for documenting the 1996–1997 transition to a higher rate of pCO2 increase reported by the previous investigators. Furthermore, since the amplitudes for the pCO2 seasonal variability (about 60 μatm) and are much greater than the mean decadal rate of pCO2 increase (14 to 34 μatm decade−1), a trend change is difficult to detect unless observations over a sufficiently long period are available. If our data are linearly regressed for the pre- and post-1996 periods separately, they yield a rate of 9.8 ± 4.7 μatm decade−1 (N = 29) for 1970–1996, and 15.1 ± 10.3 μatm decade−1 (N = 18) for 1996–2004. These rates are statistically indistinguishable owing to the large uncertainties which result from undersampling, large seasonal variability and our broader 10° × 10° sampling area. Although our data are not sufficient to resolve the 1996–1997 trend change, they allow us to obtain a mean rate of pCO2 change over the 3 decades.

[25] Second, Dore et al. [2003] and Keeling et al. [2004] both computed (pCO2)SW using an inorganic chemistry model, whereas we measured the pCO2 directly using an air-water equilibration method. In inorganic models, the ionization effects of organic acids are neglected. Organic acids may affect the carbonate equilibria by the ionic dissociation (i.e., organic acid alkalinity) and/or by forming complex ions with other ions. Considering the fact that the total dissolved organic carbon (DOC) ranges from about 60 to 100 μmol kg−1, the organic carbon alkalinity could be of an order of 10% of DOC concentration. This must be subtracted from the total (or titration) alkalinity in order to compute the carbonate alkalinity, and hence it alters the computed pCO2 values by as much as 5% [Millero et al., 2002]. However, the organic alkalinity cannot be determined since the compositions of organic acids within DOC and their dissociation constants are not known. Hence the information needed for evaluating the systematic differences between the computed and measured (pCO2)SW values are not available presently. Keeling et al. [2004] cited Dore's personal communication that the computed values are systematically greater than a limited number of directly measured pCO2 values by 3.6 ± 5.7 μatm. However, whether the differences are distributed randomly or correlated with salinity is not clear from their descriptions. Thus we are unable to evaluate whether the pCO2 trend change in association with salinity changes is real or due to changes in organic alkalinity associated with changes in salinity and water masses.

4.3. Area (50°N–60°N and 180°–170°W) in the Central Bering Sea

[26] The 1973–2002 data in this box (Figure 6) are entirely from the southern Bering Sea, which is free of ice cover year round. The open circles indicate the deseasonalized mean monthly values, which are used for computing the mean rate of change using a linear regression. In contrast to the two previous areas in the open Pacific, the surface water pCO2 decreases with time at a mean rate of −17 ± 12 μatm decade−1. The mean SST, meanwhile, stayed virtually unchanged (Figure 6, middle panel): 0.1 ± 0.2 °C decade−1 based on our measurements made concurrently with pCO2 and −0.01 ± 0.05 °C decade−1 on the basis of the data of Reynolds et al. [2003]. Using the pCO2 data corrected to a constant temperature of the area mean 5.76°C, we estimate that pCO2 decreased at a rate of −19 ± 13 μatm decade−1 (Figure 6, bottom panel) as a result of changes in chemistry. If this change were assumed to be due solely to TCO2 change, a TCO2 decrease of 8 ± 5 μmol kg−1 decade−1 is expected using a Revelle factor of 14.5 (see Table 2 in section 4.5). Similar decreasing trends are observed in other areas located within the Bering Sea and in waters just outside the Okhotsk Sea. We discuss below various factors which contribute to the observed trend.

Figure 6.

Surface water pCO2 and SST observations in Box 55°N–175°W located in the central Bering Sea. The dots indicate individual observations, and the solid lines indicate the linear regression lines computed using the deseasonalized mean monthly values (open circles). (top) The pCO2 at SST, (middle) the SST measured concurrently with pCO2, and (bottom) the pCO2 values normalized to a constant temperature of 5.76°C, the mean of the monthly mean values.

[27] On the basis of the Coastal Zone Color Scanner (CZCS; 1979–1986) and Sea-viewing Wide Field-of–view Sensor (SeaWiFS; 1997–2003) ocean color measurements from satellites, Gregg et al. [2003] reported that the ocean primary production increased substantially in the Bering and Okhotsk Seas over the decade. However, their results could be uncertain because of persistent cloud cover in the region, which could bias the CZCS and SeaWiFS data differently. Although their observations reflect the gross production rather than the net community production which is relevant to changes in pCO2 and TCO2 in seawater, the reported trends are considered to be a factor that contributes to the observed pCO2 trend. The increase in the productivity may reflect changing supplies of nutrients into the Bering Sea caused by changes in land hydrology or by increases in airborne or river inputs of nutrients mediated by anthropogenic activities.

[28] In the southeastern Bering Sea, the mixed layer depth decreased and stratification increased as a result of warming and freshening of surface waters from May 2001 through October 2004 [Wirts and Johnson, 2005]. Significant warming was also observed at a mooring (56.8°N and 164°W) in the eastern shelf area in 1995–2003 [Overland and Stabeno, 2004]. This may be due to the combined effects of weak wind forcing and increased inflow of the warmer, nutrient-rich Alaskan Stream water. As a result, while the supply flux of nutrients and CO2 into the surface layer by the wintertime vertical mixing was decreased, lateral supply of nutrients into the Bering Sea was increased. This, combined with an increase in productivity due to earlier ice melt, may have caused the decadal decrease in surface water pCO2. Even if the community production per unit ocean surface area remains unchanged, a reduction in mixed layer depths due to warming (i.e., a smaller volume of photic waters per unit area) could cause a greater lowering of pCO2 and TCO2 in the mixed layer.

[29] The observed decrease in (pCO2)SW in the Bering Sea translates to a pH increase of about 0.02 ± 0.01 decade−1, suggesting that the seawater has become more alkaline and that the H+ ion concentrations have decreased by 14% for the past 30 years. This is in contrast to the acidification of −0.1 pH unit in open ocean that is estimated for surface ocean waters in equilibrium with atmospheric pCO2 of 280 μatm during the pre-industrial period and today's 380 μatm [Feely et al., 2004]. Thus the Bering Sea may provide an interesting environment for investigating the contrasting pH effects on marine ecosystems.

4.4. Decadal Change of pCO2 in the North Pacific Surface Waters

[30] The mean decadal rates of change in (pCO2)SW observed in 43 box areas are tabulated in Table 1. As mentioned earlier, all of these boxes have observations in 6 or more months during the most recent 7-year period, 1997–2004. However, since many of the boxes have been sampled at irregular intervals with varying degrees of sampling density during the study period of 1970–2004, the reliability of our method for computing mean rates of change needs to be tested. For this purpose, we established the following criteria using the available SST data. We compute the decadal mean rate of change for SST in each box using the complete monthly set (1° × 1° spatial resolution) obtained by Reynolds et al. [2003] for 1981–2002 (Table 1, third column), and compare it with that computed using the SST values measured concurrently with pCO2 (Table 1, fourth column). The mean for the 43 pairs of the differences between columns 3 and 4 is 0.18°C decade−1 with a standard deviation of 0.85°C decade−1 (Table 1, fifth column). The boxes which have SST change rates that differ from the climatological SST rates of the third column by more than 0.85 °C decade−1 are considered to be unreliable, and hence 11 boxes are rejected (as indicated with bold numbers in fifth and sixth columns), and the remaining 32 boxes are accepted for further study.

Table 1. Decadal Mean Rate of Change of Surface Water pCO2 and SST in 10° × 10° Box Areasa
N. Lat. °NLong. °E or °WClimatological Mean SST Rate, °C decade−1pCO2 Data Mean SST Rate, °C decade−1Difference, °C decade−1Surface Water pCO2 Mean Rate of Change, μatm decade−1
IncreaseDecrease
  • a

    Climatological Mean SST Rate is the SST change rates computed using the complete monthly data for 1981–2002 by Reynolds et al. [2003]; pCO2 Data Mean SST Rate is the SST change rates computed using the observations made concurrently with pCO2 measurements. Difference is the difference between climatological mean SST rate and pCO2 data mean SST rate, and the boxes which exceed a difference of ±0.85°C decade−1 are rejected as insufficient observations. The box areas to be rejected are marked with bold numbers. Surface Water pCO2 Mean Rate of Change is obtained by a linear regression of the deseasonalized mean monthly pCO2 values: the positive rates are on the left side column, and the negative rates are in italics listed on the right side column.

  • b

    Eastern boundary coast area.

15 ± 5135E ± 50.22 ± 0.040.25 ± 0.10−0.0314.3 ± 1.4 
15 ± 5145E ± 50.25 ± 0.040.27 ± 0.16−0.0212.6 ± 1.6 
15 ± 5165W ± 5−0.07 ± 0.04−0.20 ± 0.100.1312.4 ± 1.8 
15 ± 5155W ± 5−0.17 ± 0.04−0.16 ± 0.11−0.018.5 ± 2.7 
15 ± 5145W ± 5−0.21 ± 0.050.08 ± 0.23−0.2913.0 ± 2.1 
15 ± 5125W ± 50.20 ± 0.050.71 ± 0.490.918.4 ± 4.6 
15 ± 5115W ± 5−0.10 ± 0.04−0.20 ± 0.560.104.5 ± 6.0 
15 ± 5105W ± 50.09 ± 0.040.38 ± 0.34−0.2914.0 ± 3.2 
15 ± 595W ± 50.06 ± 0.04−0.47 ± 0.510.5325.5 ± 9.9b 
25 ± 5135E ± 50.28 ± 0.05−0.06 ± 0.190.3411.9 ± 1.6 
25 ± 5145E ± 50.36 ± 0.05−0.13 ± 0.300.4912.3 ± 2.0 
25 ± 5165E ± 50.35 ± 0.050.35 ± 0.580.0016.2 ± 3.4 
25 ± 5165W ± 50.26 ± 0.050.46 ± 0.34−0.2017.6 ± 2.0 
25 ± 5155W ± 50.04 ± 0.040.10 ± 0.25−0.0613.0 ± 2.2 
25 ± 5135W ± 50.23 ± 0.051.51 ± 0.441.284.0 ± 3.9 
25 ± 5125W ± 5−0.16 ± 0.06−0.19 ± 0.600.0313.4 ± 2.9 
25 ± 5115W ± 50.15 ± 0.081.09 ± 0.921.2413.2 ± 5.4 
35 ± 5145E ± 50.64 ± 0.060.41 ± 0.431.059.9 ± 2.1 
35 ± 5155E ± 50.54 ± 0.070.22 ± 0.430.328.8 ± 2.7 
35 ± 5165E ± 50.41 ± 0.07−0.07 ± 0.470.488.1 ± 2.0 
35 ± 5175E ± 50.31 ± 0.07−0.09 ± 1.100.407.3 ± 4.5 
35 ± 5175W ± 50.34 ± 0.081.67 ± 0.652.018.9 ± 3.3 
35 ± 5165W ± 50.35 ± 0.080.93 ± 0.741.288.4 ± 4.0 
35 ± 5155W ± 50.29 ± 0.083.24 ± 0.843.535.7 ± 4.0 
35 ± 5145W ± 50.10 ± 0.07−0.69 ± 0.390.7915.4 ± 2.3 
35 ± 5135W ± 5−0.01 ± 0.060.46 ± 0.35−0.4717.0 ± 2.9 
35 ± 5125W ± 5−0.11 ± 0.060.33 ± 0.46−0.442.0 ± 7.7b 
45 ± 5145E ± 50.31 ± 0.071.13 ± 0.58−0.82 −4.0 ± 6.7
45 ± 5155E ± 50.23 ± 0.060.91 ± 0.661.149.9 ± 4.2 
45 ± 5165E ± 50.26 ± 0.070.86 ± 0.491.1218.6 ± 3.4 
45 ± 5175E ± 50.22 ± 0.070.50 ± 0.49−0.2811.8 ± 3.4 
45 ± 5175W ± 50.15 ± 0.060.53 ± 0.44−0.3811.4 ± 3.3 
45 ± 5165W ± 50.09 ± 0.060.08 ± 0.370.015.3 ± 6.1 
45 ± 5155W ± 50.10 ± 0.091.25 ± 0.671.1514.9 ± 3.2 
45 ± 5145W ± 50.03 ± 0.091.11 ± 0.251.0814.8 ± 1.6 
45 ± 5135W ± 5−0.04 ± 0.07−0.44 ± 0.270.4010.1 ± 2.4 
45 ± 5125W ± 50.07 ± 0.070.01 ± 0.210.069.7 ± 5.4 
55 ± 5175E ± 5−0.01 ± 0.050.40 ± 0.14−0.41 −14.2 ± 4.2
55 ± 5175W ± 5−0.01 ± 0.050.11 ± 0.17−0.12 −17.0 ± 11.8
55 ± 5165W ± 50.13 ± 0.07−0.02 ± 0.260.15 −9.0 ± 14.7
55 ± 5155W ± 5−0.02 ± 0.06−0.10 ± 0.260.087.9 ± 9.2 
55 ± 5145W ± 5−0.11 ± 0.07−0.07 ± 0.10−0.0419.9 ± 2.3b 
55 ± 5135W ± 50.03 ± 0.07−0.05 ± 0.450.0813.2 ± 3.6 
TotalsClimatological Mean SST Rate, °C decade−1pCO2 Data Mean SST Rate, °C decade−1Difference, °C decade−1Surface Water pCO2 Mean Rate of Change, μatm decade−1
IncreaseDecrease
Basin Mean0.11−0.060.1812.0−11.1
Standard Deviation±0.21±0.80±0.85±4.8±5.7
Number of Box Areas434343284
Standard Error±0.03±0.12 ±0.92±2.9

[31] Figure 7 shows the geographical distribution of (pCO2)SW change rates in these 32 boxes, of which 4 have negative rates and 28 have positive rates. The three boxes with negative rates are located in the Bering and one is outside the Okhotsk Sea along the Kuril chain, and possible causes for the negative trends have been discussed in section 4.3). The four boxes listed in the sixth column of Table 1 give a mean rate of −11.1 ± 5.7 μatm decade−1, whereas the mean of the 28 open ocean boxes is 12.0 ± 4.8 μatm decade−1 (Table 1). The map shows that a western temperate zone between 30°N–−40°N and 150°E–180° appears to have slower rates (7.3–8.8 μatm decade−1) than the open ocean mean. They are located in the areas of the Kuroshio Current between the polar and subtropical fronts and also are influenced by the Oyashio Current which contains the outflow from the Okhotsk Sea with negative pCO2 change rates. Furthermore, according to Chen and Wang [1999] and Chen et al. [2004], the alkalinity in the East China Sea is high as a result of the input from the two major rivers draining the Chinese mainland and from the oxidation of organic debris in the broad shelf sediments. Hence we speculate that the East China Sea waters would have a negative rate of pCO2 change caused by an increasing influx of alkalinity, and hence that entrainment of these waters into the Kuroshio Current might lower the rate of pCO2 increase. A long-term monitoring of chemical properties of the river and shelf waters is needed for our improved understanding of the influence of river waters and continental processes on the open ocean water carbon chemistry.

Figure 7.

Mean decadal rate of change of surface water pCO2 (at SST) in the North Pacific. These changes in pCO2 include the effects of changes in SST and other factors, and represent actual rates of change in the ocean pCO2. The numbers in bold letters indicate the rate of increase in μatm decade−1, the light italic numbers indicate the rate of decrease in μatm decade−1, and the values in parentheses indicate the uncertainty in the same unit. The gray curves in the western Pacific show the approximate locations of the Polar Front (PF) [Belkin et al., 2002] and Subtropical Front (STF).

[32] Coastal areas along the North and Central Americas exhibit widely varying rates ranging from 2.0 ± 7.7 μatm decade−1 off the Oregon coast, to 19.9 ± 2.3 μatm decade−1 in the Gulf of Alaska and 25.5 ± 9.9 μatm decade−1 in the Guatemala Basin. In these eastern boundary areas, the rate of pCO2 change depends largely on vertical and lateral circulation rates of water, and is influenced primarily by basin- and local-scale meteorological and climatic conditions rather than the sea-air CO2 flux [e.g., Friederich et al., 2002; Hales et al., 2005; van Geen et al., 2000]. Although our SST data in these areas give decadal rates consistent with those from Reynolds et al. [2003], the wide range of decadal (pCO2)SW rates is most likely due to undersampling of highly variable coastal systems.

[33] The 1970–2003 mean rate of (pCO2)SW increase for all 28 box areas is 12.0 ± 4.8 μatm decade−1 (Table 1). If the above three boxes located in the North American coastal upwelling areas (footnote b in Table 1) are excluded, the remaining 25 open ocean boxes give 12.0 ± 3.8 μatm decade−1. These rates are some what slower than the 1970–2003 mean rate of atmospheric CO2 increase of about 15 μatm decade−1 (or 1.5 ppm yr−1) observed at the Mauna Loa Observatory over the 33-year period [Keeling and Whorf, 1994]. We are thus unable to resolve unequivocally whether the North Pacific mixed layer waters as a whole are taking up CO2 from the atmosphere at the same or slower rate as the atmospheric CO2 increase. The ocean mean rate of 12.0 μatm decade−1 is, however, consistent with the results of a simple time-dependent box model for the sea-air CO2 exchange with a constant gas transfer rate of 20 mol CO2 m−2 yr−1 [Broecker et al., 1986]: The surface water pCO2 lags behind the atmospheric pCO2 by about 3 years, and the rate of increase in surface water pCO2 should be only a few percent slower than that for the atmospheric increase rate. Greater differences between the increase rates of atmospheric and surface ocean pCO2 would imply changes in ocean circulation, wind intensity, and/or marine ecosystems such as a shift in phosphate-nitrate dynamics observed in the North Pacific [Karl, 1999; Karl et al., 2001].

4.5. Decadal Change of Surface Water CO2 Chemistry

[34] Large effects of temperature on (pCO2)SW may be removed by normalizing it to a constant temperature (the mean SST of each box area), so that the variation in chemical properties in seawater may be identified. For this purpose, we use a constant (∂ ln pCO2/∂ T) of 0.0423 °C−1 [Takahashi et al., 1993] and its integrated form, pCO2 (at mean SST) = pCO2 (observed) × Exp [0.0423 × (mean SST − observed SST)]. The bottom panels of Figures 4, 5 and 6 show the (pCO2)SW values that are normalized to the mean SST in respective box areas. The significance of the temperature-normalized pCO2 values may be illustrated by the following cases. Consider that a parcel of surface seawater (with a constant alkalinity and salinity) is always in equilibrium with increasing atmospheric CO2 at a rate of 15 μatm decade−1. If the water is warmed at a rate of 1°C decade−1, its pCO2 should increase at a rate of about 15 μatm decade−1, same as atmospheric CO2. Hence this water has the same pCO2 as in the atmosphere CO2, inferring that no net transfer of CO2 occurs throughout the decade, and the TCO2 in the water remain unchanged. The (pCO2)SW values that are normalized to the initial temperature remain unchanged through the decade. On the other hand, if the water is kept at a constant temperature, the pCO2 in water increases by taking up atmospheric CO2, and also its TCO2. Conversely, a decrease in the temperature-normalized pCO2 values indicates a reduction in TCO2 in seawater with time. In real oceans, the rate of change of temperature-normalized pCO2 depends not only on the sea-air CO2 flux, but also on many other processes that govern the carbon dynamics in the mixed layer. These processes include seasonal and interannual changes in SST, alkalinity, ecosystem structure (which affects the Redfield ratios and carbon export from the mixed layer), nitrification of atmospheric nitrogen, depths and rate of upwelling waters (which affect preformed concentrations of CO2 and nutrients) and river water flux.

[35] In Table 2, the decadal mean rate of change of the temperature-normalized pCO2 in each box area is listed, and its geographical distribution is shown in Figure 8. The rates in open ocean areas are positive indicating that TCO2 is increasing, and show some geographical variability. In the western North Pacific areas north of the Subtropical Front (STF), the observed rates of increase (3.5 to 11.4 μatm decade−1) appear to be smaller than the 1970–2003 mean atmospheric CO2 increase rate of about 15 μatm decade−1. Increasing flux of nutrients from the East China Sea might cause an increase in net community production, and hence a reduction in TCO2. In contrast, the rates tend to be greater than the atmospheric rate in the northeastern areas. This may be related to change in the rate and source depth for upwelling waters. The negative rates of change observed in the Bering Sea and outside the Okhotsk Sea have been discussed previously. A negative rate is also observed in a box area (35 ± 5°N) along the west coast of North America. The upper layer dynamics and primary production in this area are affected significantly by changes in wind regimes and ocean current structure associated with El Niño events [Goes et al., 2004]. Because of the complexity in the area, a simple answer for the negative trend cannot be offered on the basis of the available data.

Figure 8.

Mean decadal rate of change of temperature-normalized pCO2 (at the mean SST in each box) in the North Pacific surface water. By the temperature normalization, the effects of changes in SST are removed. Hence these rates of change reflect primarily changes in ocean water chemistry. The numbers in bold letters indicate the rate of increase in μatm decade−1, the light italic numbers indicate the rate of decrease in μatm decade−1, and the values in the parentheses indicate the uncertainty. The gray curves in the western Pacific show the approximate locations of the Polar Front (PF) [Belkin et al., 2002 and Subtropical Front (STF).

Table 2. Mean Decadal Rate of Change for TCO2 in Mixed Layer Estimated on the Basis of the Rate of Change of pCO2, Normalized to the Mean SST in Each Box Areaa
Latitudes, °NLongitudes, °E or °WMean SST 1997–2003, °CNumber Mon. Obs.(pCO2)T Change, μatm decade−1Mean Revelle FactorMean TCO2, μmol kg−1Mean (pCO2)sw, μatmMean Rate of TCO2 Change, μmol kg−1decade−1
  • a

    Number Mon. Obs is the total number of months during which observations were made in the 1997–2004 period; (pCO2)T indicates the mixed layer pCO2 normalized to the mean SST in each box area. Revelle factor, mean TCO2, and mean pCO2 in each box are estimated from available data. The decadal mean rate of change for the TCO2 in mixed layer is computed using the information in each box and is listed in the last column. Note that the rate of pCO2 change in box 35°N and 125°W in Table 1 changed its sign to negative in Table 2 as a result of temperature normalization. Hence the number of positive rates is reduced to 27 and that for negative rates is increased to 5 in Table 2, while the total number of box areas remains same at 32.

  • b

    Weighted average is given according to uncertainties: Σ(Xi/σi2)/Σ (1/σi2), where Xi are observed values and σi are uncertainties.

  • c

    This is area weighted mean.

15 ± 5135E ± 528.63 ± 0.232611.5 ± 1.38.819023487.2 ± 0.8
15 ± 5145E ± 528.66 ± 0.18189.0 ± 2.48.819033525.6 ± 1.5
15 ± 5165W ± 526.74 ± 0.26815.1 ± 1.99.119483579.1 ± 1.2
15 ± 5155W ± 525.73 ± 0.271610.9 ± 2.79.119373546.6 ± 1.6
15 ± 5145W ± 526.35 ± 0.38912.1 ± 2.89.319433557.1 ± 1.7
15 ± 5115W ± 527.16 ± 0.38127.6 ± 6.39.419563644.3 ± 3.6
15 ± 5105W ± 528.12 ± 0.52128.7 ± 5.99.219373754.9 ± 3.3
15 ± 595W ± 528.01 ± 0.441435.2 ± 19.99.8192639217.7 ± 10
25 ± 5135E ± 525.60 ± 0.583314.1 ± 2.08.919253418.9 ± 1.3
25 ± 5145E ± 525.76 ± 0.771215.1 ± 3.49.119413439.3 ± 2.1
25 ± 5165E ± 524.96 ± 0.771311.9 ± 6.29.119613427.5 ± 3.9
25 ± 5165W ± 524.54 ± 0.88711.1 ± 5.19.219723446.9 ± 3.2
25 ± 5155W ± 524.76 ± 0.471411.5 ± 3.09.519803476.9 ± 1.8
25 ± 5125W ± 520.70 ± 0.651718.2 ± 9.210.119693709.6 ± 4.9
35 ± 5155E ± 520.06 ± 0.81395.6 ± 6.910.619963303.2 ± 3.9
35 ± 5165E ± 518.62 ± 0.83308.5 ± 7.110.419903364.8 ± 4.1
35 ± 5175E ± 516.71 ± 0.962011.4 ± 17.210.519943396.4 ± 9.7
35 ± 5145W ± 517.16 ± 0.741423.9 ± 5.610.8198535512.4 ± 2.9
35 ± 5135W ± 517.55 ± 0.37229.8 ± 4.710.919803605.0 ± 2.4
35 ± 5125W ± 515.95 ± 0.30294.6 ± 12.511.119743542.3 ± 6.3
45 ± 5145E ± 510.40 ± 0.804725.0 ± 12.113.1207232612.1 ± 5.9
45 ± 5175E ± 58.78 ± 0.94225.1 ± 8.312.320413672.3 ± 3.8
45 ± 5175W ± 59.34 ± 0.81223.5 ± 7.912.420293561.6 ± 3.6
45 ± 5165W ± 58.11 ± 0.66364.5 ± 7.412.320093462.1 ± 3.5
45 ± 5135W ± 511.78 ± 0.632916.5 ± 4.311.219613438.4 ± 2.2
45 ± 5125W ± 512.28 ± 0.473510.9 ± 6.611.419573455.4 ± 3.3
55 ± 5175E ± 55.81 ± 0.671620.1 ± 5.214.020633777.9 ± 2.0
55 ± 5175W ± 55.76 ± 0.542019.3 ± 13.114.520763647.6 ± 5.2
55 ± 5165W ± 55.71 ± 0.49228.3 ± 15.513.520343653.4 ± 6.4
55 ± 5155W ± 57.34 ± 0.702011.3 ± 10.814.020483554.6 ± 4.4
55 ± 5145W ± 57.91 ± 0.682322.1 ± 3.313.620353639.1 ± 1.4
55 ± 5135W ± 510.64 ± 0.771715.9 ± 7.612.619913497.2 ± 3.5
Totals(pCO2)T Change, μatm decade−1Mean Revelle FactorMean TCO2, μmol kg−1Mean (pCO2)sw, μatmMean Rate of TCO2 Change, μmol kg−1decade−1
Mean (Open North Pacific Ocean)12.6b10.6c1975.5c353.5c7.4b
Standard Deviation±6.6±1.8±47.5±14.0±3.3
No. of Boxes2732323227
Mean (Bering and Okhotsk Seas)−18.1b   −7.5b
Standard Deviation±8.6   ±3.9

[36] The mean rate of increase for 27 box areas in the open Pacific (Table 2) is 12.6 ± 6.6 μatm decade−1. If this is assumed to be caused by change in TCO2 without temporal changes in alkalinity and salinity in each box area, these pCO2 rates may be converted to TCO2 changes using the Revelle factor, which is estimated for each box using the available pCO2 and TCO2 data (see Table 2). The mean rate of TCO2 increase in the surface layer of the North Pacific thus estimated is 7.4 ± 3.3 μmol kg−1 decade−1. This is somewhat lower but is indistinguishable from the area-weighted mean rate of increase of 8.1 ± 1.0 μmol kg−1 decade−1 for the North Pacific surface waters that are assumed to be in equilibrium with the current atmospheric CO2 increase rate of 15 μatm decade−1 and have no decadal change in the alkalinity.

4.6. Comparison of TCO2 Changes With Other Studies

[37] Our estimate of TCO2 increase may be compared with the rate estimated based on direct measurements of TCO2 in the northeastern Pacific. On the basis of direct TCO2 measurements made during seven expeditions spanning from 1973 to 2000, Feely et al. [2003] and Sabine et al. [2004b] estimated a mean rate of TCO2 increase of 13 ± 2 μmol kg−1 decade−1 for surface mixed layer waters over 30°N–50°N and 140°W–180°. Analyzing the depth profile data, Sabine et al. [2004b] used a multiple linear regression analysis for the station data, and reported an average TCO2 increase rate of 7.9 ± 4 μmol kg−1 decade−1 for the North Pacific waters below the mixed layer down to 1250 m. This difference is noteworthy because the surface water data yielded a rate much greater than 8 μmol kg−1 decade−1 expected from equilibration with atmospheric CO2 increasing at a mean rate of 15 μatm decade−1.

[38] Using the data listed in Table 2, we computed a mean rate of increase in TCO2 of 6.0 ± 4.5 μmol kg−1 decade−1 for the same area studied by Feely et al. [2003] and Sabine et al. [2004b]. While this is about one half of their surface water value, this is consistent with their rate based on the depth profile data. The strength of Feely-Sabine approach is that they measured TCO2 directly. On the other hand, their measurements were made during seven expeditions that took place over several years during nonwinter months, and the effects of seasonal changes in mixing and biological utilization were corrected using a multiple linear regression with salinity and nutrient data and a fixed P:N:C stoichiometry ratio for net production. Their surface TCO2 increase may be affected by the lack of full-seasonal cycle observations. In contrast, the pCO2 data used in this study have more extensive seasonal and area coverage than the TCO2 data. However, the pCO2 values are converted to TCO2 assuming that the alkalinity (normalized to a constant salinity) is unchanged through time. To account for the discrepancy between their TCO2 and our pCO2 data, it would require an alkalinity increase of about 7 μeq kg−1 decade−1, which could be caused by a reduction in the production rate of CaCO3 shells. Observations are equivocal presently. The surface water salinity, which commonly correlates tightly with alkalinity, has remained nearly constant over the past 30 years at a mean of 32.6 ± 0.2 in this area, suggesting a constancy of the alkalinity. On the other hand, Wong et al. [2002] observed over the entire subarctic and some subtropical areas an increasing trend for the alkalinity (normalized to a constant salinity of 35) for the 2-year period 1995–1997. However, since their alkalinity data exhibit a varying degree of scatter as large as 75 μeq kg−1, the temporal trend has not been firmly established.

[39] The agreement between the Feely-Sabine 1250-m mean rate with our pCO2-based rate for TCO2 increase could be fortuitous since the subsurface waters were formed elsewhere outside the area where the samples were collected, or it may be attributable to the fact that the subsurface TCO2 data represent values at the time of winter convective mixing, during which large seasonal variability of TCO2 and nutrient concentrations in the mixed layer is filtered out.

5. Systematics of Winter Surface Water pCO2 Variation

[40] Increased vertical mixing caused by the winter cooling of surface waters brings up respired CO2 and nutrients from deep to surface regimes, and thus sets up the initial condition for phytoplankton growth during the following spring through summer. Because of low sun angles, short daylight hours and low temperatures, biological activities are minimum during the winter months. Therefore the distribution of pCO2 in SST minimum waters is governed primarily by physical processes, and represents an annual initial condition for the ocean. In this section, we present systematic relationships of (pCO2)SW with the SST and density of mixed layer waters observed during winter time when SST is minimum. These relationships may be used for spatial interpolation of pCO2 and TCO2 data. In addition, the wintertime pCO2-TCO2 relationships for the time of minimum biological activities may be useful for testing the validity of physical process formulations used in biogeochemical carbon cycle models.

5.1. The pCO2 and SST Relationships in Winter Surface Waters

[41] The wintertime pCO2, and SST values in each box area are obtained from the combined 1997–2004 observations, and hence they represent an average for the 7-year period. In order to remove the effects of variation in salinity and temperature, the observed pCO2 values are normalized to a constant SST and salinity of 14.3°C and 34.0 (an area weighted mean for box areas located north of 10°N) using the relationship

equation image

where ΔSal and ΔT are respectively the difference between the observed and the normalization salinity and temperature values. The effect of salinity on pCO2 is calculated using (∂ ln pCO2/∂ ln Sal) = 0.94 (or (pCO2)sal=34 = (pCO2)obs [34.0/(Sal)obs]0.94) [Takahashi et al., 1993]. This salinity effect includes the effects of salinity on the CO2 solubility and dissociation constants in seawater as well as changes in alkalinity, TCO2 and borate concentrations that are proportional to salinity. The observed pCO2 values are also corrected to a constant temperature using the effect of temperature on pCO2, (∂ ln pCO2/∂ T) = 0.0423 °C−1. It should be pointed out that the temperature values used were measured concurrently with (pCO2)SW using a calibrated sensor, whereas the salinity values listed in Table 3 were selected for the minimum temperature month identified from the 45-year climatological mean values [Young et al., 1995]. Hence the salinity values do not correspond exactly in space and time with the (pCO2)SW data. The climatological salinity values are used because the salinity values measured concurrently with (pCO2)SW are subject to errors caused by drifts and lack of frequent calibration of the underway thermosalinograph systems.

Table 3. List of the Mean Monthly Data for SST, Salinity, Sigma-t Density, Revelle Factor, and pCO2 in Surface Waters for the Month of Minimum SSTa
Latitudes, °NLongitudes, °E, °WTmin Mo.(SST)min, °C(pCO2)SW at Tmin, μatmSalinity at TminSigma-t at Tmin, kg m−3Revelle Factor(pCO2)T,S, μatmln(pCO2)T,S
  • a

    Tmin Mo. means the month of minimum SST (1 = January, 2 = February, etc.); (pCO2)T,S indicates the wintertime (pCO2)SW values normalized to the constant SST and salinity values of 14.3°C and 34.0 using equation (1). Two boxes (15°N ± 5, 95°W ± 5; and 55°N ± 5, 165°W ± 5) which are located within coastal upwelling areas are listed at the bottom of the tabl e and are not plotted in Figure 9.

15 ± 5135 ± 5327.2 ± 0.0335.6 ± 0.034.5 ± 0.222.3 ± 0.08.8189.55.244
15 ± 5145 ± 5327.3 ± 0.4346.1 ± 8.734.7 ± 0.222.4 ± 0.18.8193.65.266
15 ± 5195 ± 5225.6 ± 0.1351.2 ± 1.734.4 ± 0.222.7 ± 0.09.1213.85.365
15 ± 5205 ± 5324.3 ± 0.2339.9 ± 2.534.4 ± 0.123.1 ± 0.19.1219.35.391
15 ± 5215 ± 5224.2 ± 0.1337.3 ± 1.734.2 ± 0.223.0 ± 0.09.3219.85.393
15 ± 5235 ± 5424.0 ± 1.4333.4 ± 4.734.5 ± 0.123.3 ± 0.49.6216.45.377
15 ± 5245 ± 5325.9 ± 1.8347.4 ± 1534.0 ± 0.322.3 ± 0.69.3212.25.358
15 ± 5255 ± 5123.5 ± 0.2372.4 ± 8.433.8 ± 0.222.9 ± 0.19.2254.55.539
25 ± 5135 ± 5421.5 ± 1.4312.8 ± 9.234.9 ± 0.124.2 ± 0.49.0223.55.409
25 ± 5145 ± 5120.6 ± 1.1300.7 ± 5.834.9 ± 0.124.5 ± 0.39.1223.05.407
25 ± 5165 ± 5219.2 ± 0.2308.4 ± 3.135.3 ± 0.225.2 ± 0.19.1240.95.485
25 ± 5195 ± 5220.1 ± 0.4317.7 ± 1.535.2 ± 0.124.9 ± 0.19.3238.05.472
25 ± 5205 ± 5220.1 ± 0.5321.2 ± 2.335.1 ± 0.224.8 ± 0.19.4241.65.487
25 ± 5225 ± 51218.4 ± 0.0349.9 ± 0.034.8 ± 0.225.0 ± 0.010.0286.45.657
35 ± 5145 ± 5212.6 ± 5.0328.6 ± 2534.5 ± 0.326.0 ± 0.911.4349.05.855
35 ± 5155 ± 5412.6 ± 2.7317.9 ± 1834.6 ± 0.226.2 ± 0.611.1336.55.819
35 ± 5165 ± 5410.2 ± 1.0346.2 ± 5.634.6 ± 0.226.6 ± 0.211.5405.86.006
35 ± 5175 ± 549.7 ± 1.5339.6 ± 4.634.6 ± 0.226.7 ± 0.311.8406.66.008
35 ± 5185 ± 5410.2 ± 0.5330.6 ± 5.034.6 ± 0.226.6 ± 0.111.1388.55.962
35 ± 5195 ± 5410.6 ± 0.4324.1 ± 4.634.4 ± 0.326.4 ± 0.111.1374.75.926
35 ± 5205 ± 5311.3 ± 1.0313.6 ± 3.434.3 ± 0.426.2 ± 0.211.1352.85.866
35 ± 5215 ± 5412.9 ± 0.5324.4 ± 5.834.2 ± 0.525.8 ± 0.111.0343.15.838
35 ± 5225 ± 5414.8 ± 0.6343.0 ± 3.133.6 ± 0.525.0 ± 0.111.1339.35.827
35 ± 5235 ± 5314.6 ± 1.0339.8 ± 8.133.1 ± 0.224.6 ± 0.211.5343.65.840
45 ± 5145 ± 532.3 ± 2.4382.8 ± 5133.1 ± 0.526.4 ± 0.313.7646.96.472
45 ± 5155 ± 521.7 ± 0.5407.5 ± 2533.4 ± 0.326.7 ± 0.015.2702.26.554
45 ± 5165 ± 522.4 ± 0.0385.2 ± 0.033.5 ± 0.426.7 ± 0.013.7644.56.469
45 ± 5175 ± 533.2 ± 0.4375.8 ± 9.433.4 ± 0.426.6 ± 0.014.1607.76.410
45 ± 5185 ± 533.9 ± 0.3355.0 ± 1133.3 ± 0.526.5 ± 0.013.8558.66.325
45 ± 5195 ± 533.9 ± 1.2348.8 ± 2533.0 ± 0.526.2 ± 0.112.9552.86.315
45 ± 5205 ± 535.6 ± 1.2343.3 ± 5.033.1 ± 0.426.1 ± 0.212.7504.06.223
45 ± 5215 ± 536.3 ± 0.6340.2 ± 3.132.9 ± 0.325.9 ± 0.112.1488.36.191
45 ± 5225 ± 548.2 ± 0.6339.9 ± 5.832.8 ± 0.225.5 ± 0.111.9451.86.113
45 ± 5235 ± 528.8 ± 1.1367.7 ± 3932.3 ± 0.525.0 ± 0.211.9482.46.179
55 ± 5175 ± 532.4 ± 0.2427.7 ± 1433.1 ± 0.126.4 ± 0.015.2717.56.576
55 ± 5185 ± 532.2 ± 0.4412.3 ± 2032.9 ± 0.326.3 ± 0.015.1701.76.553
55 ± 5205 ± 533.1 ± 0.4402.2 ± 2532.5 ± 0.425.9 ± 0.014.3664.46.499
55 ± 5215 ± 533.7 ± 0.9394.4 ± 2532.8 ± 0.126.0 ± 0.113.6632.06.449
55 ± 5225 ± 536.3 ± 0.6367.1 ± 1232.2 ± 0.325.3 ± 0.113.0533.86.280
 
Coastal Areas Along North America
15 ± 5265 ± 51223.1 ± 0.5445.6 ± 2133.9 ± 0.823.1 ± 0.110.1308.25.731
55 ± 5195 ± 533.1 ± 0.2455.6 ± 1631.4 ± 0.725.0 ± 0.014.5766.96.642

[42] The wintertime temperature-salinity-normalized pCO2 values (hereafter expressed as (pCO2)T, S) are listed in the last two columns in Table 3, and reflect primarily changes in TCO2 but also include contributions from alkalinity deviated from the linear dependence of alkalinity on salinity. Figure 9a shows that the natural logarithm of wintertime (pCO2)T,S increases linearly with increasing winter water density. This reflects that deeper and denser waters that contain greater CO2 concentrations are mixed upward into higher latitude mixed layer waters. Greater scatters observed for the winter density plot (Figure 9a) may be due to the fact that the density values are computed using the observed SST and climatological mean salinity data. It also shows that the trend for colder waters north of 35°N is displaced above that for warmer waters south of 35°N, and that, for a given density, the colder subarctic waters contain greater TCO2 than the subtropical waters. The double trends are attributed to the fact that the colder and lower salinity subarctic waters contain distinctly greater CO2 (and alkalinity) than the warmer and higher salinity subtropical waters [Takahashi et al., 1980, 1993], which have a same density as the corresponding cold subarctic waters. In the North Pacific, the 18°C isotherm runs along 35°N nearly all the way across the North Pacific during winter, while the salinity is high in the west (∼34, the Kuroshio Current) and lower (∼33, the Eastern Subarctic Pacific waters). The zonal gradient of salinity is represented by the density (x axis) displacement of the two trend lines along the constant TCO2 of 2050 μmol kg−1.

Figure 9.

Wintertime pCO2 in surface waters versus (a) the sigma-t density of wintertime mixed layer waters and (b) seasonal minimum SST in each box area. The natural logarithm of box mean wintertime surface water pCO2 values that are normalized to constant SST and salinity values of 14.3°C and 34.0 is plotted. Note that the two trends seen in the winter density plot in Figure 9a collapse into a single trend in Figure 9b. Open circles indicate subarctic waters located north of 35°N, and pluses subtropical waters located south of 35°N. Data from the two boxes (15°N, 95°W; and 55°N, 165°W) are excluded from this plot since the values are strongly influenced by coastal upwelling. The TCO2 scale shown on the right vertical axis is computed using a constant alkalinity of 2250 μeq kg−1 and hence is only approximate.

[43] It is interesting to see that these two trends merge into a single trend with a much less scatter when ln(pCO2)T,S is plotted against the minimum SST (Figure 9b). This suggests that the surface water CO2 chemistry is regulated primarily by temperature, which affects not only the air-sea CO2 flux via the effects on pCO2 and CO2 solubility, but also the rates of biological processes, the depth of mixed layer and the vertical mixing rate. The data yield a R2 = 0.97 using a linear regression (not shown), and R2 = 0.98 using a quadratic regression. The quadratic fit gives a σ of ±5.7% for (pCO2)T,S (= 0.057 for ln (pCO2)T,S) and a standard error of ±0.9%. For the analysis to follow, we use the quadratic fit, in which (SST)min is taken as independent variable,

equation image

where (SST)min is the winter minimum SST in °C.

5.2. Relationship Between pCO2 and TCO2

[44] In the previous section, we have found that the (pCO2)T,S values decrease approximately as a logarithmic function of the winter SST. Since the winter-time TCO2 is an important quantity especially for the areas of water mass formation, we attempt to find an empirical relationship between (pCO2)T,S and TCO2 in this section.

[45] The pCO2 at a constant temperature, salinity and alkalinity (TA) is related to TCO2 by the Revelle factor, γ = (∂ ln pCO2/∂ ln TCO2) T, S, TA. If γ, T salinity and TA are constant, an integration of this equation yields

equation image

where the superscript ° indicates a constant value at reference state. TCO2 values depend on salinity, but are independent of temperature. For the reference values for TCO2 and pCO2, the area-weighted mean of winter values over the North Pacific (north of 10°N) may be used. Since γ varies from about 8 in tropical waters and 14 in subarctic waters high in TCO2, and since TA deviates from the proportionality with salinity, γ° is a basin mean value including the effects of change in the alkalinity/salinity ratio. Accordingly, since (pCO2)T,S is given by equation (2) as a function of (SST)min, the winter-time (TCO2)S may be computed using equation (3), if the three reference parameter values are known. Using about 400 wintertime TCO2 measurements available over the North Pacific, an empirical equation that relates (pCO2)SW with TCO2 can be obtained.

[46] Three to 22 TCO2 measurements are available for the three winter months (January–March) in each of 18 box areas out of the 41 boxes, where we have pCO2 data. The TCO2 values are first normalized to a constant salinity of 34.0 (the area weighted mean for the box areas located north of 10°N), and a box mean is computed. An area-weighted basin mean (pCO2)T,S of 386.2 μatm is computed using mean box values normalized to SST = 14.3°C and salinity = 34.0 as explained earlier, and is accepted for (pCO2T,S. The relevant values are listed in Table 4. Substitution of equation (2) into equation (3) gives a form which relates (TCO2)S with (SST)min. The values for (TCO2)S° and γ° are adjusted so that the computed (TCO2)S values are in agreement with the observed values. As shown in Figure 10, a 1:1 correspondence between the observed and calculated TCO2 values is obtained for (TCO2)S° = 2025 μmol kg−1 and γ° = 8.33. Substituting these values and equation (2) into equation (3), we obtain (TCO2)S = 34.0 as a function of (SST)min,

equation image

This equation gives winter (TCO2) values with a σ of ±21 μmol kg−1 and a standard error (σ/N1/2) of ±5 μmol kg−1 at (SST)min from 2.0 to 27.5°C. The data points that have large scatters in Figure 10 are from the areas, where spatial variability is large within 10° × 10° box areas (such as the Kuroshio-Oyashio confluence).

Figure 10.

Comparison of observed (TCO2)S = 34.0 in 18 box areas in the North Pacific with those calculated using equation (4). The best fit between these quantities is achieved with a reference Revelle factor (γ°) = 8.33, (TCO2S=34.0 = 2025 μmol kg−1 and (pCO2S=34.0,T=14.3 = 386.2 μatm. The error bars for the observed TCO2 values include the accuracy of measurements (±2 μmol kg−1) and the time-space variability, whereas the error bars for the calculated pCO2 values indicate ±0.9% standard error for the ln(pCO2)-SST regression, equation (2). Since the calculated TCO2 is precise but not accurate, this is used as a dependent variable in the linear regression.

Table 4. Measured and Calculated Mean Monthly TCO2 Concentrations in Surface Waters During the Month of Minimum SST in 10° × 10° Box Areasa
Latitudes, °NLongitudes, °E, °WTmin Mo.(SST)min, °C(pCO2)sw at Tmin μatm(TCO2)ob, S = 34.0, μmol kg−1SalinitySigma-t(TCO2)calc, Sal = 34.0, μmol kg−1
  • a

    Tmin Mo. is the month of minimum SST indicated by 1 = January, 2 = February, 3 = March. (TCO2)ob is an observed value normalized to a constant salinity of 34.0. (TCO2)calc is computed using equation (4).

15 ± 5205 ± 5324.26 ± 0.2339.9 ± 2.51915 ± 0.034.40 ± 0.1423.11 ± 0.061891.6
15 ± 5215 ± 5224.19 ± 0.07337.3 ± 1.71917 ± 0.034.23 ± 0.1523.00 ± 0.021892.1
25 ± 5145 ± 5120.59 ± 1.07300.7 ± 5.81928 ± 0.034.92 ± 0.1024.54 ± 0.291918.9
25 ± 5165 ± 5219.15 ± 0.19308.4 ± 3.11894 ± 4.735.25 ± 0.1825.17 ± 0.051931.4
25 ± 5195 ± 5220.14 ± 0.43317.7 ± 1.51942 ± 0.635.20 ± 0.1424.88 ± 0.111922.7
25 ± 5205 ± 5220.12 ± 0.53321.2 ± 2.31937 ± 2035.10 ± 0.1524.81 ± 0.141922.8
35 ± 5145 ± 5212.58 ± 5.04328.6 ± 251978 ± 4634.49 ± 0.2725.96 ± 0.942002.1
35 ± 5205 ± 5311.33 ± 0.96313.6 ± 3.41999 ± 1534.32 ± 0.4126.18 ± 0.192018.2
45 ± 5145 ± 532.25 ± 2.43382.8 ± 512180 ± 0.033.11 ± 0.4526.40 ± 0.252163.4
45 ± 5155 ± 521.66 ± 0.48407.5 ± 252182 ± 2133.41 ± 0.3126.72 ± 0.032174.6
45 ± 5175 ± 533.20 ± 0.37375.8 ± 9.42134 ± 1033.37 ± 0.4326.57 ± 0.032145.7
45 ± 5185 ± 533.88 ± 0.32355.0 ± 112149 ± 1.933.31 ± 0.4726.45 ± 0.032133.4
45 ± 5195 ± 533.86 ± 1.15348.8 ± 252134 ± 1332.96 ± 0.4926.17 ± 0.112133.8
45 ± 5205 ± 535.64 ± 1.15343.3 ± 52067 ± 2833.09 ± 0.3526.08 ± 0.162103.1
45 ± 5235 ± 528.81 ± 1.09367.7 ± 392076 ± 3.932.26 ± 0.4625.00 ± 0.172053.4
55 ± 5205 ± 533.08 ± 0.41402.2 ± 252142 ± 1432.46 ± 0.4025.85 ± 0.042147.9
55 ± 5215 ± 533.69 ± 0.87394.4 ± 252120 ± 4.532.75 ± 0.1126.02 ± 0.092136.8
55 ± 5225 ± 536.30 ± 0.61367.1 ± 122107 ± 0.032.24 ± 0.3425.33 ± 0.082092.2

[47] Some explanations are necessary for the values obtained for the reference Revelle factor (γ°) and (TCO2)° values in order to achieve the 1:1 correspondence between the observed and calculated TCO2 values. The slope of the linear correlation line shown in Figure 10 depends primarily on γ°, and the intercept on (TCO2)°. The γ° value of 8.33 obtained for the best fit is significantly smaller than the area-weighted mean of 10.6 for the γ values computed for each of the 37 box areas in the North Pacific surface water assuming a constancy of alkalinity and salinity in each box (Table 3). Hence these γ values do not reflect changes in these properties from one box to another. On the other hand, since the γ° of 8.33 is obtained by fitting the box data from subtropical and subarctic regimes, it includes the effects of geographical variation of the alkalinity and salinity. Over the entire North Pacific, the alkalinity/salinity ratio is not constant, and instead, it increases in the subarctic surface waters due to greater vertical mixing of high alkalinity subsurface waters. Carbonate chemistry equilibria calculations show that an increase of alkalinity from 2200 μeq kg−1 to 2400 μeq kg−1 should decrease γ from 11.8 to 8.5 while TCO2, pCO2 and salinity are kept constant at 2000 μmol kg−1, 330 μatm and 35.0 respectively [Takahashi et al., 1980]. This accounts for the low mean γ value obtained by curve fitting the North Pacific data. The best fit value of 2025 μmole kg−1 for (TCO2S=34.0 is consistent with the area-weighted mean of 2030 μmol kg−1 (with a standard error of ±25 μmol kg−1) for the observed (TCO2)S=34.0 in the 18 box areas.

6. Summary and Conclusion

[48] Approximately 327,000 surface water pCO2 observations over 3 decades from 1970 to 2004 have been used to determine for mean decadal variability in thirty-two 10° × 10° box areas over the North Pacific Ocean north of 10°N. Our findings are summarized below.

[49] 1. Of the 32 box areas studied, 28 open ocean box areas exhibit increasing decadal trends for (pCO2)SW, whereas four box areas located in the southern Bering Sea and outside the Okhotsk Sea exhibit decreasing trends.

[50] 2. The surface ocean pCO2 in the open North Pacific has increased with an area-weighted mean rate of 12.0 ± 4.7 μatm decade−1. This is somewhat smaller than but is statistically indistinguishable from the mean atmospheric CO2 increase rate of about 15 μatm decade−1. Although this suggests that surface ocean waters of the North Pacific as a whole follow the atmospheric CO2 increase, geographical variation of the observed rates indicates that the rates are also controlled by local oceanographic processes including vertical and lateral circulation and biological activities.

[51] 3. The coastal areas along the North American continent have a wide range of rates from 2 to 25 μatm decade−1. This may be caused by local differences in lateral and vertical transport rates of ocean waters.

[52] 4. Lower rates of 5 to 9 μatm decade−1 found in the western temperate Pacific may be due to the influx of high-alkalinity waters from the East China Sea which receives alkalinity from two major rivers draining the Chinese mainland [Chen and Wang, 1999; Chen et al., 2004].

[53] 5. Surface water pCO2 appears to decrease with time in the southern Bering Sea and outside the Okhotsk Sea at a mean rate of −11.1 ± 5.7 μatm decade−1. This may be attributed to the combined effects of an increase in primary production [Gregg et al., 2003], changes in nutrient supply by earlier ice melt, river and land runoff and anthropogenic fallouts, as well as changes in lateral and vertical circulation of ocean waters [Wirts and Johnson, 2005]. The observed decrease in pCO2 corresponds to an increase of pH by 0.02 decade−1 or a decrease of H+ concentration by 14% for the past 3 decades. This is in contrast to the acidification of −0.1 pH unit (or a 25% increase in H+) in open ocean that is estimated for changes in surface ocean waters since the industrial revolution of the 1800s.

[54] 6. Assuming that the increase in (pCO2)SW is due only to an increase in TCO2, we estimate that the rate of TCO2 increase in subarctic Pacific surface waters for the area 30°N–50°N and 140°W–180° to be 6 ± 4.5 μmol kg−1 decade−1. Our estimate is about one half the mean rate of increase of 13 ± 2 μmol kg−1 decade−1 estimated for the same area by Feely et al. [2003] and Sabine et al. [2004b] based upon TCO2 measured in mixed layer waters during 7 non–winter month expeditions spanning from 1973 to 2000. This discrepancy may be accounted for by an increase in alkalinity of 7 μeq kg−1. On the other hand, our value is in agreement with a mean rate of 7.9 ± 4 μmol kg−1 decade−1 estimated by Sabine et al. [2004b] for deep waters down to 1250 m based also upon TCO2 measurements. This could mean that the water mass formation processes occurring in winter homogenized and filtered seasonal changes.

[55] 7. Wintertime surface water pCO2 is regulated primarily by physical processes, since biological activities are minimum in winter. The pCO2 values at seasonal SST minimum in each box area is correlated with the winter SST with a standard error of ±0.9% pCO2 (see equation (2)).

[56] 8. Using about 400 TCO2 measurements made in winter, the observed TCO2 values are compared with TCO2 values calculated using the wintertime pCO2-SST relationship. An empirical equation which gives winter TCO2 values (normalized to 34.0 salinity) as a function of winter SST is presented in equation (4). This equation yields the TCO2 values which are consistent with the observed values with a standard error of ±5 μmol kg−1.

Acknowledgments

[57] This work has been supported by the grants to Lamont-Doherty Earth Observatory from NOAA (DC NOAA NA030AR4310 and NA16GP2001) and NASA (NAG 5-11357). We acknowledge C. S. Wong of IOS, Canada, H. Y. Inoue and M. Ishii of JMC, Japan, and Y. Nojiri, NIES, Japan, for their contributions to the data set used in this study. We appreciated constructive suggestions offered by two anonymous reviewers. This is a LDEO contribution 6859.

Ancillary