Report

Decadal Variation of the Surface Water PCO2 in the Western and Central Equatorial Pacific

See allHide authors and affiliations

Science  31 Oct 2003:
Vol. 302, Issue 5646, pp. 852-856
DOI: 10.1126/science.1088570

Abstract

The equatorial Pacific Ocean is one of the most important yet highly variable oceanic source areas for atmospheric carbon dioxide (CO2). Here, we used the partial pressure of CO2 (PCO2), measured in surface waters from 1979 through early 2001, to examine the effect on the equatorial Pacific CO2 chemistry of the Pacific Decadal Oscillation phase shift, which occurred around 1988 to 1992. During the decade before the shift, the surface water PCO2 (corrected for temperature changes and atmospheric CO2 uptake) in the central and western equatorial Pacific decreased at a mean rate of about –20 μatm per decade, whereas after the shift, it increased at about +15 μatm per decade. These changes altered the CO2 sink and source flux of the equatorial Pacific significantly.

The equatorial Pacific Ocean is known to undergo significant changes on interannual time scales (e.g., El Niño and La Niña) and on decadal time scales. This area is a major source of CO2 to the atmosphere during non–El Niño periods (15) but near neutral during strong El Niño periods (25). Over decadal time scales, the North Pacific Ocean has undergone major physical and biological changes commonly called the Pacific Decadal Oscillation (PDO) (69). The most recent and major shifts of 1977 and 1988 to 1992 have been documented on the basis of extensive physical and biological information (68). Although the causes and effects of PDO have been investigated in recent years, its effects on CO2 chemistry have not yet been identified. Here, we describe the changes in CO2 chemistry in the Pacific equatorial waters associated with the 1988 to 1992 PDO phase shift (68) using the partial pressure of CO2 (PCO2) in seawater, which is a measure of the escaping tendency of CO2 from seawater to the overlying atmosphere.

Because of the heightened interest in the El Niño–Southern Oscillation (ENSO) events and the carbon cycle, PCO2 in surface ocean waters has been measured more frequently since 1979 over the equatorial Pacific Ocean. In this study, we assembled a database containing about 100,000 PCO2 observations made in surface waters of the equatorial belt, 5°N to 5°S, between April 1979 and March 2001 (10). We classified these into three groups: (i) the Niño 3.4 area (lat 5°S to 5°N, long 170°W to 120°W) during non–El Niño periods (which include La Niña periods); (ii) the Niño 3.4 area during El Niño periods; and (iii) the western warm pool area (lat 5°N to 5°S, long 175°E to 135°E). We identified six El Niño periods in the Niño 3.4 area on the basis of sea surface temperature (SST) anomalies (11). The remaining data were considered as occurring during non–El Niño periods.

In the Niño 3.4 area, observations during non–El Niño periods were made more or less evenly throughout the year along a total of 11 meridional transects across the equator during the pre-1990 period, and 14 transects during the post-1990 period. In this area, no measurements were made during the 1982 to 1983 El Niño event. The subsequent El Niño events were sampled by five transects during the pre-1990 period and nine transects during the post-1990 period. However, only the early to middle phases were sampled during the three events between 1992 and 1995. In the western warm pool, measurements for the non–El Niño periods were made along 8 and 13 transects during the pre- and post-1990 periods, respectively, and those for the El Niño periods were made along 9 and 10 transects during the pre- and post-1990 periods, respectively. Observations before 1990 were made mostly during the first quarter of the year, whereas observations were made more or less evenly throughout the year after 1990. Here, the seasonal variability is very small, so the first quarter bias in the first decade should not bias our analysis. As a whole, we believe that the sampling density, although not ideal, is adequate for our analysis (12).

Figure 1A shows the time trend for PCO2 during non–El Niño periods in the Niño 3.4 area. We have plotted all of the observations but used the monthly mean values for the time-trend analysis. Because of the high PCO2 in the upwelling subsurface waters, there is a large seasonal variability in PCO2, ranging from 330 to 500 μatm. To estimate the time trends before and after the PDO phase shift occurred around 1990, the data were split into the pre– and post–phase shift groups, and each group was linearly regressed to obtain a mean rate of change. Because our record is only two decades long and the seasonal variability is large, the time trend is affected by choices of the onset for phase shifts. Accordingly, accepting the 1988 to 1992 phase shift reported by others (68), we first computed the mean rate of PCO2 change using each of the 6 years from 1987 to 1992 for the phase shift. The mean rate and standard deviation for these six rates are computed and reported as a mean time trend for the pre– and post–phase shift. We obtained a mean rate of –0.9 ± 1.3 μatm year–1, which includes no change, for the pre–phase shift decade, and +1.8 ± 0.7 μatm year–1 for the post–phase shift decade. A Student's t test gives 92% probability that the preshift mean rate will be different from the postshift rate and indicates that the PCO2 increases at a significantly greater rate after the phase shift.

Fig. 1.

(A) PCO2 at SST in surface seawater during non–El Ninño periods in the Ninño 3.4 area, lat 5°N to 5°S, long 120°W to 170°W. The measured values are indicated with dots and the monthly mean values with open squares. The monthly values are used for estimating the decadal mean by linear regression: –0.9 ± 1.3 μatm year–1 for the pre–phase shift decade and +1.8 ± 0.7 μatm year–1 for the post–phase shift decade. The uncertainties are expressed in terms of 1 SD for six mean rates of change evaluated by using each of the phase-shift years of 1987, 1988, 1989, 1990, 1991, and 1992. (B) The SST during non–El Ninño periods in the Ninño 3.4 area. The mean rate of change is –0.03° ± 0.04°C year–1 for the preshift decade and –0.07° ± 0.04°C year–1 for the postshift decade. (C) Surface water PCO2 values normalized to a constant temperature of 28°C in the same area. The solid lines indicate the decadal mean rate of change, which is fitted to the mean monthly values for 28°C PCO2: –0.3 ± 1.7 μatm year–1 for the preshift decade and +3.2 ± 0.6 μatm year–1 for the postshift decade. The dashed line indicates the rate of +2.8 μatm year–1 (Table 1), which is determined from the rates of change for PCO2 at SST (A) and SST (B) for the postshift decade. The rates for the preshift decade, which are determined by the two different methods, are indistinguishable.

Figure 1B shows the non–El Niño period SST measured concurrently with seawater PCO2. As described earlier, we obtain a mean rate of SST change for six phase-shift years: –0.03° ± 0.04°C year–1 for the pre–phase shift decade and –0.07° ± 0.04°C year–1 for the post–phase shift decade. This indicates that the surface water cooled after 1990 at a significant rate. These changes are consistent with the more complete time series SST data of Reynolds et al. (12, 13).

Our analysis above suggests that, whereas PCO2 and SST remained nearly unchanged during the preshift decade, PCO2 increased as SST deceased during the postshift decade. A nonparametric Kendall's τ (14, 15) is used to test further the validity of the PCO2-SST correlation after the phase shift. It indicates that the PCO2 is anticorrelated significantly with SST with τ = –0.59 (i.e., 79% of pairs are concordant) with a very high (>99%) confidence level (Z statistic = –4.9) (16). Because the subsurface waters upwelling in the eastern Pacific are cooler with higher PCO2 than the surface water, these results may be interpreted as a result of increased upwelling rates during the post–phase shift period of 1990 to 2001.

The observed rate of change in seawater PCO2 includes the effects of changes in the SST as well as seawater chemistry. The effect of temperature can be estimated with the mean rate of change of SST that we obtained and the temperature effect on PCO2 for a parcel of isochemical seawater (∂ ln PCO2/∂T) of 0.0423°C–1, where T is temperature (17). This yields –0.4 ± 0.5 μatm year–1 for the preshift, and –1.0 ± 0.6 μatm year–1 for the postshift period (Table 1). Applying this as a correction to the PCO2 changes above, we obtain –0.6 ± 1.4 μatm year–1 for the preshift, and +2.8 ± 0.9 μatm year–1 for the postshift decade. These changes are attributed to changes in seawater chemistry, which include the total concentration of CO2 and the alkalinity (i.e., ionic charge balance) in seawater. The chemical changes may also be estimated by analyzing the individual PCO2 values that are normalized to a constant temperature of 28°C with (∂ ln PCO2/∂T) of 0.0423°C–1 (17). The individual and mean monthly values thus normalized are shown in Fig. 1C. A linear regression for the monthly mean values gives a mean rate of change of –0.3 ± 1.7 μatm year–1 for the preshift decade, and +3.2 ± 0.6 μatm year–1 for the postshift decade (Table 1). These decadal mean rates are consistent with the respective rates estimated with PCO2 at SST and the mean rates of SST changes (Table 1).

Table 1.

Rates of change observed in the atmospheric CO2 concentration and PCO2 in surface waters of the western and central equatorial Pacific Ocean. sw, seawater.

Year periods 1979 to 1990 ± 3 1990 ± 3 to 2001
Non—El Niño periods, Niño 3.4 area in the equatorial Pacific, lat 5°N to 5°S, long 120°W to 170°W
Rate of atmospheric CO2 increaseView inline +1.49 ± 0.01 ppm year-1 +1.49 ± 0.01 ppm year-1
Rate of SST change -0.03° ± 0.04°C year-1 -0.07° ± 0.04°C year-1
SST effect on (PCO2)sw -0.4 ± 0.5 μatm year-1 -1.0 ± 0.6 μatm year-1
Rate of change in (PCO2)sw at SST -0.9 ± 1.3 μatm year-1 +1.8 ± 0.7 μatm year-1
[(PCO2)sw at SST] - [SST effect] -0.6 ± 1.4 μatm year-1 +2.8 ± 0.9 μatm year-1
Rate of change in (PCO2)sw at 28°C -0.3 ± 1.7 μatm year-1 +3.2 ± 0.6 μatm year-1
[(PCO2)sw at SST] - [SST effect] - [atmospheric CO2 rate] -2.1 ± 1.4 μatm year-1 +1.3 ± 0.9 μatm year-1
Number of mean monthly values usedView inline 20, 23, 25, 28, 31, 34 40, 37, 35, 32, 29, 26
Western warm pool waters of the equatorial Pacific, lat 5°N to 5°S, east of long 175°E View inline
Rate of SST change +0.10° ± 0.01°C year-1 +0.02° ± 0.02°C year-1
SST effect on (PCO2)sw +1.5 ± 0.13 μatm year-1 +0.3 ± 0.3 μatm year-1
Rate of change in (PCO2)sw at SST +0.5 ± 0.3 μatm year-1 +3.4 ± 0.4 μatm year-1
[(PCO2)sw at SST] - [SST effect] -1.0 ± 0.3 μatm year-1 +3.1 ± 0.5 μatm year-1
Rate of change in (PCO2)sw at 28°C -1.0 ± 0.3 μatm year-1 +3.0 ± 0.2 μatm year-1
[(PCO2)sw at SST] - [SST effect] - [atmospheric CO2 rate] -2.5 ± 0.3 μatm year-1 +1.6 ± 0.5 μatm year-1
Number of mean monthly values usedView inline 15, 20, 24, 25, 31, 33 47, 42, 38, 37, 31, 29
  • View inline* Mean rate of increase for a period of 1976 to 1995 over 4°N to 4°S, based on (18).

  • View inline The number of mean monthly values used for the computation of the mean decadal rate of change before and after the following transition years in the order of 1987, 1988, 1989, 1990, 1991, and 1992.

  • View inline All values are the mean for the combined data set for the non—El Niño and El Niño periods.

  • A portion of these changes must also reflect the seawater uptake of atmospheric CO2, which has increased at a rate of +1.49 ± 0.01 parts per million (ppm) year–1 over the period of 1976 to 1995 (18). If the source waters for the equatorial area are assumed to be nearly equilibrated with atmospheric CO2, the PCO2 in source waters would have increased at a rate of about +1.5 μatm year–1. If we apply this correction to the rates of change for seawater PCO2, we obtain respectively –2.1 ± 1.4 μatm year–1 for the preshift decade and +1.3 ± 0.9 μatm year–1 for the postshift decade (Table 1). The seawater PCO2, an intrinsic chemical property of seawater, tends to decrease during the preshift decade (a cold PDO phase), whereas it increases during the postshift decade (a warm PDO phase).

    Now we turn to El Niño periods. The peak-to-peak amplitude for PCO2 variation during El Niño is about 125 μatm (Fig. 2) and is about two-thirds of that observed for the non–El Niño periods. The smaller amplitude is due primarily to weaker upwelling of high PCO2 subsurface waters. Because of the lack of the pre-1987 record, a trend change cannot be determined. Instead, a mean rate of increase is estimated to be +1.7 ± 1.4 μatm year–1 for the period 1987 to 1998, which is similar to +1.8 ± 0.7 μatm year–1, which has been estimated for the postshift decade in the non–El Niño periods. This suggests that, during the postshift period, the mean rate of PCO2 change for the El Niño periods is indistinguishable from that for the non–El Niño periods.

    Fig. 2.

    PCO2 at SST in surface ocean water during El Ninño periods in the Ninño 3.4 area. The measured values are indicated with dots, and the monthly mean values with open squares. The mean rate of change for the preshift decade is +1.7 ± 1.4 μatm year–1. The uncertainty is expressed in terms of (sxy/sx2)½), where sxy is the covariance of x and y and sx2 is the sample variance of x.

    We observe that the PCO2 values during the El Niño periods are systematically smaller than those observed during the non–El Niño periods. On the basis of the linear regressions shown in Figs. 1A and 2, we estimate that the seawater PCO2 for the post–phase shift El Niño periods is on average 20 ± 5 μatm smaller than that for the corresponding non–El Niño periods. This difference, which includes the effect of warming of 1.5° ± 0.5°C during the El Niño periods in the Niño 3.4 area, indicates that a sea-air PCO2 difference of about +65 μatm (normalized to the mean atmospheric PCO2 of 347 μatm in 1995) for non–El Niño conditions (1) is reduced by about 25% to about +45 μatm during El Niño events. If the sea-to-air CO2 transfer rates over the area are unchanged, the net sea-air CO2 flux over the Niño 3.4 area would be reduced on average by about 25% during El Niño events. Because wind speeds during El Niño periods are normally slower than those during non–El Niño periods (3), the sea-to-air CO2 flux during non–El Niño periods of 0.22 Pg C year–1 for the Niño 3.4 area (1) would be expected to be reduced by more than 25% during El Niño periods. The effect of El Niño on the Niño 3.4 area in the central equatorial Pacific contributes to about 3% increase in the global ocean uptake of about 2 Pg C year–1.

    The western Pacific warm pool waters are significantly warmer (∼30°C) and less saline (<34.5 salinity) than the eastern waters, and they are nearly depleted of nutrients (5, 19). These waters also have low biological productivity, low chlorophyll concentrations (∼0.1 mg chlorophyll a m–3), low zooplankton concentration (∼400 mg m–2), and a nearly constant SST (30° to 31°C) (19). Figure 3 shows the surface water PCO2 values in the western warm pool during 1982 to 2001. The El Niño values appear to be indistinguishable from the non–El Niño values. Therefore, we applied a linear regression to the combined data set and computed a mean rate of PCO2 change for each period before and after a phase-shift year ranging from 1987 to 1992. The mean rate of change is +0.5 ± 0.3 μatm year–1 for the preshift decade, and +3.4 ± 0.4 μatm year–1 for the postshift decade. These rates are different with a probability of 99.9% on the basis of a t test. Correcting for the effect of changes in SST and atmospheric CO2 (Table 1), we calculated that the PCO2 intrinsic to seawater decreased at a rate of –2.5 ± 0.3 μatm year–1 for the first decade and increased at a rate of +1.6 ± 0.5 μatm year–1 for the second decade.

    Fig. 3.

    PCO2 at SST in surface ocean water observed between 5°N and 5°S in the warm water areas west of 175°E. The black dots indicate the measurements made during non–El Ninño periods, and the gray dots indicate those during El Ninño periods. The mean monthly values are shown with open squares and are used for the regression calculations. The mean rate of change is computed as it was in Fig. 1: +0.5 ± 0.3 μatm year–1 for the preshift decade and +3.4 ± 0.4 μatm year–1 for the postshift decade. The two monthly points (between 400 and 450 μatm) observed early in 1989 are from a narrow westward-flowing current of upwelled waters located just west of 175°E near the equator, and, hence, are excluded from the regressions.

    These rates are statistically indistinguishable from those observed for the Niño 3.4 area during the respective decades (Table 1). This suggests that the PCO2 (intrinsic to the seawater chemistry) in the entire central and western Pacific from 135°E to 120°W decreased at a rate of about –2 μatm year–1 (or –20 μatm per decade) for the preshift decade, and increased by +1.5 μatm year–1 (or +15 μatm per decade) in the postshift decade. In the western warm water area, the sea-air PCO2 difference has been reported to be nearly 0 μatm, on average (1, 3). However, as a result of the seawater PCO2 increase that occurred during the 1990s, the area has become a weak source for atmospheric CO2. Because the net sea-air flux depends on the time-space distribution of PCO2, SST, and wind speed (which affects the gas transfer rate) over the equatorial areas, further work is necessary to estimate the CO2 flux changes. Additionally, it is important to know whether the decadal PCO2 trend might reverse synchronously with the next PDO phase change, or whether it might behave differently. Continued measurements in the future will be necessary to answer these questions.

    Guilderson and Schrag (20) measured the 14C concentration in corals grown between 1957 and 1983 in the Galapagos Archipelago (lat 1°S, long 91°W), which is located about 3000 km east of the Niño 3.4 area. They found that the 14C concentration in corals for the upwelling period (July to September) increased substantially (by about 20 per mil in Δ14C) after 1976 through 1983 (the most recent sample studied). This coincided with a warming of SST by about 1°C. They interpreted this as a result of a decrease after 1976 of the equatorial upwelling of subthermocline waters, which are colder and have lower 14C concentrations. Because the Galapagos area (600 km west from Ecuador coasts) is influenced with waters from the cold Peru Current and the Peru/Ecuador coastal upwelling system, the area may not be strictly comparable with the Niño 3.4 area. Nevertheless, their observations are consistent with the decrease in PCO2 observed during the 1980s in our study. Combined, these observations suggest that the decreasing trend of the deepwater upwelling continued through the 1980s and changed to an increasing trend sometime around 1990.

    Gu and Philander (21) developed a simple idealized atmosphere-ocean model and demonstrated that climate fluctuations of decadal time scales may be caused by the entrainment of subducted surface waters of the subtropical North and South Pacific into the upwelling equatorial waters. McPhaden and Zhang (22) investigated changes in the meridional overturning circulation in the upper layers of the Pacific Ocean using a circulation model in association with hydrographic and wind data obtained over the central and eastern Pacific. They proposed that the general warming that occurred in the tropical Pacific from the 1970s through the decade of the 1990s was a manifestation of the slowing of the meridional circulation of the pycnocline waters (50 to 400 m). A reduction in the rate of equatorial upwelling is a consequence of this basinscale slowdown. Their conclusion is consistent with our observations up to 1990 and with the 14C increase in corals (20). However, our 1990s observations of increasing PCO2 and decreasing SST, which indicate an increasing in upwelling, contradict with McPhaden and Zhang (22). This could be a result of undersampling for PCO2 in the study area, but is not likely by the reason stated in (12). Alternatively, the observed change in PCO2 may be due to changes in processes local to the equatorial belt rather than basin-scale phenomena.

    In support of local origin of the equatorial signals, Schneider et al. (23) analyzed seawater temperature data observed in the upper 400 m of the North Pacific during 1972 through 1994 and concluded that SST anomalies in the tropical Pacific were caused by local forcing and were independent of the arrival of thermal signals from the central North Pacific. Dore et al. (24) reported that seawater PCO2 at the time-series station ALOHA (lat 22.7°N, long 158°W), located in the temperate North Pacific gyre, increased at a mean rate of 2.5 ± 0.3 μatm year–1 from 1989 to 2001. They attribute this increase primarily to a 3% decadal increase of salinity, which is caused by hydrological cycle changes. This rate of PCO2 increase is similar to that observed in our study of equatorial waters, thus suggesting causal connections between them. However, in the Niño 3.4 area, the surface water salinity is nearly constant from 1987 through 2001, and hence the PCO2 increase in the equatorial waters is not caused by a salinity increase and appears to be unrelated to that observed in the northern temperate Pacific. Therefore, whether decadal changes of the equatorial water CO2 chemistry are caused by basinwide or local-scale processes is not clearly understood.

    Our study shows that the seawater PCO2 over the Pacific equatorial zone appears to have changed substantially during the past two decades in coincidence with the PDO phase shift that occurred between 1988 to 1992. After the PDO phase shift, the in situ PCO2 increased at a mean rate of 34 ± 4 μatm per decade in the western equatorial Pacific and 18 ± 7 μatm per decade in the central equatorial Pacific (Table 1). Because the atmospheric PCO2 was increasing at a rate of 15 μatm per decade during this period, the sea-air PCO2 difference increased by 19 μatm (equal to 34–15 μatm) in the western area by the year 2001. Thus, the area changed from a neutral to a CO2 source of about 0.05 Pg C year–1. This corresponds to about 7% of the equatorial Pacific flux of 0.7 Pg C year–1(1) or 2.5% of the net global ocean carbon flux of 2 Pg C year–1. In contrast, in the central area, the in situ PCO2 increased at a similar rate to that in the atmosphere, and the sea-air PCO2 difference, and hence the net sea-to-air CO2 flux, stayed nearly unchanged. This is because the lowering effect on PCO2 of cooling is nearly balanced by the PCO2 increase from changes in the chemistry of seawater, and the effect of atmospheric CO2 increase is uncompensated. Whether the CO2 source intensity from this area would change in the future depends on the competing effects of changes in SST and the chemistry of seawater. Further observations and model studies are needed for understanding the major carbon cycle perturbations described in this paper.

    References and Notes

    View Abstract

    Stay Connected to Science

    Navigate This Article