Decadal variability of the air-sea CO2 fluxes in the equatorial Pacific Ocean



[1] In order to determine the interannual and decadal changes in the air-sea carbon fluxes of the equatorial Pacific, we developed seasonal and interannual relationships between the fugacity of CO2 (fCO2) and sea surface temperature (SST) from shipboard data that were applied to high-resolution temperature fields deduced from satellite data to obtain high-resolution large-scale estimates of the regional fluxes. The data were gathered on board research ships from November 1981 through June 2004 between 95°W and 165°E. The distribution of fCO2sw during five El Niño periods and four La Niña periods were documented. Observations made during the warm boreal winter-spring season and during the cooler boreal summer-fall season of each year enabled us to examine the interannual and seasonal variability of the fCO2sw-SST relationships. A linear fit through all of the data sets yields an inverse correlation between SST and fCO2sw, with both interannual and seasonal differences in slope. On average, the surface water fCO2 in the equatorial region has been increasing at a rate similar to the atmospheric CO2 increase. In addition, there appears to be a slight increase (∼27%) in the outgassing flux of CO2 after the 1997–1998 Pacific Decadal Oscillation (PDO) regime shift. Most of this flux increase is due to increase in wind speeds after the spring of 1998, although increases in fCO2sw after 1998 are also important. These increases are coincident with the recent rebound of the shallow water meridional overturning circulation in the tropical and subtropical Pacific after the regime shift.

1. Introduction

[2] The equatorial ocean plays an important role in the global carbon cycle, as it is the ocean's largest natural source of CO2 to the atmosphere. The annual contribution of CO2 to the atmosphere from the oceanic equatorial belt is estimated between 0.6 and 1.5 Pg C [Tans et al., 1990; Takahashi et al., 1997, 1999, 2002]. This region, particularly the central and eastern equatorial Pacific, exhibits a large amount of spatial and temporal variability in surface water carbon dioxide fugacity (fCO2sw) due to interannual and decadal variability [see Takahashi et al., 2003, and references therein; Gloor et al., 2003; Le Quéré et al., 2000, 2002]. The central and eastern equatorial Pacific is a major source of CO2 to the atmosphere during non–El Niño periods; it is near neutral during strong El Niño periods, and a weak source during weak El Niño periods [Etcheto et al., 1999; Chavez et al., 1999; Cosca et al., 2003; Feely et al., 1995, 1997, 1999, 2002, 2004; Takahashi et al., 2002]. On decadal timescales, the Pacific Ocean has undergone major physical and biological regime shifts commonly referred to as the Pacific Decadal Oscillation (PDO), which has been documented on the basis of extensive physical and biological data [Trenberth and Hoar, 1996; Hare and Mantua, 2000; McPhaden and Zhang, 2002; Chavez et al., 2003; McPhaden and Zhang, 2004]. While the causes and effects of these regime shifts have been investigated in recent years, only one study of its effect on CO2 chemistry in the equatorial Pacific has been conducted [Takahashi et al., 2003] which has shown consistent agreement between decadal changes in SST and fCO2sw. Here we expand on the earlier work of Takahashi et al. [2003] to describe the short- and long-term variations of the CO2 flux in the central and eastern equatorial Pacific over the past 23 years.

[3] The mean circulation of the equatorial Pacific Ocean is characterized by upwelling that brings cold nutrient- and carbon-rich water to the surface along the equator east of about 160°E during non–El Niño periods. The primary source of the upwelled water along the equator is the narrow Equatorial Undercurrent (EUC), which flows eastward across the basin. This mean circulation and its seasonal variations are significantly modulated on interannual and decadal timescales by two prominent modes of natural variability: (1) the El Niño–Southern Oscillation (ENSO) cycle; and (2) the Pacific Decadal Oscillation (PDO). The warm El Niño phase of the ENSO cycle is characterized by a large-scale weakening of the trade winds, decrease in upwelling of CO2 and nutrient-rich subsurface waters and a corresponding warming of the sea surface temperature (SST) in the eastern and central equatorial Pacific [McPhaden et al., 1998]. El Niño events occur roughly once every 2–7 years and typically last about 12–18 months. The opposite phase of the ENSO cycle, called La Niña, is characterized by strong trade winds, cold tropical SSTs, and enhanced upwelling along the equator. El Niño and La Niña are associated with dramatic shifts in the atmospheric pressure difference between eastern and western Pacific (referred to as the Southern Oscillation) that have major impacts on the climate variability worldwide [McPhaden, 1999].

[4] In contrast, the PDO has been characterized either in terms of fluctuations over a broad band of periods between 10 and 70 years [Minobe, 2000], or in terms of abrupt temporal “regime” shifts in climate conditions and ecosystems over large parts of the basin [Mantua et al., 1997]. The most recent of these regime shifts occurred in 1976–1977, 1988–1989, and 1998 [Trenberth and Hoar, 1996; Watanabe and Nitta, 1998; Beamish et al., 1999; Hare and Mantua, 2000; McPhaden and Zhang, 2002; Chavez et al., 2003; Takahashi et al., 2003; McPhaden and Zhang, 2004]. For example, Hare and Mantua [2000] found evidence for the 1989 regime shift from time series of a number of physical, biological and chemical parameters in the North Pacific. In the tropics, the 1976–1977 regime shift was characterized by a slowdown of the shallow meridional overturning circulation and a warming of the sea surface by nearly 1°C in the cold tongue region of the eastern and central equatorial Pacific Ocean [McPhaden and Zhang, 2002]. The most recent shift, which occurred in the 1997–1998 period, was characterized by an enhancement of the meridional transport and a slight decrease in SST [McPhaden and Zhang, 2004]. On the other hand, on the basis of a coupled atmosphere-ocean model, Rodgers et al. [2004] suggested that nonlinearities in ENSO variability can play an important role in determining the structure of tropical Pacific variability on decadal timescales. Thus it is still open to debate whether the decadal modulation of ENSO is a cause of, or effect of, the PDO.

[5] Because of the importance of the ENSO events on the global carbon cycle, fCO2 measurements in surface waters have been obtained since 1968 throughout the equatorial Pacific Ocean. In this study, we have assembled a large database containing approximately 200,000 fCO2sw observations made in the equatorial Pacific surface waters spanning the region from 5°N to 10°S and 165°E to 95°W from 1979 through June 2004 (Figure 1). We classified these data sets into two periods: (1) the El Niño period; and (2) non–El Niño periods (which include La Niña periods), on the basis of sea surface temperature (SST) anomalies in the Niño 3.4 area (5°N to 5°S, 120°W to 170°W) [Trenberth, 1997]. Our flux estimates of CO2 across the air-sea interface are based on measurements of the degree of their supersaturation or undersaturation of CO2 in the surface ocean mixed layer, combined with estimates of the gas transfer velocity, k. The flux of a CO2 across the air-sea boundary is given by

equation image

where s is the solubility coefficient of the gas, fCO2sw is the fugacity of CO2 in surface seawater, and fCO2a is the fugacity of CO2 in the atmosphere above the interface. The fugacity is similar to the partial pressure, but is corrected for the nonideality for the CO2–CO2 and CO2–H2O molecular interactions in the gaseous state. The solubility coefficient for CO2 as a function of temperature and salinity is summarized by Weiss [1974] and Weiss et al. [1982]. The fCO2sw and fCO2a values were corrected using equilibrium water vapor pressures computed at SST and equilibration temperature, respectively. Both fCO2sw and fCO2a are generally measured with a high degree of precision (±0.5 μatm) and accuracy (±2 μatm). Consequently, the major uncertainty in the local (or instantaneous) flux estimate is associated with the estimate of the gas transfer velocity. For this study we chose to use the widely used relationship of Wanninkhof [1992] since it is compatible with most previous observational and modeling results.

Figure 1.

Cruise tracks for the TAO buoys. The data were continuously collected along the latitudinal transects containing the TAO buoys (small circles) as part of the biannual servicing of the TAO array in the equatorial Pacific. These track lines are repeated twice per year. The rectangular box shows the study region for flux estimates given in Table 4.

2. Sampling and Analytical Methods

[6] During the 91 cruises (Figure 1 and Table 1) onboard the research ships underway fCO2sw measurements were made using shipboard pumping systems which continuously pumped seawater from the ship's sea chest to the oceanographic laboratory following the gas equilibration methods that are similar to those described by Feely et al. [1998], Wanninkhof and Thoning [1993], and Takahashi et al. [1993]. All reference standards undergo a precruise and postcruise calibration against standards certified by the World Meteorological Organization (WMO). Typical standard deviations for air values are ±0.1 ppm and for equilibrated headspace they are ±0.3 ppm. The fCO2sw at in situ conditions was calculated using the measured CO2 mole fraction in dry air, atmospheric pressure, sea surface temperature, and salinity, and equilibrator temperature employing the equations outlined by Weiss [1974], Weiss and Price [1980], and Murphy et al. [1998].

Table 1. List of Cruises for This Study
Cruise IDDates of Cruise (mm/yy)LatitudeLongitude
GPASsfc11/7313°N175°E to 180°
GPASsfc12/738°S to 8°N180° to 165°W
GPASsfc05/74–06/7413°S to 11°N125°W
FGGEsfc04/7915°S to 11°N150°W
FGGEsfc05/7912°N to 14°N160°W
FGGEsfc07/7911°S to 11°N150°W to 160°W
FGGEsfc11/79–12/7912°S to 11°N150°W to 160°W
N044sfc03/7915°S to 12°N95°W to 100°W
N044sfc08/79–09/7915°S to 13°N150°W to 160°W
N044sfc11/79–12/7915°S to 14°N150°W to 160°W
FGGEsfc01/80–02/8014°S to 14°N150°W to 160°W
FGGEsfc05/8013°N to 15°S160°W to 150°W
N044sfc03/80–04/8014°S to 14°N150°W to 160°W
N044sfc05/80–06/8015°S to 14°N150°W to 160°W
INOUsfc02/8111°N to 5°N135°E
INOUsfc02/8211°N to 2°N135°E
INOUsfc02/8313°N to 3°N135°E
INOUsfc12/8314°N to 14°S145°E to 155°E
LILLsfc12/833°N to 13°S155°W to 175°W
PSEAsfc12/8314°N to 14°S105°W to 85°W
INOUsfc02/8414°N to 1°N135°E
LILLsfc01/8415°S to 13°S175°W
LILLsfc07/84–08/849°N to 10°S150°W to 170°W
PSTRsfc12/8410°N to 11°S105°W to 90°W
RSHRsfc05/84–07/8411°N to 1°S150°W
INOUsfc02/8512°N to 0°S135°E
LILLsfc01/8514°S to 10°N175°E to 165°W
LILLsfc03/85–04/8514°S to 11°N180° to 160°W
LILLsfc08/85–09/8515°S to 14°N180° to 160°W
PSTRsfc04/8514°S to 11°N175°W to 160°W
EPCSsfc04/86–05/8614°N to 6°N110°W to 80°W
INOUsfc02/8613°N to 8°N135°E
LILLsfc01/86–02/8614°S to 10°N165°E to 175°W
LILLsfc04/8614°S to 13°N180° to 160°W
LILLsfc08/8613°S to 1°S180° to 170°W
LILLsfc10/86–11/8614°S to 13°N180° to 160°W
PRC_sfc12/867°N to 14°S135°E to 160°E
OC1986_EP8605/86–06/863°S to 18°N156°W to 91°W
INOUsfc02/87–03/8713°N to 14°N135°E to 160°W
INOUsfc06/87–07/8713°N to 14°N155°E to 135°E
INOUsfc11/87–12/8715°N to 14°S150°W to 105°W
KAIYsfc11/87–12/8715°N to 5°N150°E to 135°E
LILLsfc03/8714°S to 12°N180° to 160°W
LILLsfc05/8715°S to 14°N175°W to 140°W
LILLsfc08/8714°S to 9°N180° to 160°W
LILLsfc10/8714°S to 13°N180° to 160°W
DI1993_17002/93–03/9310°S to 2°N171°W to 168°W
MB1993_Spr02/93–03/938°S to 31°N152°W to 88°W
MB1993_L205/932°S to 8°N126°W to 125°W
MB1993_L104/93–05/935°S to 6°N142°W to 140°W
P18_sfc04/94–05/9415°S to 13°N105°W to 110°W
RYOFsfc02/9415°N to 5°N135°E
RYOFsfc08/94–09/9414°N to 14°N135°E to 145°E
DI1994_11003/94–04/9427°S to 32°N123°W to 103°W
DI1994_17011/948°S to 10°N170°W
DI1994_18011/948°S to 10°N180° to 177°W
MB1994_Fall08/94–09/948°S to 32°N126°W to 75°W
MB1994_Spr04/94–06/949°S to 32°N141°W to 80°W
RT95sfc11/9513°N to 14°S160°W
RYOFsfc02/9512°N to 3°N135°E
RYOFsfc09/9513°N to 12°N135°E to 140°E
P15_sfc03/9614°S to 7°S170°W
RYOFsfc02/96–03/9612°N to 14°N135°E to 165°E
RYOFsfc08/9614°N to 12°N135°E
RYOFsfc11/96–12/9612°N to 12°N165°E to 145°E
KA1996_0306/96–07/968°S to 20°N180°
KA1996_0407/96–08/968°S to 19°N180°
KA1996_0508/96–09/968°S to 31°N155°W to 118°W
KA1996_0609/96–10/968°S to 30°N117°W to 95°W
KA1996_0711/96–12/968°S to 22°N171°W to 149°W
MB199511/96–12/9614°S to 30°N178°E to 180°
MB1996_L105/968°S to 8°N111°W to 95°W
MB1996_L206/96–07/968°S to 19°N156°W to 104°W
RYOFsfc02/9713°N to 5°N135°E
RYOFsfc06/97–07/9714°N to 13°N145°E to 135°E
RYOFsfc10/97–11/9715°N to 14°N165°E to 145°E
KA1997_0102/97–03/978°S to 27°N132°W to 95°W
KA1997_0203/97–04/972°S to 22°N158°W to 121°W
KA1997_0305/97–06/975°S to 8°N171°W to 170°W
KA1997_0406/97–07/978°S to 8°N180°
KA1997_0507/97–08/978°S to 19°N111°W to 95°W
KA1997_0609/97–10/979°S to 27°N158°W to 120°W
KA1997_0711/97–12/978°S to 8°N171°W to 155°W
RYOFsfc02/9813°N to 9°N135°E
RYOFsfc07/9813°N to 15°N145°E to 135°E
RYOFsfc10/98–11/9813°N to 13°N165°E to 145°E
SKAUsfc02/98–03/9812°N to 8°N160°W to 165°W
SKAUsfc03/9815°S to 13°N155°E to 145°E
KA1998_0102/98–03/988°S to 30°N117°W to 95°W
KA1998_0204/98–05/982°S to 32°N156°W to 118°W
KA1998_0306/98–07/988°S to 20°N171°W to 155°W
KA1998_0407/98–08/988°S to 10°N180°
LILLsfc12/8714°S to 13°N165°E to 165°W
PRC_sfc01/87–02/8711°S to 10°S140°E to 135°E
OC1987_TW8707/8712°S to 8°N153°E to 155°E
APL_sfc02/8814°N to 14°N145°E
HAKUsfc02/8813°N to 14°S150°E to 155°E
HAKUsfc03/8814°S to 7°N165°E to 170°E
INOUsfc02/88–03/8814°N to 13°N135°E to 155°E
INOUsfc11/888°N to 14°S145°E to 160°E
LILLsfc01/8812°N to 13°S150°W to 175°W
LILLsfc02/8815°S to 13°N165°E to 165°W
OC1988_EP8805/88–06/8812°S to 16°N171°W to 140°W
APL_sfc02/8914°N to 13°N145°E
INOUsfc01/890–3/8913°N to 1°N160°W to 175°E
RYOFsfc02/8914°N to 6°N135°E
DI1989_RT1_8902/89–03/8926°S to 21°N111°W to 105°W
DI1989_RT2_8903/89–04/8961°S to 17°S150°W to 105°W
DI1989_RT3_8904/8915°S to 47°N149°W to 126°W
HAKUsfc09/90–12/9015°N to 13°N180° to 145°E
N044sfc03/90–04/9015°S to 14°N170°W to 165°W
RYOFsfc02/9015°N to 4°N135°E
RYOFsfc07/90–08/9011°N to 13°N155°E to 135°E
EW91sfc05/91–06/9114°S to 14°S145°W to 150°W
P17_sfc08/917°S to 13°S135°W
RYOFsfc02/9113°N to 8°N135°E
RYOFsfc06/91–07/9112°N to 14°N155°E to 135°E
TUN2sfc08/916°S to 13°S135°W
WEISsfc07/91–08/9115°N to 11°S135°W
WEISsfc09/91–10/9115°S to 14°N150°W to 155°W
RYOFsfc02/9213°N to 5°N135°E
RYOFsfc07/92–08/9214°N to 14°N155°E to 135°E
TT07sfc02/92–03/9212°N to 12°S140°W to 135°W
TT11sfc08/92–09/9212°N to 11°S140°W to 135°W
DI1992_110_9511/92–12/9214°S to 27°N117°W to 95°W
DI1992_12509/92–10/9210°S to 10°N126°W to 125°W
DI1992_14009/927°S to 7°N141°W to 140°W
DI1992_17008/928°S to 8°N171°W to 170°W
MB1992_L102/92–03/928°S to 17°N153°W to 82°W
MB1992_L304/92–05/9210°S to 9°N141°W to 140°W
MB1992_L204/9210°S to 10°N171°W to 170°W
N071sfc10/93–11/9314°S to 13°N165°E to 150°E
P19_sfc04/9314°S to 13°N85°W to 90°W
RYOFsfc02/9313°N to 3°N135°E
RYOFsfc07/93–08/9313°N to 14°N155°E to 135°E
WEISsfc04/9315°S to 7°N85°W to 80°W
DI1993_15502/93–03/9310°S to 2°N171°W to 155°W
KA1998_0509/98–10/988°S to 21°N158°W to 125°W
KA1998_0710/98–11/9817°S to 18°N180°
KA1998_0811/98–12/982°N to 8°N180°W to 172°W
RB1998_0610/98–11/988°S to 35°N123°W to 81°W
RYOFsfc02/9912°N to 7°N135°E
RYOFsfc07/99–11/9913°N to 14°N145°E
KA1999_0101/99–02/995°S to 9°N142°W to 125°W
KA1999_0204/99–06/998°S to 9°N111°W to 95°W
KA1999_0306/99–07/998°S to 5°N180°W to 180°
RB1999_0811/99–12/998°S to 47°N125°W to 95°W
RYOFsfc02/0512°N to 5°N135°E
RYOFsfc07/0513°N to 14°N135°E
RYOFsfc10/04–11/0414°N to 15°N165°E to 135°E
KA2000_0204/04–05/048°S to 12°N111°W to 95°W
KA2000_0306/04–07/048°S to 20°N180°
KA2000_0407/04–08/048°S to 19°N180°
KA2000_0508/04–10/048°S to 13°N148°W to 125°W
KA2000_0610/04–11/048°S to 18°N180°
KA2000_0811/04–12/048°S to 20°N180°
KEIFsfc02/04–03/0414°N to 14°N165°E to 140°E
KEIFsfc06/04–12/0413°N to 14°N135°E to 140°E
PS02sfc11/04–12/0413°N to 15°S165°W to 165°E
RYOFsfc02/0514°N to 7°N135°E
RYOFsfc11/04–12/0414°N to 13°N165°E to 145°E
KA2001_0101/04–02/048°S to 32°N147°W to 118°W
KA2001_0203/04–05/048°S to 32°N118°W to 96°W
KEIFsfc02/04–03/0412°N to 15°N165°E to 145°E
KEIFsfc05/04–08/0412°N to 12°N135°E
KEIFsfc12/059°N to 14°N135°E
PS02sfc04/0510°S to 6°N80°W to 95°W
RYOFsfc02/0514°N to 5°N135°E
RYOFsfc10/04–11/0412°N to 13°N165°E to 145°E
KA2002_0103/04–04/048°S to 12°N113°W to 91°W
KA2002_0204/04–05/049°S to 21°N158°W to 105°W
KA2002_0305/04–06/0414°S to 20°N180°W to 180°
KA2002_0508/04–09/049°S to 5°N141°W to 140°W
KA2002_0710/055°S to 18°N180°
ka2003_0101/059°S to 14°N150°W to 125°W
KA2003_0203/04–04/048°S to 17°N112°W to 95°W
ka2003_0508/04–09/049°S to 21°N158°W to 125°W
ka2003_0710/04–11/0414°S to 18°N180°
ka2003_0811/04–12/048°S to 21°N180°
ka2004_0103/04–04/048°S to 31°N117°W to 95°W
ka2004_0204/04–05/049°S to 18°N155°W to 106°W
ka2004_0306/04–07/0414°S to 18°N180°W to 180°
ka2004_0407/04–08/048°S to 19°N180°

3. Results and Interpretations

3.1. Decadal Trends of fCO2 in the Equatorial Pacific

[7] The surface ocean fCO2 data obtained in the equatorial Pacific between 1974 and 2004 are summarized in Figure 2 for the Niño 3.4 area and in Figure 3 for the Western Pacific Warm Pool region (WPWP; 5°N to 5°S, west of 175°E). As shown in Figures 2 and 3, the fCO2sw data have a wide range of variability (as large as 150 μatm) reflecting highly variable current structures and upwelling in the equatorial zone. The observed spatial variability is much greater than the decadal rate of changes being sought. Furthermore, the data distribution and sampling density are irregular in space and time. Therefore, in order to minimize the effects of uneven density of observations, we have chosen to use monthly mean values instead of individual observations for our time-trend analysis. The mean decadal rate of change and its uncertainty that are reported in this paper are based on 95 mean monthly values for the Niño 3.4 area and 77 for the WPWP region. The top panels in Figures 2 and 3 show all the fCO2sw data at in situ temperature (solid dots) and the mean monthly values (open circles). The red line represents a linear regression line computed using all the monthly mean values, yielding a mean rate of increase of fCO2sw at SST of 11.3 ± 3.1 μatm decade−1 for the Niño 3.4 area and 19.1 ± 2.2 μatm decade−1 for the WPWP area over the 30-year period (Table 2). During the same period, the atmospheric fCO2 has been increasing with mean decadal rates in a range of 16 ± 3 μatm decade−1 as a result of anthropogenic emissions [GlobalView-CO2, 2005]. Accordingly, on the 30-year average, the surface water fCO2 in these two regions has been increasing at a rate similar to the atmospheric CO2 increase owing most likely to air-sea transfer of atmospheric CO2.

Figure 2.

Surface water fCO2 at SST observations in the Niño area 3.4 (5°N–5°S, 170°W–120°W). The solid dots indicate individual observations, and the solid lines indicate the linear regression lines computed using the mean monthly values (red open circles). (a) All data. (b) El Niño data. (c) Non–El Niño data. The mean decadal rate of change in each property is shown along the top of each panel, and N indicates the number of the mean monthly values used. Uncertainties for the rates of change listed in are expressed in terms of ±[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.

Figure 3.

Surface water fCO2 at SST observations in the Western Pacific Warm Pool (west of 175°E). The solid dots indicate individual observations, and the solid lines indicate the linear regression lines computed using the mean monthly values (open circles). (a) All data. (b) El Niño data. (c) Non–El Niño data. The mean decadal rate of change in each property is shown along the top of each panel, and N indicates the number of the mean monthly values used. Uncertainties for the rates of change listed in are expressed in terms of ±[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. The four black circles above the trend lines between 1988 and 1991 represent a very narrow sliver of cold waters, which appeared near the equator. Since the cold SSTs suggest an upwelled waters from the east, these four monthly mean values have been excluded from the rate calculations (see Figures 5b and 5c).

Table 2. Summary of the Mean Decadal fCO2sw Change Rates Over the Equatorial Zone of Pacific, 5°N–5°Sa
 Number of MonthsfCO2 at SST, μatm decade−1SST Increase, °C decade−1fCO2 at 28°C, μatm decade−1
  • a

    The rates for the El Niño periods represent the mean over six El Niño events which occurred between 1982 through 2003, and those for the non–El Niño periods represent the 30-year mean from 1974 through 2004.

Niño 3.4 Region
All years9511.3 ± 3.1−0.1 ± 0.213.2 ± 5.6
Pre-1990.536−11.1 ± 9.60.7 ± 0.6−21.2 ± 16.9
Post-1989.56017.5 ± 7.40.1 ± 0.416.2 ± 13.4
Non–El Niño6310.4 ± 2.9−0.2 ± 0.213.8 ± 5.3
Pre-1990.529−1.5 ± 9.10.0 ± 0.5−0.3 ± 15.4
Post-1989.52517.0 ± 7.90.3 ± 0.511.9 ± 15.3
El Niño (1987–2003)3225.0 ± 8.6−0.7 ± 0.435.8 ± 13.5
Western Warm Pool
All years7719.1 ± 2.20.5 ± 0.111.6 ± 2.7
Pre-1990.5273.9 ± 7.50.9 ± 0.3−9.3 ± 8.3
Post-1989.55223.2 ± 4.60.5 ± 0.215.6 ± 5.7
Non–El Niño5322.2 ± 2.60.5 ± 0.114.2 ± 3.1
Pre-1990.51812.2 ± 5.00.7 ± 0.31.7 ± 4.6
Post-1989.53723.9 ± 5.90.3 ± 0.218.2 ± 7.1
El Niño (1987–2003)249.4 ± 3.70.3 ± 0.24.4 ± 5.0

[8] However, the rate of change of fCO2sw appears to have varied during the period. Takahashi et al. [2003] demonstrated that there was a measurable rate change of fCO2sw at SST starting about 1990 ± 2 years, which they suggested was related to a regime shift in the PDO. Our results for a longer time series show a similar rate change from −11.1 ± 9.6 μatm decade−1 for the pre-1990 period to 17.5 ± 7.4 μatm decade−1 for the post-1990 period in the Niño 3.4 area (Figure 2a); and from 3.9 ± 7.5 μatm decade−1 for the pre-1990 to 23.2 ± 4.6 μatm decade−1 for the post-1990 period in the WPWP area (Figure 3a). These values show that during the pre-1990 period, fCO2sw was nearly unchanged or slightly decreased with time, whereas it increased after 1990 at rates similar to or exceeding the atmospheric CO2 increase rate (Table 2). These results are consistent with a corresponding change in the rate of increase of SST that also occurred at about 1990 in both regions (Table 2). These results suggest that the rate of change in surface water fCO2sw was governed not only by sea-air gas exchange, but also by oceanographic conditions including changes in vertical and lateral flow fields associated with regime shifts in climate modes such as PDO. The observed small rates of fCO2sw change during the pre-1990 period indicate that the sea-air fCO2 differences decreased with time, whereas they remained nearly constant during the post-1990 period. This implies that if sea-air gas transfer rates over the equatorial Pacific remained unchanged with time, the equatorial Pacific waters became a weaker CO2 source during the pre-1990 period, and remained nearly unchanged during the post-1990 period. Similar increases in the anomalies of dissolved inorganic carbon were observed in the early 1990s in the western Subtropical Pacific, which was correlated with a corresponding decrease in SST [Midorikawa et al., 2006].

[9] In order to explore the sensitivity of the choice of the transition year, we changed the transition year by 1-year steps between 1986 and 1993 and computed mean rate of fCO2sw changes for the period before and after an assumed transition year using a linear regression method. The results are summarized in Figure 4, with the Western Pacific Warm Pool region represented in Figure 4a and the Niño 3.4 region represented in Figure 4b. Figure 4a shows that when the transition year is selected at 1986 and 1987, respectively, the posttransition rates (open circles) are distinctly higher than the pretransition rates. The error bars indicate the uncertainties in the rate (computed as shown in Figure 2 caption), and the numbers associated with each point indicate the number of monthly mean values used. Each pair of prerates and postrates is separated by several times the magnitude of the uncertainty bars prior to the transition year. The mean rate of fCO2sw changes for posttransition year periods varies from +1.0 to +2.2 μatm yr−1 with a mean rate of +1.8 ± 0.7 μatm yr−1. For pretransition periods the rate varies from −1.0 to 1.4. On the other hand, when the transition year is moved to 1989 and thereafter, the pretransition rates become statistically indistinguishable from the posttransition rates as indicated by overlapping of uncertainty bars. This suggests that the actual transition occurred between 1989 and 1990 for the Western Warm Pool region. Similarly, Figure 4b shows the mean rate of change for fCO2sw (at SST) as a function of the selected transition year for the Niño 3.4 region. Here the determination of the transition year depends on subtle differences between the uncertainties in the pretransition and posttransition rates. Note that the rates of fCO2sw change for the posttransition period (open circles) are distinctly higher than those for the pretransition periods (solid circles) in 1986 and 1987, as evidenced by the separation of the error bars. The “student's t-test” indicates that the posttransition rate is different from the pretransition rate with a probability of 99%. In contrast, for the transition year which is chosen to be 1989 and thereafter, the error bars for the posttransition and pretransition rates overlap each other, indicating that they are indistinguishable with much lower probabilities.

Figure 4.

Effect of different phase shift years chosen on the mean rate of change in fCO2sw for the preshift and postshift decade. (a) Western Pacific Warm Pool region. (b) Niño 3.4 region. The error bars indicate the uncertainties in the rate, and the numbers associated with each point indicate the number of data used in the analysis.

3.2. Effect of ENSO on the Decadal Trends of fCO2sw

[10] The mean rate of change of fCO2sw at SST also differs from the El Niño to non–El Niño periods. In the Niño 3.4 area, the mean rate was 25.0 ± 8.6 μatm decade−1 for the El Niño periods between 1987 and 2003 El Niño periods and was 17.0 ± 7.9 μatm decade−1 for the non–El Niño periods between 1990 and 2004 (Figures 2b and 2c and Table 2). In the WPWP area, the mean rate of 9.4 ± 3.7 μatm decade−1 for the El Niño periods between 1983 and 2003 was lower than that of 22.2 ± 2.6 μatm decade−1 for the non–El Niño periods between 1982 and 2004 (Figures 3b and 3c and Table 2). This means that, during the El Niño periods, fCO2sw increased at a much faster rate in the central and eastern equatorial Pacific than the western warm pool waters, whereas during the non–El Niño periods, fCO2sw in the western warm pool waters increased at a somewhat faster rate than the central and eastern Pacific waters. Such behaviors imply that the carbon chemistry in the equatorial Pacific is controlled by complex dynamic interactions between ocean circulation, biological activities and sea-air gas transfer processes.

[11] To remove the effect of interannual changes in SST, on the fCO2sw values, the values were normalized to a constant temperature of 28°C (the mean SST for the central and western equatorial Pacific) using a temperature coefficient (∂lnfCO2/∂T) of 0.0423°C−1 [Takahashi et al., 1993], and the rates of change are listed in the last column of Table 2. These rates indicate fCO2sw changes that are caused entirely by changes in seawater chemistry. The overall mean rate of 13.2 ± 5.6 μatm decade−1 for the Niño 3.4 and 11.6 ± 2.7 μatm decade−1 for the WPWP waters are indistinguishable from the mean atmospheric CO2 increase rate of about 16 ± 3 μatm decade−1 (or 1.6 ± 0.3 μatm yr−1). If the alkalinity is assumed to be constant over this period, the observed fCO2sw increase rate should correspond to a total carbon dioxide (TCO2) increase rate of 10 μmol kg−1 decade−1. This is consistent with the rate that is expected for the sea-air CO2 equilibrium with a 16 ± 3 μatm decade−1 atmospheric CO2 increase rate.

[12] During non–El Niño periods, we observe in both the western and central areas that the rate of change in fCO2sw due to chemistry change was negative or nearly zero for the pre-1990 period (Table 2), and hence TCO2 decreased with time or stayed nearly unchanged. On the other hand, it was positive for the post-1990 period ranging from 11.9 to 18.2 μatm decade−1 indicating an increase in a TCO2 at a rate similar to that for equilibrium with atmospheric CO2. The observed increase during the post-1990 period may be attributed to a number of processes that include: (1) an increase in the upwelling rate of deep waters rich in CO2, (2) change in the source for upwelling waters to higher CO2 content waters, and/or (3) a reduction in the net community production/carbon export from the mixed layer. All of these processes would be expected to be affected by changes in climate parameters including wind fields and ocean circulation. However, the relative importance of the individual processes affecting the carbon chemistry of the equatorial waters cannot be determined on the basis of our data alone.

[13] In contrast to the non–El Niño periods, the rate of increase for the 28°C fCO2sw (35.8 ± 13.5 μatm decade−1) during the El Niño events between 1987 and 2003 for the Niño 3.4 area was much higher than the atmospheric CO2 increase rate. The high rate of increase suggests that influx of waters with higher CO2 concentrations increased during the El Niño periods in 1987–2003. Concurrently, the mean SST decreased at a rate of −0.7 ± 0.4 °C decade−1 (Table 2), suggesting an increasing influx of colder and CO2-rich waters into the central equatorial Pacific. Such waters may be transported into the area (1) via increased upwelling of deep waters or (2) via increased lateral flow of colder subtropical waters driven by equatorward winds. The wind data shown in Figures 5a and 6b suggest that minimum wind speeds (4 to 5 m s−1) did increase slightly with progressing years of El Niño events after 1998. Hence the data support the possibility of increased lateral transport of subtropical waters into the region. Precise pathways by which this water of subtropical origin may make its way to the equator requires further investigation, but tropical-subtropical interactions mediated by the general circulation offer a range of possibilities [Kleeman et al., 1999; Izumo et al., 2002].

Figure 5.

Time-longitude distribution of (a) zonal winds (in m s−1), (b) SST (°C), and (c) SST anomaly from 1995 thru 2003 and the (d) ENSO and (e) PDO indexes for the North Pacific. The analysis for Figures 5a, 5b, and 5c are based on monthly averages between 2°S and 2°N from the TAO/TRITON time series array in the equatorial Pacific. The ENSO and PDO indexes are from

Figure 6.

Estimated (a) SST, (b) FSU Subjective Analysis wind speeds, (c) fCO2sw, and (d) CO2 flux between 90°W and 165°E, 5°N to 10°S, from 1982 through June 2004. Surface water fCO2 was by calculated by applying the interannual and seasonal fCO2sw-SST relationships in Table 3 to Reynolds SST data, and CO2 flux was calculated with the FSU Subjective Analysis wind speeds and the Wanninkhof [1992] gas exchange coefficient.

[14] The increased upwelling may be substantiated by additional observations such as nutrient concentrations: increasing upwelling of nutrient-rich subsurface waters would have increased nutrient concentrations, whereas influx of nutrient-depleted subtropical surface waters would not. However, owing to the lack of systematic nutrient data, no firm conclusion may be drawn at this point. On the other hand, WPWP waters had a low increase rate of 4.4 ± 5.0 μatm decade−1 for the El Niño years. This is probably due to the lateral convergence of very warm off equatorial waters of lower TCO2 concentration in the far western Pacific as a result of the intensification of the westerly winds.

3.3. The fCO2-SST Relationships

[15] The data were collected during the period from April 1979 through July 2004, which included five El Niño periods (1982–1983, 1986–1987, 1991–1994 and 1997–1998 and 2002–2003) and four La Niña periods (1984–1985, 1988–1989, 1995–1996 and 1998–2000). Figure 5 shows the regional variations of zonal wind speed, SST and SST anomaly based on the TAO data from 1986 through 2003. During the strong 1997–1998 El Niño the cold waters of the eastern equatorial Pacific disappear whereas during the weaker El Niños of 1991–1994 and 2002–2003, the equatorial cold tongue is present but less pronounced. The strong El Niño of 1997–1998 has SST anomalies exceeding 4°C and weak winds throughout most of the eastern equatorial Pacific from about 95°W to 140°W. In contrast, the El Niños of 1991–1994 and 2002–2003 have SST anomalies of about 1°–2°C and near normal zonal winds in the eastern equatorial Pacific [McPhaden, 2004]. The shipboard fCO2sw and SST data were separated into El Niño and non–El Niño time periods, and then additionally separated into a warm season (January through June) when the winds are seasonally less intense and a cool season (July through December) when the winds are seasonally stronger. Table 1 gives a list of the cruises used during this study.

[16] We developed fCO2sw-SST relationships for the three periods: 1979–1989, 1990–1997, and 1998–2004.5, corresponding to the three regime shifts in the Pacific. The SST and fCO2sw data, spanning the region between 5°N to 10°S and 165°E to 95°W, were binned into one-degree-latitude intervals and averaged. Linear fits were applied to the averaged data sets for the three time periods. The results are shown in Table 3. The relationships have an appreciable uncertainty as indicated by the root mean standard deviation of the calculated fCO2sw (values in brackets following fCO2sw in the above equations; also see the discussion of the uncertainties by Feely et al. [2004]. A linear fit applied to each of the data sets for the three periods indicates an inverse correlation between fCO2sw and SST with an average slope of approximately −11.4 μatm °C−1, which is a smaller negative slope than was previously observed by Cosca et al. [2003] for a data set that spanned a shorter time domain. Separating the non–El Niño data by warm and cool seasons yields significantly larger negative slopes during the warm season (Table 3), which suggests that seasonal variations are significant for all three periods. Under non–El Niño conditions, the fCO2sw values in the eastern equatorial Pacific are significantly lower for a given temperature in the boreal fall than in the boreal spring. This decrease in fCO2sw may be the result of increased biological activity in the boreal fall in the eastern Pacific [Takahashi et al., 2002; Foley et al., 1997; Wang et al., 2006a, 2006b]. Significant seasonal differences in the slopes also are observed during weak El Niño periods, but not during strong El Niños. These results suggest that weak El Niño periods are sensitive to seasonal changes in the biological uptake of carbon between 95°W and 125°W, whereas strong El Niño periods are less so.

Table 3. The fCO2sw-SST Equations for the Equatorial Pacific Geographical Region: 95°W to 165°E, 5°N to 10°S Time Periods Analyzed: 1979 Through 1989, 1990 Through 1997, 1998 Through Mid-2004
Time PeriodfCO2sw-SST Equation
1979 Through 1989
All (n = 828)fCO2(±27.9) = −11.47(±0.55)T + 702.4(±15.4)r2 = 0.348p < .0001
El Niño
   All (n = 330)fCO2(±23.5) = −9.72(±1.10)T + 646.2(±32.0)r2 = 0.193p < .0001
   Warm Season (n = 157)fCO2(±24.3) = −21.06(±2.33)T + 980.8(±68.0)r2 = 0.345p < .0001
   Cool Season (n = 173)fCO2(±19.7) = −6.26(±1.05)T + 542.0(±30.8)r2 = 0.171p < .0001
Non–El Niño
   All (n = 498)fCO2(±29.9) = −10.03(±0.77)T + 666.3(±21.0)r2 = 0.257p < .0001
   Warm Season (n = 348)fCO2(±27.9) = −13.42(±0.99)T + 765.1(±27.3)r2 = 0.347p < .0001
   Cool Season (n = 150)fCO2(±29.8) = −8.08(±1.15)T + 601.7(±30.8)r2 = 0.251p < .0001
1990 Through 1997
All (n = 1218)fCO2(±30.9) = −11.76(±0.43)T + 719.6(±11.7)r2 = 0.385p < .0001
El Niño
   All (n = 500)fCO2(±23.9) = −12.78(±0.59)T + 745.7(±16.5)r2 = 0.487p < .0001
   Warm Season (n = 299)fCO2(±21.8) = −20.36(±1.10)T + 963.6(±31.0)r2 = 0.537p < .0001
   Cool Season (n = 201)fCO2(±22.7) = −10.88(±0.65)T + 686.6(±18.0)r2 = 0.587p < .0001
Non–El Niño
   All (n = 718)fCO2(±34.7) = −10.76(±0.62)T + 694.4(±16.6)r2 = 0.299p < .0001
   Warm Season (n = 265)fCO2(±36.1) = −20.29(±1.60)T + 966.8(±42.9)r2 = 0.381p < .0001
   Cool Season (n = 453)fCO2(±28.1) = −9.16(±0.54)T + 642.3(±14.6)r2 = 0.386p < .0001
1998 Through Mid-2004
All (n = 1469)fCO2(±32.9) = −11.04(±0.37)T + 708.1(±10.2)r2 = 0.377p < .0001
El Niño
   All (n = 345)fCO2(±26.5) = −9.81(±0.81)T + 669.8(±23.1)r2 = 0.301p < .0001
   Warm Season (n = 144)fCO2(±23.3) = −12.16(±2.77)T + 736.7(±80.4)r2 = 0.120p < .0001
   Cool Season (n = 201)fCO2(±28.5) = −9.46(±0.92)T + 661.1(±26.2)r2 = 0.346p < .0001
Non–El Niño
   All (n = 1124)fCO2(±34.6) = −10.98(±0.44)T + 707.2(±12.0)r2 = 0.352p < .0001
   Warm Season (n = 508)fCO2(±31.9) = −19.61(±1.09)T + 960.6(±30.0)r2 = 0.389p < .0001
   Cool Season (n = 616)fCO2(±29.1) = −11.17(±0.41)T + 698.9(±11.0)r2 = 0.541p < .0001

3.4. CO2 Flux Estimates

[17] In order to obtain high-resolution CO2 flux maps, we have combined the satellite temperature and various wind products for the timeframe between November 1981 and June 2004 with the fCO2sw-SST relationships described in Table 3. The SST values are based on an optimally interpolated combination of the Advanced Very High Resolution Radiometer (AVHRR) as described by Reynolds and Smith [1994]. Figure 6 shows the application of the fCO2sw-SST equations with the AVHRR SST data, with Figure 6a showing the interpolated SST distributions, Figure 6b showing the Florida State University (FSU) Subjective Analysis wind speeds, Figure 6c showing the resulting fCO2sw between 5°N to 10°S and 165°E to 95°W, and Figure 6d showing the corresponding CO2 fluxes calculated from the derived fCO2sw using the FSU winds for the same region. The Reynolds SST weekly data were gridded into 1° by 1° pixels and combined with the fCO2sw-SST equations in Table 3 to provide the monthly estimates of the fCO2sw distributions (Figure 6c). As can be seen in Figure 6a, warm SST water (>28°C), which is normally confined to west of 160°W during non–El Niño periods, are found as far east as 95°W during the 1982–1983 and 1997–1998 El Niños. During the 1986–1987, 1991–1994 and 2002–2003 El Niño periods, 28°C water is also found east of its normal longitudinal range, for example, to about 130°W during 1986–1987. During these El Niño events, fCO2sw values drop below 400 μatm west of 150°W. Conversely, during strong La Niña events (i.e., 1988–1989 and 1996–1997), fCO2sw values in excess of 460 μatm extend all the way from 95°W to at least 125°W. Also evident in Figures 6c and 7is the strong seasonal variation of fCO2sw, with lower fCO2sw values in the late boreal fall-winter months relative to the boreal spring and early summer months. This is most dramatic in the eastern equatorial Pacific from approximately 95°W to 125°W, where the seasonal differences in fCO2sw are on the order of 30–50 μatm, consistent with the earlier results of Takahashi et al. [2002], who suggested this was due to enhanced biological activity in the boreal fall. These conclusions are also consistent with the recent modeling results of Wang et al. [2006a, 2006b] which show enhanced primary productivity in the boreal fall due to increased upwelling of both micro (i.e., Fe, trace elements) and macro nutrients. Farther to the west at about 140°W–160°W, the seasonal patterns are weaker based on the satellite-based analyses.

Figure 7.

Distribution of shipboard sea-air fCO2 difference (ΔfCO2, in μatm) for the region between 90°W and 160°E, 10°N to 10°S, from January 1992 through July 2004.

[18] Under non–El Niño conditions in the eastern Pacific, the zonal trade winds tend to be strongest during the latter part of the year, and sea surface temperature is correspondingly lowest during this period (Figure 5). On the basis of the shipboard observations alone (Figure 7) for the 1992–2004 period, the ΔfCO2 levels are highest in the southeastern equatorial Pacific between 0° and 3°S where upwelling is strong and the thermocline is at its shallowest depth [McPhaden, 1999]. In the western portion of the basin, the thermocline is deeper, sea surface temperatures are generally higher than 28°C (Figure 5), and surface waters have near-equilibrium ΔfCO2 values (Figure 7). During the 2002–2003 El Niño event, the easterly trade winds in the eastern equatorial Pacific remained strong and the thermocline was not greatly depressed there [McPhaden, 2004]. Consequently, the decrease in ΔfCO2 levels during 2002–2003 was primarily observed in the western and central regions of equatorial Pacific.

[19] Figure 8 shows the average monthly mean wind speed, fCO2sw and ΔfCO2 levels, and corresponding CO2 fluxes for the entire region of the equatorial Pacific (5°N–10°S and 165°E to 95°W) based on the Reynolds-Smith SST; several wind speed products (TAO, ECMWF (ERA-40), FSU Subjective Analysis, NCEP/DOE AMIP-II Reanalysis 2 and QSCAT); and the fCO2sw-SST relationships of Table 3 for the region from 5°N to 10°S and 165°E to 95°W, spanning an area of approximately 18 × 106 km2. Also shown in Figure 8c are the 6-month average CO2 flux values based on the shipboard measurements of fCO2 (open squares with crosses). All wind speed products were corrected to 10 m height above the seawater surface to correspond with the gas transfer velocity [Wanninkhof, 1992]. The results show reasonably good agreement between the direct measurements and the satellite data–based estimates, although The NCEP/DOE AMIP-II Reanalysis 2 winds show slightly lower overall fluxes than the other wind products (Table 4). The best agreement is between the TAO winds and the FSU Subjective Analysis winds. This wind product, which is based on stresses computed from individual observations, also agrees well with the satellite-based QSCAT winds, which are of shorter overall duration. The uncertainties on the monthly fluxes are primarily due to the uncertainty of interpolated fCO2sw and the uncertainty in the wind speed gas exchange relationship [Feely et al., 2004], with the uncertainty of fCO2sw being generally higher by a factor of about 2 in equatorial Pacific. Large differences in the regional efflux of CO2 are observed between the fully developed strong El Niño events of 1982–1983 and 1997–1998 (Flux < 1.0 moles m−2 yr−1) and the La Niña events of 1984–1985, 1988–1989, 1995–1996 and 1998–2000 (e.g., flux >2.5 moles m−2 yr−1) for this study region. Intermediate fluxes (from ∼1.0 to 1.5 moles m−2 yr−1) are observed during weak El Niño events.

Figure 8.

(a) Monthly mean wind speed, (b) fCO2sw and ΔfCO2, and (c) CO2 sea-air flux between 165°E and 95°W, 5°N to 10°S, from November 1981 through June 2004. Surface water fCO2 was by calculated by applying the interannual and seasonal fCO2sw-SST relationships in Table 3 to the Reynolds SST data; and the CO2 fluxes were calculated with TAO, ECMWF ERA-40, QSCAT, and NCEP Reanalysis 2 wind speeds calculated for 10 m above the sea surface and the Wanninkhof [1992] wind speed–gas exchange relationship.

Table 4. Temporal Changes in Wind Speeds, fCO2, ΔfCO2, and CO2 Fluxes in the Central and Eastern Equatorial Pacific for the Region 5°N–10°S, 95°W–165°Ea
  • a

    Temporal changes are monthly averages. Wind speeds are corrected to 10 m height above the sea.

TAO wind speed, m s−15.91 ± 0.506.16 ± 0.43
NCEP Reanalysis 2 wind speed, m s−15.90 ± 0.505.53 ± 0.525.73 ± 0.41
ECMWF ERA-40 wind speed, m s−15.68 ± 0.465.44 ± 0.405.92 ± 0.43
FSU Subjective Analysis wind speed, m s−15.73 ± 0.575.67 ± 0.476.27 ± 0.50
QSCAT wind speed, m s−16.17 ± 0.43
fCO2 sw, μatm388 ± 11396 ± 10411 ± 9
ΔfCO2, μatm56 ± 1053 ± 1156 ± 11
TAO CO2 flux, moles m−2 yr−11.63 ± 0.511.88 ± 0.53
NCEP Reanalysis 2 CO2 flux, moles m−2 yr−11.72 ± 0.471.44 ± 0.471.60 ± 0.43
ECMWF ERA-40 CO2 flux, moles m−2 yr−11.60 ± 0.471.38 ± 0.401.85 ± 0.39
FSU CO2 flux, moles m−2 yr−11.64 ± 0.491.51 ± 0.461.97 ± 0.61
QSCAT CO2 flux, moles m−2 yr−11.78 ± 0.46

[20] Over the time duration of the measurements, the region was characterized by a significant increase in average fCO2sw levels, variable wind speeds with a minimum in the 1990–1997 timeframe, and a small increase in CO2 fluxes starting near the middle of 1998 and extending to the end of 2001 (Figure 8). The fCO2sw and CO2 flux increases are consistent with the increase in anthropogenic CO2 in the surface waters [Takahashi et al., 2003] and the regime shift from the warm phase to the cold phase with corresponding enhancement of upwelling [McPhaden and Zhang, 2004].

4. Discussion

4.1. Long-Term Trends of the Sea-to-Air CO2 Fluxes

[21] The 23-year record of CO2 fluxes in the equatorial Pacific indicates that interannual processes associated with ENSO events account for the major portion of the flux variability in the region; variations in the biological uptake of carbon and decadal variations in meridional circulation and wind stress account for most of the remaining the variability [Cosca et al., 2003; McPhaden and Zhang, 2004; Wang et al., 2006a, 2006b]. In their earlier work, Takahashi et al. [2003] observed a break in the slopes of both the rate of increase of seawater pCO2 levels and rate of increase of SST in the eastern and western equatorial Pacific that occurred at about the same time as the regime shift between 1988 and 1992. During the decade of the 1980s, prior to the regime shift, the temperature-corrected fCO2sw levels were slightly decreasing or approximately steady with time. After 1990 the temperature corrected fCO2sw levels increased significantly. More recent analysis of the updated surface water fCO2 data suggests that the mean rate of increase of the temperature-corrected seawater fCO2 for the combined El Niño and non–El Niño periods in the Niño 3.4 region for the entire period of record is of 13.2 ± 5.6 μatm decade−1 is not inconsistent with the atmospheric CO2 increase rate of about 16 μatm decade−1 based on the NOAA CMDL GobalView data set. The seawater fCO2 increased during El Niño events (1986–2003) at a faster rate of 35.8 ± 13.5 μatm decade−1 than 13.8 ± 5.3 μatm decade−1 for the non–El Niño periods (Table 2). This means that the seawater fCO2 increase in the central equatorial Pacific is due to a faster increase of seawater fCO2 during El Niño events.

[22] After the 1997–1998 regime shift the CO2 flux increase appeared to roughly coincide with the observed rebound of the meridional overturning circulation [McPhaden and Zhang, 2004]. This is the result of both local increases in wind speed, which started about 1996 in the eastern Pacific (Figure 8), and increases in ΔfCO2, which reached a maximum at about the beginning of 2000. Apparently, the increased overturning circulation resulted in a slight increase in the amount of CO2-enriched water that was upwelled to the surface at the equator.

[23] Our results show significant changes in the slopes of the fCO2sw-SST relationships before and after the 1988–1992 regime shift, particularly during the non–El Niño warm season (Table 3). However, no major changes in the slopes occurred after the 1997–1998 regime shift. Instead, the average wind speeds over the region continued to grow stronger between 1998 and 2002, which accounted for a significant portion of the CO2 flux increase over the region after mid-1998. During the decade from 1990 to 1998.5, the mean combined El Niño and non–El Niño CO2 flux for the region was 1.51 ± 0.46 moles m−2 yr−1, or approximately 0.34 ± 0.10 Pg C yr−1 for the study region from 5°N to 10°S and 165°E to 95°W. In contrast, during the period from 1998 to 2004 the mean CO2 flux was 1.97 ± 0.61 moles m−2 yr−1, or about 0.43 ± 0.13 Pg C yr−1, an increase of about 27% over the previous 8 years. This increase in the mean CO2 flux is primarily due to the increase in average wind speeds. These results, suggesting a slight trend toward stronger CO2 outgassing after the 1997–1998 regime shift, are tentative because the differences do not exceed the overall uncertainty in the flux estimates and because a longer data record is required to ensure that they indeed represent a significant regime shift. There still remains the real possibility that these shifts are related to nonlinearities in ENSO variability [Rodgers et al., 2004]. Nevertheless, it is interesting to note that if these changes are in the same direction elsewhere in the tropical and subtropical Pacific, then they will contribute to the higher year-to-year increases in atmospheric CO2 that has been observed during the last few years (Pieter Tans, personal communication, 2006).

5. Conclusions

[24] Our surface seawater fCO2 measurements in the equatorial Pacific over the past 23 years show that the surface water CO2 has been increasing at a rate similar to the atmospheric CO2 increase, suggesting that the equatorial waters are actively exchanging CO2 with the atmosphere. However, the fCO2sw data combined with SST and wind observations indicate strong interannual ENSO variability throughout the region and a weaker, but significant, seasonal variability in the eastern Pacific. There is also some evidence for a slight increase in the outgassing flux of CO2 after the 1997–1998 regime shift. Most of this recent increase is due to an increase in average wind speeds, and a lesser contribution from an additional increase in fCO2sw. These increases are consistent with the recent rebound of the shallow water meridional overturning circulation in the tropical and subtropical Pacific after the regime shift.


[25] This work was sponsored by the NOAA/OGP Global Carbon Cycle COSP Programs under the leadership of Mike Johnson and Kathy Tedesco of the NOAA Office of Global Programs. We also thank the officers and crew of the NOAA ships Ronald H. Brown and Ka'imimoana for logistics support. Robert Castle of AOML maintained the underway fCO2 system on the Ronald H. Brown and reduced the fCO2 data. We also thank the many scientists that have contributed their data to the World Data Centers. This publication is partially supported by the Joint Institute for the Study of the Atmosphere and Ocean (JISAO) under NOAA Cooperative Agreement NA17RJ1232, contribution 1128. This is also PMEL contribution 2796. This works has also been supported by the grants from the NOAA to Lamont-Doherty Earth Observatory, NOAA NA030AR4310 and NA16GP2001.