Seasonal and interannual variability of ... - Semantic Scholar

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 106, NO. C12, PAGES 31,197–31,215, DECEMBER 2001

Seasonal and interannual variability of phytoplankton, nutrients, TCO2, pCO2, and O2 in the eastern subarctic Pacific (ocean weather station Papa) Sergio R. Signorini,1 Charles R. McClain,2 James R. Christian,3 and C. S. Wong4 Abstract. A coupled, one-dimensional ecosystem/carbon flux model is used to simulate the seasonal and interannual variability of phytoplankton, nutrients, TCO2, O2, and pCO2 at ocean weather station Papa (OWS P at 50⬚N, 145⬚W). The 23-year interannual simulation (1958 –1980) is validated with available data and analyzed to extend seasonal and interannual variations beyond the limited observational records. The seasonal cycles of pCO2 and sea-air CO2 flux are controlled by a combination of thermodynamics, winds, and biological uptake. There is ingassing of CO2 during the fall-winter months when SSTs are colder and wind forcing is vigorous, while there is a much smaller ingassing of CO2 during the summer when sea surface temperatures are warmer and wind speeds are reduced. Biological production plays a major role in maintaining the air-sea equilibrium. An abiotic simulation showed that OWS P would be a source of atmospheric CO2 (1.41 mol C m⫺2 yr⫺1) if the biological sink of CO2 were removed. The peak net community production in summer compensates for the increased temperature effect on pCO2, which prevents large outgassing in summer. Oxygen anomalies relative to the temperaturedetermined saturation value show that there is a seasonal cycle of air-sea flux, with ingassing in winter and outgassing in summer. The net surface oxygen flux is positive (0.8 mol m⫺2 yr⫺1), indicating that OWS P is a source of oxygen to the atmosphere. The average primary production is 167 g C m⫺2 yr⫺1. The 1960 –1980 (1958 and 1959 spin-up years removed) mean carbon flux is ⫺1.8 mol C m⫺2 yr⫺1, indicating that the ocean at OWS P is a sink of atmospheric carbon. The sea-air CO2 flux ranges from ⫺1.2 to ⫺2.3 mol C m⫺2 yr⫺1 during the 21-year simulation period. This finding emphasizes the need for long-term observations to accurately determine carbon flux budgets. A series of sensitivity experiments indicate that the seasonal variability and overall (21 years) mean of TCO2, pCO2, ⌬pCO2, and air-sea CO2 flux are strongly dependent on the gas transfer formulation adopted, the total alkalinity near the surface, and the bottom (350 m) value adopted for TCO2. The secular atmospheric pCO2 upward trend is manifested in the TCO2 concentration within the upper 100 m by an increase of 15 mmol m⫺3 in 20 years, consistent with observations at other locations [Winn et al., 1998; Bates, 2001].

1.

Introduction

The global oceans contain ⬃50 times as much CO2 in dissolved forms as that in the atmosphere, while the land biosphere including the biota and soil carbon contains ⬃3 times as much carbon (as CO2) as that in the atmosphere [Sundquist, 1985]. Thus the spatial and temporal variability of CO2 fluxes over the ocean are crucial for projecting the future level of atmospheric CO2. One of the most important parameters of the oceanic CO2 system is the partial pressure of dissolved carbon dioxide, pCO2, in the surface ocean [Wong and Chan, 1991; Takahashi et al., 1993]. The difference between pCO2 in surface seawater 1

SAIC General Sciences Corporation, Beltsville, Maryland, USA. NASA Goddard Space Flight Center, Greenbelt, Maryland, USA. 3 Earth System Science Interdisciplinary Center, University of Maryland at College Park, College Park, Maryland, USA. 4 Climate Chemistry Laboratory, Institute of Ocean Sciences, Fisheries and Oceans Canada, Sidney, British Columbia, Canada. 2

Copyright 2001 by the American Geophysical Union. Paper number 2000JC000343. 0148-0227/01/2000JC000343$09.00

and in the overlying atmosphere defines the source and sink areas of CO2 over the global oceans. Since the high-latitude waters are probably undersaturated with respect to CO2 in the summer [Keeling, 1968], these oceanic areas could play an important role in climate-CO2 feedback processes by removing large quantities of CO2 from the atmosphere [Wong and Chan, 1991]. Temperate and polar oceans of both hemispheres are the major sinks for atmospheric CO2, whereas the equatorial oceans are the major sources for CO2 [Takahashi et al., 1997]. Thus the evaluation of the air-sea exchange of CO2 is crucial to determine local and global balances of carbon in the atmosphere-ocean system. The evaluation of the atmosphere-ocean CO2 exchange is regulated by the gradient of pCO2 across the air-sea interface, the gas transfer velocity (or piston velocity), and the solubility of CO2 in water. There are a few methods available for evaluating the transfer velocity of CO2 air-sea transfer, which can be obtained by field measurements [Broecker and Peng, 1974; Wanninkhof, 1992; Wanninkhof and McGillis, 1999] or in the laboratory [Liss, 1988]. Laboratory methods using wind/wave tunnel experiments can be used to relate the wind speeds with the measured transfer velocities [Liss and Merlivat, 1986]. These methodologies have been used to study the variability of

31,197

31,198

SIGNORINI ET AL.: OCEAN WEATHER STATION P CARBON BUDGET

the oceanic CO2 system in the subarctic North Pacific [Gordon et al., 1971, 1973; Peng et al., 1974; Murphy et al., 1998]. The seasonal and interannual variations of CO2 in the surface oceans are not only affected by the air-sea exchange physical processes but also by the photosynthetic uptake of CO2 by phytoplankton. For instance, spring phytoplankton blooms in the surface waters of the North Atlantic Ocean can cause a precipitous reduction of surface water pCO2 and nutrients in a span of a couple of weeks. The mechanisms that drive this large biogeochemical variability were modeled by previous investigators [e.g., Taylor et al., 1991]. In contrast, seasonal changes in CO2 and nutrients are more gradual in the North Pacific, and macronutrients are only partially consumed in the surface waters of the subarctic North Pacific Ocean [Takahashi et al., 1993]. In this study, we describe the model functionality and analyze its application to the seasonal and interannual variations of phytoplankton, nutrients, pCO2, and CO2 concentrations in the eastern subarctic Pacific at ocean weather station Papa (OWS P) (50⬚N, 145⬚W). We use the one-dimensional ecosystem model of McClain et al. [1996], coupled with newly incorporated carbon flux, carbon chemistry, and iron components, to simulate 23 years (1958 –1980) of pCO2 and CO2 variability at OWS P. This relatively long period of simulation verifies and extends the findings of previous studies [Wong and Chan, 1991; Archer et al., 1993; Antoine and Morel, 1995a, 1995b] using an explicit approach for the biological component and using realistic coupling with the carbon flux dynamics. Our study shows that realistic simulations of the biogeochemical variability can be achieved with a prognostic biological model. The slow currents and the horizontally homogeneous ocean in the subarctic Pacific make OWS P one of the best available candidates for modeling the biogeochemistry of the upper ocean in one dimension [Archer et al., 1993].

2.

Model Description

The model consists of a four-component ecosystem model [McClain et al., 1996] coupled with newly added carbon dioxide and oxygen components [Signorini et al., 2000]. The ecosystem model is physically forced by sea surface temperature (SST) and mixed layer depth values originating from the mixed layer model of Garwood [1977]. A description of the mixed layer model validation and application to the forcing of the ecosystem model is given by McClain et al. [1996] and is not reproduced here. The ecosystem model has some modifications, which are explicitly defined by McClain et al. [1998, 1999]. In addition to these modifications, two new ecosystem components, iron (Fe) and dissolved organic carbon (DOC), were added to the model. Previous studies [Martin et al., 1989; Boyd et al., 1996; Maldonado et al., 1999] showed that phytoplankton growth in the subarctic Pacific is generally iron limited. We added the iron component to obtain more accurate simulations of nitrate, phytoplankton, and community production. The DOC component was added to more accurately reproduce the seasonality of the net community production via the respiration of DOC by bacteria during the fall-winter period. In addition, the light penetration model uses a set of spectral water absorption (a w ) and phytoplankton specific absorption (a *p ) coefficients rather than a broadband photosynthetically active radiation (PAR) attenuation formulation. Figure 1 shows the spectral variability of these coefficients. The clear sky irradiance is modified to account for the observed cloud cover by

Figure 1. Spectral light absorption coefficient for sea water ( a w ( ␭ ) ) and specific chlorophyll absorption coefficient (a *p ( ␭ )) adopted in the spectral downwelling irradiance model formulation.

applying a power law correction [McClain et al., 1996] tuned to yield the observed climatological monthly mean surface irradiances [Dobson and Smith, 1988]. The cloud attenuation formulation is given by E O ⫽ E clear 关1 ⫺ 0.45Cld0.5兴.

(1)

The coefficient for the cloud attenuation was changed from 0.53 [McClain et al., 1996] to 0.45 to better match the climatological cloudy sky radiation (E O ) of Dobson and Smith [1988]. This new coefficient also provided an improved match of the simulated SST with the observed values. The cloud cover (Cld, in oktas) was obtained from OWS P observations. The following time(t) and depth( z) dependent system of coupled differential equations simulates the dynamics of phytoplankton nitrogen (P), zooplankton nitrogen (Z), ammonium (NH4), nitrate (NO3), iron (Fe), dissolved organic carbon (DOC), total carbon dioxide (TCO2), and oxygen (O2) stocks within the upper ocean:

冋册冋册

⭸P ⭸P ⭸SP P ⭸P ⭸ Kv ⫽ 共G ⫺ m ⫺ rP 兲 P ⫺ IZ, ⫹ we ⫹ ⫺ ⭸t ⭸z ⭸z ⭸z ⭸z

(2)

⭸Z ⭸Z ⭸Z ⭸ K ⫽ 共1 ⫺ ␥兲IZ ⫺ 共 g ⫹ hZ兲 Z ⫺ rZ Z, ⫹ we ⫺ ⭸t ⭸z ⭸z v ⭸z (3) ⭸NH4 ⭸NH4 ⭸ ⫹ we ⫺ ⭸t ⭸t ⭸z

冋

Kv

册

⭸NH4 ⫽ 共a Pm ⫹ r P ⫺ ␲ 1G兲 P ⭸z

⫹ 关aZ共 g ⫹ hZ兲 ⫹ rZ ⫹ cpel␥ I兴Z ⫺ A n ⫹ k rc DOC (N/C)Red, (4) ⭸NO3 ⭸NO3 ⭸ ⫹ we ⫺ ⭸t ⭸z ⭸z ⭸Fe ⭸Fe ⭸ ⫹ we ⫺ ⭸t ⭸z ⭸z

冋

冋

Kv

Kv

册

⭸NO3 ⫽ ⫺␲ 2GP ⫹ A n, ⭸z

(5)

冉冊

(6)

册

⭸Fe Fe ⫽ ␦ 共 z兲FFe ⫺ N P , ⭸z N


Figure 2. model.

Flowchart showing the principal components of the ecosystem/carbon flux one-dimensional

冋

⭸DOC ⭸DOC ⭸ ⫹ we ⫺ ⭸t ⭸z ⭸z

Kv

⭸DOC ⭸z

⫽ 共a⬘PmP ⫹ a⬘Z共 g ⫹ hZ兲 Z兲

冋冉冊

⭸TCO2 ⭸TCO2 ⭸ ⫹ we ⫺ ⭸t ⭸z ⭸z ⫽ ␦ 共 z兲FCO2 ⫺ N P ⭸O2 ⭸O2 ⭸ ⫹ we ⫺ ⭸t ⭸z ⭸z

31,199

冋

Kv

C N

冉冊 C N

册 ⫺ k rc DOC,

(7)

Red

⭸TCO2 Kv ⭸z

册

D e ⫽ ␲ 共2K v /f 兲 1/ 2

(10a)

or

,

冉冊

⭸O2 O2 ⫽ ␦ 共 z兲FO2 ⫹ N P ⭸z N

冋冉冊冉冊册冋冉冊册冋

R共 z兲 ⫽

. Red

(9) A flowchart of the coupled model, showing the interaction between its major components, is given in Figure 2. Table 1 summarizes the parameters used in (2)–(9) and the derived quantities. The net community production is denoted by N p . The ratios (C/N)Red and (O2/N)Red are the carbon-to-nitrogen (106:16) and oxygen-to-carbon (138:16) Redfield ratios, respectively. The revised C/N and O2/N ratios given by Anderson and Sarmiento [1994], 117:16 and 170:16, respectively, were also tested, but the ⬃10% upward adjustment imparted in the primary production by these revised ratios causes a decrease in pCO2 and TCO2, which decreases the quality of the agreement with the observed values. The iron-to-nitrogen ratio (Fe/N) is assumed to be 24 ␮mol mol⫺1. The Kroenecker delta (␦[z ⫽ 0] ⫽ 1; ␦[z ⬎ 0] ⫽ 0) is used to denote that the carbon dioxide and oxygen fluxes (FCO2 and FO2, respectively) are only applied at the air-sea interface. Details of the Ekman upwelling (w e ) and vertical diffusion (K v ) formulations, the numerical method to solve the coupled differential equations, and the various parameters used by the ecosystem model components are given by McClain et al. [1996]. The Ekman upwelling pro-

4.3W , 关sin 共兩 ␾ 兩兲兴 0.5

De ⬇

(8)

Red

册

file function was modified to account for depth attenuation below the Ekman depth (D e ). The formulation for D e [Pond and Pickard, 1978] as a function of wind speed W and latitude ␾, Corolis parameter f the shape function R( z), and the vertical velocity are given by

1 ⫺ cos

R共z兲 ⫽ 1 ⫺

␲z 2D e

z ⫺ De zmax ⫺ D e

exp

1/5

cos

⫺␲ z De

(10b) z ⱕ D e,

␲ 共 z ⫺ D e兲 2共 z max ⫺ D e兲

册

(11) z ⬎ D e, (12)

w e ⫽ w DeR共 z兲.

(13)

Equation (10b) [Pond and Pickard, 1978] produces values of D e nearly identical to the standard Ekman formula given in (10a). The standard formula introduced numerical instabilities, so (10b) was used. The shape function R( z) is constructed based on McClain et al. [1996] for the portion above D e (equation (11)). For z ⬎ D e , R( z) is formulated such that continuity at the inflection points z ⫽ D e and z ⫽ z max (350 m in this case) is maintained and R( z) decreases to zero monotonically toward z ⫽ z max (equation (12)). The net community production, N p , is defined as the primary production minus remineralization, which is equivalent to the summation of all sources and sinks of nitrate and ammonium in (3) and (4): N p ⫽ GP ⫺ 共a pm ⫹ r p兲 P ⫺ 关a z共 g ⫹ hZ兲 ⫹ r z ⫹ c pel␥ I兴Z ⫺ k rc DOC (N/C)Red.

(14)

31,200


Table 1. Model Variables and Input Parametersa Symbol

Value

Definition

P Z NH4 NO3 Fe DOC TCO2 O2 ␲1 ␲2 K NO3 K NH4 K Fe I Np m G rP k gp ␩ k rp Ik max Ik max pk SP Chl-a:N g h Rm ⌳ TH ␥ r zo k rz ap az a⬘p a⬘z k rc An D min KD c pel Kv K vbot we FCO2 FO2 FFe C dust Sv Fe C Fe S Fe

0.2 0.1 0.1 Climatology 50/500 10/0 2050/2250 320/60

phytoplankton concentration (mmol N m⫺3) zooplankton concentration (mmol N m⫺3) ammonium concentration (mmol N m⫺3) nitrate concentration (mmol N m⫺3) iron concentration (pM) dissolved organic carbon (mmol C m⫺3) total CO2 concentration (mmol CO2 m⫺3) oxygen concentration (mmol O2 m⫺3) regenerated production fraction new production fraction half saturation for NO3 uptake (mmol N m⫺3) half saturation for NH4 uptake (mmol N m⫺3) half saturation for Fe uptake (pM) herbivore grazing term (Ivlev [1955]) net community production (mmol C m⫺3 d⫺1) phytoplankton death rate (d⫺1) phytoplankton growth rate (0.5899 d⫺1 at 0⬚C) respiration rate for phytoplankton (d⫺1) temperature sensitivity of algal growth (⬚C⫺1) respiration coefficient temperature sensitivity of algal respiration (⬚C⫺1) maximum photoacclimation parameter (␮mol m⫺2 s⫺1) maximum photoacclimation parameter (␮mol m⫺2 s⫺1) ammonium inhibition of NO3 uptake phytoplankton sinking rate (1.0 m d⫺1 maximum) chlorophyll to nitrogen ratio linear coefficient of zooplankton mortality (d⫺1) quadratic coefficient of zooplankton mortality (mmol⫺1 m3 d⫺1) maximum zooplankton grazing rate (d⫺1) Ivlev constant (m⫺3 mmol N) maximum carbon threshold for Z grazing (mg C m⫺3) unassimilated zooplankton ingestion ratio respiration rate for zooplankton (r Z) at 0⬚C (d⫺1) temperature sensitivity of Z respiration (d⫺1) fraction of dead P converted to NH4 fraction of dead Z converted to NH4 fraction of dead P converted to DOC fraction of dead Z converted to DOC respiration rate of DOC into CO2 (d⫺1) rate of nitrification (2.0 mmol d⫺1 maximum) minimum inhibition dosage for nitrification (W m⫺2) half saturation dosage for nitrification photoinhibition (W m⫺2) fecal pellet remineralization fraction vertical eddy diffusion (m2 d⫺1) minimum bottom eddy diffusion (m2 d⫺1) Ekman upwelling velocity (m d⫺1) air-sea CO2 flux air-sea O2 flux eolian iron flux atmospheric dust concentration (␮g kg⫺1 of air) iron scavenging ratio iron mass fraction soluble iron fraction

0.5 0.1 35 0.05 0.0633 0.02 0.0633 50 (⬍60 m) 250 (⬎60 m) 5 1 0 0.35 4 0.8 5 0.3 0.019 0.15 0.2 0.2 0.1 0.1 0.01 0.0095 0.036 0.8 17.3

Climatological 1000 0.035 0.1

a For variables (P, Z, NH4, NO3, Fe, DOC, TCO2, and O2) the column of values contains initial mixed layer concentration values (for Fe, DOC, TCO2, and O2, deep reservoir concentration values are also provided). The remaining values are the values assigned to the various parameters.

The effective growth is a function of the maximum growth (G O), light limitation (L lim), and nutrient limitation (N lim): G ⫽ ␤ G Oe 共kgpT兲,

(15)

␤ ⫽ L lim min 共N lim, Felim兲,

(16)

L lim ⫽ 1 ⫺ exp

冉

冊

E , ⫺ Ik max

Fe Felim ⫽ , 共K Fe ⫹ Fe兲

(17) (18)

N lim ⫽ NH4lim ⫹ NO3lim, NO3lim ⫽

(19)

共1 ⫺ NH4兲 NO3 , 共K NO3 ⫹ NO3) 共K NH4 ⫹ NH4兲

(20)

NH4 , 共K NH4 ⫹ NH4兲

(21)

NH4lim ⫽

where K Fe, K NO3, and K NH4 are Fe, NO3, and NH4 concentrations, respectively, where growth is half maximal and E is the ambient irradiance (␮mol photons m⫺2 s⫺1).


31,201

Figure 3. Time series of precipitation rate and eolian iron flux for 1970 –1980. Precipitation data are from the National Centers for Environmental Prediction at 50.5⬚N, 144.4⬚W.

3.

Model Forcing and Boundary Conditions

The Neumann boundary condition, ⭸X/⭸ z ⫽ 0, is applied at both the surface and lower (350 m) model domain boundaries for P and Z. Initial profiles of temperature and NO3 were obtained from winter and annual climatologies, respectively. Depth-independent initial concentrations of P and Z are 0.2 and 0.1 mmol N m⫺3, respectively. Observed fall profiles of NH4 were used as initial conditions [Frost, 1993]. At the lower boundary, fixed values equal to 0.0 and 41.5 mmol N m⫺3 were applied to NH4 and NO3, respectively. The Neumann condition was applied at the surface. For TCO2 and O2, observed values at the surface and at the bottom were used to construct the initial linear profiles (surface TCO2 ⫽ 2050 mmol m⫺3, and bottom TCO2 ⫽ 2250 mmol m⫺3; surface O2 ⫽ 320 mmol m⫺3, and bottom O2 ⫽ 60 mmol m⫺3). The oxygen data originate from Martin et al. [1989], and the TCO2 data were collected during 1996 –1997 monthly cruises in the vicinity of OWS P. A fixed value equal to the initial condition was applied at the lower boundary. The following formulations for the CO2 and O2 gas transfer were applied in the form of flux boundary conditions (FCO2 and FO2 in mol m⫺2 yr⫺1) at the sea-air interface: FCO2 ⫽ K O␣ ⌬pCO2

(22)

FO2 ⫹ K O[O2(sat) ⫺ O2],

(23)

where, K O is the piston velocity, in m d⫺1, which is a function of water temperature and wind speed [Wanninkhof, 1992; Liss and Merlivat, 1986], ␣ is the CO2 solubility in seawater (in mmol m⫺3 ␮atm⫺1), which is a function of temperature and salinity [Weiss, 1974], ⌬pCO2 (in ␮atm) is the difference between air and sea pCO2, and O2(sat) is the oxygen saturation concentration (in mmol m⫺3) in seawater, which is a function of temperature and atmospheric pressure [Weiss, 1970]. The CO2 gas transfer coefficient is defined as K ⫽ FCO2/⌬pCO2 ⫽ K O␣ . The eolian iron flux, FFe, is given by

(24)

FFE ⫽

C dustSvFePrC FeS Fe , 55.84

(25)

where, C dust is the seasonal atmospheric dust concentration in ng/(kg of air) from Tegen and Fung [1994], SvFe ⫽ 1000 is the scavenging ratio [Duce, 1995], Pr is the daily National Centers for Environmental Prediction precipitation in kg m⫺2 d⫺1, C Fe ⫽ 0.035 is the iron mass fraction in the dust, and S Fe ⫽ 0.1 is the soluble iron fraction. The numerical factors are used to convert dust concentration to an iron flux (FFe) in nmol m⫺2 d⫺1. The seasonal variability of precipitation and the iron flux for the period of 1970 –1980 is shown in Figure 3. The 30-year pCO2 atmospheric time series used to force the model is based on a least squares fit to the 15 years (1978 – 1993) of pCO2 observations in Cold Bay, Alaska, the closest long-term monitoring site at roughly the same latitude as OWS P. The atmospheric pCO2 time series is given by pCO2air ⫽ A O ⫹ A 1t ⫹ A 2 sin ⫹ A 4 sin

冉

冊

冉

冊冉

2␲t ⫹ A3 12

2␲t ⫹ A 5 ⫹ A 6 sin 6

(26)

冊

2␲t ⫹ A7 , 4

where t is the time in months, A 0 ⫽ 292.23 ␮atm is the intercept, A 1 ⫽ 0.134 ␮atm month⫺1 is the slope, and the amplitudes A 2 , A 4 , and A 6 are 6.61, 3.01, and 0.87 ␮atm, respectively. The phases A 3 , A 5 , and A 7 are 0.77, ⫺15.2, and 0.15 rad, respectively. Figure 4 shows the time series of atmospheric pCO2 at Cold Bay and at Mauna Loa and shows the synthesized signal (dashed line) using the analytical formula (equation (26)). Note that the long-term trends at Cold Bay and Mauna Loa are essentially identical, while there is a significant difference in the seasonal amplitudes. The fact that the trends are similar and the amplitudes are different can be attributed to atmospheric mixing processes. Specifically, the atmospheric seasonal mixing is imparted preferentially along latitude lines by virtue of stronger zonal flows in the atmosphere, allowing the meridional gradients of pCO2 to be much

31,202


Figure 4. Time series of Mauna Loa and Cold Bay atmospheric pCO2. The dashed line represents the time series of pCO2 used to force the model. stronger than the zonal gradients. Conversely, the long-term trend in pCO2 distribution is latitudinally more uniform because the meridional mixing timescale is short when compared to the pCO2 trend due to anthropogenic sources. Sensitivity tests were conducted off-line using the air-sea flux formulation implemented in the model. Four formulations for K O were tested: Liss and Merlivat [1986], Tans et al. [1990], Wanninkhof [1992], and Wanninkhof and McGillis [1999]. Monthly averaged (1973–1978) winds, SST, salinity, and ⌬pCO2 from Wong and Chan [1991] were used in these calculations. The Liss and Merlivat [1986] method provides the smallest sea-air flux (⫺0.79 mol C m⫺2 yr⫺1), while the methods of Tans et al. [1990] and Wanninkhof [1992] provide very similar results (Figure 5). The largest sea-air flux (⫺2.78 mol C

m⫺2 yr⫺1) was provided by the Wanninkhof and McGillis [1999] method since it is a cubic function of the wind speed (W): K O ⫽ 关1.09W ⫺ 0.333W 2 ⫹ 0.078W 3兴共Sc/660兲 ⫺1/ 2,

(27)

where Sc is the Schmidt number of CO2 [Wanninkhof, 1992] given as a function of temperature T in degrees Celsius as Sc ⫽ 2073.1 ⫺ 125.62T ⫹ 3.6276T 2 ⫺ 0.043219T 3.

(28)

Equation (27) is used for the stand-alone flux calculations using climatological winds. For the actual model simulations we use the formula suggested by Wanninkhof and McGillis [1999] for short-term (⬍1 day) winds given by K O ⫽ 0.0283W 3共Sc/660兲 ⫺1/ 2.

(29)

We adopted the Wanninkhof and McGillis [1999] method for the final interannual run (1958 –1980). We conducted several sensitivity runs (1970 –1980) using different methods to derive the gas transfer coefficient. Figure 6 shows the ⌬pCO2 and air-sea flux seasonal plots simulated by the model for three different gas transfer coefficient methods and shows the effect of biological drawdown by removing the biological CO2 uptake (abiotic run). The surface CO2 fluxes obtained with the Liss and Merlivat [1986], Wanninkhof [1992], and Wanninkhof and McGillis [1999] methods are ⫺1.83, ⫺1.92, and ⫺1.92 mol C m⫺2 yr⫺1, respectively. The negative sign indicates that the flux is from the atmosphere to the ocean. The role of biology in the CO2 air-sea exchange is striking, judging from the relatively large outgassing of CO2 (1.41 mol C m⫺2 yr⫺1) simulated by the abiotic run. We discuss more details of these results in section 5.

4.

Oceanic pCO2 Submodel

To calculate the pCO2 concentration in seawater, we must first understand the thermodynamics of the CO2 system. The total CO2 concentration in seawater, TCO2, can be written as TCO2 ⫽ [CO2] ⫹ [HCO3⫺] ⫹ [CO2⫺ 3 ],

Figure 5. Comparison between sea-air CO2 flux calculations using (top) climatological winds and seasonal ⌬pCO2 from 1973–1978 data [Wong and Chan, 1991] for (bottom) four different parameterizations for the gas transfer coefficient.

(30)

where all quantities in brackets are stoichiometric concentrations; [CO2] is the dissolved carbon dioxide, [HCO⫺ 3 ] is the bicarbonate ion, and [CO2⫺ 3 ] is the carbonate ion. Another quantity that influences the calculation of pCO2 is the total alkalinity, TA, which can be written as the sum of its major terms:


31,203

ing 1973–1978 [Wong and Chan, 1991] were used to validate and formulate model parameters. In addition, surface salinity and temperature were collected together with underway pCO2 measurements for the periods 1973–1978, 1988 –1989, and 1996 –1996. The data sampling was conducted daily at approximately hourly intervals but with a significant number of gaps during the 10 years of sampling. On the basis of these salinity, S, data we derived an analytical formulation to provide an uninterrupted sequence of hourly values to force the model. A regression of salinity versus temperature in degrees Celsius (r 2 ⫽ 0.4, n ⬇ 12,800) yields the linear relationship S ⫽ 32.8066 ⫺ 0.01638T.

(33)

A few daily profiles of TA, TCO2, S, and T (Figure 7) were taken in 1996 (February 29, March 2, May 20, and August 29) and in 1997 (February 21, June 16, and September 3). A linear regression of TA versus salinity, using upper 200-m TA and salinity from these profiles (46 points), provided a statistically significant (r 2 ⫽ 0.92) TA expression to force the model, TA ⫽ ⫺1267.46 ⫹ 107.14S.

(34)

To produce hourly values of TA for the model, we expressed TA as a function of the predicted SST by substituting (33) into (34), TA ⫽ 2247.4 ⫺ 1.755 SST.

(35) ⫺1

Figure 6. Seasonal variability of sea-air (top) ⌬pCO2 and (bottom) surface flux predicted by the model using three different gas transfer coefficients. Seasonal variabilities of sea-air ⌬pCO2 and surface flux from an abiotic run using the Wanninkhof and McGillis [1999] gas transfer coefficient are also shown.

TA ⫽ [HCO3⫺] ⫹ 2[CO32⫺] ⫹ [B(OH)4⫺] ⫹ [OH⫺] ⫺ [H⫹],

(31)

⫺ ⫹ where [B(OH)⫺ 4 ] is the borate and [OH ] and [H ] are the products of H2O dissociation. The partial pressure of dissolved CO2 is defined by the relationship

pCO2 ⫽ [CO2]/ ␣ ,

(32)

where [CO2] represents the carbon dioxide in solution and ␣ (in mmol m⫺3 ␮atm⫺1) is the solubility of carbon dioxide in seawater [Weiss, 1974]. We can calculate [CO2] from TCO2 and TA (the values for the hydrogen ion concentration, H⫹, and the carbonate alkalinity, CA, are also computed). We follow the recursive method of Peng et al. [1987] to estimate the dissolved [CO2] concentration. Peng et al. [1987] provide a detailed description of the method in their Appendix. Therefore the details of the method are not reproduced here. The concentrations of dissociated species of carbonic, boric, silicic, and phosphoric acids in seawater are related by the respective apparent dissociation constants. The method uses the carbonic acid dissociation constants of Mehrbach et al. [1973], as refit by Dickson and Millero [1987], together with the boric acid dissociation constants of Dickson [1990]. The surface phosphate and silicate concentrations are kept at climatological values (1.0 and 15.0 ␮M, respectively, from Conkright et al. [1998]), since the model does not calculate PO4 and SiO2 concentrations. Surface and subsurface biogeochemical data collected dur-

Equation (35) yields a TA of 2238.4 ␮mol kg for March (SST ⫽ 5.1⬚C) and 2224.8 ␮mol kg⫺1 for September (SST ⫽ 12.9⬚C). At the bottom boundary we used a constant bottom TCO2 value of 2250 ␮mol kg⫺1, obtained from the observed profiles of Figure 7.

5. Model-Data Comparison and Sensitivity Experiments Table 2 provides a comparison between model and observed [Wong and Chan, 1991] biogeochemical parameters. This particular model run uses the gas transfer coefficient, K, from Wanninkhof [1992]. Both model and observed values are averaged over years 1973–1978. Primary production (0 – 80 m) and chlorophyll and nitrate (0 –25 m) are depth-averaged. The temperature, nitrate, and chlorophyll climatological means for 1973–1978 were obtained from the average of monthly means. The primary production climatological mean for 1973–1978 was obtained from the average of seasonal means. The model primary production is within 20% of the observed value. Chlorophyll and nitrate are within ⫾8% of the observed values. The predicted mean SST is within one tenth of a degree of the observed values. The salinity is obtained via regression with SST; thus it is not independent of the observations and should not be evaluated as a model parameter. Figure 8 shows a comparison between model (solid lines) and observed (dotted lines) seasonal temperature profiles, averaged over the overlapping model-data years (1958 –1966). The model profiles represent the seasonal variability of the temperature stratification and SST at OWS P reasonably well. The model TCO2 is only 0.1% (1.5 mmol m⫺3) lower on average than the observed TCO2. The mean mixed layer pCO2 is within 0.4% of the observed value. The largest uncertainty lies in the calculation of the air-sea CO2 flux (FCO2). The surface flux is the most elusive quantity in the oceanic CO2 measurements and modeling. It depends on accurate measure-

31,204


Figure 7. Observed daily profiles (four in 1996 and three in 1997) of (left) total carbon dioxide and (right) total alkalinity at OWS P. ments of ⌬pCO2 and on a robust formulation for the gas transfer coefficient, K; methods for direct measurements of the air-sea flux are still experimental. As we have shown in section 4, there are multiple gas transfer coefficient formulations in the literature which yield a wide range of values for a given wind speed. For example, as shown in Figure 5, FCO2 can change significantly depending on which formulation for K is used. In view of this, we conducted a series of sensitivity runs with the model to assess the effects of different formulations for K. We conducted five sensitivity runs (1970 –1980) and one final interannual run (1958 –1980) as follows: (1) LM86: baseline run with the best parameter set for the biology components and K from Liss and Merlivat [1986]; (2) W92: same run as in LM86 but using K from Wanninkhof [1992]; (3) WM99: same run as in W92 but using K from Wanninkhof and McGillis [1999]; (4) same run as in WM99 but with no iron limitation (nitrogen limited); (5) same run as in WM99 but with removal of biological drawdown of CO2 (abiotic); and (6) same run as in WM99 but as an interannual run. Table 2. Comparison Between Model and Observed Parameters (1973–1978) ⫺2

⫺1

PP, g C m yr Chlorophyll, mg m⫺3 Nitrate, ␮M SST, ⬚C Sea surface salinity TCO2, mmol m⫺3 Air pCO2, ␮atm Ocean pCO2, ␮atm ⌬pCO2, ␮atm a

From Wong and Chan [1991]. From McClain et al. [1996].

b

Model

Observed

167 0.33

140 0.41a 0.32b 10.6 8.2 32.67 2042.1 332.2 316.7 ⫺15.5

9.9 8.2 32.68 2040.6 332.3 317.9 ⫺14.4

The results are summarized in Table 3. A comparison between the sensitivity runs for K (runs 1, 2, and 3) reveals that as K changes with the different algorithms, ⌬pCO2 changes in the opposite direction so that FCO2 changes are more subtle or gradual. For example, the mean values for K, ⌬pCO2, and FCO2 from run 1 are 0.057 mol m⫺2 ␮atm yr⫺1, ⫺30.2 ␮atm, and ⫺1.83 mol C m⫺2 yr⫺1, respectively. The mean values for K, ⌬pCO2, and FCO2 from run 2 are 0.124 mol m⫺2 ␮atm yr⫺1, ⫺14.4 ␮atm, and ⫺1.92 mol C m⫺2 yr⫺1, respectively. Thus, for run 2, K is ⬃2 times larger, ⌬pCO2 is ⬃0.5 times smaller, and FCO2 is only ⬃5% larger than for run 1. This means that there is a dynamic negative feedback in the model that prevents the surface flux from assuming extremely large values no matter how large the value of K is. However, we use the Wanninkhof and McGillis [1999] gas transfer coefficient because it provides the least ⌬pCO2 bias (model minus observed ⬃0.3 ␮atm). The nitrogen-limited run (run 4; without iron limitation) provides a surface flux 1.3 times larger than the iron-limited run (run 3), whereas the abiotic run (run 5) reverses the flux direction to ⫹1.41 mol C m⫺2 yr⫺1. The relative influence of light and nutrient limitation on the phytoplankton growth simulated by the model is shown in the climatological vertical profiles of light limitation (L lim), nitrate and ammonium limitation (N lim), and iron limitation (Felim) in Figure 9. The macro-nutrient limitation factor (N lim) is shown for both iron-limited and nitrogen-limited runs. For the ironlimited run (run 3), within the top 50 m, iron is the limiting factor, while below 50 m, and within the euphotic zone, light is the limiting factor. The primary production in this case is 167 g C m⫺2 yr⫺1. For the nitrogen-limited run (run 4), the primary production is 195 g C m⫺2 yr⫺1, which is very close to the value (190 g C m⫺2 yr⫺1) reported by McClain et al. [1996], predicted by their nitrogen-limited model for OWS P. Figure 10 shows a comparison between modeled and observed profiles of NO3, Fe, O2, and TCO2. The observed values


31,205

Figure 8. Comparison between model (solid lines) and observed (dotted lines) seasonal temperature profiles averaged over the period 1958 –1966.

Table 3. Comparison Between Five Different Sensitivity Runs (1970 –1980) and the Final Interannual Run (1958 –1980)a

Run Type Run 1: LM86 1970–1980 Run 2: W92 1970–1980 Run 3: WM99 1970–1980 Run 4: WM99 1970–1980 no iron Run 5: WM99 1970–1980 abiotic Run 6: WM99 1958–1980 Observed

PP, g C m⫺2 yr⫺1

Chl-a, mg m⫺3

NO3, ␮M

Ocean pCO2, ␮atm

TCO2, mmol m⫺3

⌬pCO2, ␮atm

K ⫽ K0␣, mol m⫺2 ␮atm yr⫺1

FCO2, mol C m⫺2 yr⫺1

167

0.33

9.9

302.0

2032.0

⫺30.2

0.057

⫺1.83

167

0.33

9.9

317.9

2040.6

⫺14.4

0.124

⫺1.92

167

0.33

9.9

316.5

2040.0

⫺15.8

0.113

⫺1.92

195

0.35

3.9

309.8

2036.3

⫺22.5

0.113

⫺2.54

0

38.8

354.2

2058.1

21.9

0.113

⫹1.41

167

0.33

10.9

309.4

2033.7

⫺14.0

0.120

⫺1.84

140

0.32

10.6

316.7

2042.1

⫺15.5

0.05b

⫺0.7b

0

a All values for the sensitivity runs were obtained by averaging the model-data overlapping years (1973–1978). The values for the interannual run were averaged over the period 1960 –1980, with spin-up initial years 1958 and 1959 eliminated from the average. The observed values are derived from Wong and Chan [1991] for comparison. b Calculated by Wong and Chan [1991] using gas transfer algorithm of Liss and Merlivat [1986].

31,206


Figure 9. Climatological profiles of light (L lim) and nutrient (N lim) limitation predicted by the model. Both iron-limited and nitrogen-limited profiles of N lim are shown. The Felim profile is also shown for comparison.

are from the Vertical Transport and Exchange (VERTEX) cruise in the Gulf of Alaska, which included a cast at OWS P on August 5, 1987 [Martin et al., 1989]. The observed TCO2 profile was from a cast taken on August 29, 1996, at OWS P. The agreement is very good within the euphotic zone ( z ⬍ 100 m), where the air-sea gas flux and biogeochemical interactions take place, but the model departs somewhat from observations at middle depth (100 m ⬍ z ⬍ 300 m). The observed profiles at middle depth indicate a water mass with more nitrate, more iron, more TCO2, and less oxygen than predicted by the model. These departures from middle-depth observations emphasize some aspects of the OWS P variability that are not captured with one-dimensional dynamics. This conclusion is supported by the annual climatologies for temperature, salinity, oxygen, nitrate, silicate, and phosphate in the Gulf of Alaska at 200 m (from Conkright et al. [1998], not shown). These climatological fields show that OWS P lies in a region of relatively weak horizontal gradients of temperature and salinity but is located at the fringe of strong middle-depth horizontal gradients of oxygen and nutrients that are maintained by large-scale processes not simulated by the onedimensional model. Figure 11 shows the seasonal variations of SST, sea surface salinity, nitrate, chlorophyll, total carbon dioxide, in situ pCO2, and pCO2 normalized to 10⬚C obtained from the model simulation (run 3; solid lines) using K from Wanninkhof and McGillis [1999] and the equivalent seasonal variations obtained from observations (circles and triangles). The chlorophyll data

Figure 10. Observed profiles of nitrate, iron, oxygen (from the Vertical Transport and Exchange cruise, August 5, 1987), and TCO2 (recorded by C. S. Wong, August 29, 1996). Observed profiles (dashed lines) are compared with the predicted profiles (solid lines) for August averaged over the period 1960 –1980.

come from two sources: the Wong and Chan [1991] data for the period 1973–1978 (triangles) and the National Oceanic Data Center (NOAA/NODC) data set for the period 1959 –1980 (circles). All parameters simulated by the model are in good agreement with observations. Figure 12 compares the model and observed pCO2 interannual variability for the period 1973–1978. The amplitude and phase of the pCO2 seasonal cycle predicted by the model (solid lines) compares very well with the observed values (black dots).

6.

Seasonal Variability

Figures 13, 14, and 15 show the seasonal variability (from 1960 –1980 monthly averages) of the physical and biogeochemical model parameters. These are the vertical velocity, vertical eddy diffusivity, temperature, photosynthetically available radiation (PAR), nitrate, ammonium, zooplankton, phytoplankton, total carbon dioxide, oxygen, and iron. The vertical velocity is very weak in general (maximum of 3 cm d⫺1), with upwelling peaks in the spring and fall when the wind stress curl is largest. Maximum downwelling occurs in December. The vertical advection effects are minimal when compared to vertical diffusion K v . Maximum surface K v values ranging from 900 to 1000 m2 d⫺1 occur in late fall and winter. The depth penetration of large K v values follows the seasonal changes of the mixed layer depth: largest in winter (120 m) and smallest in


Figure 11. Model (solid lines) versus observed (circles and triangles) seasonal variability of sea surface temperature and salinity, nitrate and chlorophyll averaged over the upper 25 m, surface total carbon dioxide, pCO2 at in situ temperature, and pCO2 normalized to 10⬚C.

summer (20 – 40 m). Downwelling light intensity and penetration are larger during April–August, peaking in May–June. The warmer temperatures are confined to the top 50 m and the time period May–October, with a peak in August–September. The seasonal variability of phytoplankton, nitrate, and iron are strongly correlated, with a peak in phytoplankton abundance and nitrate and iron depletion in the upper 50 m extending from May through October. This is consistent with biologically mediated export of nutrients during the high growth season. Total carbon dioxide and oxygen are strongly correlated with temperature because of the strong dependence of solubility on temperature. Oxygen anomalies relative to the temperature-determined saturation value (Figure 16) show that there is a seasonal cycle of air-sea flux, with ingassing in winter and outgassing in summer. The oxygen anomaly in the abiotic case is consistently less than in the full coupled model, suggesting that net community production (N P ) is positive at all times of the year. The net biological O2 production is 8.6 mol m⫺2 yr⫺1, which is balanced by a sea-air outgassing of 0.8 mol m⫺2 yr⫺1, DOC oxidation of 1.0 mol m⫺2 yr⫺1, and bottom export of 6.7 mol m⫺2 yr⫺1. In the abiotic case the air-sea flux reverses to ingas-

31,207

Figure 12. Comparison between model and observed pCO2 (in situ and normalized to 10⬚C) for the period 1973–1978.

sing of 1.7 mol m⫺2 yr⫺1. The difference between the biotic and abiotic cases is greatest in summer, with a maximum of 3.3 mmol m⫺3 in July, reflecting the seasonal cycle of N P . This summer maximum in the N P was hypothesized by Wong and Chan [1991] to explain the absence of a strong seasonal supersaturation of pCO2. The oxygen supersaturation in the summer months is 1.9 –2.8 times the value expected from thermal forcing alone. This is consistent with the oxygen supersaturations determined by Emerson et al. [1993], which range from 1.4 to 2.6 times the values for argon (an inert tracer of abiotic effects on oxygen). Iron controls primary production via eolian deposition and bottom flux (350 m). Our model results show that ⬃40% of the 25 ␮mol m⫺2 yr⫺1 of iron consumed by biological uptake originates from eolian deposition, while the remaining ⬃60% comes from bottom flux via mixing and upwelling. The seasonal variability of ⌬pCO2 and air-sea CO2 flux are represented in Figure 6, previously discussed in section 3. We focus our conclusions on the results of run 3 (dotted lines), which uses the gas transfer coefficient [Wanninkhof and McGillis, 1999] that provided the best agreement with observed pCO2 variations. The influence of SST and wind intensity on the seasonal cycle of ⌬pCO2 and air-sea carbon flux is evident.

31,208


Figure 13. Seasonal variability of (top to bottom) vertical velocity, vertical eddy diffusivity, temperature, and downwelling irradiance simulated by the model. Dashed lines in top panel correspond to negative values of vertical velocity (contour levels are ⫺0.5 cm d⫺1). There is ingassing of CO2 during the fall–winter months when SSTs are colder and wind forcing is vigorous, while there is a much smaller ingassing of CO2 during the summer when SSTs are warmer and wind speeds are reduced.

7.

Interannual Variability

The interannual variability of the major parameters simulated by the model is illustrated in the profile time series shown in Plates 1 and 2 and in the surface time series shown in Figure 17. The temperature profile shows distinct warm and cold periods. Two warm periods occur in the series, one during 1960 –1965 and another during 1976 –1980. The period 1966 – 1975 exhibits colder (⬃0.8⬚C) temperatures in the top 150 m. The oxygen profile series shows higher (⬃5 ␮mol kg⫺1) concentrations during the cold period because of increased solubility. The phytoplankton concentration in the upper 20 m was higher (⬃0.1 mg m⫺3) during the warmest temperature events (1961, 1963, and 1979). The composite time series of ⌬pCO2 and apparent oxygen utilization (AOU) (top panel in Figure 17) demonstrates the combined influence of thermodynamics and biological uptake on the ⌬pCO2 variability. The two effects have a tendency to compensate each other. The peak net

Figure 14. Seasonal variability of (top to bottom) nitrate, ammonium, phytoplankton, and zooplankton concentrations simulated by the model.

community production drawdown in summer, shown by a positive AOU maximum, compensates for the increased temperature effect on pCO2 and prevents large outgassing of CO2 in summer. The highest drawdown in the iron, nitrate, and carbon dioxide concentrations occurred during 1961–1963, a period of warmest euphotic zone temperatures during which solubility was lowest and phytoplankton growth was highest. Time series of wind speed, gas transfer coefficient, and seaair flux are shown in Figure 18. The wind speed exhibits a distinct seasonal cycle which is reflected in the seasonal variability of the gas transfer coefficient and consequently, in the sea-air flux. The winds are weakest in July (7.3 ⫾ 3.4 m s⫺1) and strongest in November (12.3 ⫾ 5.1 m s⫺1). Figure 19 shows the observed and predicted SST, the depth-averaged (0 –100 m) temperature and TCO2 variability, and the surface CO2 flux and oceanic pCO2 variability for the 1960 –1980 yearly averages. The interannual changes and trends are highlighted in Figure 19. Note the cooling trend in the temperature during the 1960s, followed by a warming trend during the middle to late 1970s. The sea-air CO2 flux varies from ⫺1.2 to ⫺2.3 mol C m⫺2 yr⫺1, which corresponds to interannual changes in SST and winds. The pCO2 trends upward almost monotonically, showing an increase of ⬃30 ␮atm in 20 years (⬃1.5 ␮atm yr⫺1). This linear trend in the ocean pCO2 follows approxi-


31,209

increase, which is gradually reduced with depth but still noticeable at depths of up to 250 m, is a response to the secular increase of atmospheric pCO2 [Winn et al., 1998; Bates, 2001].

8.

Carbon Flux Budget

Figure 20 represents the mean ecosystem carbon flux balance for 1960 –1980. The main compartments in this flowchart are the phytoplankton, zooplankton, respired DOC, and total CO2 carbon stocks. The numbers indicate the carbon flux between ecosystem components in mol C m⫺2 yr⫺1. The ammonium and nitrate boxes were replaced by the net community production (6.9 mol C m⫺2 yr⫺1), which corresponds to all sources and sinks of macronutrients combined (equation (14)). Note that the bottom carbon flux (4.3 mol C m⫺2 yr⫺1) and air-sea flux (1.8 mol C m⫺2 yr⫺1) into the model domain are balanced by the particulate organic carbon plus DOC export (5.3 mol C m⫺2 yr⫺1) and the fecal pellet loss (0.8 mol C m⫺2 yr⫺1) through the bottom (sedimentation). The total gross production (uptake) is 167 g C m⫺2 yr⫺1 (13.9 mol C m⫺2 yr⫺1). About half of the total carbon required for photosynthesis comes from recycling within the plankton community, and about half (6.9 mol C m⫺2 yr⫺1) comes from the TCO2 pool (Figure 20). Most of the supply of TCO2 comes from the flux across the lower model boundary (4.3 mol C m⫺2 yr⫺1), with smaller fractions from air-sea flux (1.8 mol C m⫺2 yr⫺1) and DOC remineralization (0.8 mol C m⫺2 yr⫺1).

9.

Figure 15. Seasonal variability of (top to bottom) temperature, total carbon dioxide, oxygen, and iron concentrations simulated by the model. mately the 1.61 ␮atm yr⫺1 trend in the atmospheric pCO2 (see (26)), indicating that the atmosphere and ocean are tightly coupled at OWS P. The upward trend in the ocean pCO2 is manifested in the profile series of total carbon dioxide concentration in Plate 1. The total carbon dioxide increases within the upper 100 m at a rate of ⬃15 mmol m⫺3 in 20 years. This

Figure 16. Seasonal variability of surface oxygen anomaly predicted by the model. Results from two runs are presented: an abiotic run (dotted line) and a biotic run (solid line). Seasonal averages are based on the 1973–1978 time period.

Summary and Conclusions

A coupled ecosystem/carbon flux model was developed and used to simulate biogeochemical parameters at OWS P. A series of sensitivity runs revealed that the most significant improvement in the simulation of the surface pCO2 (reduced mean bias between model and observations) is obtained by the use of a recently developed cubic formulation for the gas transfer coefficient [Wanninkhof and McGillis, 1999]. This result shows that the ocean and atmosphere are much more closely coupled at high (⬎10 m s⫺1) wind speeds than is indicated by the previous formulations [Liss and Merlivat, 1986; Tans et al., 1990; Wanninkhof, 1992]. All biogeochemical parameters, when averaged over the period of available concurrent observations (1973–1978), are within 5% of the observed values. The only exception is the sea-air CO2 flux, which can depart more significantly from previously reported values, depending on which K formulation is used. For example, for the 1973–1978 time period our CO2 flux estimate is ⫺1.8 mol C m⫺2 yr⫺1 when we use the Liss and Merlivat [1986] gas transfer coefficient. Wong and Chan [1991], also using the Liss and Merlivat [1986] gas transfer coefficient, report a value of ⫺0.7 mol C m⫺2 yr⫺1. The reason for the factor of ⬃3 increase between the flux estimate of Wong and Chan [1991] and the model estimate using the same gas transfer coefficient is that with the Liss and Merlivat [1986] algorithm the model predicts a much larger ⌬pCO2 (⫺30.2 ␮atm) than the observed value (⫺15.5 ␮atm). The best agreement between model (⫺15.8 ␮atm) and observed (⫺15.5 ␮atm) mean (1973–1978) ⌬pCO2 is achieved with the Wanninkhof and McGillis [1999] gas transfer coefficient. Our CO2 flux value is ⫺1.9 mol C m⫺2 yr⫺1 when we use the Wanninkhof and McGillis [1999] gas transfer coefficient. Thus the apparent difference in ⌬pCO2 in the simulation using the Liss and Merlivat [1986] gas transfer coefficient is a result of the weaker air-sea

31,210


Plate 1. Interannual variability of temperature, vertical eddy diffusivity, total carbon dioxide, and oxygen simulated by the model.


Plate 2. Interannual variability of phytoplankton, iron, nitrate, and ammonium simulated by the model.

31,211

31,212


Figure 17. Time series of simulated ⌬pCO2 and apparent oxygen utilization (AOU), SST and mixed layer depth, total carbon dioxide, and iron, nitrate, and ammonium. Two vertical lines mark the months of March and September to facilitate the seasonal comparison among the plotted variables. coupling using this K formulation, which feeds back into a smaller ocean pCO2 value. Global estimates of air-sea CO2 flux derived using air-sea flux models to extrapolate from observed pCO2 are limited because they are obtained from a relatively small number of discrete data points. In the short term, fully prognostic models may give the best estimates, provided that they achieve sufficient agreement with other observable quantities such as temperature, salinity, alkalinity, chlorophyll, and TCO2. Ultimately, the role of the ocean in the global carbon cycle balance will only be resolved by long-term measurements of TCO2. Given the large size of the ocean pools and the fact that highly accurate methods have only been available for a short time, this will take a systematic approach for global observations of key parameters. The Wong and Chan [1991] hypothesis that biological production maintains the near-equilibrium relationship between air and sea pCO2 at OWS P in summer is supported by our model results. Significant interannual variability in the subarctic Pacific has been observed in recent years [Freeland et al., 1997; Whitney et al., 1998]. While we make no claim that our model captures all of the mechanisms responsible for this variability, we believe

Figure 18. Time series of wind speed, gas transfer coefficient, and sea-air flux.


Figure 19. Yearly averaged time series of SST, upper-ocean temperature (0 –100 m mean), sea-air carbon dioxide flux, surface pCO2, and TCO2 (0 –100 m mean) simulated by the model for the period 1960 –1980. Black dots superimposed on the SST time series are observed values. The dashed line in the TCO2 tier shows the upward trend of ⬃15 mmol m⫺3/20 years.

Figure 20. Flowchart of simulated climatological carbon flux balance showing the principal carbon exchange compartments in the coupled ecosystem/carbon flux model.

31,213

31,214


that it is a significant step forward relative to earlier approaches. Our model is the first model of OWS P that couples the carbon chemistry and air-sea exchange of CO2 both to mixed layer physics and to a fully prognostic biological model, whereas both Archer et al. [1993] and Antoine and Morel [1995b] specified important aspects of the biological component from historical observations. While Antoine and Morel [1995b] based their model on satellite observations of surface ocean chlorophyll concentration that were expected to be available in near real time (which has since been realized), they specified the f ratio from seasonal climatologies of field observations. This is the critical parameter that relates photosynthesis estimated from satellites to net community production and therefore to carbon biogeochemistry. The depth-averaged (0 – 150 m) f ratio predicted by the model averages 0.48 ⫾ 0.07 for the period 1960 –1980. It has a distinctive seasonal cycle with a minimum of 0.39 ⫾ 0.02 in September and a maximum of 0.60 ⫾ 0.05 in February. Our model specifies nothing except the surface forcing and the lower boundary conditions, which change slowly. Our model simulations show that the OWS P site was a sink of CO2 (⫺1.8 mol C m⫺2 yr⫺1) for the period of simulation (1960 –1980), with interannual variability of ⫺1.2 to ⫺2.3 mol C m⫺2 yr⫺1 mostly attributable to changes in SST and winds. Our air-sea CO2 flux estimate for the period 1973– 1978, during which most of the data were available, is ⫺1.9 mol C m⫺2 yr⫺1, indicating that the OWS P site was a sink of CO2 for that time period. Antoine and Morel [1995b] report a CO2 flux of ⫺0.6 mol C m⫺2 yr⫺1 for the period 1975–1977; our estimate for that same time period is ⫺1.9 mol C m⫺2 yr⫺1 (same as the 1973–1978 flux estimate). We attribute this difference to our explicit modeling of the biology, the different gas transfer algorithm, and the more realistic TCO2 concentration value for the bottom boundary condition in our model. While there are important aspects of interannual variability that are not locally forced, our model can capture and isolate those aspects that are driven by local atmospheric forcing, including those resulting from atmospheric teleconnections with oceanographic processes elsewhere in the Pacific [Trenberth and Hurrell, 1994; Zhang et al., 1999]. Acknowledgments. We acknowledge the support provided by NASA’s Ocean Biogeochemistry Program (NASA RTOP 622-51-30). We are thankful to Paulette Murphy for enlightening discussions regarding the chemical thermodynamics aspects of this work. Finally, we would like to acknowledge Ina Tegen for the atmospheric dust model results and acknowledge Moss Landing Marine Labs for the iron data.

References Anderson, L. A., and J. L. Sarmiento, Redfield ratios of remineralization determined by nutrient data analysis, Global Biogeochem. Cycles, 8, 65– 80, 1994. Antoine, D., and A. Morel, Modelling the seasonal course of the upper ocean pCO2, I, Development of a one-dimensional model, Tellus, Ser. B, 47, 103–121, 1995a. Antoine, D., and A. Morel, Modelling the seasonal course of the upper ocean pCO2, II, Validation of the model and sensitivity studies, Tellus, Ser. B, 47, 122–144, 1995b. Archer, D., S. Emerson, T. Powell, and C. S. Wong, Numerical hindcasting of sea surface pCO2 at Weathership OWS P, Prog. Oceanogr., 32, 319 –351, 1993. Bates, N. R., Interannual variability of oceanic CO2 and biogeochemical properties in the Western North Atlantic subtropical gyre, Deep Sea Res., Part II, 48, 1507–1528, 2001. Boyd, P. W., D. L. Muggli, D. E. Varela, R. H. Goldblatt, R. Chretien, K. J. Orians, and P. J. Harrison, In vitro iron enrichment experi-

ments in the NE subarctic Pacific, Mar. Ecol. Prog. Ser., 136, 179 – 193, 1996. Broecker, W. S., and T.-H. Peng, Gas exchange rates between air and sea, Tellus, 26, 21–35, 1974. Conkright, M., S. Levitus, T. O’Brien, T. Boyer, J. Antonov, and C. Stephens, World Ocean Atlas 1998 CD-ROM Data Set Documentation, Tech. Rep. 15, 16 pp., National Oceanographic Data Center, Silver Spring, Maryland, 1998. Dickson, A. G., Thermodynamics of the dissociation of boric acid in synthetic sea water from 273.15 to 318.15 K, Deep Sea Res., Part A, 37, 755–766, 1990. Dickson, A. G., and F. J. Millero, A comparison of the equilibrium constants for the dissociation of carbonic acid in sea water media, Deep Sea Res., Part A, 34, 1733–1743, 1987. Dobson, F. W., and S. D. Smith, Bulk models of solar radiation at sea, Q. J. R. Meteorol. Soc., 114, 165–182, 1988. Duce, R. A., Sources, distributions, and fluxes of mineral aerosols and their relationship to climate, in Aerosol Forcing of Climate, edited by R. J. Charlson and J. Heintzenberg, pp. 43–72, John Wiley, New York, 1995. Emerson, S., P. Quay, and P. A. Wheeler, Biological productivity determined from oxygen mass balance and incubation experiments, Deep Sea Res., Part I, 40, 2351–2358, 1993. Freeland, H., K. Denman, C. S. Wong, F. Whitney, and R. Jacques, Evidence of change in the winter mixed layer in the northeast Pacific Ocean, Deep Sea Res., Part I, 44, 2117–2129, 1997. Frost, B. W., A modeling study of the processes regulating plankton standing stock and production in the open subarctic Pacific Ocean, Prog. Oceanogr., 32, 101–135, 1993. Garwood, R. W., An oceanic mixed-layer model capable of simulating cyclic states, J. Phys. Oceanogr., 7, 455– 471, 1977. Gordon, L. I., P. K. Park, S. W. Hager, and T. R. Parsons, Carbon dioxide partial pressures in North Pacific surface waters—time variations, J. Oceanogr. Soc. Jpn., 27, 81–90, 1971. Gordon, L. I., P. K. Park, J. J. Kelley, and D. W. Hood, Carbon dioxide partial pressures in North Pacific surface waters, 2, General late summer distributions, Mar. Chem., 1, 191–198, 1973. Ivlev, V. S., Experimental Ecology of the Feeding of Fishes, 302 pp., Pischepromazat, Moscow, 1955. Keeling, C. D., Carbon dioxide in surface ocean waters, 4, Global distribution, J. Geophys. Res., 73, 4543– 4554, 1968. Liss, P. S., Tracers of air-sea gas exchange, Philos. Trans. R. Soc. London, Ser. A, 325, 93–103, 1988. Liss, P. S., and L. Merlivat, Air-sea exchange rates: Introduction and synthesis, in The Role of Air-Sea Exchange in Geochemical Cycling, NATO Sci. Ser., Ser. C, vol. 185, edited by P. Buat-Me´nard, pp. 113–128, NATO, Brussels, 1986. Maldonado, M. T., P. W. Boyd, P. J. Harrison, and N. M. Price, Co-limitation of phytoplankton growth by light and Fe during winter in the NE subarctic Pacific Ocean, Deep Sea Res., Part II, 46, 2475– 2485, 1999. Martin, J. H., S. E. Fitzwater, W. W. Broenkow, VERTEX: Phytoplankton/iron studies in the Gulf of Alaska, Deep Sea Res., Part A, 36, 645– 680, 1989. McClain, C. R., K. Arrigo, K.-S. Tai, and D. Turk, Observations and simulations of physical and biological processes at ocean weather station P, 1951–1980, J. Geophys. Res., 101, 3697–3713, 1996. McClain, C. R., S. R. Signorini, K.-S. Tai, K. Arrigo, and R. Murtugudde, An ecosystem model for the simulation of physical and biological oceanic processes—IDAPAK User’s guide and applications, NASA Tech. Memo, TM-1998-206856, 1998. McClain, C. R., R. Murtuggude, and S. R. Signorini, A simulation of biological processes in the equatorial Pacific Warm Pool at 165⬚E, J. Geophys. Res., 104, 18,305–18,322, 1999. Mehrbach, C., C. H. Culberson, J. E. Hawley, and R. M. Pytkowicz, Measurement of apparent dissociation constants of carbonic acid in seawater at atmospheric pressure, Limnol. Oceanogr., 18, 897–907, 1973. Murphy, P. P., D. E. Harrison, R. A. Feely, T. Takahashi, R. F. Weiss, and R. H. Gammon, Variability of ⌬pCO2 in the subarctic North Pacific: A comparison of results from four expeditions, Tellus, Ser. B, 50, 185–204, 1998. Peng, T.-H., T. Takahashi, and W. S. Broecker, Surface radon measurements in the North Pacific Ocean OWS P, J. Geophys. Res., 79, 1772–1780, 1974. Peng, T.-H., T. Takahashi, and W. S. Broecker, Seasonal variability of

SIGNORINI ET AL.: OCEAN WEATHER STATION P CARBON BUDGET carbon dioxide, nutrients and oxygen in the northern North Atlantic surface water: Observations and a model, Tellus, Ser. B, 39, 439 – 458, 1987. Pond, S., and G. L. Pickard, Introductory Dynamic Oceanography, 241 pp., Pergamon, New York, 1978. Signorini, S. R., C. R. McClain, J. R. Christian, and C. S. Wong, One-dimensional coupled ecosystem-carbon flux model for the simulation of biogeochemical parameters at Ocean Weather Station P, NASA Tech. Memo, TP-2000-209892, 2000. Sundquist, E. T., Geological perspectives on carbon dioxide and carbon cycle, in The Carbon Cycle and Atmospheric CO2: Natural Variations Archean to Present, Geophys. Monogr. Ser., vol. 32, edited by E. T. Sundquist and W. S. Broecker, pp. 5–59, AGU, Washington, D. C., 1985. Takahashi, T., J. Olafsson, J. G. Goddard, D. W. Chipman, and S. C. Sutherland, Seasonal variations of CO2 and nutrients in the highlatitude surface oceans: A comparative study, Global Biogeochem. Cycles, 7, 843– 878, 1993. Takahashi, T., R. A. Feely, R. F. Weiss, R. H. Wanninkhof, D. W. Chipman, S. C. Sutherland, and T. T. Takahashi, Global air-sea flux of CO2: An estimate based on measurements of sea-air pCO2 difference, Proc. Natl. Acad. Sci. U.S.A., 94, 8292– 8299, 1997. Tans, P. P., I. Y. Fung, and T. Takahashi, Observational constraints on the global atmospheric CO2 budget, Science, 247, 1431–1438, 1990. Taylor, A. H., A. J. Watson, M. Ainsworth, J. E. Robertson, and D. R. Turner, A modeling investigation of the role of phytoplankton in the balance of carbon at the surface of the North Atlantic, Global Biogeochem. Cycles, 5, 151–171, 1991. Tegen, I., and I. Fung, Modeling of mineral dust in the atmosphere: Sources, transport, and optical thickness, J. Geophys. Res., 99, 22,897–22,914, 1994. Trenberth, K. E., and J. W. Hurrell, Decadal atmosphere-ocean variations in the Pacific, Clim. Dyn., 9, 303–319, 1994. Wanninkhof, R., Relationship between wind speed and gas exchange over the ocean, J. Geophys. Res., 97, 7373–7382, 1992. Wanninkhof, R., and W. R. McGillis, A cubic relationship between

31,215

air-sea exchange and wind speed, Geophys. Res. Lett., 26, 1889 –1892, 1999. Weiss, R. F., The solubility of nitrogen, oxygen and argon in water and seawater, Deep Sea Res., 17, 721–735, 1970. Weiss, R. F., Carbon dioxide in water and seawater: The solubility of a non-ideal gas, Mar. Chem., 2, 203–215, 1974. Whitney, F. A., C. S. Wong, and P. W. Boyd, Interannual variability in nitrate supply to surface waters of the northeast Pacific Ocean, Mar. Ecol. Prog. Ser., 170, 15–23, 1998. Winn, C. D., Y. H. Lee, F. T. Mackenzie, and D. M. Karl, Rising surface ocean dissolved inorganic carbon at the Hawaii Ocean Timeseries site, Mar. Chem., 60, 33– 47, 1998. Wong, C. S., and Y.-H. Chan, Temporal variations in the partial pressure and flux of CO2 at ocean station P in the subarctic northeast Pacific Ocean, Tellus, Ser. B, 43, 206 –223, 1991. Zhang, R.-H., L. M. Rothstein, and A. J. Busalacchi, Interannual and decadal variability of the subsurface thermal structure in the Pacific Ocean: 1961–90, Clim. Dyn., 15, 703–717, 1999. J. R. Christian, Earth System Science Interdisciplinary Center, University of Maryland, College Park, MD 20742, USA. (jrc@barolo. essic.umd.edu) C. R. McClain, NASA Goddard Space Flight Center, SeaWiFS Project Office, Laboratory for Hydrospheric Processes, Code 970.2, Greenbelt, MD 20771, USA. ([email protected]) S. R. Signorini, NASA Goddard Space Flight Center, Oceans and Ice Branch, SeaWiFS Project, Code 970.2, Building 28, Greenbelt, MD 20771, USA. ([email protected]) C. S. Wong, Climate Chemistry Laboratory, Institute of Ocean Sciences, Fisheries and Oceans Canada, P.O. Box 6000, 9860 West Saanich Road, Sidney, B.C. V8L 4B2, Canada. ([email protected]) (Received March 30, 2000; revised July 18, 2001; accepted July 20, 2001.)

31,216