Methods for the Measurement of pCO2 and TCO2 in Seawater for
JGOFS Southern Ocean Process Cruises 2 and 4
R.V. Nathaniel B. Palmer Cruises 97/1 and 97/8
Taro Takahashi, David Chipman, Stephany Rubin,
Colm Sweeney, and S.C. Sutherland
Lamont-Doherty Earth Observatory (CO2 Group)
61 U.S. Route 9W
P.O. Box 1000
Palisades, NY 10964-8000
U.S.A.
TOTAL CO2 in Seawater:
Except for the way we introduced the sample, the coulometric analysis
system used to measure the total CO2 (TCO2) concentration in seawater samples
during these cruises is the same as described in Chipman, D. W., Marra, J. and
Takahashi, T. (Primary production at 47°N and 20°W in the North Atlantic Ocean:
A comparison between the 14C incubation method and the mixed layer carbon
budget. Deep-Sea. Res., 40 , 151-169, 1993). The system and procedures used
are summarized below.
This system consists of a Model 5011 coulometer, manufactured by UIC
Inc., Jolliet, IL, and a sample introduction/CO2 extraction system of a Lamont-
Doherty Earth Obs. (CO2 Group) design. It differs from the Single Operator
Multiparameter Metabolic Analyzer (SOMMA) system used by other participants
of the DOE/CO2 program. In the LDEO system, a precisely known volume of
seawater sample is introduced manually into a CO2 extraction vessel using a
calibrated syringe instead of the automated pipette used by the SOMMA system.
The syringe is a hand-ground Pyrex glass medical syringe with two firm
reference stops which allow the quantitative sampling of seawater. Our
experience with syringe and pipette methods is that accumulation of coatings
formed on the glass surfaces from the repeated filling and emptying with
seawater will not significantly affect the volume contained of either, but that it will
affect the volume delivered by the passively drained pipette in the SOMMA
system. The positive displacement of the plunger of a syringe will keep the
delivered volume constant, even in the presence of surface coatings.
Additionally, since the water sample in the syringe has no air space, changes in
TCO2 due to gas exchange with the air in head space are eliminated.
Samples for TCO2 analysis were drawn from the Niskin bottles of the
rosette casts directly into 250 ml glass reagent bottles with ground standard-
taper stoppers, sealed with silicone vacuum grease and pressed in using two
strong rubber bands. Immediately after sample collection, 200 ul of 50%
saturated mercuric chloride solution was added to prevent biological alteration of
the TCO2. A small head space (~5 ml) was left in the bottle to prevent thermal
expansion of the water from causing a leak or breaking the bottle. Samples were
normally analyzed within 24 hours of collection. For analysis, a water sample
was sucked into a syringe, and a calibrated volume (19-20 ml) of water sample
was introduced into a CO2 extraction chamber through a rubber septum. The
mass of the seawater sample delivered was determined from the density of
seawater, calculated using the measured salinity, the temperature at the time of
injection, and the International Equation of State of Seawater (Millero, et al.,
1980). Prior to the expedition, the volume of each sampling syringe between two
reference stops was determined by repeatedly weighing aliquots of distilled,
deionized water dispensed. The measurements were corrected for the buoyancy
of air displaced by the water, amounting to about 0.1% of the weight of the water.
The volume was then computed using the density of pure water at the
temperature of the measurement. Repeated measurements gave a precision of
±0.03% or better.
The seawater sample in the extraction vessel was acidified with ~1 ml of
8.5% phosphoric acid introduced through a sidearm of the extraction chamber.
The evolved CO2 was stripped from the sample and transferred into the
electrochemical cell of the CO2 coulometer by a stream of CO2-free air. In the
coulometer cell, the CO2 was quantitatively absorbed by a solution of
ethanolamine in dimethylsulfoxide (DMSO). Reaction between the CO2 and the
ethanolamine formed the weak hydroxyethylcarbamic acid. The pH change of
the solution associated with the formation of this acid resulted in a color change
of the thymolphthalein pH indicator in the solution. The color change, from deep
blue to colorless, was detected by a photodiode which continuously monitored
the transmissivity of the solution. The electronic circuitry of the coulometer, in
detecting the change in the color of the pH indicator, caused a electrical current
to flow through the cell, generating hydroxyl (OH-) ions from a small amount of
water in the solution. The OH- generated then titrated the acid, returning the
solution to its original pH and color, at which point the current flow was stopped.
The product of current passed through the cell and time was related by the
Faraday constant to the number of moles of OH- generated, and hence to the
number of moles of CO2 absorbed to form the acid. A thermostated, double
walled titration cell was used during titration, to eliminate the shifting of the
endpoint of the titration due to a change in the temperature of the cell solutions.
The coulometer was calibrated using research grade CO2 gas (99.998%
pure) introduced into the carrier gas line upstream of the extraction chamber
alternately using two fixed-volume sample loops on a gas sampling valve. The
loops were vented to the atmosphere, and the ambient atmospheric pressure in
the laboratory was measured using a high precision electronic barometer with an
accuracy of better than 0.05%. The loop temperatures were measured to
±0.05°C with a thermometer calibrated against one traceable to the NIST. The
non-ideality of CO2 was incorporated into the computation of the loop contents.
Prior to the expeditions, the volumes of the loops were determined by the
difference in weight between the loop injection valve assembly when empty and
filled with water. Repeated measurements gave a precision of ±0.02%. During
the expedition the coulometer was calibrated several times a day using this gas
sampling system.
The calibration factor, which represents the ratio between the number of
moles of CO2 in the loop and the reading of the coulometer, changes during the
use of a titration cell. Depending on the condition of the solution in the titration
cell, this factor varies around the ideal ratio of unity by a few tenths of a percent.
It commonly starts from less than unity when the cell solution is new and
increases to greater than unity as increasing amounts of carbon are titrated. This
change can be represented by a quadratic equation relating values of calibration
factor with the total amount of carbon titrated in a given cell. The CO2
concentration in each seawater sample was corrected using a factor estimated
from the equation fit to the calibration data for each cell. Generally a cell had to
be cleaned and filled with fresh solution after about 40 samples. After this
number the cell began to behave erratically with unreliable analytical results.
Analyses of Certified Reference Solutions:
For the purpose of quality control of total CO2 determinations, SIO
Reference Solutions Batch Numbers 33 and 34 were run through our analytical
system at sea as unknowns. Our shipboard analyses compare with the SIO
manometric values as follows:
SIO Manometric LDEO Coulometric (SIO-LDEO)
Process CRM TCO2 TCO2 at sea Difference
Cruise Batch (umol/kg) (umol/kg) (umol/kg)
2 (NBP 97/1) 33 2009.85±0.85 2012.75±1.29 -2.90
(N=11) (N=150)
4 (NBP 97/8) 34 2061.52±1.62 2061.47±1.40 +0.05
(N=15) (N=90)
N is the number of analyses.
The two attached figures show the results graphically.
Figure for 97/1
Figure for 97/8
While the mean values for Batch #34 agree well within one standard deviation for
each set of measurements, those for Batch #33 agree only within two standard
deviations. Since we do not understand the source(s) of this discrepancy, we are
reporting both observed and CRM-adjusted values in this data set.
Determination of pCO2 in Discrete Seawater Samples:
A fully automated equilbrator-gas chromatograph (g.c.) system was used
during the expedition for the determination of partial pressure of CO2 exerted by
the seawater samples. Its design has been described by Chipman, D. W., Marra,
J. and Takahashi, T. (Primary production at 47°N and 20°W in the North Atlantic
Ocean: A comparison between the 14C incubation method and the mixed layer
carbon budget. Deep-Sea. Res., 40 , 151-169, 1993) and is summarized below.
The system consisted of a pair of circulation pumps plumbed to recirculate
air in a closed system through porous plastic gas dispersers immersed in two
separate seawater samples. Electrically driven Valco 4-port valves were used to
isolate each of the equilibrators during the initial equilibration. Manually operated
2-way and 3-way Whitey valves allowed part of the water in each equilibrator to
be replaced with air of known initial CO2 concentration, creating the necessary
headspace for equilibration. A drain line in each equilibrator insured that the
volume of water was constant at 500 ml. This allowed us to make accurate
corrections for the effect of the perturbation of the sample by the headspace air.
The equilibrators were open to the laboratory air through isolation coils, keeping
them at atmospheric pressure, and we measured lab atmospheric pressure with
a high precision electronic barometer with an accuracy of better than 0.05%.
Since the partial pressure of CO2 is strongly affected by temperature, the
equilibration flasks were kept immersed in a water bath maintained at
4.00±0.01°C. An electrically driven Valco 6-port valve allowed the entire
equilibration system to be isolated, while simultaneously connecting a calibration
gas selection valve. A 2-way normally-closed Skinner solenoid valve on the
output of the of the calibration selection valve allowed the gas flow to be
controlled by the system controller. It also provided a necessary second means
of stopping the flow of the calibration gases to prevent accidental loss of the
calibration gases in case of a control malfunction.
The analysis of the CO2 in the equilibrated air or calibration gases was
performed using a Shimadzu Mini-2 gas chromatograph, equipped with a flame
ionization detector. A 1 ml sample loop, a pre-column, and an analytical column
were attached to an electrically driven Valco 10-port valve within the column
oven of the gas chromatograph. The pre-column, 0.2 m length, and analytical
column, 2.0 m length, were both packed with Chromosorb 102. Ultra-high purity
hydrogen gas served as the carrier gas. This carrier gas is electrolytically
generated by a hydrogen generator and purified by diffusion through a palladium
foil membrane. The use of hydrogen as a carrier gas allowed the CO2 to be
converted quantitatively to methane in a catalytic converter prior to quantification
by the flame ionization detector. Our system used a catalyst of ruthenium metal
on Chromosorb W support, and did not require a palladium pre-catalyst to
remove oxygen from the carrier gas stream. Hydrocarbon-free air to support the
combustion in the flame ionization detector was provided by an Aadco Model 737
chromatographic air purifier. A Shimadzu Model CR6A Chromatopac computing
integrator controlled the equilibration and calibrations procedure, and integrated
the output signal from the gas chromatograph.
The analytical procedure follows. Water samples were drawn from the 10-
liter Niskin bottles of a rosette directly into 500 ml narrow-necked volumetric
Pyrex flasks. These served as both sample containers and equilibration vessels.
The samples were immediately inoculated with 200 ul of 50% saturated mercuric
chloride solution, sealed airtight with screw caps with conical plastic liners to
prevent biological modification of the pCO2, and stored in the dark until
measurement. Measurements were normally performed within 24 hours of
sampling. A headspace of 3 to 5 ml was left above the water to allow for thermal
expansion during storage. Prior to analysis the sample flasks were brought to
the water bath temperature of 4.00°C in the constant temperature bath, and
about 45 ml of water was displaced with air of known CO2 concentration. The air
in the flasks and in the tubing connecting them to the g.c. sampling loop was
recirculated continuously for about 20 minutes through a gas disperser immersed
in the water. This provided a large surface area for gas exchange, and
equilibrium for CO2 was attained in 15 minutes.
The equilibrated air samples were saturated with water vapor at the
temperature of equilibration and had the same pCO2 as the water. By injecting
the air aliquot into the g.c. without removal of the water vapor, the partial
pressure of CO2 is directly determined by the relationship below:
pCO2 (uatm) = [Cmeas (ppm)] X [total pressure of equilibration (atm)]
where Cmeas is the mole fraction concentration of CO2 in equilibrated moist air.
The total pressure of equilibrated air was measured by having the head space in
the equilibrator flask always at atmospheric pressure. We measured the latter
with an electronic barometer at the time each equilibrated air sample was
injected into the g.c. for CO2 determination. Since water vapor was NOT
removed from the sample, we do not have to know the water vapor pressure.
Cmeas was determined by using a quadratic equation fit to all 3 of the standards.
Although a quadratic equation was used, the CO2 concentration was very close
to a linear function of the g.c. detector output signal.
The gas chromatograph was calibrated using gas mixtures of CO2 in air, which
have been calibrated against a set of primary standard gas mixtures in our
laboratory. The primary standard gases used are traceable to the WMO
reference scale through analysis in the laboratory of C. D. Keeling of SIO (in the
case of the lower two of our secondary gas mixtures) or to a primary standard of
CO2 in nitrogen, which was calibrated in our laboratory using gravimetrically-
prepared sodium carbonate solutions (highest gas concentration only). The
values of the secondary standard gas mixtures used during the Process 2 and 4
cruises and the method used to determine their CO2 concentrations are as
follows:
233.8 ppm CO2, measured by infra-red analyzer,
555.3 ppm CO2, measured by g.c., and
856.5 ppm CO2, mean of measurements by g.c. and coulometry (0.14%
difference between those two determinations).
Corrections were made to account for the change in pCO2 of the sample
water due to the transfer of CO2 between the water and circulating air during
equilibration. We know the pCO2 in equilibrated, perturbed water and the total
CO2 (TCO2) by coulometry before the equilibration. We can also calculate the
change in TCO2 in the water based on the change in pCO2 between the post-
equilibrium value and the known concentration in the pre-equilibrium
displacement gas. With the pre-equilibrium TCO2 plus the perturbation in TCO2
during equilibration, we obtain the post-equilibrium TCO2 value. Using the post-
equilibrium TCO2 and measured pCO2 values, we calculate total alkalinity (TALK)
at the end of the equilibration, knowing the temperature, salinity, phosphate, and
silicate data. Since the perturbation does NOT change the TALK, we calculate
the pre-equilibrium pCO2 from the pre-equilibrium TCO2, the calculated TALK,
and the other parameters of temperature, salinity, etc. This is the value that we
report as pCO2, the pre-equilibrium calculated value. The magnitude of this
correction is generally less than 2 uatm.
The pCO2 values reported in this data set are in the unit of
microatmospheres at the temperature of 4.00°C at which they were measured.
In order to obtain pCO2 values at other temperatures, we recommend using the
following temperature effect determined by Takahashi, et al. (Seasonal variations
of CO2 and nutrients in the high latitude surface oceans: A comparative study.
Global Biogeochemical Cycles, 7, 843-878, 1993).
dln(pCO2)/dt = (0.0433 - 8.7 x 10-5 t) (C^-1)
where t = temperature in degrees Celsius
This relationship is independent of salinity within the normal open ocean range.
We estimate the precision of the pCO2 measurement for a single
hydrographic station to be about ±0.15% based on the reproducibility of replicate
equilibrations. We estimate the station-to-station reproducibility to be about
±0.5%.