Wat. Res. Vol.29, No. 5, pp.

1231-1234,1995
ElsevierScienceLtd. Printedin Great Britain
Pergamon 0043-1354(94)00270-3

AUTOMATIC METHOD FOR THE DETERMINATION OF

T O T A L CO 2 IN N A T U R A L W A T E R S

M. I. ABDULLAH* and E. EEK

Marine Chemistry Section, Biology Institute, University of Oslo, PB 1064 Blindern, Oslo, Norway

(First received April 1994; accepted in revised form October 1994)

Abstract--An automatic method is described for the determination of total CO2 in sea, estuarine and fresh
waters based on infrared detection. The method allows for the determination of total CO2 in the range
of 10/zM-3mM CO2 and avoids errors inherent in alternative methods based on total alkalinity
measurements.

Key words---carbon dioxide, seawater, freshwater, automatic method

INTRODUCTION (Johnson et al., 1985; Chipman et al., 1993) methods,

which are precise but elaborate and time consuming.
Accurate knowledge of carbon dioxide content in The method described here is based on the acidifica-
natural waters is important to the study of global tion of the sample, purging the liberated C02
carbon cycle as well as aquatic biochemical processes. and analysis by IR COz analyser. Such a method
The most common method of estimating total COs is has recently been tested (Goyet and Snover, 1993)
based on calculation from the carbonate alkalinity and, using discreet samples and NDIR quantification
after measurement of the total alkalinity of samples of total CO2, have demonstrated accuracy and
either by potentiometric titration (Dyrssen, 1965; precision comparable to the coulometric method.
Bradshaw et al., 1981) or single acid addition method Successful automation of this method using seg-
(Anderson and Robinson, 1946; Culberson et al., mented flow was achieved by the use of a high
1970) where excess acid is calculated from the final efficiency gas stripper with minimum mixing along
pH. For this, accurate values for pKl and pK2 for the the segmented flow is described here. The method is
carbonic acid and the hydrogen ion activity co- capable of giving unattended real-time analysis of
efficient for different salinities are required. While total CO2.
good values for these constants exist for ocean
waters, and indeed small changes of carbon dioxide
EXPERIMENTAL
content can be estimated in sea water (see Talling,
1973), these must be determined or interpolated for Reagents
coastal, estuarine and fresh waters. Account must Hydrochloric acid: 0.5 M
also be taken of other ions (borate, phosphate, Standards. Solutions of 0.22.4 mM Na2CO3 were made
from AR grade reagent dried overnight at 180°C and
ammonium, silicate, etc.) which can be protonated, in dissolved in freshly distilled water. Because of the possible
addition to the hydrogen ion-pair formation with change in the standards concentration, these were made up
sulphate and fluoride. "Surplus alkalinity"--the al- daily and kept in acid washed volumetric flasks with mini-
kalinity fraction attributed to unknown and/or or- mum headspace.
Concentrated sulphuric acid and magnesium perchlorate
ganic acids--is the main source of error in calculating (solid) were used in driers 1 and 2 respectively.
total CO2 from alkalinity measurements. In the case Nitrogen gas. High purity grade.
of estuarine and fresh waters, relatively high concen-
trations of organic acids and colloidal minerals which Instruments
cannot be readily determined, may contribute signifi- ChemLab segmented-flowAA2 system(sampler and peri-
staltic pump) coupled to Analytical Development non-
cantly to alkalinity measurements (Hollibaugh et al.,
dispersive IR CO2 analyser and a strip chart recorder was
1991). It is clear that determination of total COs from used. The IR was fitted with a 12 in, gas flowcelL Other
the indirect carbonate alkalinity measurement may models may be fitted with shorter length flow cell.
have several sources of errors.
Description of the method
Direct methods for CO 2 determination include
Figure 1 shows the main components of the automated
gas chromatographic (Weiss, 1981) and coulometric system. The analytical manifold was made up of Tygon
pumping tubes made as short as possible while all other
connections were of 2 mm glass tubing with flexiblejoints
*Author to whom all correspondence should be addressed. from 2 mm wall Tygon tubing.

1231
1232 M. 1. Abdullah and E. Eek

9 Sampler N2carrier
Mixingcoil Mixingcoil ,L

Water 0 ,~ I I~
N2 0 O, >
Acid
00 0 > IR Recorder
o
i 0 0 > Waste
0
Fig. 1. Schematic diagram of the analytical manifold.

Samples were taken at the rate of 0,4 ml/min and mixed the liquid was converted into a fine spray. This spray was
with distilled water (0.8 ml/min), and then 0.1 ml/min of then condensed in the column chamber (ca. 40 ml volume)
0.5 M HCI was added and allowed to mix. The stream was which also acted as a phase separator. Purging of CO2 was
then introduced into the gas stripper. The nitrogen carrier found to be complete since re-introduction of the purged
gas used to purge CO2 was passed at a rate of 800 ml/min. liquid back through the sample input showed no detectable
The liquid phase was pumped to waste while the gas phase CO2
was first passed over concentrated H2SO4 and then through The ChemLab sampler was supplied fitted with a "wash"
a column of magnesium perchlorate for drying before chamber into which the sampling tube withdraws at the end
introduction into the IR analyser. Both the distilled water of the sampling period thus introducing wash solution for
and the 0.5 M HCI solution were continuously purged with the method blank. In the present method, instead of circu-
nitrogen. Higher sensitivity may be obtained by changing lating "wash" solution, clean nitrogen was introduced which
the sample and distilled water rates to 0.8 and 0.4 ml/min gave a blank value between each sample. This blank was the
respectively. This will be more suitable for some fresh water true reagent blank of the method, therefore, the peaks do
where the total CO: is low (less than 1 mM). For samples not require correction for the "wash" blank.
of very high total CO2 such as sediment pore water, smaller Sampling and "wash" times were 1.5 min each which
samples may be taken, e.g. sample and distilled water rates allows for 20 determination per hour.
of 0.2 and 1.0 ml/min respectively.
The gas circuit, made up from Cu and glass tubing as
appropriate, consists of two lines: one, pressure-regulated,
cleaned by passing through a column of "Carbosorb" (BDH RESULTS AND DISCUSSION
Chemicals) to remove traces of CO 2 and used as a purge and
carrier gas; the second was divided into 4 lines, cleaned and The analytical peaks, Fig. 2, show rapid rise a n d
used for reagent purging, segmentation and blank. The return to baseline a n d fairly fiat tops indicating the
carrier gas flow rate was maintained constant throughout high resolution o f the segmented flow system, mini-
the analysis since variation in this will alter the CO2 m u m along-stream mixing a n d the rapid a t t a i n m e n t
concentration in the gas phase.
The gas stripper was made from Pyrex glass and consisted o f m a x i m u m signal for a sampling/wash interval o f
of a series of 3 glass sintered discs through which the 90 s. It should be possible to reduce this to 60 s thus
segmented flow was forced by the carrier gas (N2) and where achieving a rate of 30 samples/h w i t h o u t loss of

8 0.80

0.40

,.~ 4 0.20 0.16 mM

0.16
8. 0.08

2 ent blank

Fig. 2. Analytical peaks for various carbonate standards (mM).

 Automatic method for total CO2 determination 1233
% dilution In the absence of coulometric or manometric
10 20 30 40 50 150 70 80 90
200 I I I I I I I I I method of independent assessment of the accuracy of
180
the present method, we used the total alkalinity
determination to calculate "carbonate alkalinity"
,.-. 160 B A
and hence total CO2 of standard sea water (Ocean
Scientific International Ltd, Wormley, U.K.), estuar-
}
• 1110
ine and river waters. Total alkalinity was determined
by the single acid addition method (Anderson a n d
8O
Robinson, 1946; Culberson et aL, 1970). The need for
accurate values for the activity coefficient of hydro-
gen ion in order to calculate excess acid, particularly
for estuarine and fresh water samples, was elimi-
0 . . . . , . . . . , '- - -' -'
l 2
nated; excess acid was determined by back titration
mM CO2 using the Gran method for locating the equivalence
point (Gran, 1952)• Values for seawater were cor-
Fig. 3. Calibration plots for (A) low sensitivity and (B)
high sensitivity manifolds, together with a plot of rected for the borate ion contribution. No correction
the various dilutions of sea water (C). The line is that of is required for ion-pairs formation with sulphates
best fit. and fluoride using the back titration method• Correc-
tion for "surplus alkalinity" and contribution from
organic acids could not be made as these parameters
precision. Calibration (Fig. 3A) using the low sensi- are usually unknown.
tivity manifold, was linear between 0 and 3 mM total Compared with values calculated from alkalinity
CO 2 (as Na2CO3). For the high sensitivity manifold measurements, Table 1, the IR method gave consist-
(Fig. 3B), linearity was also confirmed up to 1 mM. ently lower total CO2 in the samples except for those
Correlation coefficient for both was better than with high salinity, e.g. SSW. Discrepancies arise from
r =0.999 (n = 14 and 12). Various dilutions (by two sources. First, the unaccounted for components
volume) of seawater from the Oslofjord (sal- of alkalinity, e.g. "surplus alkalinity", which is in-
inity = 27.41) were analysed and a plot of total CO2 cluded as carbonate, will give an over-estimation of
against dilution (Fig. 3C) show linear relationship total CO2. Differences between the calculated (from
(r = 0.9998, n = 12). alkalinity) and measured (gasometric/coulometric)
Replicate analysis of sea, estuarine and fresh CO 2 were reported to be about 21/~M for Atlantic
waters are listed in Table 1. For fresh water, the high water (Bradshaw and Brewer, 1988) and as high as
sensitivity manifold was used. The standard deviation 300/~M for the Black Sea (Goyet et al., 1991).
(S.D.) varies between 0.002 mM for fresh water and Second, the calculated total CO2 (from carbonate
0.005 mM for sea water which gives a precision of 0.1 alkalinity) may contain errors associated with pH
and 0.2%. This should be acceptable for a rapid measurements of samples of widely different salinities
automatic method for total CO2 determination in (due to definition of pH scale and other operational
waters of varied salinity. parameters, Dickson, 1984). We calculate that an

Table I. Replicate analysis o f standard seawater (SSW), estuarine water (ESTW)

and fresh water ( F W ) for total C O 2, and the analytical results for river, fjord
and standard sea waters. The carbonate alkalinity and the calculated total CO2 ~2~
(CO3 + HCO3 + H2CO3 + CO2) are also listed for comparison
Total C O 2
Sample (mM) S.D. n
SSW 2.212 0.005 8
ESTW 1.236 0.005 8
FW 0.236 0.002 10
Carbonate
Total C O 2 alkalinity Total CO2 ~2~
Sample Salinity (mM) (m.equiv./I) (raM)
FWI < 1 0.116 0.091 0.147
FW3 < 1 0.106 0.085 0.136
FW4 < 1 0.130 0.127 0.196
FWI 1 < 1 0.239 0.144 0.258
FWI2 < I 0.160 0.110 0.190
FWI3 < 1 0.240 0.193 0.281
EST6 15.68 I. 137 1.079 1.049
ESTI6a 16.74 1.163 1.182 1.173
EST16c 30.73 1.955 2.060 2.013
EST7a 26.13 1.773 1.778 1.773
EST7c 30.95 2.041 2.089 2.033
SSW 35.00 2.212 2.411 2.182
1234 M. I. Abdullah and E. Eek

error of 0.02 p H at p H 8.2 would result in an error of the mixed layer carbon budget. Deep Sea Res. H 40,
0.025 m M in total CO2. Such error is very c o m m o n 151-169.
Culberson C., Pytkowicz R. M. and Hawley J. E. (1970)
in all routine p H measurements where the precision Seawater alkalinity determination by the pH method. J.
is no better than 0.02 pH units (SCOR, 1985). Mar. Res. 28, 15-21.
The direct I R method will not be influenced by any Dickson A. G. (1984) pH scales and proton-transfer reac-
of these errors and the automatic version described tions in saline media such as sea water. Geochim. Cos-
mochim. Acta 48, 2299-2308.
here should be capable of rapid, accurate and inex- Dyrssen, D. (1965) A Gran titration of seawater on board
pensive analysis of total CO2 in natural waters. Such Sagitta. Acta chem. stand. 19, 1265.
a method should also be capable of real-time and Goyet C. and Snover, A. K. (1993) High-accuracy measure-
even online analysis of total dissolved CO2. Where ments of total dissolved inorganic carbon in the ocean:
organic acids or surplus alkalinity is suspected to be comparison of alternate detection methods. Mar. Chem.
44, 235-242.
high, the calculation of the carbonate alkalinity from Goyet C., Bradshaw A. L. and Brewer P. G. (1991) The
total CO2 will yield a more accurate value with the carbonate system in the Black Sea. Deep Sea Res. 38
constraint of accuracy of pH and equilibrium con- (Suppl. 2), S1049-1068.
stants, than those obtained by indirect total alkalinity Gran G. (1952) Determination of the equivalence point in
potentiometric titrations, II. Analyst 77, 661-671.
measurements. Hollibaugh J. T., Buddemeier R. W. and Smith S. V. (1991)
Contribution of colloidal and high molecular weight
dissolved material to alkalinity and nutrient concen-
REFERENCES trations in shallow marine and estuarine systems. Mar.
Chem. 34, 1-27.
Anderson D. H. and Robinson R. J. (1946) Rapid electro- Johnson K. M., King A. E. and Sieburth J. M. (1985)
metric determination of the alkalinity of sea water using Coulometric TCO2 analyses for marine studies; an intro-
a glass electrode. Ind. Eng. Chem. 18, 767-773. duction. Mar. Chem. 16, 61-82.
Bradshaw A, L. and Brewer P. G. (1988) High precision SCOR (1985) Oceanic CO2 measurements. Report of the
measurements of alkalinity and total carbon dioxide in third meeting of the Working Group 75, Les Houches,
seawater by potentiometric titration--1. Presence of un- France, October 1985.
known protolyte(s)? Mar. Chem. 23, 69-86. Talling, J. F. (1973) The application of some electro-
Bradshaw A. L., Brewer P. G., Shafer D. and Williams R. chemical methods to the measurement of photosynthesis
T. (1981) Measurements of total carbon dioxide and and respiration in fresh waters. Freshwat. Biol. 3,
alkalinity by potentiometric titration in the GEOSEC 335-362.
program. Earth Planet Sci. Len. 55, 99-115. Weiss R. F. (1981) Determination of CO2 and methane by
Chipman D. W., Marra J. and Takahashi T. (1993) Primary dual catalyst flame ionization chromatography and ni-
production at 47°N and 20°W in the North Atlantic trous oxide by electron capture chromatography. J. Chro-
Ocean: a comparison between 14C incubation method and matogr. Sei. 19, 611-616.