ORIGINAL PAPER

A flow injection analyser conductometric coupled systemfor the field analysis of free dissolved CO2 and total dissolvedinorganic carbon in natural waters

Valter Martinotti & Marcella Balordi & Giovanni Ciceri

Received: 11 October 2011 /Revised: 5 December 2011 /Accepted: 18 January 2012 /Published online: 30 January 2012# Springer-Verlag 2012

Abstract A flow injection analyser coupled with a gasdiffusion membrane and a conductometric microdetectorwas adapted for the field analysis of natural concentrationsof free dissolved CO2 and dissolved inorganic carbon innatural waters and used in a number of field campaigns formarine water monitoring. The dissolved gaseous CO2 presentsnaturally, or that generated by acidification of the sample, isseparated by diffusion using a hydrophobic semipermeablegas porous membrane, and the permeating gas is incorporatedinto a stream of deionised water and measured by means of anelectrical conductometric microdetector. In order to make thesystem suitable and easy to use for in-field measurementsaboard oceanographic ships, the single components of theanalyser were compacted into a robust and easy to use system.The calibration of the system is carried out by using standardsolutions of potassium bicarbonate at two concentrationranges. Calibration and sample measurements are carried outinside a temperature-constant chamber at 25 °C and in an inertatmosphere (N2). The detection and quantification limits ofthe method, evaluated as 3 and 10 times the standard deviationof a series of measurements of the matrix solution were 2.9and 9.6 μmol/kg of CO2, respectively. Data quality for dis-solved inorganic carbon was checked with replicate measure-ments of a certified reference material (A. Dickson, ScrippsInstitution of Oceanography, University of California, SanDiego), both accuracy and repeatability were −3.3% and10%, respectively. Optimization, performance qualification

of the system and its application in various natural watersamples are reported and discussed. In the future, the calibra-tion step will be operated automatically in order to improvethe analytical performance and the applicability will be in-creased in the course of experimental surveys carried out bothin marine and freshwater ecosystems. Considering the presentstage of development of the method, it can only be applied forstudying of the carbon cycle in oxic environments.

Keywords Flow injection analysis . Free dissolved CO2.

Total dissolved inorganic carbon . Seawater . Carbon captureand storage . CO2 baseline fluxes monitoring

AbbreviationsACT Alliance for Coastal TechnologiesCCS Carbon capture and storageCRM Certified reference materialDIC Dissolved inorganic carbonDL Detection limitFIA Flow injection analysisGD Gas diffusionPMMA PolymethylmethacrylatePTFE PolytetrafluoroethyleneQL Quantification limit

Introduction

The increase of anthropogenic emissions raises the CO2 con-centration in the atmosphere and leads to climate changes thatimpact ecosystems and human infrastructures. CCS in subsur-face geological formations, both on- and off-shore, is consid-ered as one of the major technology solutions for reducinganthropogenic greenhouse gas emissions [1].

Published in the special issue Euroanalysis XVI (The EuropeanConference on Analytical Chemistry) with guest editor Slavica Ražić.

V. Martinotti (*) :M. Balordi :G. CiceriRSE SpA—Environment and SustainableDevelopment Department,Via Rubattino 54,20134 Milan, Italye-mail: valter.martinotti@rse-web.it

Anal Bioanal Chem (2012) 403:1083–1093DOI 10.1007/s00216-012-5762-8

Therefore, in order to realize a safe geological CO2 storage,the subsurface characteristics and the behaviour of the injectedCO2 in the subsurface and in surface environments need to beknown. About the dynamic interaction of the injected CO2, itis important to establish appropriate monitoring tools andprocedures to get an early warning of leaks, for example atthe sediment–water interface in marine coastal environments[2]. The knowledge of the natural baseline concentrations andbenthic fluxes, before the CO2 injection, will allow the com-parison with data obtained after the storage operations toverify if an unexpected CO2 leakage occurs.

Due to the seawater characteristics (pH about 8.1) and inthe presence of small fluxes, CO2 should be mainly present inthe aqueous phase as DIC. So, it is essential to determine thischemical form, other than the gaseous CO2 (as bubbles or freesolvated), in order to keep into account the CO2 speciation inseawater. At present, in surface ocean water, the concentra-tions of the dissolved CO2 are low and range from 5 to15 μmol/kg [3], whereas, the DIC concentrations range from1.8 mmol/kg in warm surface waters to 2.4 mmol/kg in deepwaters [4]. Moreover, the determination of pH and alkalinityin the same water sample would complete the definition of themarine carbonate system. The knowledge of the carbonatesystem in seawater, together with information about the sed-imentary compartment (pore water and solid phases for otherchemical parameters), is useful to better understand the earlydiagenetic processes responsible of the observed CO2 fluxes.

Several methods of measuring the dissolved inorganiccarbon in seawater exist on the market. An evaluation ofcommercial-ready and emerging open-ocean, coastal, andfreshwater sensing technologies for CO2 instruments, in-cluding laboratory and field tests over short- and long-termdeployments, was carried out by the Alliance for CoastalTechnologies (ACT; www.act-us.info) [5]. Among thesetechniques, those based on the equilibration of CO2 inseawater diffusing across a gas-permeable membrane in asensing aqueous solution are usually used [6–16]. In general,these systems have good sensitivity and precision but they arevery big, complicated and difficult to operate [17].

On the contrary, those based on GD method coupled withFIA were proved to be simple, sensitive, rapid and easy tooperate, because they exploit the useful inherent characteristicsof FIA, which are automation, high throughput, better preci-sion, low reagents and sample consumption [18, 19]. Thediffused CO2 can be detected measuring the changes ofphysico-chemical properties of the receptor carrier, for exampleby spectrophotometry [20, 21], potentiometry [22], and con-ductometry [21, 23]. Conductometric detection is, in general,recommended because it does not involve complex reactionand potential sources of interference, typical of spectrophoto-metric method, do not affect the conductometric measurements;furthermore, it is more reproducible and sensitive [21, 24].

The GD-FIA/conductometry system, presented in this study,was designed and built to measure the background naturalconcentrations of dissolved CO2 and DIC present at the sedi-ment–water interface in marine coastal environments. The gas-eous CO2 presents in the sample is separated by diffusion usinga hydrophobic semi permeable gas porous membrane (PTFEtape), and the permeating gas is incorporated into a stream ofdeionised water and measured by means of an electrical con-ductometric microdetector. The optimization of the system, theperformance qualification, the applicability and its adaptationfor a use aboard of ships, are reported in the following sections.

Further investigations focused on the improvement of theanalytical performance of the proposed system and on itsapplicability in field will be carried out in the future.

Considering the present performance characteristics, themethod is applicable for study of the carbon cycle at thesediment–water interface, while its application to studies onthe ocean acidification may result in more inaccurate resultsthan those obtained with alternative methods proposed inthe literature. As regards the possible interferences that mayoccur when using this method in anoxic condition [23], thisproblem is not at the moment taken into account because theapplicability is proposed only for oxic environment.

Experimental

Reagents and solutions

All solutions were prepared with analytical grade reagents (C.Erba RPE-ACS and RPE-ISO). High purity deionised water(carrier solution) with a resistivity of about 5.56 MΩ cm at25 °C was obtained with a Milli-Q (Millipore Corp.) waterpurification system. A stock standard solution of DIC in NaCl0.7 m was prepared from potassium hydrogen carbonate(KHCO3), which was dried at 120 °C for 4 h before use, asNaCl reagent. All solutions were N2 purged for 5 min and keptin a sealed bottle inside a constant-temperature chamber andin N2 atmosphere, before the use. Working standard solutionswere prepared by serial dilution, on the mass bases, of thestock standard solution with NaCl 0.7 m as a solvent, justbefore use. HCl 6 M was used for the acidification of thesamples or the standard solutions.

Apparatus and experimental set-up

A schematic lay-out of the apparatus is shown in Fig. 1 anda photograph of the lay-out of the components into thehousing is shown in Fig. 2.

The procedure is based on previous work of [24] and [25]with some modifications. Themethod is based on the diffusion

1084 V. Martinotti et al.

of dissolved gaseous CO2 through a hydrophobic semi perme-able gas porous membrane (PTFE commercial tape: Lachatpart no. 50331; 3.8 cm width×8.0 cm length; 0.45 μm poros-ity) that is placed in a Flow-Thru DIALYZER™ unit (HarvardApparatus, no. 74-1302;) (MSU). The entire unit is made ofPTFE and has two separate chambers with serpentine channels(600 μL total sample volume; 700 mm length) superimposedon each other and separated by a membrane. Once assembled,the unit is placed into a metal clamp and tightened with treescrews. The transit time across the length of the membrane isabout 30 s. Once selected, a single membrane can be used for atleast 1 month, without drastic change in sensitivity andreproducibility.

The high ƒCO2 of the water sample (naturally present orafter acidification) and the low ƒCO2 in the carrier (C) drivequantitative exchange of CO2 across the membrane. Thegenerated pressure gradient is proportional to the concentra-tion of CO2 in the water sample. The permeating gas iscollected in a stream of freshly deionised water (carriersolution) that flows at a constant rate (1.5 mL/min) towards

an electrical conductivity micro flow cell (custom designedtwo-electrode cell) (20 μL volume; gold electrodes; 500 μmi.d.), put in a cubic (3.3 cm of side) body made in PMMA.The CO2 reacts with the deionised water (carrier solution)producing the carbonic acid: H2CO3* (H2CO3+CO2aq), that,in turn, dissociates generating H+, HCO3

− and CO32− ions.

This causes an increase in the conductivity of the carrierstream receiving the diffused CO2. The conductivity contrastforms the basis of the conductometric CO2 analysis.

The conductivity meter (D+SA) output is connected to apersonal computer and a squirrel data logger for the dataacquisition (DA). Data are also stored in a MMCmemory card(1 Gb) located inside the housing. A peristaltic pump (P) (ColeParmer mod. Masterflex CL pump tubing; two channels andvariable speed) was used for propelling the sample (or thestandard solutions) and the carrier solution at 1.5 mL/minconstant flow-rate. A ten-port rotary valve (V) (Valco mod.VICI AISI 316L) was used for select the various analyticalsteps (washing, loading and injection). Connections among thecomponents of themanifold are realised with PVC (2mm i.d.),except for those connecting the port of the rotary valve, the loop(in PTFE) to the rotary valve and the inlet/outlet of the MSU,that are in PTFE (0.3 mm i.d.). The path from the loop to theMSU is an 18 cm length of PTFE tube (0.3mm i.d.). The transittime is about 0.5 s. The path from the outlet of MSU to the D isa 23-cm length of PVC tube (14 cm; 1 mm i.d.) and PTFE tube(9 cm; 0.3 mm i.d.). The transit time is about 5 s. Sensitivitywas adjusted by changing the conductivity meter control (SA).

All the above components are contained into a compact(22×43×58 cm) and very light (about 10 kg) housing made inaluminium, easy to be transported in the field. To achievetemperature control, the system and all solutions used wereenclosed in an insulated constant-temperature chamber (glovebox), able to maintain an inert atmosphere (N2). Analysis andcalibration operations were carried out at 25 °C. The systemallows the injection of samples without any treatment for thedetermination of the dissolved CO2 and the injection of acidi-fied samples for the determination of DIC.

Fig. 1 Schematic lay-out of theGD-FIA/conductometry sys-tem. C carrier stream; P peri-staltic pump; S sample; MSUmembrane separation unit; Wwaste; L loop; SA signal atten-uator; D electrical conductivitymicro flow cell; V ten-port ro-tary valve; DA data acquisition(squirrel data logger and per-sonal computer)

Fig. 2 Details of the components into the housing. 1 peristaltic pump(P); 2 carrier container (C); 3 ten-port rotary valve (V); 4 loop (L); 5membrane separation unit (MSU); 6 electrical conductivity micro flowcell (D); 7 signal attenuator (SA); 8 memory card; 9 air-conditioningsystem; 10 temperature (external and internal) and conductometricacquisition signals panel

A flow injection analyser conductometric coupled system 1085

Calibration

In the laboratory, the calibration of the system was carriedout by means of a headspace-free reactor containing KHCO3

standard solutions. Two CO2 concentration ranges wereconsidered:

– CO2 low-level range, 10–100 μmol/kg in NaCl 0.7 m,at four levels (10; 20; 50; 100 μmol/kg) in triplicate, fordissolved CO2 determination;

– CO2 high-level range, 1.75–5.00 mmol/kg in NaCl0.7 m, at five levels (1.75; 2.00; 2.50; 3.41;5.00 mmol/kg) in triplicate, for DIC determination.

The preparation of the working standard solutions for thecalibration of the systemwas made in a headspace-free reactor(60 mL total volume) by addition, with a micropipette, ofknown volumes of the two stock standard solutions (low-and high-level ranges), to a known weight of the saline solu-tion (NaCl 0.7 m), in order to obtain concentrations of thediluted standard expressed on the mass bases. All operationswere carried out inside the constant-temperature chamber andin N2 atmosphere.

Then, an aliquot (30 μL for low-level range and 110 μL forhigh-level range) of HCl 6 M was added, in order to reach apH<3.5 and to convert all inorganic dissolved carbon (HCO3

−

and CO32−) forms to dissolved gaseous CO2. The solution was

mixed for 5 min and then an aliquot of 2.5 mLwas transferredby a syringe to the ten-port rotary valve in order to load theloop for the subsequent analysis step.

Although small, a salt effect of about +4% due to NaCl0.7 m ionic strengths, compared to high purity deionisedwater, is present (Fig. 3). Then, for the highest accuracy,standard solutions should be made in a comparable matrix tosamples [26].

As the measurement of the acidified saline solution (NaCl0.7 m) gave a conductometric signal, in order to keepinto account this influence, a triplicate measurement of thissaline solution was carried out and the obtained mean signalwas subtracted to those of the working standard solutions.

This correction was made only for the CO2 low-levelrange because seawater samples for dissolved CO2 de-termination are injected directly in the system without anytreatment and because the conductometric signal of the salinesolution significantly affects those of the working standardsolutions. For CO2 high-level range, this correction has notbeen made because seawater samples are acidified beforeinjection in the system as the working standard solutions andbecause the conductometric signals of the working standardsolutions are not significantly affected by that of the salinesolution.

The washing of the loop, between the analyses of twoconsecutive samples, was performed by exploiting part ofthe final volume of the carrier solution that displaces thesample loaded in the loop during the injection step and partof the initial volume of the next sample during the loadingstep. The complete calibration procedure for replicate meas-urements of standard and saline solutions lasts about 2 h and15 min.

GD-FIA/conductometry operation in the field

The GD-FIA/conductometry system and the constant-temperature chamber were located onboard of the ship andarranged for the calibration and analysis steps. Seawatersamples are collected by using a benthic chamber. Benthicchambers are open-bottom containers that enclose an area ofsediment and overlying water; the chambers are deployedon the sediments to capture gas and solute movement be-tween the sediment and the overlaying water column.

Before the recovery of seawater samples from the benthicchamber, the system was calibrated following the abovecalibration procedure adopted in laboratory but only onemeasurement for considered concentration level, except forthe saline solution. The complete calibration procedure formeasurements of standard and saline solutions lasts about1 h and 10 min.

Then, seawater samples collected from the benthic cham-ber were filtered on PTFE syringe filter (Sartorius mod. Min-isart; 0.45 μm porosity, 25 mm diameter) to removeparticulate matter and divided in two fractions. The fractionfor the dissolved CO2 determination (5 mL) was first ana-lysed. Then, an aliquot of 110μL of HCl 6Mwas added to thesecond fraction (60 mL) for DIC determination, in order toreach a pH<3.5.

The washing of the loop, between the analyses of twoconsecutive samples, was performed by exploiting part ofthe final volume of the carrier solution that displaces thesample loaded in the loop during the injection step and partof the initial volume of the next sample during the loadingstep. At the end of analyses of the two fractions, anothercalibration of the system was carried out to verify its stabilityover the time.

3,0

4,0

5,0

6,0

7,0

1,00 1,50 2,00 2,50 3,00 3,50 4,00CO2 concentration (mmol/kg)

Pea

k h

eig

ht

(arb

itra

ry u

nit

s)

NaCl 0.7 m deionised water deionised water NaCl 0.7 m

Fig. 3 Salt effect on the conductivity response

1086 V. Martinotti et al.

Result and discussion

Performance criteria

The development of the instrument has considered the fol-lowing criteria:

& Detection of the baseline natural concentration of dis-solved CO2 at the sediment–water interface in marinecoastal environment

& Measurement of both dissolved CO2 and DIC& Analysis of samples onboard of oceanographic ships

Optimization of the GD-FIA/conductometry system

Optimization approach

The optimization approach of the GD-FIA/conductometrysystem took into account variables that could affect theperformance of the system in terms of sensitivity. Consid-ered variables were: the pH of acidification, the injectionsample volume, the carrier flow-rate, the reaction time (acid-ification), and the temperature.

Only chemical species present in natural waters that areable to permeate the PTFE membrane and dissociate caninterfere with the dissolved CO2 and DIC determination.The effect of possible interferents, like ammonia (NH3)and volatile organic bases (alkyl amines) in the case ofdissolved CO2 determination (given that the sample pH isnot adjusted to values below 3.5 prior to the measurements)and like volatile organic and inorganic acids (formic acid,acetic acid and H2S) and oxides (NO2 and SO2) in the caseof DIC determinations, able to diffuse through the hydro-phobic membrane and dissociate, were not considered inthis study. Usually, these compounds are present in such lowconcentrations if compared to the inorganic carbon speciespresent in natural waters, that they pose no problem; how-ever, in specific cases (i.e. under anoxic conditions in natu-ral waters with production of high concentration of NH3 andH2S), they need to be kept into account [23].

Chemical variables

pH of acidification

The effect of the pH of acidification on the yield of conversionof the total dissolved carbonate (CO3

2− and HCO3−) in dis-

solved gaseous CO2 was studied at two CO2 concentrationlevel (low-level range, 20 μmol/kg in NaCl 0.7 m and high-level range, 2 mmol/kg in NaCl 0.7 m) and by using HCl atdifferent molarities (0.12, 1.2 and 6 M). The considered pHrange was from about 2.0 to 7.0 pH units. The highest analyt-ical responses, in terms of peak height, were obtained for pH

values <3.5 units, then this pH value, obtained by addition of30 or 110 μL of HCl acid 6 M, respectively to a low-level andhigh-level standard solutions of about 60 mL, was chosen.

The result obtained for the synthetic samples was alsoconfirmed by adding the aliquot of 110 μL of HCl acid 6 Mto the CRM of Dickson (batch no 92, 2008; DIC, 1996.35±0.46 μmol/kg, in natural seawater).

Physical variables

Sample volume

The effect of the injected volume of the sample on the con-ductometric response was studied by considering the followingloop volume, 100, 500 and 1,000 μL. The results obtained foran acidified 2 mmol/kg in NaCl 0.7 m CO2 standard solutionshowed the highest analytical responses, in terms of peakheight, for the 500-μL loop, then this loop was adopted inpractical use.

Carrier flow-rate

The effect of the carrier flow-rate on the conductometricsignal (peak height) was examined analysing three CO2

standard solutions (10 μmol/kg in NaCl 0.7 m, 1 mmol/kgin NaCl 0.7 m, and 10 mmol/kg in NaCl 0.7 m). Two carrierflow-rates were considered, 1.5 and 2.5 mL/min. The resultsare shown in Fig. 4.

The flow-rate of 1.5 mL/min was adopted to obtain thebest analytical throughput and sensitivity.

Reaction time (acidification)

The reaction time of the acidification of the sample for theconversion of the total dissolved carbonate (CO3

2− andHCO3

−) in dissolved gaseous CO2 was investigated for twostandard solutions (2 mmol/kg in NaCl 0.7 m and 10mmol/kgin NaCl 0.7 m). The range of reaction time considered was

0,0

1,0

2,0

3,0

4,0

5,0

6,0

7,0

8,0

9,0

1,5

1,5

1,5

1,5

1,5

2,5

2,5

2,5

2,5

2,5

Carrier flow-rate(mL/min)

Pea

k h

eig

ht

(arb

itra

ry u

nit

s)

1 mmol/kg 10 mmol/kg

Fig. 4 Effect of the carrier flow-rate

A flow injection analyser conductometric coupled system 1087

from 1 to 5 min. On the basis of the obtained analyticalresponses, in terms of peak height, a reaction time of 5 minwas chosen in order to assure the complete conversion of thetotal dissolved carbonate.

Temperature

The effect of the temperature on the measurements wasstudy by using a constant-temperature chamber with inertatmosphere (N2) at 10±1 °C and 25±1 °C, measuring thefollowing standard solutions of KHCO3 (n03 each concen-tration of the same solution), 10; 20; 50; 100 μmol/kg inNaCl 0.7 m (low-level range), for dissolved CO2 and, 1; 2;5; 10 mmol/kg in NaCl 0.7 m (high-level range), for DIC.Figure 5 shows the obtained results.

Only values of working standard solutions for the CO2

low-level range were subtracted by the contribution fromsaline solution, as indicated in the “Calibration” section.

It is evident from the graphs that at 25±1 °C the response ofthe analyser, in terms of peak height, is better than thatobtained at 10±1 °C. This could be a consequence of the

increase of the permeation rate through the membrane andthe decrease of the solubility of CO2 with increase of temper-ature; therefore, keeping the system at constant temperature isnecessary for an adequate control of this variable [23, 27].

In order to further investigate the effect of the temperatureof measurement, replicate (n010) analyses of a real seawatersample were performed at the following temperature, 10, 15,20, 25 and 30 °C. Figure 6 shows the obtained results.

As shown in the above figure, the temperature of 25±1 °C gives the best performance both in terms of peak heightand repeatability.

Method evaluation

Performance qualification

Operating the system under optimized conditions (pH ofacidification, <3.5; acidification volume, 30 μL of HCl6 M for low-level and 110 μL of HCl 6 M for high-level;sample volume, loop 500 μL; carrier flow-rate, 1.5 mL/min;reaction time, 5 min; temperature of analysis, 25±1 °C andin N2 atmosphere), the following performance parameterswere evaluated.

System stability

The stability of the whole system was evaluated observingthe trend of the baseline obtained analysing the carriersolution (Milli-Q deionized water) for a period of 4 h, bothfor the low-level range (signal attenuation, 10 kΩ) and high-level range (signal attenuation, 500 kΩ) instrumental con-ditions. The drift was assumed as parameter to evaluate thestability of the system. The baseline drift was not significantover the investigated period of time and was included within60 mV and 10 mV, respectively, for the low- and high-levelrange instrumental conditions.

0,0

1,0

2,0

3,0

4,0

5,0

6,0

7,0

8,0

0 10 20 30 40 50 60 70 80 90 100 110

CO2 concentration (µmol/kg)

Pea

k h

eig

ht

(arb

itra

ry u

nit

s)

10 °C 25 °C

0,0

2,0

4,0

6,0

8,0

10,0

12,0

0 1 2 3 4 5 6 7 8 9 10 11

CO2 concentration (mmol/kg)

Pea

k h

eig

ht

(arb

itra

ry u

nit

s)

10 °C 25 °C

Fig. 5 Effect of the temperature on the measurements of CO2 standardsolutions: (upper plot) CO2 low-level range, (lower plot) CO2 high-level range

0,0

0,5

1,0

1,5

2,0

2,5

3,0

3,5

4,0

4,5

1 2 3 4 5 6 7 8 9 10

Measurement number

Pea

k h

eig

ht

(arb

itra

ry u

nit

s)

20 °C 15 °C 10 °C 25 °C 30 °C

Fig. 6 Effect of the temperature on the measurements of a real sea-water sample

1088 V. Martinotti et al.

Detection and quantification limits

The DL of the method, corresponding to three times of thestandard deviation of the baseline variability (n020 meas-urements) obtained analysing the matrix solution (NaCl0.7 m acidified with 30 μL of HCl 6 M), was 2.9 μmol/kgof CO2 (see Table 1).

The QL of the method, corresponding to ten times of thestandard deviation of the baseline variability (n020 meas-urements) obtained analysing the carrier solution (NaCl0.7 m acidified with 30 μL of HCl 6 M), was 9.6 μmol/kgof CO2 (see Table 1).

Calibration linearity

The linearity of the calibration graphs was evaluated ana-lysing standard solutions of KHCO3 over the two followingconcentration ranges:

& CO2 low-level range, 10–100 μmol/kg in NaCl 0.7 m,for dissolved CO2 determination (10, 20, 50 and100 μmol/kg levels) and n03 replicate measurementeach concentration of the same solution);

& CO2 high-level range, 1–10 mmol/kg in NaCl 0.7 m, forDIC determination (1,75; 2,00; 2,50; 3,41; and5,00 mmol/kg levels) and n03 replicate measurementeach concentration of the same solution).

Values of working standard solutions for the CO2 low-levelrange were subtracted by the contribution from saline solution,as indicated in the Calibration section. The R2 (regressioncoefficient) value was considered for evaluate the linearity.The regression equations of the calibration graphs were: y00.069x+0.157 (R200.996), for dissolved CO2 and y01.028x+2.783 (R200.985), for DIC, where y is the electrical conduc-tometric signal measured as peak height. The calibrationgraphs were linear over the both CO2 concentration ranges.

Calibration stability

The stability of the calibration over the time was evaluatedcarrying out calibration of the system for the dissolved CO2

(CO2 low-level range, see “Calibration linearity” section) andfor DIC (CO2 high-level range, see “Calibration linearity”section) in three (straight) days a week (inter-repeatability),applying the optimized procedure. In Table 2, the statisticalcharacteristics of the calibration graphs are reported.

The CO2 standard solutions (both low- and high-levelranges), used to make the first calibration graph, were readboth on the same calibration graph (that is 24 and 48 h later)and on the second and third calibration graphs obtained inthe successive days (that is 24 and 48 h later, respectively),in order to evaluate the concentration offset over the time(see Table 3).

Acquired data indicate that the offsets obtained from lec-tures of CO2 standard solutions, both for CO2 low-level andCO2 high-level ranges, on the first calibration graph after 24 hare similar to those ones obtained from lectures on the secondcalibration graph. Lectures on the first and on third calibrationgraphs performed 48 h later, show that the offsets are worse(dissolved CO2) and quite similar (DIC) for the first case. So,it was assumed that the calibration graph could be used theday of preparation until the successive day (that is 24 h later)with offset for the calculated concentrations less than about12% for both dissolved CO2 and DIC, as absolute value. Acalibration of the system before the samples measurement is,however, preferable and recommended.

Accuracy

For dissolved CO2, the accuracy (mean value minus nomi-nal reference value, keeping into account the sign) wasevaluated by replicate (n010 each concentration of the samesolution) measurements of KHCO3 standard solutions attwo different concentration levels, 10 μmol/kg in NaCl0.7 m and 20 μmol/kg in NaCl 0.7 m.

Accuracy for dissolved CO2 was +0.73 μmol/kg (+7.3%)and +2.75 μmol/kg (+13.8%), respectively. For DIC, theaccuracy (mean value minus nominal reference value, keep-ing into account the sign) was evaluated by replicate (n010each concentration of the same solution) measurements ofKHCO3 standard solutions at two different concentrationlevels, 1.75 mmol/kg in NaCl 0.7 m and 5.00 mmol/kg inNaCl 0.7 m. Accuracy for DIC was +0.08 mmol/kg (+4.6%)and +0.04 mmol/kg (+0.8%), respectively.

Furthermore for DIC, the accuracy (mean value minuscertified reference value, keeping into account the sign) was

Table 1 Detection andquantification limits CO2 concentration

(μmol/kg)

DL 2.9

QL 9.6

Table 2 Statistical characteristics of the calibration graphs obtained indifferent days for dissolved CO2 and DIC

First calibrationgraph

Second calibrationgraph

Third calibrationgraph

Dissolved CO2

y00.063x+0.211 y00.072x+0.063 y00.066x+0.193

R200.996 R200.999 R200.993

DIC

y00.934x+2.891 y01.002x+2.657 y00.784x+3.769

R200.981 R200.988 R200.988

A flow injection analyser conductometric coupled system 1089

evaluated by replicate (n010 each concentration of the samesolution) measurements of CRM of Dickson. Accuracy forDIC was −66 μmol/kg (−3.3%).

Repeatability

Data obtained for accuracy were used to calculate the re-peatability as absolute (micromoles per kilogram or milli-moles per kilogram) and relative (%) standard deviation(95% confidence level or p00.05).

Repeatability for dissolved CO2 was ±1.69 μmol/kg(±16.9%) and ±3.06 μmol/kg (±15.3%), respectively.

Repeatability for DIC was ±0.07 mmol/kg (±4.0%) and±0.11 mmol/kg (±2.2%), respectively.

Furthermore for DIC, the repeatability as absolute (micro-moles per kilogram or millimoles per kilogram) and relative(%) standard deviation (95% confidence level or p00.05), wascalculated using the data obtained from the accuracy of the

CRM of Dickson. Repeatability for DIC was ±0.20 mmol/kg(±10%). Figure 7 shows the dispersion of the replicate meas-urements for the CRM of Dickson.

Table 4 summarizes the all obtained results.

Robustness

The robustness of the method was estimated by comparingthe slopes of all calibration graphs obtained more times aday and in different days, both for the CO2 low-level (10–100 μmol/kg in NaCl 0.7 m) and CO2 for high-level (2–10 mmol/kg in NaCl 0.7 m) concentration ranges. For theCO2 low-level range, the slopes varied between 0.062 and0.070 with an average value of 0.066 and a relative standarddeviation of 6.1%. For the CO2 high-level range, the slopesvaried between 0.948 and 0.752 with an average value of0.850 and a relative standard deviation of 8.1%. Figure 8shows graphically the obtained results for CO2 low-leveland CO2 high-level ranges, respectively.

1,0

1,2

1,4

1,6

1,8

2,0

2,2

2,4

2,6

1 2 3 4 5 6 7 8 9 10

Measurement number

CO

2 co

nce

ntr

atio

n (

mm

ol/

kg)

nominal concentration measured concentration 25 % offset

Fig. 7 Dispersion of the replicate measurements for the CRM ofDickson

Table 3 Concentration offset over the time for dissolved CO2 and DIC standard solutions

Readings (performed 24 h later)on first calibration graph

Readings (performed 48 h later)on first calibration graph

Readings (performed 24 h later)on second calibration graph

Readings (performed 48 h later)on third calibration graph

Calc. conc.(μmol/kg)

Conc.offset (%)

Calc. conc.(μmol/kg)

Conc.offset (%)

Calc. conc.(μmol/kg)

Conc.offset (%)

Calc. conc.(μmol/kg)

Conc.offset (%)

Dissolved CO2

10.5 −5.0 6.2 38.3 8.6 14.0 7.4 25.9

19.7 1.5 23.0 −15.1 20.8 −3.8 20.7 −3.3

55.1 −10.3 55.8 −11.7 47.9 4.2 50.2 −0.4

112.2 −12.2 102.8 −2.8 88.6 11.4 94.5 5.5

DIC

1.53 12.4 1.25 28.6 1.57 10.3 1.80 −2.6

1.93 3.5 1.84 7.9 2.24 −11.9 2.50 −25.2

3.83 −12.2 3.27 4.0 3.46 −1.6 3.80 −11.5

4.99 0.1 4.66 6.9 4.76 4.7 5.18 −3.6

Table 4 Repeatability for dissolved CO2 and DIC standard solutionsand for CRM of Dickson

Sample Concentration Repeatability

Dissolved CO2

standard solution10 μmol/kg ±1.69 μmol/kg

(±16.9%)

20 μmol/kg ±3.06 μmol/kg(±15.3%)

DIC standard solution 1.75 mmol/kg ±0.07 mmol/kg(±4.0%)

5.00 mmol/kg ±0.11 mmol/kg(±2.2%)

CRM 1996.35±0.46 μmol/kg ±0.20 mmol/kg(±10%)

1090 V. Martinotti et al.

Measurement frequency

Given that the analysis of a filtered seawater sample last 4 minfor dissolved CO2 determination and 9 min for DIC determi-nation, the throughput of the system, evaluated applying theoptimized procedure, resulted of 15 samples/h for dissolvedCO2 determination and 8 samples/h for DIC determination.

Applicability and case study

The proposed system was applied to the determination ofboth dissolved CO2 and DIC in various natural water sam-ples, such as freshwaters (tap water, industrial water andlacustrine water) and seawater. In Table 5, the obtainedresults for freshwaters are reported.

0,0

1,0

2,0

3,0

4,0

5,0

6,0

7,0

8,0

0 10 20 30 40 50 60 70 80 90 100 110

CO2 concentration (µmol/kg)P

eak

hei

gh

t (a

rbit

rary

un

its)

3,0

4,0

5,0

6,0

7,0

8,0

9,0

1 2 3 4 5 6

CO2 concentration (mmol/kg)

Pea

k h

eig

ht

(arb

itra

ry u

nit

s)

Fig. 8 Calibration graphsobtained during the same dayand in different days: (upperplot) CO2 low-level range,(lower plot) CO2 high-levelrange

Table 5 Results of analyses of DIC in different type of freshwaters

Watersample

Mean valuea

mmol/kgStandard deviationmmol/kg (%)

Calibrationrange mmol/kg

Tap water 5.76 0.52 (9.0) 3.41–6.80

Industrial water 5.44 0.48 (8.8) 3.41–6.80

Lacustrine water 2.52 0.12 (4.8) 1.75–3.41

a n010 replicates

Table 6 Results of analyses of dissolved CO2 and DIC in filteredseawater samples collected in four station of the Adriatic Sea

Parameter Station

A B C D

Dissolved CO2

(μmol/kg)<QLa 27.95 37.05 16.14

DIC (μmol/kg) 2411.8 2373.9 2454.9 2471.0

a QL of the method09.6 μmol/kg

A flow injection analyser conductometric coupled system 1091

Regarding seawater samples, in the course of someoceanographic surveys, carried out in the Northern andCentral Adriatic Sea (Italy), the proposed GD-FIA/con-ductometry system was used to determine the baselineconcentration of dissolved CO2 and DIC at the sediment–water interface of coastal seafloors. Table 6 shows theobtained results.

Conclusions

The proposed GD-FIA/conductometry system is rapid, in-expensive, compact and easy to operate onboard. It workswell for natural water samples over the concentration rangestypically observed.

The most important advantage of this system is the pos-sibility to perform inorganic carbon speciation (dissolvedCO2 and DIC) in natural waters simply analysing the non-perturbed and the acidified samples.

The selected operative conditions are for the pH of acid-ification: below 3.5 pH; for the acidification volume, 30 μLof HCl 6 M for CO2 low-level range and 110 μL of HCl 6 Mfor CO2 high-level range; for the sample volume, loop500 μL; for the carrier flow-rate, 1.5 mL/min; for the reac-tion time, 5 min; for the temperature of analysis, 25±1 °Cand in N2 atmosphere

The measurement frequency is enough for carry out amonitoring suitable to characterise the daily variability of theCO2 natural levels due to production/respiration processes.

Even if the current version of the system is to be consid-ered suitable for the fixed objectives, in order to improve itsperformance in terms of accuracy and precision, the calibra-tion step and the DIC analysis will be operate automaticallyin the future. Moreover, attaining lower detection limit is leftfor further development.

Considering the present performance characteristics, themethod is applicable for study of the carbon cycle at thesediment–water interface, while its application to studies onthe ocean acidification may result in more inaccurate resultsthan those obtained with alternative methods proposed inthe literature. As regards the possible interferences that mayoccur when using this method in anoxic condition, this prob-lem is not at the moment taken into account because theapplicability is proposed only for oxic environment.

Acknowledgements This work was financed by the Research Fund forthe Italian Electrical System under the Contract Agreement between RSE(formerly known as ERSE) and theMinistry of Economic Development—General Directorate for Nuclear Energy, Renewable Energy and EnergyEfficiency stipulated on July 29, 2009 in compliance with the Decree ofMarch 19, 2009. Thanks to Mrs. Flavia Tulli for her contribution incarrying out the experimental activity and to Dr. Alberto Novo for hiscontribution in the statistical and graphical elaboration of data.

References

1. IPCC (2005) Special Report on Carbon Dioxide Capture andStorage. Prepared by Working Group III of the Intergovern-mental Panel on Climate Change. Metz B, Davidson O, deConinck HC, Loos M and Meyer LA (Eds.), Cambridge UniversityPress

2. Golmen LG, Sundfjord A, Berge JA, Caramanna G (2007) NIVAReport SNO 5478-2007. Geological storage of CO2: the marinecomponent. Impact on sediments, seawater and marine biota fromleaks. A literature review. pp.43

3. Kim JM, Lee K, Shin K, Kang JH, Lee HW, Kim M, JangPG, Jang MC (2006) The effect of seawater CO2 concentra-tion on growth of a natural phytoplankton assemblage in acontrolled mesocosm experiment. Limnol Oceanogr 51(4):1629–1636

4. Hales B, Chipman D, Takahashi T (2004) High-frequency mea-surement of partial pressure and total concentration of carbondioxide in seawater using microporous hydrophobic membranecontactors. Limnol Oceanogr Methods 2:356–364

5. Tamburi MN, Johengen TH, Atkinson MJ, Shar DWH, RobertsonCY, Purcell H, Smith GJ, Pinchuk A, Buckley EN (2011) Alliancefor Coastal Technologies: advancing moored pCO2 instruments incoastal waters. Mar Technol Soc J 45(1):43–51

6. DeGrandpre MD (1993) Measurement of seawater pCO2 using arenewable-reagent fiber-optic sensor with colorimetric detection.Anal Chem 65:331–337

7. LeFevre N, Ciabrini JP, Michard G, Brient B, DuChauffat M,Merlivat L (1993) A new optical sensor for pCO2 measurements inseawater. Mar Chem 42:189–198

8. DeGrandpre MD, Hammar TR, Smith SP, Sayles FL (1995) Insitu measurements of seawater pCO2. Limnol Oceanogr 40:969–975

9. DeGrandpre MD, Hammar TR, Wallace DWR, Wirick CD(1997) Simultaneous mooring-based measurements of seawaterCO2 and O2 off Cape Hatteras, North Carolina. Limnol Oceanogr42:21–28

10. DeGrandpre MD, Hammar TR, Wirick CD (1998) Short termpCO2 and O2 dynamics in California coastal waters. Deep-SeaRes II 45:1557–1575

11. DeGrandpre MD, Wanninkhof R, McGillis WR, Strutton P (2004)A Lagrangian study of pCO2 dynamics in the eastern equatorialPacific Ocean, submitted to Journal Geophysical Research 109:C08S07, doi:10.1029/2003JC002089

12. Hales B, Burgess L, Emerson SR (1997) An absorbance-basedfiber-optic sensor for CO2(aq) measurement in porewaters of seafloor sediments. Marine Chem 59:51–62

13. Zhao P, Cai W-J (1997) An improved potentiometric pCO2 micro-electrode. Anal Chem 69:5052–5058

14. Cai W-J, Zhao P, Wang Y (2000) pH and pCO2 micro-electrodes measurement and diffusive behaviour of carbondioxide species in coastal marine sediments. Mar Chem 70:133–148

15. Ishiji T, Chipman DW, Takahashi T, Takahashi K (2001) Amper-ometric sensor for monitoring of dissolved carbon dioxide inseawater. Sensors Actuat B76:265–269

16. Wang Z, Cai W-J, Wang Y (2003) Spectrophotometric pCO2

measurements based on a long path length liquid-core waveguidein the South Atlantic Bight. Mar Chem 84:73–84

17. Ishii M (2000) Bunseki 311:65818. Ruzicka J, Hansen EH (1988) Flow injection analysis, 2nd edn.

Wiley, New York19. Mirò M, Estela JM, Cerdà V (2003) Application of flowing stream

techniques to water analysis. Part I Ionic species: dissolved

1092 V. Martinotti et al.

inorganic carbon, nutrients and related compounds. Talanta60:867–886

20. Motomizu S, Toei K, Kuwaki T, Oshima M (1987) Anal Chem59:2930

21. Kuban V, Dasgupta PK (1993) Comparison of photometry andconductometry for the determination of total carbonate by gaspermeation flow injection analysis. Talanta 40(6):831–840

22. Linares MD, Luque de Castro D, Valcarcel M (1985) Anal ChimActa 225:443

23. Aoki T, Fujimura Y, Oka Y, Fujie K (1993) Continuous-flowmethod for the determination of total inorganic carbonate in water.Anal Chim Acta 284:167–171

24. Pasquini C, De Faria LC (1987) Flow-injection determination ofammonia in Kjeldahl digests by gas diffusion and conductometry.Anal Chim Acta 193:19–27

25. Almeida FV, Guimarães JR, Jardim WF (2001) Measuring the CO2

flux at the air/water interface in lakes using flow injection analysis.J Environ Monit 3:317–321

26. Hall POJ, Aller RC (1992) Rapid, small-volume, flow injectionanalysis for ΣCO2, and NH4

+ in marine and freshwaters. LimnolOceanogr 37(5):1113–l 119

27. Carlson RM (1978) Automated separation and conductimetricdetermination of ammonia and dissolved carbon dioxide. AnalChem 50(11):1528–1531

A flow injection analyser conductometric coupled system 1093