DF

Ma

b

c

a

ARA

KDCCSW

1

adsirtmobdHt

UP

0d

Aquacultural Engineering 49 (2012) 10– 17

Contents lists available at SciVerse ScienceDirect

Aquacultural Engineering

j o ur nal homep age : www .e lsev ier .com/ locate /aqua- onl ine

irect and continuous dissolved CO2 monitoring in shallow raceway systems:rom laboratory to commercial-scale applications

aria-Teresa Borgesa,b,∗, Jorge O. Dominguesa, João M. Jesusa, Carlos M. Pereirac

CIIMAR – Centro Interdisciplinar de Investigac ão Marinha e Ambiental, Universidade do Porto, Rua dos Bragas 289, 4050-123 Porto, PortugalDepartamento de Biologia, Faculdade de Ciências da Universidade do Porto, Edifício FC4, Rua do Campo Alegre S/N, 4169-007 Porto, PortugalCentro de Investigac ão em Química and Departamento de Química, Faculdade de Ciências da Universidade do Porto, Rua do Campo Alegre 687, 4169-007 Porto, Portugal

r t i c l e i n f o

rticle history:eceived 10 November 2011ccepted 19 January 2012

eywords:issolved carbon dioxideO2 meterontinuous monitoringhallow racewayater quality

a b s t r a c t

Direct and continuous measurement of dissolved CO2 (dCO2) is crucial for intensive aquaculture, espe-cially in shallow raceway systems (SRS). In this work the performance of a portable dissolved CO2 probeanalyzer was tested for the effects of different aqueous solutions, pure oxygen injection and agitation. Lab-oratory results showed significant (p < 0.05) solution effects on probe performance for low (10–20 mg L−1)and high (30–50 mg L−1) dCO2 concentrations. Globally performance was better in deionized water, fol-lowed by marine fish farm water and artificial seawater. Accuracy and response time were the parametersmost affected by the type of solution tested. Linearity was always observed (R2 = 0.995–0.999). The probewas sensitive to 1 mg L−1 dCO2 increments for concentrations <6 mg L−1 in artificial seawater. Pure oxy-gen injection did not affect probe readouts, and agitation was needed for better accuracy and responsetime. In real marine SRS with tanks in series dCO2 dynamics was revealed using the probe coupled toa developed flow cell. A prototype SRS was built and used to study dCO2 dynamics without endanger-

ing cultivated fish. Generally, results obtained indicate that the probe tested although precise, is bettersuited for discrete, single-point dCO2 monitoring, being a limited resource for the special needs of shal-low raceway systems. As SRS represent a paradigm change in aquaculture, new water quality monitoringstrategies and instrumentation are needed, especially for dCO2. Fiber optic sensors can be a solution forcontinuous, multipoint monitoring, thus contributing to the understanding of water quality dynamics inhyperintensive aquaculture systems.

. Introduction

The current environmental concerns put high pressure on thequaculture industry forcing the development of cleaner pro-uction processes. Recirculating aquaculture systems (RAS) andhallow raceway systems (SRS) are two recent developments aim-ng at achieving greener production coupled with better naturalesources management, which is both economic and environmen-al beneficial. It is consensual that RAS enables the re-use of water

ultiple times before discharge, using different water treatmentptions (e.g. filters and biofilters), while SRS reduces water usagey lowering tank water column, maximizing at the same time fish

ensity and productivity (Øiestad, 1999; Timmons et al., 2002).owever, increases in fish density and water recirculation lead

o increasing concentrations of metabolites in the water, which

∗ Corresponding author at: Departamento de Biologia, Faculdade de Ciências daniversidade do Porto, Edifício FC4, Rua do Campo Alegre S/N, 4169-007 Porto,ortugal. Tel.: +351 220402808; fax: +351 220402709.

E-mail address: mtborges@fc.up.pt (M.-T. Borges).

144-8609/$ – see front matter © 2012 Elsevier B.V. All rights reserved.oi:10.1016/j.aquaeng.2012.01.003

© 2012 Elsevier B.V. All rights reserved.

negatively affect fish performance and the environment. DissolvedCO2 (herein named dCO2) is one of the metabolites whose con-centration increases, not only due to fish respiration but also tobacterial activity in piping systems and biofilters (Summerfeltand Sharrer, 2004). Additionally, salinity negatively affects CO2removal from water, a fact that can further increase dissolved CO2concentrations in marine aquaculture systems (Moran, 2010). Man-aging dCO2 concentration is an important issue in aquaculture. Therequirement for controlling dissolved CO2 concentration, so thatit is not too low for the nitrification processes (Azam et al., 2005)and hence for the quality of recirculated water, or too high, riskingto induce negative effects in fish (Gil Martens et al., 2006), makescontinuous, multipoint, dissolved CO2 monitoring a clear necessity.This is of particular importance in shallow raceways, consideredhyperintensive culture systems, where the very high fish densitiesused, combined with multiple tanks with small water volumes,result in working “on the edge”, without much time for correc-

tive measures in case of system failure (Øiestad, 1999). Recently,under the scope of the EU funded Raceways project, real-time waterquality monitoring (including dCO2) in a commercial SRS fish farmwas addressed, and the importance and difficulties of this task

ural E

rbmØtqciadonaatbsciime

dltits

2

2

Oltwddstuiatiptbiamrstcwasd

M.-T. Borges et al. / Aquacult

ecommended further research (Borges, unpublished data). Itecame also evident that there is not much published work docu-enting important SRS aspects like design and operation (besidesiestad, 1999 and Labatut and Olivares, 2004) and on the effects

hat changes in hydrodynamic characteristics may have in wateruality, as pointed out by Cripps and Poxton (1993) for other fishulture tanks. Defining a SRS prototype and studying dCO2 behav-or in it can be very useful for the aquaculture industry. However,ssessing dCO2 concentration is difficult. The standard APHA proce-ures (Clesceri et al., 1992) are very limited, having a large marginf error when applied to saline water (Pfeiffer et al., 2011) andot enabling direct and continuous monitoring. A commerciallyvailable solution for direct dCO2 evaluation, specified for use inquaculture, is the OxyGuard Portable CO2 Analyzer®. Some impor-ant features of this instrument were evaluated at laboratory scaley Moran et al. (2010) and a comparison with other dCO2 mea-uring methods can be found in Pfeiffer et al. (2011). However,onsidering the present uniqueness of this market solution, themportance of dCO2 evaluation in aquaculture and more specificallyn shallow raceways operation, further research with this instru-

ent is needed, covering situations closer to or in real aquaculturenvironments.

The present study intends to extend knowledge on theCO2 probe recommended for aquaculture, testing parameters

ike type of aqueous solution, pure O2 injection and agita-ion, and to give insights on the utilization of this probe forn situ continuous operation, especially in shallow raceway sys-ems, either in a laboratory prototype or in real aquaculturecenarios.

. Materials and methods

.1. Dissolved CO2 probe characteristics and operation

All the tests described in this work were performed using anxyGuard Portable Dissolved CO2 Analyzer® with built-in data

ogging capability. This instrument is based in infrared gas absorp-ion and was purchased with a CO2 gas pressure sensitive probeith a thermostat for temperature compensation (referred asCO2 probe in this work), a battery powered transmitter withisplay, a battery charger and calibration accessories (beaker,tirrer and appropriate chemicals). Calibration was performed inhe calibration beaker according to manufacturer instructions,sing calibration fluid and stirring. In brief, calibration steps

nclude checking calibration water for CO2 content, generating known amount of CO2 by a chemical reaction and adjustinghe instrument accordingly for 0 and slope. General operationnstructions refer the need of a non-specified flow under therobe, an accuracy dependent on calibration, with expected “prac-ical accuracy up to ±1 mg L−1” (OxyGuard, 2007), maximumattery operating time of 72 h, 15,000 values of recording capac-

ty and output for connection to a PC (G02C2PLOG). It is alsodvised to calibrate for sample salinity and values close to theaximum dCO2 expected. During this work the manufacturer

ecommendations were followed with some modifications, con-isting in preparing standard solutions of known dCO2 based onhe reaction of sodium carbonate (Na2CO3, Merck®, p.a.) withitric acid (C6H8O7, Merck®, p.a.) plus dilutions with Millipore

ater, and addition of solutions with a micropipette for better

ccuracy. This procedure was repeated for the different aqueousolutions tested, using these solutions as solvents for all the stan-ards.

ngineering 49 (2012) 10– 17 11

2.2. Assessment of dCO2 probe performance

2.2.1. Experimental set-up for small scale laboratory testsSmall scale tests were done using either the OxyGuard calibra-

tion beaker (100 mL useful volume) or a laboratory glass vessel(300 mL working volume) and stirring. Further laboratory testswere conducted in a rectangular tank 32 cm long, 60 cm wide, and5 cm water depth, kept uncovered to simulate typical SRS condi-tions. For discrete evaluations known concentrations of dCO2 wereobtained adding sodium carbonate and citric acid until pH < 4.5.For dynamic studies, compressed CO2 aquarium bottles wereemployed (Dupla CO2 Set Delta 400, 500 g bottles) and a flow cellwas specially constructed to hold the dCO2 probe.

2.2.2. Probe calibrationCalibrations for 20 and 50 mg L−1 dCO2 in deionized water or

saline water (NaCl in deionized water, according to the man-ufacturer) were tested. These concentrations were chosen asthey correspond to the maximum advisable for fish cultivation(Timmons et al., 2002) and the maximum range of the instrument,respectively.

2.2.3. Precision, accuracy and response time in different aqueoussolutions

A series of tests was carried out in the calibration beaker withstirring, at room temperature, to evaluate precision (coefficient ofvariation of measurements, CV (%) = SD/mean × 100, after stabiliza-tion), accuracy (difference between known and measured values,mg L−1), response time (time required to achieve a stable reading)and linearity, in the following solutions: deionized water, artifi-cial seawater (after Aminot and Chaussepied, 1983, 3.4% salinity,without NaHCO3) and marine SRS fish farm water (2.2% salinity,after elimination of the carbonate system by acidification, N2 flush-ing and pH restoration). Known dCO2 concentrations tested variedfrom 5 to 50 mg L−1, with 3 independent runs for each concentra-tion, and data acquisition was made every second for up to 15 min.The time to achieve 80, 95 and 99% of dCO2 value obtained after sta-bilization was also calculated. The dCO2 results obtained with theOxyGuard probe were also checked for accuracy using a Gas Sens-ing Electrode (GSE) from Mettler-Toledo. These measurements arebased in the electrode internal pH changes due to CO2 diffusionthrough the electrode membrane and it was calibrated and oper-ated following manufacturer procedures. A specific experimentwas done to compare the simultaneous response of the two instru-ments for dCO2 concentrations from 5 to 35 mg L−1 in deionizedwater.

2.2.4. Probe sensitivity to small dCO2 incrementsSensitivity of the dCO2 probe, defined as the ability to detect

small variations in dCO2 concentrations, was tested in artificial sea-water (formulated as referred above), using the calibration beakerand stirring. Concentrations were varied by the incremental addi-tion of 1 mg L−1 dCO2 every 5 min up to 6 mg L−1 dCO2, using aknown standard solution of sodium carbonate and citric acid. Theprobe was calibrated to 20 mg L−1 dCO2 in NaCl solution (3.5%) andmeasurements were logged at 1 s intervals for 30 min.

2.2.5. Effect of pure oxygen injectionIn intensive aquaculture systems, like SRS, dissolved O2 is kept

at super saturation using pure O2 injection, which can lead to achange in dissolved gas pressures. To assess the possible effect ofhigh dissolved oxygen concentrations in the quality of the probe

measurements, small volume tests were conducted in deionizedwater with stirring, at room temperature, by simultaneously inject-ing compressed O2 and CO2. Dissolved CO2 (dCO2 probe, calibratedfor 20 mg L−1) and pH (pH electrode SenTix 41, WTW multi 340i)

1 ural Engineering 49 (2012) 10– 17

w1

2

dip

2r

uimMlsartt3(ttdiifPaaeaTmdww1h

2s

2

uaSØt1twah10pj∅

F∅

the

Oxy

Gu

ard

Port

able

CO

2A

nal

yser

®p

erfo

rman

ce

(mea

n

±

SD, n

=

3)

in

dei

oniz

ed

wat

er, a

rtifi

cial

seaw

ater

and

mar

ine

fish

farm

wat

er

at

dif

fere

nt

dis

solv

ed

CO

2(d

CO

2)

con

cen

trat

ion

s.

−1)

Aqu

eou

s

solu

tion

Dei

oniz

ed

wat

er

(0%

)

Art

ifici

al

seaw

ater

(3.4

%)

Fish

farm

wat

er

(2.2

%)

M.C

.P

|A|

R.T

.M

.C.

P|A

|R

.T.

M.C

.

P

|A|

R.T

.

6.24

±

0.21

1.12

±

0.62

1.24

±

0.21

4.71

±

0.24

5.73

±

0.13

1.46

±

1.27

0.73

±

0.13

4.64

±

0.44

–

–

–

–10

.91

±

0.43

1.38

±

0.50

0.91

±

0.43

4.88

±

0.36

10.0

3

± 0.

07

0.60

±

0.71

0.04

±

0.06

7.38

±

0.59

11.6

9

±

0.40

1.19

±

0.54

1.69

±

0.40

5.59

±

0.26

15.7

9

±

0.18

1.20

±

0.35

0.79

±

0.18

4.52

±

0.16

14.4

2 ±

0.11

0.46

±

0.43

0.58

±

0.11

6.26

±

0.96

15.0

2

±

0.24

0.96

±

0.58

0.17

±

0.13

5.94

±

0.05

20.1

8

±

0.64

0.96

±

0.19

0.42

±

0.44

5.61

±

0.78

19.2

0 ±

0.11

0.88

±

1.04

0.80

±

0.11

7.86

±

0.55

19.0

8

±

0.36

0.65

±

0.23

0.92

±

0.36

6.19

±

0.68

30.7

4

±

0.78

1.01

±

0.57

0.74

±

0.78

5.98

±

0.98

29.0

3

±

0.64

0.41

±

0.36

0.97

±

0.64

7.34

±

0.48

29.4

9

±

0.30

0.85

±

0.43

0.51

±

0.30

6.21

±

0.54

40.1

6

±

0.40

0.90

±

0.16

0.32

±

0.20

5.53

±

1.62

37.5

4

±

0.10

0.83

±

0.29

2.46

±

0.10

6.65

±

0.78

40.6

9

±

0.85

0.86

±

0.13

0.69

±

0.85

6.85

±

0.54

50.7

5

±

0.41

0.82

±

0.13

0.75

±

0.41

5.37

±

0.40

46.4

9

±

0.24

0.85

±

0.22

3.51

±

0.24

8.27

±

0.60

47.8

8

±

0.92

0.93

±

0.15

2.12

±

0.92

7.93

±

0.38

ed

con

cen

trat

ion

(mg

L−1);

P,

pre

cisi

on

(CV

%),

|A|, a

ccu

racy

(abs

olu

te

valu

es, m

g

L−1);

R.T

.,

resp

onse

tim

e

(min

);

–

not

det

erm

ined

.

2 M.-T. Borges et al. / Aquacult

ere simultaneously monitored for 20 min, with data logging at min intervals.

.2.6. Effect of agitationTo test the dCO2 probe response in the absence of agitation,

iscrete measurements were conducted without turning the cal-bration beaker stirrer on and comparing the data obtained withrevious results acquired under the same conditions with stirring.

.3. Continuous dCO2 measurement in a commercial shallowaceway system

A series of tests was performed in a real aquaculture facilitysing recirculation and shallow raceways for flatfish cultivation,

n order to evaluate the possibility of in situ and continuous dCO2onitoring. A detailed description of this system can be found inatos et al. (2011) and for this work a set of two nursery shal-

ow raceways in series was chosen and four sampling points wereelected, corresponding to the inlet and outlet of the selected topnd bottom tanks. The top tank receives recycled farm water (90%ecycling), and the bottom tank receives reused water from theop tank plus 10% new seawater from a salt water well. Tempera-ure, oxygen, pH (electrodes TetraCon® 32, SenTix 41, WTW multi40i), salinity (YSI Instruments refractometer) and dissolved CO2OxyGuard Portable CO2 Analyzer®) were simultaneously moni-ored. Grab samples were also taken to confirm dCO2 results usinghe Mettler-Toledo GSE. Initially, continuous dCO2 monitoring wasone placing the dCO2 probe directly inside the tanks and initiat-

ng data logging. Afterwards, a portable flow cell was constructedn order to hold the probe and to provide a suitable water flowor continuous measurements. It consisted of an inverted T shapedVC tube with a probe holder at the center and a connection to

small aquarium pump (Eheim Compact 1000, 150–1000 L h−1)t one end. Water flow can be regulated by an adjustable valve,nsuring probe immersion in the holder up to the built-in temper-ture sensor and an adequate water velocity close to the membrane.ests were conducted to assess the reliability of the probe measure-ents when connected to this flow cell (data not shown). With this

evice, the dynamics of dCO2 was followed in the two shallow race-ays in series described above. Additionally, a 1.5 h logging testas done near the outlet of a 3000 cm length, 300 cm width, and

7 cm height shallow raceway tank, temporarily used to stock fish,eavily loaded with turbot above 25 cm length in size.

.4. Dissolved CO2 dynamics in a prototype shallow racewayystem

.4.1. Prototype design and constructionIn order to avoid uncontrolled factors and to be able to sim-

late different situations without endangering cultivated fish, prototype SRS was needed. As no strict guidelines exist forRS construction and operation, besides the information given iniestad (1999) and the real SRS monitored, a compromise solu-

ion was envisaged. As raceways commonly have a L:W ratio of0:1 (Timmons et al., 2002) and most SRS referred in the litera-ure are adaptations of different sized raceways working with lowater depth, we constructed a prototype shallow raceway (here-

fter referred as pSRS) of 200 cm length, 20 cm width, and 20 cmeight, using PVC. The whole system comprehended the pSRS, a00 L pump sump tank with two recirculation pumps (Eheim Series8111, 2500–5000 L min−1) and appropriate tubing. As a turbulentlug flow regime was needed, the water inlet was designed to avoid

et currents. The best configuration tested was a T shaped 6 cm tube, with one horizontal slit, directed to the tank back wall.

inal design for water outlet consisted of three holes (4 cm internal) in the tank bottom, opposite to the water inlet region. During Ta

ble

1Su

mm

ary

of

dC

O2

(mg

L

5 10 15

20 30 40 50

M.C

.,

mea

sur

ural Engineering 49 (2012) 10– 17 13

ooptwwmf

2

sF4atCt2c(a(vmptcwSuia2

2

1MbgT

3

3a

cctorww(dmportDs

M.-T. Borges et al. / Aquacult

peration, water depth inside the tank was varied changing theutlet flow by adjusting the position of ball-valves existent inipes connected to the three outlet holes. These pipes ended athe bottom of the sump tank, to restrict contact of the outflowater with the atmosphere. With this configuration, the maximumater velocity achieved was 9.8 cm s−1 (determined using the floatethod for open channels according to Labatut and Olivares, 2004)

or 7 cm water height and a flow of 83 L min−1 in a 0.4 m2 pSRS.

.4.2. Dissolved CO2 dynamics studiesAs CO2 source, an aquarium type compressed CO2 delivering

ystem was used (Dupla CO2 Set Delta 400, with manual control).or gas injection, a special cylindrical diffuser (Dupla CO2-Reactor00) connected to an Eheim HJ-731, 550 L h−1 pump, was placedfter the water inlet pipe, near the bottom of the tank, parallel tohe left wall. As no fish were used during these tests, a simulation ofO2 production was done using turbot as a model, as it is consideredhe most studied fish in SRS (Øiestad, 1999; Labatut and Olivares,004) and was the main species reared in the real SRS fish farm. Cal-ulations based on experimental data for turbot O2 consumptionImsland et al., 1995), corrected for temperature, and on gener-lly used O2 consumption and CO2 excretion data for salmonidsTimmons et al., 2002; Summerfelt et al., 2004), conducted to aalue of 22 mL min−1 of CO2 (g) to be used. For dCO2 continuousonitoring the developed portable flow cell was placed near the

SRS outlet, also along its left wall. The datalogger was programmedo acquire data at 1 min intervals, starting with CO2 injection. Dis-rete pH monitoring (pH electrode SenTix 41, WTW multi 340i)as done at the same sampling point as above at 5 min intervals.

everal runs were repeatedly done after pSRS flow stabilizationnder different situations of water depth, water velocity and gas

njection. All experiments were performed in a controlled temper-ture room (19 ± 1 ◦C), using artificial seawater (Sera Marine Salts,.9–3.0% salinity).

.5. Statistical methods

Statistical analysis were performed using software Statistica vs.0, StatSoft Inc. Shapiro–Wilk normality test, Linear regression,ann–Whitney U, and Kruskal–Wallis (KW) analysis of variance

y ranks (followed by post hoc multiple comparison of treatmentroups, Siegel and Castellan, 1989) were used where appropriate.he significance level used was 0.05.

. Results

.1. Probe precision, accuracy and response time in differentqueous solutions

A summary of dCO2 probe performance for dissolved CO2 con-entrations from 5 to 50 mg L−1 in different aqueous solutionsan be found in Table 1. Results show that for deionized water,he probe precision ranged from 0.82 to 1.38%; dCO2 values werever read, with minimum accuracy for 5 mg L−1 (1.24 mg L−1) andesponse time ranged from 4.71 to 5.98 min. In artificial sea-ater, precision varied from 0.41 to 1.46%, dCO2 concentrationsere generally under read, with minimum accuracy for 50 mg L−1

3.51 mg L−1) and response time increased when compared toeionized water (minimum of 4.64 min for 5 mg L−1 and maxi-um of 8.27 min at 50 mg L−1). Using marine fish farm water,

recision ranged from 0.65% to 1.19%, accuracy changed fromver readings (maximum of 1.69 mg L−1 for 10 mg L−1) to under

eadings (maximum of 2.12 mg L−1 for 50 mg L−1) and responseime varied from 5.59 (at 10 mg L−1) to 7.93 min (at 50 mg L−1).ata from the 5 mg L−1 experience in this solution were lost and

o are not presented. Some effect of dCO2 concentration in the

Fig. 1. Linearity of OxyGuard Portable CO2 Analyzer® for known dissolved CO2

concentrations (dCO2) in different aqueous solutions.

extreme ranges tested was apparent from the present results, butthe low number of observations per sample (n = 3) precluded fur-ther data analysis. In order to extract more useful informationfrom these results data were clustered into a low concentrationgroup (dCO2 = 10–20 mg L−1, less dangerous range to fish) and ahigh concentration group (dCO2 = 30–50 mg L−1, dangerous rangeto fish). As data deviated from normality even after transforma-tion, non parametric KW analysis of variance was employed. Forthe low concentration group, KW test showed no significant group(aqueous solution) differences for the parameters probe precision(p = 0.08) and accuracy (p = 0.26). Considering response time, signif-icant group differences were found (p < 0.001) and post hoc analysisshowed deionized water mean ranks to be significantly differ-ent from artificial seawater (p < 0.001). For the high concentrationgroup, results were similar for precision (p = 0.33), but significantgroup differences were found for accuracy (p = 0.02) and responsetime (p < 0.01). In both cases, post hoc analysis showed significantmean rank differences only for the pairs deionized water-artificialseawater, with p = 0.01 for accuracy and p < 0.001 for response time.The time needed to reach 80, 95 and 99% of the dCO2 concentra-tion obtained after stabilization is shown in Table 2. Data weregrouped in low and high dCO2 concentration groups as above andKW test was run for solution effects. In the low dCO2 set signif-icant group differences were found for t80%, t95% and t99% of span(p < 0.001). Generally, post hoc tests showed deionized water meanranks differentiating from both saline solutions (except t95%, wereit differentiated only from artificial seawater). For the high dCO2concentration group, results were similar, with significant meanrank differences in all cases between deionized water and the twosaline solutions. A linear regression between known and measureddCO2 concentrations in the range 5–50 mg L−1 dCO2 was obtainedin every situation analyzed in the laboratory (Fig. 1). Calculatedcorrelation coefficients were R2 = 0.999 in both deionized and arti-ficial seawater, and R2 = 0.995 for fish farm water. Results of theperformance test comparisons between the dCO2 probe and GSEMettler-Toledo showed no significant differences between the twomethods (U = 23.0, n1 = n2 = 7, p = 0.90).

3.2. Probe sensitivity to small dCO2 increments

The dCO2 probe revealed an appropriate sensitivity response tolow levels of dissolved CO2 concentrations (2–6 mg L−1 in artifi-cial seawater), as well as a short stabilization time (1–2 min) after1 mg L−1 increments in the concentration of this gas. Neverthe-

less, time needed for stabilization slowly increased in time, andaccuracy decreased, with increasing dCO2 concentrations (Table 3).This test was done under standard operation conditions in thelow concentration range, but it is expected that this stabilization

14 M.-T. Borges et al. / Aquacultural Engineering 49 (2012) 10– 17

Table 2Time (min) required for the OxyGuard Portable CO2 Analyzer® to achieve 80, 95 and 99% of the dissolved CO2 (dCO2) values, obtained after stabilization, for concentrationsranging from 5 to 50 mg L−1 in deionized water, artificial seawater and fish farm water (mean ± SD, n = 3).

dCO2 (mg L−1) t80% t95% t99%

D.W. A.S. F.F.W D.W. A.S. F.F.W D.W. A.S. F.F.W

5 1.80 ± 0.22 2.52 ± 0.15 – 3.56 ± 1.20 3.93 ± 0.35 – 4.67 ± 0.02 5.35 ± 1.00 –10 1.80 ± 0.20 2.57 ± 0.26 2.43 ± 0.14 3.49 ± 1.32 4.92 ± 0.40 4.18 ± 0.72 3.21 ± 0.48 6.86 ± 0.34 5.20 ± 0.3015 1.84 ± 0.13 2.47 ± 0.16 2.34 ± 0.16 3.47 ± 0.32 4.39 ± 0.13 4.05 ± 0.39 4.26 ± 0.50 5.37 ± 0.16 5.29 ± 0.5020 1.86 ± 0.09 2.61 ± 0.05 2.21 ± 0.14 3.65 ± 0.61 5.15 ± 0.13 4.08 ± 0.36 4.57 ± 1.02 7.83 ± 0.56 6.03 ± 0.9930 1.88 ± 0.16 2.96 ± 0.16 2.48 ± 0.12 3.44 ± 0.47 4.93 ± 0.24 4.70 ± 0.25 5.50 ± 1.72 7.03 ± 0.77 6.23 ± 0.6440 2.09 ± 0.43 2.63 ± 0.19 2.55 ± 0.04 3.83 ± 0.72 5.13 ± 0.57 4.38 ± 0.25 4.63 ± 0.48 6.84 ± 0.56 8.09 ± 0.6650 1.92 ± 0.03 2.73 ± 0.15 2.64 ± 0.14 3.68 ± 0.07 5.18 ± 0.25 4.89 ± 0.75 5.09 ± 0.15 7.06 ± 0.28 7.88 ± 0.94

D.W., deionized water (0%); A.S., artificial seawater (3.4%); F.F.W., fish farm water (2.2%); – not determined.

racew

tabmrma

3

igOe

TSs

Fig. 2. Dynamics of dissolved CO2 (dCO2) in two tanks of a commercial shallow

ime will further increase under more realistic situations, and/orfter the addition of larger amounts of dissolved CO2. This sta-ilization time is of crucial importance in a continuous in situonitoring application, as it translates the lag time between a

eal increase in dissolved CO2 concentration and probe measure-ent and display, therefore further delaying possible corrective

ctions.

.3. Effect of pure oxygen injection

The dCO2 probe is a gas pressure sensitive device but no

nformation was found on its behavior in the presence of otherases in the water besides CO2. As in intensive fish farming pure2 is used to supersaturate the water, an effect of this gas, orlse, of the associated increment in total gas pressure, on the

able 3ensitivity of the OxyGuard Portable CO2 Analyzer® to 1 mg L−1 increments of dis-olved CO2 at 5 min intervals in artificial sea water (3.4% salinity).

Known dCO2

(mg L−1)Measured dCO2

(mg L−1)Accuracy(mg L−1)

Stabilizationtime (mm:ss)

2.00 1.75 0.15 –2.99 2.50 0.49 1:263.97 3.50 0.47 1:214.94 4.25 0.69 1:415.90 4.75 1.15 2:16

ay system. Tanks were stacked, with water cascading from top to bottom tank.

diffusion process through the membrane should be tested. Toconfirm changes in dCO2 concentration pH was used, and so it wasexpected that if a decrease in the measured dCO2 was observedwithout a concomitant increase in pH, this would indicate O2hindrance in the probe response. It was observed that the simulta-neous injection of pure O2 and CO2 led to a drop in dCO2 readings,which could not be attributable to probe functioning problemsbut to stripping of CO2 out of the solution, as confirmed by the pHbehavior, which was not kept constant but increased at the sametime.

3.4. Measurements without agitation

Tests performed in artificial seawater without agitation near theprobe membrane showed a negative effect on measurement accu-racy and response time. At 20 mg L−1 dCO2, 60 min were neededfor a stable reading without agitation (only 6.92 min with agita-tion) and for 50 mg L−1 it took 90 min to obtain a stable record(only 6.52 min with agitation). This is not in accordance withthe information provided by the manufacturer, which indicatesa response time in still water of 15 min. Besides, it is not men-tioned any difference in accuracy as a result of this situation. In

fact, for 20 mg L−1 dCO2 accuracy decreased from 1.00 mg L−1 withagitation to 4.25 mg L−1 without it. For 50 mg L−1 dCO2, accuracydecreased from 3.69 mg L−1 to 14.25 mg L−1, with and without agi-tation, respectively.

M.-T. Borges et al. / Aquacultural Engineering 49 (2012) 10– 17 15

Fig. 3. Example of Portable OxyGuard CO2 Analyzer® graphical output for continuous dCO2 monitoring in a real turbot shallow raceway system. Discrete measurementswere done with a Mettler-Toledo GSE (control point A) and with OxyGuard calibration beaker with stirring (control point B).

3

iseSofist1pcbti3dpatdstotwivs2aw

afilt

.5. In situ continuous dCO2 measurement in real SRS

In this experiment it was intended to test the probe at workn a real aquaculture facility, in a pair of stacked 15 m longhallow raceways working in series, at the beginning of the nurs-ry phase (low system biomass) and before feeding the fish.ampling points (4) were strategically chosen to show inlet-utlet and top-down differences in dCO2, irrespective of thesh biomass present. Placing the dCO2 probe directly at eachampling point and starting data logging showed that, althoughhe probe was sensitive to dCO2 in this system, readings took2–15 min to stabilize at each point. Grab samples taken at the sameoints, analyzed using the calibration beaker and stirring,onfirmed the concentrations obtained after only 3–5 min of sta-ilization. Observed dCO2 values increased by 1 mg L−1 from topank inlet to bottom tank outlet, but no trends could be detected forntermediate points. Registered tank water velocities varied from.3 cm s−1 (top tank) to 4.5 cm s−1 (bottom tank). In situ continuousCO2 measurements were acquired using the probe coupled to theortable flow cell developed. This device performed as expectednd reliable records (confirmed using the Mettler-Toledo GSE) onhe evolution of dCO2 in SRS tanks in series were obtained, asepicted in Fig. 2. Temperature and salinity remained stable at all 4ampling points, with values of 20.4 ± 0.3 ◦C and 2.4 ± 0.1%, respec-ively. pH decreased from the inlet value of 7.81 to the outlet valuef 7.50. Oxygen levels decreased from the inlet (14.23 mg L−1) tohe outlet (13.13 mg L−1) of top tank; the disturbance caused byater falling from point 2 (top tank outlet) to point 3 (bottom tank

nlet) further decreased oxygen levels to 10.60 mg L−1 and a finalalue of 9.62 mg L−1 was registered at the bottom tank outlet. Dis-olved CO2 increased from system inlet to the outlet (top tank, from.68 to 3 mg L−1; bottom tank, from 2.5 to 4 mg L−1), as predicted,lthough bottom tank was influenced by extra dCO2 input from saltater well make-up water.

In situ, continuous, long duration (90 min) dCO2 monitoring was

lso performed in a 30 m long shallow raceway heavily stocked withsh, representing an extreme situation, putting additional chal-

enges to the portable flow cell and to the dCO2 probe (fish impacts,urbulence, higher metabolite levels, etc.). Prevailing water quality

characteristics were temperature = 22.2 ◦C, salinity = 2.1%, pH = 7.25and dO2 = 5.33 mg L−1 at the sampling point. An example of thetype of results obtained using the OxyGuard software can beseen in Fig. 3. The reading was stable and maximum dissolvedCO2 values were high (11 mg L−1), in consonance with the higherbiomass present. Grab samples taken after 90 min logging showedgood agreement in dCO2 values, either using the same probe andthe calibration stirrer (0.25 mg L−1 difference) or using the GSE(0.15 mg L−1 difference). Fish were curious about the device usedbut no noticeable behavioral changes were seen or was there anyvisible damage to the flow-cell.

3.6. Dissolved CO2 dynamics in a SRS prototype

Shallow raceways are fish rearing tanks not yet fully character-ized and so a prototype (named pSRS) was built to initiate studieson the effects of operational changes on water quality, more specif-ically on dCO2 dynamics. Several runs were done under differentcombinations of water depth and water velocity under constanttemperature (20 ± 1 ◦C), salinity (2.9%, artificial seawater), and con-tinuous CO2 (g) injection (22 mL L−1) and monitoring (OxyguardPortable CO2 Analyser® and constructed flow cell). The outcomeof a 15 h duration experiment can be seen in Fig. 4. In this case,water depth was kept at 10.6 cm and water velocity at 4.7 cm s−1,approaching real fish farm conditions. Results from CO2 injection(22 mL min−1, part I, Fig. 4) show that dCO2 rapidly rose to 3 mg L−1

and was kept constant throughout 7 h, while pH dropped from 8.03to 7.55. A stress test was performed doubling the CO2 (g) injection(part II, Fig. 4), simulating a situation of higher metabolic activity(as occurs after feeding or at the end of a nursery cycle, were max-imum biomass is achieved). Results showed that dissolved CO2doubled from 2.50 mg L−1 to 4.25 mg L−1 in 1 h, with a concomitantpH decrease from 7.60 to 7.42. However, this rise of dCO2 in thewater was shown faster as a drop in pH by the pH sensor, comparedto the response given 20 min later by the dCO2 probe. An analysis

of system recovery after stopping CO2 injection (part III, Fig. 4)showed that dCO2 concentration returned to the initial 0.75 mg L−1

level within 2 h and pH increased from 7.43 to 7.97 in the same timeperiod. It was also found that dCO2 accumulation in this system

16 M.-T. Borges et al. / Aquacultural En

Fig. 4. Continuous dissolved CO2 profile (solid line) in a shallow raceway prototypes −1 −1 −1

ww

wiil

4

ddwpscddm2ttewta2vt

ratccIapaseAbTcpbmii

ystem for 22 mL min (I), 44 mL min (II) and 0 mL min (III) CO2 (g) injection. pHas simultaneously recorded (dots). Prevailing hydraulic conditions were 4.7 cm s−1

ater velocity and 10.6 cm water depth.

as highly dependent on water velocity (results not shown) point-ng to the importance of maintaining adequate water velocitiesn SRS not only for tank self-cleaning but also to keep low dCO2evels.

. Discussion

The results of the present study confirmed that the portableCO2 probe enables precise and direct measurements of dCO2 inifferent aqueous solutions and dCO2 concentrations, comparingell with instruments like a gas sensing electrode. Globally, besterformance was found in deionized water, followed by the salineolutions, possibly reflecting the effects of different electrolyteompositions and salinity levels on CO2 solubility, despite stan-ard calibration procedures. Comparisons with other studies areifficult due to the paucity of published works and the differentethodologies used. Nevertheless, for precision, CV (%) values for

5 and 50 mg L−1 in deionized water (0.96% and 0.82%) were higherhan those referred by Moran et al. (2010) in a mixture of dis-illed and tap water (0.37% and 0.48%). On the other hand, Pfeiffert al. (2011) found an average CV for this probe of 4.09%, whichas higher (hence lower precision) than the values acquired in

his study (1.46% maximum, in artificial seawater). In addition, theverage time to reach 99% span in standard operation conditions at0 mg L−1 dCO2 was lower (4.57 min in deionized water) than thealue given by Moran et al. (2010) (6–7 min in distilled water plusap water, for 17 mg L−1 dCO2).

The tested analyzer proved to be a robust apparatus, usable ineal and demanding aquaculture environments like SRS. Addition-lly, the portable model, with data logger, enables data recording upo 72 h and exportation to a PC. Nevertheless, this is different fromontinuous monitoring, and as it is commercialized, this instrumentannot be used for this purpose in normal aquaculture operations.n fact, we have experienced that placing the probe directly in realquaculture tanks resulted in long stabilization time per samplingoint and a lack of detection of dCO2 dynamics. This was most prob-bly due to the relatively low water velocities of SRS and the need ofufficient agitation near this probe membrane, as shown by Morant al. (2010) and our results from the with/without agitation tests.lthough an agitation stirrer for discrete measurements is providedy the manufacturer, this equipment cannot be used inside a tank.he need of proper agitation near the probe membrane was suc-essfully overcome in the present work by the development of aortable flow cell. Moreover, calibration is lengthy (as also observed

y Pfeiffer et al., 2011) and the procedure described in the user’sanual is possibly oversimplified and should be improved. Detailed

nformation on probe performance, crucial for aquaculture users,s also lacking, but the tables given in this work can help overcome

gineering 49 (2012) 10– 17

this problem. Also, the probe performed well in oxygen supersatu-rated farm water conditions (175% saturation, according to tablesin Huguenin and Colt, 1989). Despite satisfactory performance inreal aquaculture waters, the utilization of this probe is limited toone tank only, as moving the probe to other tanks with differentconditions (salinity, in particular) results in a need for new calibra-tion and stabilization time. Besides, the user’s manual clearly statesthat moving the probe is not advisable, in order to avoid mechan-ical stress and, in this case, its use as a bench top instrument isrecommended.

Further implications of probe performance results can be fore-seen considering the specific case of marine SRS applicationswhere, according to Øiestad (1999), very high fish densities (often100–500 kg m−3), small water volumes (7 mm to 25 cm waterdepths), and several tanks arranged in racks are used. In this sit-uation, the observed response times can be too long for correctivemeasures to be taken in case of need.

On the other hand, this probe is big (14 cm height, 4 cm ∅) con-sidering the shallowness of SRS. So, for the special needs of shallowraceway systems, this probe is a limited resource.

Overall, our work clearly points out some of the challengesassociated with SRS operation using dCO2 as water quality descrip-tor, and presents some ways toward overcoming these difficulties.Prototype systems, like the one constructed here, and continuouswater quality monitoring are essential tools to help under-standing and improving hyperintensive aquaculture systems. Amultifactorial approach can provide insights on several parame-ter interactions (some yet unknown or not completely understood)that may affect fish performance simultaneously (Foss et al., 2004).This will help defining standards and good management practices(Colt, 2006), namely for dCO2. Moreover, continuous dCO2 mon-itoring can also provide further control of biofilter operation andof degassing units, characterizing their in situ efficiency, reducingcosts and ensuring performance. In addition to continuous dCO2monitoring, multipoint measurements are also necessary to detectdepth stratification of dCO2 and its effect on fish performance, aspreviously studied for dissolved O2 by Reig et al. (2007). Never-theless, such studies cannot be done using the dCO2 probe tested.Fiber optic sensors allow real time and multipoint detection of dif-ferent parameters and can be miniaturized (Caldas et al., 2008).Presently, new optical fiber based chemical sensing platforms arebeing developed under the scope of AQUAMONITOR project (Borgeset al., 2011) with promising preliminary results. We believe thesesensors will contribute to solve the new monitoring challenges inaquaculture.

Acknowledgements

The authors acknowledge the fish farm manager permission touse farm facilities. This work was funded by EC (project RACE-WAYS, COOP-CT/016869/2006-2008) and by Portuguese Scienceand Technology Foundation (Project AQUAMONITOR, FCOMP-01-0124-FEDER-013911 – Ref. FCT-PTDC/AAC-AMB/112424/2009).

References

Aminot, A., Chaussepied, E., 1983. Manuel des Analyses Chimiques en Milieu Marin.Editions CNEXO, Brest, 388 pp.

Azam, F., Gill, S., Farooq, S., 2005. Availability of CO2 as a factor affecting the rate ofnitrification in soil. Soil Biology and Biochemistry 37, 2141–2144.

Borges, M.-T., Pereira, C.M., Gouveia, C., Kovacs, B., Jorge, P.A.S., 2011. Fibre-opticTechnology to Benefit Aquaculture: Portuguese Team Develops New CO2 Sen-sors. Hatchery International, September/October 2011, p. 34.

Caldas, P., Jorge, P.A.S., Araújo, F.M., Ferreira, L.A., Marques, M.B., Rego, G., Santos,

J.L., 2008. Fiber modal Michelson interferometers with coherence addressingand heterodyne interrogation. Optical Engineering 47 (4), 044401–44411.

Clesceri, L.S., Greenberg, A.E., Eaton, A.D., 1992. Standard Methods for the Exami-nation of Water and Wastewater, 18th ed. American Public Health Association(APHA), Washington, DC.

ural E

C

C

F

G

H

I

L

M

M

M

O

Ø

P

R

S

S

S

T

sor. His research work focuses on analytical applicationsof liquid–liquid interfaces, electrochemical sensors andcharacterization of surfaces modified by nanoparticlesand composites films and ionic liquid–electrode inter-faces.

M.-T. Borges et al. / Aquacult

olt, J., 2006. Water quality requirements for reuse systems. Aquacultural Engineer-ing 34, 143–156.

ripps, S.J., Poxton, M.G., 1993. A method for the quantification and optimization ofhydrodynamics in culture tanks. Aquaculture International 1, 55–71.

oss, A., Siikavuopio, S.I., Sæther, B.-S., Evensen, T.H., 2004. Effect of chronic ammoniaexposure on growth in juvenile Atlantic cod. Aquaculture 237, 179–189.

il Martens, L., Witten, P.E., Fivelstad, S., Huysseune, A., Sævareid, B., Vikeså, V.,Obach, A., 2006. Impact of high water carbon dioxide levels on Atlantic salmonsmolts (Salmo salar L.): effects on fish performance, vertebrae composition andstructure. Aquaculture 261, 80–88.

uguenin, J.E., Colt, J., 1989. Design and Operation Guide for Aquaculture SeawaterSystems. Elsevier, New York, USA, 264 pp.

msland, A.K., Folkvord, A., Stefansson, S.O., 1995. Growth, oxygen consumptionand activity of juvenile turbot (Scophthalmus maximus L.) reared under differenttemperatures and photoperiods. Netherlands Journal of Sea Research 134 (1-3),149–159.

abatut, R.A., Olivares, J.F., 2004. Culture of turbot (Scophthalmus maximus) juvenilesusing shallow raceways tanks and recirculation. Aquacultural Engineering 32,113–127.

atos, A., Borges, M.-T., Peixe, C., Henriques, I., Pereira, C.M., Castro, P.M.L., 2011. Amolecular and multivariate approach to the microbial community of a commer-cial shallow raceway marine recirculation system operating with a Moving BedBiofilter. Aquaculture Research 42, 1308–1322.

oran, D., 2010. Carbon dioxide degassing in fresh and saline water. I: Degassingperformance of a cascade column. Aquacultural Engineering 43, 29–36.

oran, D., Tirsgård, B., Steffensen, J.F., 2010. The accuracy and limitations of a newmeter used to measure aqueous carbon dioxide. Aquacultural Engineering 43,101–107.

xyGuard, 2007. OxyGuard CO2 Portable Carbon Dioxide Analyser® . User Manualgb 0907. OxyGuard International A/S, Birkerød, Denmark, pp. 1–7.

iestad, V., 1999. Shallow raceways as a compact, resource-maximizing farmingprocedure for marine fish species. Aquaculture Research 30, 831–840.

feiffer, T.J., Summerfelt, S.T., Watten, B.J., 2011. Comparative performance ofCO2 measuring methods: marine aquaculture recirculation system application.Aquacultural Engineering 44, 1–9.

eig, L., Piedrahita, R.H., Conklin, D.E., 2007. Influence of California halibut (Par-alichthys californicus) on the vertical distribution of dissolved oxygen in araceway and a circular tank at two depths. Aquacultural Engineering 36,261–271.

iegel, S., Castellan Jr., N.J., 1989. Nonparametric Statistics for the Behavioral Sci-ences, 2nd ed. McGraw-Hill International Editions. Statistics Series, 399 pp.

ummerfelt, S.T., Sharrer, M.J., 2004. Design implication of carbon dioxide pro-duction within biofilters contained in recirculating salmonid culture systems.Aquacultural Engineering 32, 171–182.

ummerfelt, S.T., Sharrer, M.J., Hollis, J., Gleason, L.E., Summerfelt, S.R., 2004. Dis-

solved ozone destruction using ultraviolet irradiation in a recirculating salmonidculture system. Aquacultural Engineering 32, 209–223.

immons, M.B., Ebeling, J.M., Wheaton, F.W., Summerfelt, S.T., Vinci, B.J., 2002.Recirculating Aquaculture Systems, 2nd ed. Northeastern Regional AquacultureCenter. Publication No. 01-002, Cayuga AquaVentures, Ithaca, NY, 769 pp.

ngineering 49 (2012) 10– 17 17

Prof. Maria Teresa Borges obtained her PhD in Biologyfrom University of Porto in 1993 in collaboration with Uni-versity of Santiago Compostela (Spain). She is currentlyAssistant Professor. Her research work focuses on waterquality in aquaculture, more specifically in recirculationmarine systems. Studies include biological solutions foreffluent treatment, reuse and valorization (bacterial biofil-tration, microalgae, and integrated aquaculture systems)and direct and real-time system monitoring.

Jorge Domingues has a Master degree in EnvironmentalSciences and Technology from Faculty of Sciences, Uni-versity of Porto, completed in 2010. His master’s thesisfocused on dissolved CO2 monitoring in Aquaculture.

João Jesus graduated in Environmental Sciences and Tech-nology from Faculty of Sciences, University of Porto, in2008 and completed his MSc in the same area. Since 2009he has worked in dissolved CO2 continuous monitoringand is a researcher of AQUAMONITOR project.

Prof. Carlos M. Pereira obtained his PhD in Chemistryfrom University of Porto in 1997 under the supervisionof Prof. Fernando Silva. He is currently Assistant Profes-