Evaluation of a semi-automatic system for long-term seawater carbonate chemistry manipulation

The assessment of the effects of Ocean Acidifi cation (OA) on marine life has received increasing attention in recent marine research. On a mesocosmic scale, the CO2 levels in seawater can be manipulated to evaluate experimentally the consequences of OA on marine organisms (vertebrates and invertebrates). An ideal manipulation of carbonate chemistry should mimic exactly the changes to carbonate chemistry, which will occur in years to come. Although some methods have been described in the literature, here we describe in detail a simple, robust and inexpensive system to produce CO2-enriched seawater by bubbling the seawater with air-CO2 mixtures. The system uses mass fl ow controllers (MFC) to blend atmospheric air with pure C O2 to produce two pCO2 levels. The air-CO2 mixtures are delivered continuously to seawater equilibration reservoirs, a nd simultaneously to an infrared CO2 analyser to verify CO2 levels in the air-CO2 mixture delivered to the equilibration tanks. We monitored both pH and total alkalinity in the equilibration reservoirs over a period of one year in order to document the long-term performance of this system for simulating the future carbonate chemistry of seawater in a coastal laboratory. System performance was suffi cient to maintain three contrasting (e.g., 397, 709 and 1032 matm) and relatively constant (the coeffi cient of variability was 11 %, 9 % and 9 % respectively) seawater pCO2 during a year-long monitoring.


INTRODUCTION
Carbon dioxide (CO 2 ) r eleased due to human activities has caused a rise in the concentration of atmospheric CO 2 , and an increase in dissolved inorganic carbon content of oceanic sur face waters.CO 2 reacts with seawater reducing its pH (a phenomenon named Ocean Acidifi cation (OA)) and carbonate ion concentration (Caldeira & Wickett 2003;Or r et al. 2005;Raven et al. 2005).The carbonate ion reduction results in a decrease in the Aragonite (Ar) and Calcite (Ca) saturation state of seawater (Orr et al. 2005).Therefore, it is predicted that the primar y effect of OA will be the undersaturation of seawater in Ar in a vast part of the world's oceans, in particular high-latitude regions which have naturally low carbonate ion concentrations (Fabry et al. 2009;Orr et al. 2005).
The evidence that OA is occurring is now irrefutable (Orr et al. 2005).OA allows to fundamentally alter marine ecosystems in time frames that are essentially irreversible from the perspective of human societies (Raven et al. 2005).Understanding the impacts of OA in ocean ecosystems will likely be a major subject in marine science in the coming decades.In this context, developing CO 2 systems that allows emulating the future conditions of an ocean with high CO 2 levels is essential.An ideal manipulation of carbonate chemistr y (i.e.CO 2 system) should mimic exactly the changes to carbonate chemistry that will occur in years to come.During the last decade, experimental simulations of future marine carbonate chemistr y have been conducted mainly in North America and Europe, to study the effect of OA on marine organisms (Fangue et al. 2010;Gattuso et al. 2010).A growing body of experimental evidence on OA effects is mainly derived from short-term experiments of typically less than 100 days duration (Kroeker et al. 2013), which normally cor responds to a small proportion of the life span of the respective model organisms.There is therefore a need to develop and test experimental facilities that can allow longer periods of experimentation (i.e. from several months to years) covering one or several life cycles of marine invertebrates, in order to know more about the physiological acclimation and the adaptation capacity of species to OA.
Seawater chemistry can be manipulated in various ways in order to alter the carbonate system (Gattuso et al. 2010).Probably the most common carbonate system manipulation techniques are "aeration techniques" bubbling CO 2 -air mixtures or pure CO 2 to reach a target pH or pCO 2 level.Two major subgroups within these aeration techniques are (1) the "pHstat" and (2) the continuous bubbling with pre-mixed gases (Gattuso et al. 2010).The fi rst is based on a feedback system through continuous monitoring of seawater pH, and the intermittent injection of pure CO 2 or CO 2 -free air to keep pH into a target range.The second is based on the continuous bubbling with an air-CO 2 mixture of a defi ned pCO 2 .
These two aeration techniques share a common weakness: they do not compensate for changes in A T resulting from precipitation and dissolution of CaCO 3 or those changes resulting from evaporation.Furthermore the "pH-stat" is limited by inherent variability associated to the feedback system, while bubbling with premixed gases produces a less variable control on seawater pCO 2 .An accurate comparison between systems are beyond the scope of this paper, however there is clearly a wide variability in the setting and the performance of these systems, necessitating a detailed description of both methods and carbonate chemistr y (Gattuso et al. 2010).
The carbonate chemistr y experimental setup described here corresponds with the "aeration with pre-mixed gases" group.This system was implemented at Calfuco Marine Laborator y (http://www.ciencias.uach.cl/unidades/calfuco/index.php), located in southern Chile (40° S) near the city of Valdivia and under the infl uence of the Valdivia River.The coastal water supplied to the aquaria may therefore experience variations in salinity and other parameters associated with freshwater inputs, in addition to marine variations such as coastal upwelling and tidal cycles.
In this paper we will describe the first laborator y facility to study the consequences of the OA on marine invertebrates in Chile and South America.Moreover we discuss the longterm performance of this system (operative since late 2010) and others considerations in maintaining incubations at different CO 2 levels for extended time periods (i.e.robustness and operational cost of the system).

The pCO 2 generation system
This system manipul ates carbon chemistry by changing Total Carbon (C T ) at constant Total Alkalinity (A T ), using the "Aeration to target pCO 2 " technique (Gattuso et al. 2010).The pCO 2 regulation system described here (Fig. 1) has been adapted to generate air-CO 2 mixtures of 750 ppm and 1200 ppm in dry air.Those nominal levels correspond approximately with projected atmospheric levels for years 2079 and 2127 under RPC 8.5 scenario (Meinshausen et al. 2011;Riahi et al. 2007;Vuuren et al. 2011).Current (control) levels of CO 2 (approx.387-391 Fig. 1: Main characteristics of the Calfuco Marine Laboratory seawater-CO 2 equilibration system.Compressed air (1) is blended with pure CO 2 (2) using MFCs (3).The CO 2 -air mixture is split into two branches, one going to the CO 2 analyser (4) to allow fi ne regulation of the MFCs, and the other to the seawater in the equilibration tanks (5).The pCO 2 of the seawater in the mixing reservoirs can be monitored using a seawater-air equilibrator (6).The equilibrated air is pumped to a CO 2 analyser, with analogue output.The voltage signal is displayed graphically to show the evolution of seawater equilibration with CO 2 .The CO 2 analyser is calibrated with a standard CO 2 -air mixture (7) and with CO 2 -free air (CO 2 removed with soda lime, 8).The gas entering the CO 2 analyser is dried using a desiccant (9) and fi ltered (removing particles larger than 1 µm).The air-CO 2 mixture is injected into the 250 l equilibration tanks or directly into the experimental treatments.Samples for discrete analysis of pH and A T are collected periodically from the equilibration tanks to verify the functioning of the system.
ppm for the 2011 and early 2012 period at the Southern Hemisphere; http://www.csiro.au)were obtained by the equilibration of seawater with atmospheric air.For the 750 ppm and 1200 ppm treatments, we blended dry air with pure CO 2 to each target concentration using an air MFC (Aalborg, model GFC; http://www.aalborg.com)and a CO 2 MFC (Aalborg, model GFC).In total we used 4 MFC valves mounted on a board for horizontal gas fl ow.Dry and fi ltered air is generated by compressing atmospheric air (117 psi) using an oil-free, 4-piston air compressor (Schulz, model MSV12); the dry air passes through a particle fi lter rack (MTA; http://www.mta-it.com) to retain particles larger than 1µm before entering the air MFCs.Pressure in the air line is maintained at approximately 10 psi using a standard regulator before entering the MFC valves.Similarly, pure CO 2 (Research grade, INDURA; http://www.indura.cl) was regulated at 10 psi using a single stage CO 2 regulator.The air fl ow in the air MFCs was set at 5 litre min -1 for both treatments, and the CO 2 fl ows were set initially at 4.06 ml min -1 and 1.56 ml min -1 , to theoretically (assuming that the pCO 2 of the air used in the blend was already 388ppm of CO 2 ) produce a pCO 2 of 1200 ppm and 750 ppm, respectively.Finally we manually adjusted the CO 2 MFC fl ow until the required target pCO 2 of the air-CO 2 mixture was reached.The pCO 2 of the air-CO 2 mixture was continuously monitored using the system described below.

The gas blend pCO 2 monitoring system
The pCO 2 monitoring system was based on a CO 2 analyser (Qubit system, model S151), primarily for measuring the CO 2 content in the air-CO 2 mixture.
The CO 2 analyser voltage output was displayed in a PC using a multimeter interface (RadioShack, model 22-812).Based on this monitoring methodology, we manually adjusted the CO 2 MFCs to achieve the target pCO 2 , and we verifi ed that the pCO 2 remained constant.All air samples passing through the CO 2 analyser were previously dried using a Drierite desiccant column (W.H. Hammond Drierite Co.; http://www.secure.drierite.com) and fi ltered (Millipore 1 µm fi lter) (Fig. 1).The CO 2 analyser was calibrated with air-CO 2 mixture standards of 1,114 ppm manufactured by INDURA (http://www.indura.cl),while zero was obtained by passing the air through a soda lime CO 2 -removal column (Fig. 1).The fl ow rate of the CO 2 -air mixture (samples and standard) entering the CO 2 analyser was kept at approximately 150 ml min -1 using an Air-MFC (Aalborg).

The seawater m ixing reservoirs
Once the experimental treatment gases are produced, each gas is mixed with 1 µm fi ltered seawater (FSW hereafter) in a gas-mixing reservoir bucket.Each reservoir consists of a 250 litre food grade bucket (polyethylene HADAD plastics) into which the air-CO 2 mixture is injected though a large aquarium grade air diffuser.The air-CO 2 mixtures are injected at the bottom of the buckets at ca. 6-7 psi, using plastic tubing and an air-stone bubbler.
Since the air -CO 2 mix is water undersaturated due to condensation in the air compression tank, evaporation of seawater causes an increment in salinity and total alkalinity.The magnitude of this increment is variable, and depends on the intensity of the bubbling relative to the volume of the seawater to be equilibrated, as well as other factors (e.g., temperature).We expect that maximum A T increments occur when a large dry air-CO 2 mixture equilibrates with a small volume of water.In order to assess the magnitude of this effect, we measured the A T in one liter of fi ltered seawater before and after 24 hours of constant air-CO 2 bubbling.In these experimental trials the total alkalinity concentration rose 8-12 mmol l -1 d -1 (an increment of ca.0.4 % of the initial A T ).Although this increment in alkalinity does not affect the fi nal pCO 2 of the equilibrated water, it does affect other aspects of carbon chemistry (e.g., changes in Omega, usually < 1 %).It is calculated that Omega will increase by 6% for each unit of salinity increment, and therefore we used two strategies to prevent or reduce the effect of increment in salinity and alkalinity: 1) frequent changing of the equilibrated water (e.g., total or partial daily replacement of the equilibrated water at 250 L container); 2) bubbling the mixture into the water to allow the air-CO 2 mixture reach water saturation level, before injecting the air-CO 2 mixture into the experimental reservoirs.

Monitoring of pCO 2 , pH and A T in equilibration reservoirs.
The reservoir tanks were cleaned at regular intervals (approximately every 7 days) and the seawater was replaced with fresh FSW.Twelve hours after initiating the bubbling of the new seawater, we collected samples from the equilibration tanks and from the intertidal zone where the inlet of the seawater supply system is located, for analysis of the pH and A T .The equilibrated seawater pCO 2 may also be monitored by pumping the seawater to a gas exchange column or equilibrator (Mini-Module Membrane Contactor; http://www.liqui-cel.com) and measuring the pCO 2 of the equilibrated air.This last feature is particularly useful for determining the equilibration period after renewal of the seawater, i.e. when the pCO 2 -time relationship reaches a plateau (normally in the first 12 h).However this latter procedure does not play any role in the functioning or calibration of the air-CO 2 mixture generation system.pH samples were collected in 50 mL syringes and immediately transferred to a close 25 mL cell, thermostatically controlled at 25.0 °C.The pH was measured at 25.0 °C with a Metrohm 713 pH meter (input resistance > 10 13 Ohm, 0.1 mV sensitivity and nominal resolution 0.001 pH units) and a glass combined double junction Ag/AgCl electrode (Metrohm model 6.0219.100)calibrated with 8.089 Tris buffer (DOE 1994) at 25.0 °C; pH values are reported on the total hydrogen ion scale (DOE 1994).
Seawater samples for A T were poisoned with 50µL of saturated HgCl 2 solution and stored in 250 ml polypropylene bottles in darkness at room temperature until analysis.A T was determined by potentiometric titration in an open cell with 0.05M HCl (Merck Titrisol®) (Haraldsson et al. 1997).The accuracy was controlled against a certifi ed reference material (CRM, supplied by Andrew Dickson, Scripps Institution of Oceanography, San Diego, USA).The correction factor was approximately 1.002, corresponding to a difference of about 5µmol kg -1 .Every sample was analysed with 2 or 3 replicates.
Omega Ar (Ω Ar ) and Omega Ca (Ω Ca ) were estimated from the pH-A T pairs, in addition to temperature, salinity and pressure were obtained with a small CTD (Hydronaut).Carbonate system parameters calculations were performed using CO2SYS software (Lewis &Wallace 1998) available on http://cdiac.ornl.gov/ftp/co2sys/.Seawater pCO 2 , Omega Aragonite and Omega Calcite were calculated using Mehrbach solubility constants (Mehrbach et al. 1973) refi tted by Dickson and Millero (Dickson & Millero 1987).The calculations were performed on a total hydrogen ion scale (pH T ).For KSO 4 we used the constant determined by Dickson (1990).

Uses of CO 2 equilibrated seawater
Depending upon the experimental setting, the equilibrated seawater in the equilibration reservoirs was used in two distinct ways.In one type of experiments a special tubing system was used to allow seawater recirculation between the equilibration reservoirs and the rearing containers, the total volume of equilibrated water circulated through the rearing containers was approximately 10000 times the volume of the organisms.In other type of experiments the equilibrated seawater was used to fi ll the rearing containers and during the experimental rearing each container was connected with tubing to directly inject the required air-CO 2 mixture.In this last case the volume of rearing containers was typically 500-1000 times the volume of the organism.

Evaluation of the system to manipulate carbonate chemistry in the equilibration reservoirs.
Performance Aeration to achieve a target CO 2 will change the amount and speciation of dissolved inorganic carbon in equilibration tanks, in turn modifying pH (Fig. 2A, B), pCO 2 (Fig. 2E) and Ω Ar (Fig. 2F).The pCO 2 of the CO 2 -enriched water (inferred from pH and At measurements) was relatively low compared to the pCO 2 of the air-CO 2 mixtures, particularly for the highest CO 2 / air blend (1200 µatm) which results in a mean seawater pCO 2 of ca.1032 µatm, i.e. 14 % lower (see Fig. 2E).The temperature and A T remained constant between treatments (Fig. 2C, D) but there was a conspicuous variation over the course of the year (Fig. 3B,C), caused by environmental factors such as variable influence of rivers, upwelling, annual temperature cycle, etc.The seawater A T at both the intertidal (inlet) and the equilibration reser voirs was positively cor related with salinity (Fig. 4).However events of low salinity water at the intertidal of Calfuco (e.g., < 30) were rare (Fig. 4).Serie de tiempo de pH calculado a la temperatura in situ (i.e.la temperatura en los tanques de equilibración al momento del muestreo); A T y temperatura en los tanques de mezcla y en el intermareal (T1 es el tratamiento atmosférico, T2 es el tratamiento 750 ppm, T3 es el tratamiento de 1200 ppm, T4 es el intermareal donde se ubica la toma de agua del acuario) durante el 2011 y el comienzo del 2012.
Environmental variations in temperature, A T and salinity explain most of the variability in pH and Aragonite (Ω Ar ) in the CO 2 enriched treatments.For example, based on the ranges of variation in temperature, salinity and A T measured in the atmospheric treatment seawater (i.e.8.4 °C, 9 and 526 µmol kg -1 respectively), we calculated that the maximum fl uctuation in pH and Ω Ar at constant pCO 2 (i.e. at ~380 µatm), should be ~0.13pH units and ~1.7 respectively.
The lower variability of pH in the equilibration reser voirs when compared to the intertidal site (Fig. 2A) demonstrates the capacity of the system to manipulate carbonate parameters, even when critical variables (such as temperature) were not kept constant.

Robustness
The system requires little daily attention with the exception of: (1) draining the compressor reser voir (see Fig. 1) to remove condensed water, and (2) measuring pH and A T in the equilibration reservoirs.The CO 2 mole fraction remained virtually constant for several weeks or even months, and so little adjustment of the MFCs was required.After more than a year of continuous functioning of this system we have not detected any malfunctions.Even during power failures the bubbling in the equilibration tanks restarted automatically when power was restored.

Operational cost
The operational costs of the experimental system are low (less than US$ 2000 per year), consisting mainly in: (1) ultra pure CO 2 (one CO 2 cylinder per year), (2) air-CO 2 standards (one every 1 or 2 years, depending on the frequency of IR calibration), and (3) maintenance or replacement of compressor pistons (probably after 1-2 years).The most signifi cant cost is related to accurate monitoring of pH and A T , including reagents, buf fers, electrodes and reference seawater material (ca.US$ 5000 per year).

DISSCUSION
The "aeration to target pCO 2 " system described here was suffi cient to maintain different pH, pCO 2 and CaCO 3 saturation states in the treatments year round.However the air-CO 2 mixtures (pCO 2 of 750 µatm and 1200 µatm) injected into the equilibration tanks produce CO 2 -enriched seawater with a lower pCO 2 (6 % and 14 % lower, respectively; see Fig. 2E).We estimate that the dilution effect of water moisture on seawater pCO 2 can explain 1-2% of the pCO 2 reduction from the originally dry air-CO2 mixture.Seawater fully equilibrated with a dry air-CO 2 mixture with a pCO 2 = 1200 µatm should produce a maximum seawater pCO 2 of 1176 µatm at 18 °C.The rest of this discrepancy can be attributed to the incapacity of this system to fully equilibrate seawater at high pCO 2 levels in a period of 12 h.Seawater pCO 2 increment slows down as seawater pCO 2 approaches to the target pCO 2 levels, in asymptotic fashion, requiring longer equilibrations periods that the one used here (12 hours).However, in spite of the discrepancies discussed above, this system had the capacity to produce significantly different CO 2 levels in seawater (Fig. 2E) with a relatively low variability (CV= 9 %).This overall variability includes the ef fect of the natural variability in salinity, alkalinity and temperature, as well as the bias in the pH and A T measurements, and uncer tainties in the pCO 2 calculation.The error of A T analysis was constrained using reference material.The pH er ror was calculated to be lower than 0.006-0.009pH units (Torres et al. 1999) however in the few occasions when salinity dropped below 30, additional bias, associated with larger differences in the liquid junction potential between the buffer and the sample, are expected to become signifi cant (Wedborg et al. 2007).Tests on the same sample using TRIS buffer, at salinity 35 and 25, show minor discrepancies (< 0.005 pH units); we expect that even in this particular example (salinity range between 28-30) the pH error does not exceed 0.01 pH units.If we consider that maximum uncertainties of pH and alkalinity are in the order 0.01 pH units and ca. 9 in alkalinity, we expect a total error of ca.10 µatm for our pCO 2 estimations (Torres et al. 1999), which in turn corresponds with a small percentage of the pCO 2 of the CO 2 -enriched treatments (ca. 1 %).Thus most of the overall variability repor ted here arises because of changes in temperature, salinity and alkalinity over time.Despite less than full control over these parameters, the system was suffi cient for long term experimentation at contrasting pCO 2 (Fig. 2E), pH (Fig. 2A) and Omega Ar (Fig. 2F) levels.
The natural variability of pH and pCO 2 along the coast of Chile (Torres et al. 2011;Torres et al. 1999) is extreme when compared to other geographic areas (e.g., Tropical waters (Astor et al. 2005)), hence the fl uctuation in carbonate system parameters in the different treatments shown here is not unusual for marine life in this region.Moreover a fully constant chemostat (for pH or Ω Ar ) might be considered unrealistic for simulating high-CO 2 scenarios to which coastal organisms in Eastern Boundary Current systems may be exposed.Therefore we conclude that the performance of the system described here is adequate to simulate high-CO 2 scenarios for Chilean coastal waters under laboratory conditions.
The robustness of this system makes it suitable for short and long-term experiments (months to years), necessar y to adequately investigate the consequences of OA on marine inver tebrates.Using the system described here we have been able to rear egg-capsules of Concholepas concholepas (Bruguière, 1789) during almost their entire developing period until hatching (i.e., 30-60 days of rearing), and small juveniles of the same species originating from competent larvae collected in the fi eld (i.e. 1 to 2 years of rearing, (Manríquez et al. 2013)).Other species such as the mitilid Mytilus chilensis (Hupé, 1854) and the intertidal snail Acanthina monodon (Pallas, 1774) have also been reared in this system (Navarro et al. 2013).These experiments carried out so far, have shown signifi cant effects of OA on the studied species, both positive and negative (e.g., Navarro et al. 2013).This highlights that the system is well suited for long-term experiments investigating the consequences of OA on the performance of early ontogenetic stages of marine invertebrate species.Finally, it is important to highlight that the relatively lowcost maintenance (ca.US$ 20 per day) makes this system economically feasible for a wide range of marine laboratory facilities.

Fig. 2 :
Fig. 2: Statistical parameters: pH measured at 25 °C, pH calculated at in situ temperature, temperature, total alkalinity, partial pressure of CO 2 and Omega Aragonite, in the equilibration reservoirs and intertidal (T 1 for atmospheric treatment, T 2 for 750 ppm treatment, T 3 for 1200 ppm treatment, T 4 for the intertidal site where the aquarium inlet is located) during 2011 and early 2012.The box plot indicates the smallest observation (sample minimum), lower quartile (Q1), median (Q2), upper quartile (Q3), and the largest observation (sample maximum).

Fig. 3 :
Fig. 3: Time series pH calculated at in situ temperature (i.e. the temperature in the equilibration tanks at the moment of sampling); A T and temperature in the equilibration reservoirs and intertidal (T 1 for Atmospheric treatment, T 2 for 750 ppm treatment, T 3 for 1200 ppm treatment, T 4 for the intertidal site where the aquarium inlet is located) during 2011 and early 2012.

Fig. 4 :
Fig. 4: A T -Salinity relationship in the equilibration reservoirs and the intertidal (T 1 for Atmospheric treatment, T 2 for 750ppm treatment, T 3 for 1200ppm treatment, T 4 for the intertidal site where the aquarium inlet is located) during 2011 and early 2012.Relación entre A T y salinidad en los tambores de equilibración y en el intermareal (T1 es el tratamiento atmosférico, T2 es el tratamiento 750 ppm, T3 es el tratamiento de 1200 ppm, T4 es el intermareal donde se ubica la toma de agua del acuario) durante el 2011 y el comienzo del 2012.