Scottish Marine and Freshwater Science Volume 5 Number 1: UK Ocean Acidification Coastal Monitoring Network - Expanding the Network - Defra Contract C5801/ME5309

Materials and Methods

Feasibility Study

A mooring consisting of a McLane RAS (Remote Access Sampler) suspended below a subsurface buoy at depth two metres was deployed in the North Channel (54° 57.4' North 005° 51.7' West) and firmly anchored to the sea bed. The RAS sampler is a self-contained instrument designed to collect up to 48 individual water samples. Each of the 30 ml polypropylene sample holders was connected to a filter unit fitted with a GF/F paper. Addition of a preservative solution was not possible with this model of sampler. The RAS sampler was programmed to collect duplicate samples at 30 minute intervals and to coincide with the moored RAS sampler's programme a Sea-Bird discrete water sampler was deployed to take duplicate samples at depth two metres. As far as practically possible both sets of samples were treated in a similar manner. Samples collected using the Sea-Bird sampler were also filtered through a GF/F paper and stored refrigerated in 40 ml glass EPA vials fitted with septa. At the conclusion of the study the mooring was recovered and 19 samples taken in duplicate) were removed from the RAS sampler and transferred to glass EPA vials and stored refrigerated. Both sets of samples were returned to the laboratory and remained refrigerated for a period of four days prior to analysis for DIC.

Samples were analysed at AFBI for DIC using an automated Apollo 9000 TOC analyser (Teledyne Instruments, Ohio, USA). Following acidification inorganic carbon in the sample was converted to carbon dioxide and the response from a nondispersive infrared detector was compared to stored calibration data to calculate sample concentration in mmol C/l. Calibrants (solutions of sodium carbonate in the range of 0-3 mmol C/l) were run both before and after samples were analysed. Samples were analysed in triplicate.

Measurement of Discrete Samples

Sample Collection

Water samples were collected in the Minch North (Scottish West Coast) by the crew of the MV Isle of Lewis and transported via the SEPA Stornoway office to the laboratory in East Kilbride where salinity and nutrients were analysed. Samples were taken weekly from January to July 2013, however, on several occasions the crew were not able to take samples due to technical issues on the ferry. TA and DIC samples were stored at room temperature prior to being sent to NOC for analysis.

Water samples were collected weekly, weather permitting, from the MSS long term monitoring site at Stonehaven from the MRV Temora between January and July 2013, inclusive. Water samples were collected for nutrients, salinity, TA and DIC analysis. Samples for nutrient analysis were stored frozen, while salinity, TA and DIC samples were stored at room temperature. Samples were analysed for nutrients and salinity by MSS, samples for TA and DIC analysis were sent to NOC.

AFBI collected water samples while undertaking routine maintenance at the Cefas SmartBuoys at the Celtic Deep and Liverpool Bay in March, May and August 2013. TA and DIC samples were stored at room temperature prior to being sent to NOC for analysis.

Total Alkalinity ( TA) and Dissolved inorganic Carbon ( DIC)

Discrete water samples (250 ml) were collected, by individual laboratories, for the determination of DIC and TA into 250 ml glass bottles (Schott Duran) and poisoned with 50 μl saturated HgCl ₂ solution to prevent biological alteration during storage. A head-space of 2.5 mL was left to allow for water expansion and the bottles were sealed using a greased ground glass stoppers to ensure they remained gas-tight. Samples were shipped to National Oceanography Centre Southampton ( NOC) for analysis at the Natural Environment Research Council ( NERC) laboratory.

Analysis was performed using colorimetric and potentiometric open titration cell techniques. Samples were analysed using the Versatile Instrument for Analysis of Titration Alkalinity ( VINDTA 3C, Marianda, Germany) on the two NERC Ocean Biogeochemistry and Ecosystems group carbonate facility ( VINDTA units 11 and 24).

All the samples were heated to 25ºC using a water bath (F12, Julabo, Germany) immediately before analysis. The guidelines followed for analysis are outlined in Mintrop ^[18] and Hartman et al. ^[19] . DIC and TA were analysed in batches of between 9 and 24 samples. Duplicate samples were, where possible, analysed on different NOC Vindta instruments.

Dissolved Inorganic Carbon

DIC was measured by a coulometric titration (coulometer 5011, UIC, USA) following the extraction of CO ₂ from a ~20 ml sub-sample. The DIC section of the VINDTA 3C consists of two main parts where reactions take place. In the first, the sample is acidified with phosphoric acid (H ₃PO _4, 10%) and bubbled through with the inert carrier gas (nitrogen), thereby reducing the pH and converting the carbon species into CO ₂ gas. The CO ₂ gas is then transferred into the second part by the inert carrier gas being cooled to remove water vapour enroute into the coulometer cell, where it is titrated colorimetrically.

The coulometer cell consists of two chambers separated by a sintered glass frit, the cathode and the anode chamber, with a platinum and silver electrode, respectively, and connected to the coulometer to produce a current. The cathode cell is filled with a dimethylsulfoxide solution containing monoethanolamine (HOCH ₂CH ₂ NH _2,) and the pH indicator thymol blue. The produced CO ₂ reacts with the monoethanolamine to form hydroxyethyl carbamic acid (HOCH ₂CH ₂NHCOOH) causing the indicator to turn colourless, increasing the transmittance. The current is activated and the electrons titrate the hydroxethyl carbamic acid returning the pH to the value before the CO ₂ addition, returning the indicator to blue and the transmittance to 29.6%. At the beginning of each session, a blank measurement was undertaken adding only H ₃PO ₄, generating an amount of counts (ideally below 100), which then are subtracted from the end count as well as used to determine the titration endpoint. Analysis concluded after four end-points had been achieved.

Total Alkalinity

Total alkalinity ( TA) was determined by a potentiometric open-cell titration on a ~100 ml sub-sample using a pH half-cell electrode (Orion, Ross, USA) and Ag/AgCl reference electrode (Metrohm, Switzerland). The sample was titrated against hydrochloric acid (HCl, 0.1M) prepared in sodium chloride solution (NaCl, 0.7 M). HCl was added via a Metrohm titrator (0.15 ml additions) until the carbonic acid equivalence point was reached. The pH of the titration was monitored by the pair of electrodes which measured the difference in electromotive force (emf) caused by the change in pH. The emf and the amount of acid added allows the calculation of total alkalinity by a curve fitting method based on a Gran plot approach (Dickson ^[20] ) by the VINDTA software.

Quality Control

Precision

Repeat measurements on a previously analysed samples were undertaken before sample analysis each day (n>3). Instrument precision was better than ± 1.5 μM/kg for DIC and TA. The standard deviation for DIC and TA was calculated for each day of analysis (Figure 2).

The overall precision was calculated as the mean of the standard deviation of all the daily analysis. The precision of Vindta instrument 11 was 1.26 μM/kg and 1.29 μM/kg for DIC and TA, respectively and for Vindta instrument 24 the precision was 1.31 μM/kg and 1.11 μM/kg for DIC and TA, respectively. There is little difference in the standard deviations on the two instruments, showing overall good performance and coherence between both systems. The differences between duplicate samples was higher at 7.49 μM/kg and 7.48 μM/kg for DIC and TA respectively, which may be a consequence of poor sampling.

Instrument Calibration and Monitoring Analytical Performance

Reference Materials ( RM) from Dr Andrew Dickson (Scripps Institution of Oceanography) were used for calibration to assure the accuracy of the measurements. RMs were analysed at the beginning, middle and end of the sessions. The result of the first RM analysis was used in the final calculation to avoid bias in DIC due to gas exchange providing a correction factor (k) for each analysis session.

K= RM _measured/ RM _certificate

RMs were monitored on control charts for each individual instrument. RM lot number 123 was used for the duration of the project.

The mean concentration for RM on Vindta 11 was 2005 ( SD 3.42) μM/kg and 2125 ( SD 8.29) μM/kg for DIC and TA respectively, while for Vindta 24 it was 2013 ( SD 2.13) μM/kg and 2129 ( SD 3.15) μM/kg for DIC and TA, respectively ( Figure 3).

Supporting Determinand Measurements

The supporting determinands, nutrients (phosphate and total oxidised nitrogen) and salinity, were analysed by individual laboratories collecting the discrete water samples. The data was provided to NOC to permit accurate calculation of the TA and DIC concentrations.

SEPA, additionally, collected and analysed water samples for chlorophyll-a to aid data interpretation.

Correction for Salinity

The VINDTA software assigns a default salinity of 35 psu. However, if the sample salinity was known at the time of analysis the TA was corrected for this within the instrument software. The DIC was corrected for salinity post analysis as per the procedure of Friis et al. ^[21] .

CO ₂SYS Routine

The marine carbonate system can be characterised from any two of the four parameters: DIC, TA, pCO ₂ and pH. The Excel program "CO ₂SYS" can be used to calculate the partial pressure of CO ₂, pH, calcite and aragonite saturation states, Revelle factor, carbonate and bicarbonate ion concentrations ^[22] . The combination of input parameters DIC, TA, phosphate and silicate concentrations and laboratory pressure and temperature (0 dbar, 25 Cº) and output conditions (Real temperature and pressure when the sample was taken) were used to calculate the two other carbonate chemistry parameters pCO ₂ and pH.

The dissociation constants of carbonic acid (pK ₁ and pK ₂) determined in real seawater by Millero ^[23] were used as constants in the CO ₂SYS calculation. The pCO ₂ calculation comes from the inaccuracies of the thermodynamic dissociation constants (mainly pK ₁ and pK ₂) and the experimental measurements of the variables used for calculation and can be in the order of ± 7 μatm ^[23] .