Winter measurements of oceanic biogeochemical parameters in the Rockall Trough ( 2009-2012 )

Abstract. This paper describes the sampling and analysis of biogeochemical parameters collected in the Rockall Trough in January/February of 2009, 2010, 2011 and 2012. Sampling was carried out along two transects, one southern and one northern transect each year. Samples for dissolved inorganic carbon (DIC) and total alkalinity (TA) were taken alongside salinity, dissolved oxygen and dissolved inorganic nutrients (total-oxidized nitrogen, nitrite, phosphate and silicate) to describe the chemical signatures of the various water masses in the region. These were taken at regular intervals through the water column. The data are available on the CDIAC database, http://cdiac.ornl.gov/ftp/oceans/Rockall_Trough/ .


Introduction
Between February 2008 and August 2010 a pilot project to initiate research in ocean carbon processes in Irish marine waters was carried out jointly by the National University of Ireland, Galway (NUIG), and the Marine Institute, Ireland (MI).The project titled Increased Atmospheric CO 2 on Ocean Chemistry and Ecosystems was carried out under the Sea Change strategy with the support of the Marine Institute and the Marine Research Sub-Programme of the National Development Plan 2007-2013(O'Dowd et al., 2011).Through collaboration with annual MI winter surveys, a range of biogeochemical parameters were measured across the Rockall Trough in January or February of 2009February of , 2010February of , 2011February of and 2012.The Rockall Trough plays an important role in the global thermohaline circulation as it provides a pathway for warm saline waters of the upper North Atlantic to reach the Nordic Seas.There is also a complex interaction of water masses of different origins in the trough (McGrath et al., 2012b) and therefore it is an important region in oceanclimate research.The 2009 and 2010 data have recently been compared with data measured in the trough in the 1990s by the World Ocean Circulation Experiment (McGrath et al., 2012a, b).

Data provenance
All surveys were carried out on the RV Celtic Explorer; see exact dates in Table 2.While conductivity, temperature and depth (CTD) data are available for every station in Fig. 1, inorganic carbonate parameters were generally measured every second station in 2009 and 2010.In 2011, only 5 surface carbonate samples were taken for inter-laboratory comparison Published by Copernicus Publications.with samples analysed at Scripps Institution of Oceanography (SIO), while in 2012 carbonate samples were taken at every station along the southern transect.Salinity and nutrients were measured along both transects every year.Stations were approximately 27 km apart, except for along the shelf edge where there was greater horizontal sampling resolution.
3 Methods and quality control procedures

Hydrography
A Seabird CTD profiling instrument (SBE 911) with water bottles on a rosette was used on each survey.Temperature and conductivity sensors were sent to Seabird annually for calibration.On every survey there was a primary and secondary temperature and conductivity sensor set on the CTD, which were compared to ensure they match to a high level of precision.Processed salinity sensor data were also compared with the discrete water samples analysed on a Guildline Portasal salinometer (Model 8410A) at the MI.On every survey the sensor-laboratory comparison resulted in an r 2 value greater than 0.999, while the average difference between salinity sensor and bottle data was −0.010 for CE0903, −0.003 for CE10002, −0.004 for CE11001 and −0.013 for CE12001 (where the sensor data was consistently lower than the bottle data).The CTD sensor data will be calibrated with these offsets in the second version of the GLODAP database (Key et al., 2004), currently being developed.An SB43 oxygen sensor was deployed with the CTD, which was also calibrated annually with the manufacturer.On surveys where oxygen samples were taken, the oxygen sensor data was calibrated with the laboratory results.

Collection of water samples for chemical analysis
Water samples were collected in Niskin bottles attached to the CTD rosette.Two Niskin bottles were filled at every depth.At the beginning of every survey all Niskin bottles and taps were checked for leaks, which were fixed prior to sampling.Samples were drawn from the deepest depths first, and chemical parameters were drawn from the first Niskin at each depth in the following order: dissolved oxygen, dissolved inorganic carbon and total alkalinity, dissolved nutrients and salinity.

Dissolved inorganic carbon and total alkalinity
The Guide to Best Practices for Ocean CO 2 measurements (Dickson et al., 2007), which describes the standard methods now in use for the determination of these parameters, was followed for the sampling and analysis of DIC (dissolved inorganic carbon) and TA (total alkalinity).was first rinsed before filling slowly from the bottom.The water was overflowed by approximately 1 bottle volume.Using a pipette, a headspace (∼ 2 mL) was left in the top of the bottle to allow for water expansion, then 0.1 mL of saturated mercuric chloride solution was added to poison the sample.

Sampling
The glass stopper was greased with Apiezon L Grease before arriving at the station.After the sample was poisoned, excess water was wiped from the neck of the bottle and the stopper was twisted slowly into place, squeezing the air out of the grease.Finally the stopper was clamped in place using 3 thick elastic bands.The bottle was inverted several times to disperse the mercuric chloride and the sample was stored in a cool, dark location and analysed on land.Any seawater or materials contaminated with mercuric chloride during sample collection and analysis were collected and disposed of by a chemical waste disposal company in full compliance with hazardous waste regulations.In CE10002, 13 of the 95 DIC and TA samples were collected in separate containers (by the same method described above) due to insufficient borosilicate glass bottles.DIC was collected in 250 mL amber glass bottles with ground glass stoppers and TA was collected in 500 mL high-density polyethylene (HDPE) bottles with screw caps.As TA concentration is not affected by CO 2 in the atmosphere and biological activity was at a minimum due to winter sampling, it was not necessary to seal the TA samples in the same manner as DIC.These samples were consistent with the vertical profile from that year and with other surveys in the region.
7 second sample, while replicate bottles were taken for 5-10% of the total sample number from each survey.The accuracy and precision of the measurements were calculated as the average and standard deviation, respectively of the differences between duplicate samples, Table 3. Table 3 Accuracy and precision of DIC and TA in µmol kg -1 for each of the surveys, calculated as the average and standard deviation, respectively of the differences between duplicate samples, where n is the number of duplicate samples.

DIC and TA Storage Expe riment
A storage experiment was carried out to investigate if storing samples for a prolonged length of time had an effect on the DIC and TA concentration.On May 21 st 2010, twenty-six 500ml Schott Duran bottles were filled with water taken from 448m deep along the shelf edge (10.0322ºW, 55.2565ºN).All bottles were filled by the author (TMG), while a colleague poisoned, greased and sealed the bottles.Six 250ml glass (not-borosilicate) bottles, used for DIC only, were also tested to ensure these bottles did not affect the stored samples differently than the Schott Duran bottles.All samples were poisoned with mercuric chloride, stored at 4ºC in a dark fridge until they were analysed.The first set of samples (T=0) was run on the 29 th May 2010 and a new set of samples was run monthly for the first 7 months (with one exception), with subsequent analysis every 2-3 months.A duplicate of every bottle was run, and the final result for each bottle be low (Figure 3) is an average of the duplicate values.The 8-days from when the samples were drawn until T=0 should not affect the initial concentration as samples are not expected to be

Analysis
DIC was measured on a VINDTA-3C (Versatile Instrument for the Determination of Titration Alkalinity) system with UIC coulometer (Dickson et al., 2007).A known volume of sample is acidified with phosphoric acid in order to convert all dissolved inorganic carbon to CO 2 and the resulting CO 2 , forced out of the sample using nitrogen as a carrier gas, is titrated coulometrically (Johnson et al., 1987(Johnson et al., , 1993)).TA was analysed by potentiometric titration with 0.1 M hydrochloric acid, also on the VINDTA 3C (Mintrop et al., 2000).During the titration the bases in the TA definition (Dickson, 1981) are transferred to their acidic forms and the titration is monitored by a pH electrode that measures the electromotive force (emf).The process is controlled by the LabVIEW ™ software and the endpoint is determined by the change in pH against the volume of acid added to the solution.The result of the titration is evaluated with curve fitting (Mintrop et al., 2000).

Quality control
The accuracy of both DIC and TA analysis was ensured by analysing duplicate certified reference materials (CRMs) before every batch of samples.CRMs were provided by A. Dickson, Scripps Institution of Oceanography, USA (Dickson et al., 2003).If many samples (> 10) were run in a single batch, another duplicate CRM was run at the end of the day.The mean of the measured CRM results was used to calculate a CRM correction factor to adjust DIC and TA sample results for any offset in the VINDTA.
CRM correction factor = certified value/measured value  et al., 2012a).All suggest that there is no problem/offset in the TA results generated from this batch.By using the CRM correction factor, results were corrected for the larger offset in the instrument during this time.Duplicate samples from the same bottle were run every second sample, while replicate bottles were taken for 5-10 % of the total sample number from each survey.The accuracy and precision of the measurements were calculated as the average and standard deviation, respectively of the differences between duplicate samples, Table 3.

DIC and TA storage experiment
A storage experiment was carried out to investigate if storing samples for a prolonged length of time had an effect on the DIC and TA concentration.On 21 May 2010, twentysix 500 mL Schott Duran bottles were filled with water taken from 448 m deep along the shelf edge (10.0322 • W, 55.2565 • N).All bottles were filled by the author (T.Mc-Grath), while a colleague poisoned, greased and sealed the bottles.Six 250 mL glass (not-borosilicate) bottles, used for DIC only, were also tested to ensure these bottles did not affect the stored samples differently than the Schott Duran bottles.All samples were poisoned with mercuric chloride, stored at 4 • C in a dark fridge until they were analysed.The first set of samples (T = 0) was run on 29 May 2010 and a new set of samples was run monthly for the first 7 months (with one exception), with subsequent analysis every 2-3 months.A duplicate of every bottle was run, and the final result for each bottle below (Fig. 3) is an average of the duplicate values.The 8 days from when the samples were drawn until T = 0 should not affect the initial concentration as samples are not expected to be contaminated that quickly if the bottles are sealed correctly (personal communication with staff at Andrew Dickson's laboratory at SIO, 2013).The mean and standard deviations for DIC were based on the first year of storage when concentrations were within ±3 µmol kg −1 around the mean, while both years of storage results were used for TA as they were all within ±2 µmol kg −1 of the mean.
The average DIC at T = 0 was 2143 µmol kg −1 (analysed from 4 replicate sample bottles, where a duplicate was run from each bottle).The average DIC over the first full year of storage was 2142 µmol kg −1 , and variation around the mean is less than ±3 µmol kg −1 .Both Schott Duran and soft glass bottles had similar concentrations after a year.There was greater variability in DIC concentrations in the second year of storage, results from one month (July 2011) were discarded as concentrations were over 10 µmol kg −1 below the mean.
The average TA at T = 0 was 2331 µmol kg −1 , and remained constant for the 26 months of storage, with variation of less than ±2 µmol kg −1 around the mean.Results from one month (June 2010) were discarded as concentrations were 8 µmol kg −1 above all other months and appear to be a oneoff error.
Results indicate that, while TA samples can be stored for at least 2 yr, DIC samples should be analysed within 1 yr of sampling.All samples collected in the Rockall Trough were analysed well within 1 yr of sampling.

Cross validation of DIC and TA analysed at Scripps Institution of Oceanography
In the survey CE11001 across the Rockall Trough in January 2011, a batch of surface DIC and TA samples was sent to Scripps Institution of Oceanography, USA, for analysis.Five replicates of these samples were analysed by the author (T.McGrath) at NUIG, where a duplicate was run for every sample.DIC and TA concentrations from NUIG were within ±3.1 µmol kg −1 and ±1.3 µmol kg −1 , respectively, of those analysed at SIO.

Sampling
All equipment involved in the sampling and filtration of nutrient samples was acid-cleaned in 10 % hydrochloric acid prior to sampling (Grasshoff et al., 1999).Water for nutrient samples was collected from the Niskin bottle in 1 L HDPE bottles.The 1 L bottles were first rinsed three times with sample water before filling.The sample was filtered through a 0.40 µm polycarbonate filter and the filtrate was poured into two 50 mL polypropylene tubes.The tubes were immediately frozen upright at −20 • C and analysed on land.

Analysis
Seawater samples were analysed for total oxidized nitrogen (TOxN), nitrite, silicate and phosphate on a Skalar San ++ continuous flow analyser at the Marine Institute (Grasshoff et al., 1999).The Skalar San ++ System uses automatic segmented flow analysis where a stream of reagents and samples, segmented with air bubbles, is pumped through a manifold to undergo treatment such as mixing and heating before entering a flow cell to be detected.The sample is pumped into the system and split into 4 channels where it is mixed with reagents.The reagents act to develop a colour, which is measured as an absorbance through a flow cell at a given wavelength.

TOxN
The Skalar method for the determination of TOxN is based on Greenberg et al. (1980), ISO 13395 (1996), Navone (1964) and Walinga et al. (1989).The sample is first buffered at a pH of 8.2, with a buffer reagent made of ammonium chloride and ammonium hydroxide solution, and is then passed through a column containing granulated copper-cadmium to reduce nitrate to nitrite.The nitrite, originally present plus reduced nitrate, is determined by diazotizing with sulfanilamide and coupling with N-(1-naphthyl)ethylenediamine dihydrochloride to form a strong reddish-purple dye which is measured at 540 nm.

Nitrite
The Skalar method for the determination of nitrite is based on EPA (1974), Greenberg et al. (1980) andISO 13395 (1996), where the diazonium compounds formed by diazotizing of sulfanilamide by nitrite in water under acidic conditions (due to phosphoric acid in the reagent) is coupled with N-(1-naphthyl)ethylenediamine dihydrochloride to produce a reddish-purple colour which is measured at 540 nm.

Silicate
The Skalar method for the determination of silicate is based on Babulak andGildenberg (1973), ISO 16264 (2002) and Smith and Milne (1981).The sample is acidified with sulfuric acid and mixed with an ammonium heptamolybdate solution forming molybdosilicic acid.This acid is reduced with L(+)ascorbic acid to a blue dye, which is measured at 810 nm.Oxalic acid is added to avoid phosphate interference.

Phosphate
The Skalar method for the determination of phosphate is based on Boltz and Mellon (1948), Greenberg et al. (1980), Walinga et al. (1989) andISO 15681-2 (2003), where ammonium heptamolybdate and potassium antimony (III) oxide tartrate react in an acidic medium (with sulfuric acid) with diluted solutions of phosphate to form an antimony-phosphomolybdate complex.This complex is reduced to an intensely blue-coloured complex by L(+)ascorbic acid and is measured at 880 nm.
Based on the daily calibration standards, concentrations of nutrients can be quantified up to the maximum calibration standard concentration.If sample concentrations fall above this range (Table 4), they must be diluted with artificial seawater.

Nutrient's quality control
The accuracy of the nutrient analysis was ensured by running Eurofins CRMs (http://www.eurofins.dk/dk/milj0/reference-materialer.aspx) with every batch of samples and controlled using quality control (QC) charts according to best practice as per ISO 17025.The system is also calibrated in every run using seven calibration standards made up daily in the laboratory.A replicate of every sample is analysed and the relative percent difference (RPD: difference between the two values/mean × 100) of the results greater than the limit of quantification should be ≤ 10.
To assess the accuracy of the nutrient methods and procedures the MI participates in the QUASIMEME laboratory quality control programme (www.quasimeme.org).Test materials, analysed twice a year, have a large range of concen- The neck of the bottles was dried well with clean kim wipes to prevent salt crystals forming on the top.A plastic insert was then placed into the bottle to produce a tight seal to prevent evaporation, followed by closing the bottle with the screw cap.Samples were stored upright at room temperature.trations from below the detection limit to high concentrations that have to be diluted.The laboratory performance is expressed with a z score where |z| < 2 is considered acceptable, where z is the difference between the laboratory result and the assigned value divided by the total error (Cofino and Wells, 1994).Between October 2008 and May 2012 the MI participated in 8 rounds of QUASIMEME proficiency testing scheme exercises (51 samples) for nutrients in the marine environment.The average z score for all nutrients was < 0.5, see Fig. 4. The MI is accredited to ISO 17025 for nutrient analysis in seawater and is audited annually by the Irish National Accreditation Board, INAB.

Sampling
Salinity samples were collected in clear glass salinity bottles with plastic screw caps.The bottle was first rinsed three times with the sample water before filling up to the shoulder of the bottle.The necks of the bottles were dried well with clean Kimwipes to prevent salt crystals forming on the top.
A plastic insert was then placed into the bottle to produce a tight seal to prevent evaporation, followed by closing the bottle with the screw cap.Samples were stored upright at room temperature.

Analysis
Salinity was analysed on a Guildline Portasal salinometer at the MI, where 4 electrode conductivity cells suspended in a temperature-controlled bath measure the conductivity of the sample.The conductivity is related to salinity by calibration from a known standard.

Dissolved oxygen
Dissolved oxygen (DO) samples were collected and analysed as per the standard operating procedure of Dickson (1995).

Sampling
DO samples were collected in 250 mL iodine bottles with plastic stoppers.All bottle-stopper pairs are individually calibrated and exact bottle volumes are used in final oxygen calculations.The bottle was filled from the bottom using silicone tubing, care was taken to minimize bubbles when filling and the water was overflowed by 3 flask volumes.Two millimetres of the pickling reagents, MnCl 2 (no. 1) and NaOH/NaI (no.2), were added immediately to the sample, before carefully inserting the stopper and inverting the bottle several times.After the precipitate had settled at least half way, the bottle was shaken again.Samples were then stored in a cool dark location until titration, which was mostly carried out within 12 h of sampling.

Analysis
Oxygen samples were analysed using a modified Winkler method (Dickson, 1995), where the titration is carried out in the sample bottle.The sample is first acidified with sulfuric acid (H 2 SO 4 ) to a pH between 1 and 2.5, which dissolves the hydroxide precipitates, and iodide ions added by reagent no. 2 are oxidized to iodine by the manganese (III) ions, which are reduced to Mn(II) ions in the process.In the final step, the iodine is reduced to iodide by titration with sodium thiosulfate, the amount of iodine generated, which is equivalent to the amount of oxygen in the sample, is determined by the amount of thiosulfate required to reach the endpoint.A Metrohm 848 Titrino Plus, with a Metrohm combined Pt electrode was used to determine the endpoint, i.e. potentiometric endpoint determination, measuring the change in redox potential of the sample, which reaches a minimum at the endpoint (Furuya and Harada, 1995).This method of determination was also used effectively by numerous WOCE cruises in the Atlantic and Pacific oceans, and also on some Hawaii Ocean Time Series (HOT) cruises (http://www.soest.hawaii.edu/HOT_WOCE/).

Oxygen quality control
Before titration of the samples, duplicate reagent blanks were determined and duplicate standardization of the sodium thiosulfate titrant was carried out.The reagent blank should ideally be less than 0.01 mL, while the duplicate thiosulfate standardization should typically fall within 0.002 mL of each other (Dickson, 1995).Standardization of the thiosulfate is carried out in precisely the same conditions that the samples are analysed under so that any iodine lost through the volatilization or gained by the oxidation of iodide while analysing the seawater samples is compensated for with similar errors occurring during the standardization procedure (Knapp et al., 1989).Precision of the samples is estimated by running duplicate samples every 10-15 samples.

Secondary quality control
Secondary QC was carried out on all four data sets using cross-over analysis (Tanhua, 2010;T. Tanhua and S. Lauvset, personal communication, 2013) with other data sets in the northeastern Atlantic in the Global Ocean Data Analysis Project (GLODAP) (Key et al., 2004) and CARbon dioxide IN the Atlantic Ocean (CARINA) (Key et al., 2010).Crossover analysis is an objective comparison of deep water data (in this case below 1500 m) from one cruise with data T. McGrath et al.: Winter measurements of oceanic biogeochemical parameters from other cruises in the same area to quantify systematic biases in the data set (Tanhua, 2010).Results from the Rockall Trough should be approached with caution as there is natural oceanographic variability in this region.In all four Rockall surveys (2009-2012) there appears to be a negative offset in salinity (of approx.0.01 each year), along with silicate (∼ 1 µmol kg −1 ), relative to other cruises in GLODAP and CARINA.However, we believe that the change in salinity is natural and may suggest deeper water with more northerly component (lower salinity, lower nutrients and higher oxygen -all of which were observed in recent years).There were no adjustments required for the DIC and TA data, except for the 2009 TA which showed a negative offset of 11 µmol kg −1 .A very small portion of this may be related to the lower salinity observed, however such a large difference is unlikely to be true.This negative offset in the 2009 TA data had previously been observed through the water column (therefore not due to CaCO 3 formation) and so was not included in McGrath et al. (2012a).Preliminary results suggest that this parameter will be adjusted in the second version of the GLODAP database, which is currently being developed.

Figure 1 .
Figure 1.Station positions from the surveys CE0903 (February 2009), CE10002 (February 2010), CE11001 (January 2011) and CE12001 (January 2012).Stations with triangle symbols are where carbonate parameters were measured.Note in CE11001 only surface samples were taken, in other years samples were taken through the water column.

Figure 2
Figure 2 DIC and TA CRM measured values plotted against certified values for each of the surveys.Dashed lines indicate a new survey.

Figure 2 .
Figure 2. DIC and TA CRM measured values plotted against certified values for each of the surveys.Dashed lines indicate a new survey.

9Figure 3 .
Figure3.Average monthly (from at least 2 sample bottles) DIC and TA concentrations over 2 yr of storage.The mean and standard deviations for DIC were based on the first year of storage when concentrations were within ±3 µmol kg −1 around the mean, while both years of storage results were used for TA as they were all within ±2 µmol kg −1 of the mean.

13Figure 4
Figure 4 z-scores of 8 QUASIMEME rounds of nutrients in the marine environment between October 2008 and May 2012, where the green dashed lines indicate a z-score of 2. The dashed black line at a z score of ±3 designates unsatisfactory performance.Most of the Quasimeme samples were in the range of oceanic values measured in the Rockall Trough.

Figure 4 .
Figure 4. z scores of 8 QUASIMEME rounds of nutrients in the marine environment between October 2008 and May 2012, where the green dashed lines indicate a z score of 2. The dashed black line at a z score of ±3 designates unsatisfactory performance.Most of the QUASIMEME samples were in the range of oceanic values measured in the Rockall Trough.

Table 1 .
List of parameters from the Rockall Trough as seen in the CDIAC database.

Table 2 .
Details of surveys with number of chemistry samples taken on each.

Table 2
Details of surveys with number of chemistry samples taken on each.

Jan 2012). Stations with triangle symbols are where carbonate parameters were measured. Note in CE11001 only surface samples were taken, in other yea rs
samples were taken through the water column.

Table 3 .
Accuracy and precision of DIC and TA in µmol kg −1 for each of the surveys, calculated as the average and standard deviation, respectively of the differences between duplicate samples, where n is the number of duplicate samples.

Table 4 .
Limit of detection (LOD), limit of quantification (LOQ), both in µmol L −1 , and uncertainty of measurement (UCM) for the nutrient analysis.The ranges given are the linear calibration ranges, concentrations that fall above these are diluted into the linear range.
Two consecutive conductivity readings within 0.00002 units of each other must be