Sampling Methods
The samples (with the exception of surface samples) were collected for trace metal determinations at 21 stations from the RV Knorr using a custom-built US GEOTRACES trace metal clean rosette consisting of an epoxy painted Al rosette frame containing 24x12 L GO-FLO bottles (Cutter and Bruland, 2012). Immediately after the package was recovered, the tops of the GO-FLO bottles were covered with plastic bags and the bottles were removed from the frame and carried into the US GEOTRACES clean van for sub-sampling. The GO-FLO bottles were pressurized to 10 psi using 0.2 µm-filtered compressed air and samples were filtered through 0.2 µm Acropak filters. All sub-sampling was undertaken in the clean van using rigorous trace metal protocols. Surface samples are from the towed fish surface sampling system.

Analytical Methods
FIA Sample Preparation: Samples were drawn into pre-numbered 125 ml PMP bottles after three rinses and were stored in plastic bags in the dark at room temperature before determination which was usually within 12 -36 hours of collection. Prior to determination samples were acidified by the addition of 125 ul sub-boiling distilled 6N HCl (hereinafter 6N HCl) and were microwaved in groups of 4 for 3 minutes in a 900 W microwave oven to achieve a temperature of 60+10 °C. Samples were allowed to cool for at least 1 hour prior to determination. Samples were determined in groups of 8.

FIA Standards: Shipboard mixed standards (Al and Fe, Mn) were prepared in the shore-based laboratory by serial dilution of commercial Al, Fe, and Mn standards (BDH Aristar) into distilled water which was acidified with the equivalent of 4 ml sub-boiled 6N HCl. Standards for instrument calibration were prepared daily from filtered seawater by acidifying 1 L of low Fe seawater from a previous cast with 1 ml of 6N HCl and microwaving for 5 minutes to reach a temperature of 60 + 10°C. After 1 hour, 200 + 2 ml of the cooled seawater was added to each of three 250 ml PMP bottles each of which had been rinsed three times with the microwaved seawater and shaken dry. Working standards were prepared by adding 0, +100µL, +200µL spikes of the shipboard mixed standard to these bottles, to yield a standard curve of +20.77nM and +31.53nM for Al, +0.97nM and +1.94nM for Fe, +3.84nM and +7.68nM for Mn. The system blank from the addition of the acid and buffer to samples was determined by double spiking a replicate sample i.e. by adding 2 x 125 ul 6N HCl and 5 ml of sample buffer to the replicate bottle and comparing the resulting signal to the original sample.

Dissolved Al Concentrations: Dissolved Al was determined using a Flow Injection Analysis scheme with fluorometric detection. Major components were a Rabbit peristaltic pump, a Dynamax FL-1 fluorometer, a Rainin A/D board and a Macintosh G3 computer running Rainin MacIntegrator v 1.4.3 to log and reduce data. The analytical scheme produces a complex between lumogallion and dissolved Al which when excited at 484 nm produces flourescence at 552 nm. Detailed description of the methodology is published in Resing and Measures (1994).

A 3-minute pre-concentration of sample (~9 ml) onto an 8-hydroxyquinoline (8-HQ) resin column yielded a detection limit of 0.24 and a precision of 4.8% at 1.67 nM.

Dissolved Fe: Dissolved Fe was determined using a Flow Injection Analysis scheme with spectrophotometric detection (Rainin Dynamax UV-C). Major components were a Rabbit peristaltic pump, a Dynamax FL-1 fluorometer, a Rainin A/D board and a Macintosh G3 computer running Rainin MacIntegrator v 1.4.3 to log and reduce data. The spectrophotometric detection of the iron eluted from the column is achieved through its catalytic effect on the oxidation of N,N-dimethylp-phenylenediamine dihydrochloride (DPD) the oxidised product is measured at 514 nm. Detailed description of the methodology is published in Measures et al (1995).

A 3-minute pre-concentration of sample (~9 ml) onto an 8-hydroxyquinoline (8-HQ) resin column yielded a detection limit of 0.064 nM and a precision of 2.8%.

Dissolved Mn: Dissolved Mn was determined using a Flow Injection Analysis scheme with spectrophotometric detection (Rainin Dynamax UV-C). Major components were a Rabbit peristaltic pump, a Dynamax FL-1 fluorometer, a Rainin A/D board and a Macintosh G3 computer running Rainin MacIntegrator v 1.4.3 to log and reduce data. The spectrophotometric detection of the manganese eluted from the column is achieved through its catalytic effect on the formation of malachite green which is measured at 620 nm. Detailed description of the methodology is published in Resing and Mottl (1992).

A 3-minute pre-concentration of sample (~9 ml) onto an 8-hydroxyquinoline (8-HQ) resin column yielded a detection limit of 0.03 nM and a precision of 2.9% at 0.05 nM.

Calculation of FIA data: Calculation of sample concentrations was by dividing the peak height derived from sample using the A/D software by the calculated slope of the standard curve. Variations in the slope of the standard curve during a day's run were corrected by the following procedure.

The change in the value of the slope of the standard curve between each run of standards was divided by the number of samples run between those standards to provide a calculated value for the slope of the standard curve at the point each sample was run. The value of the peak height for each sample was then recalculated by the estimated ratio of the standard curve slope at the point that sample was run. The estimate of the slope at each sample run is calculated by: (Initial slope + (incremental change per sample X # of samples run since initial standard was run)). The sample concentration is then calculated from the initial standard curve slope.

Data correction with shore-based ICPMS data
Shipboard determined dissolved Fe and Mn have been corrected using the shore-based ICPMS data that was measured by co-PI Dr. Jingfeng Wu. That data will be submitted separately. The corrected FIA data was calculated using the slope and the intercept of the correlation plot between the ICPMS data and FIA data on each of the days the FIA was run.

Shipboard dissolved Fe: Shipboard calculation:
The Fe concentration of samples was determined by dividing the sample peak height by the slope of the Fe standard solution. The standard curve was produced by adding +0 µL (+0 nM), +100 µL (+0.972 nM), +200 µL (+1.944 nM) of the shipboard standard solution into 200 mL aliquots of seawater. This seawater was obtained from a 1L sample of seawater (usually from near the chlorophyll max) that had been filtered through a 0.2 µm pore-size Acropack cartridge, acidified with 1 ml of 6N HCl (sub boiling distilled) and microwaved to achieve a temperature of 60 + 10°C. A typical standard curve is shown in Figure 1.

Figure 1. Fe standard curve from the standard addition method. (click on the image to view a larger version)

The shipboard standard solution was 1.944 µM and had been prepared by gravimmetric dilution from a high purity standard solution (1000 µg/mL) at the shore-based laboratory of University of Hawaii. Overall precision is 2.8%, and the average detection limit is 0.064 nM.

Shipboard dissolved Fe: Blank calculation:
At sea the blank calculation of the data was made using the value derived from a sample that had been spiked with EDTA. This process yields a sample that should have no uncomplexed Fe thus allowing subtraction of the system blank.

Shore based adjustment of the shipboard data was made by comparison with the shore-based ICP MS data measured by J. Wu at U Miami. A plot of each station's data (FIA vs ICP MS) yielded a slope and intercept that was then used to recalculate the shipboard data. This indicated that the shipboard blank correction was underestimating the system blank.

Table 1. Slope and intercept between shipboard FIA (substituted EDTA blank) and ICPMS data
Station           Slope   Intercept   R2
USGT11-01    1.470    -0.022      0.944
USGT11-02    1.272    0.057         0.869
USGT11-03    1.144    0.205         0.948
USGT11-06    1.079    0.277       0.930
USGT11-08    1.036    0.208   0.948
USGT11-10    1.399    0.103         0.974
USGT11-12    1.114    0.441         0.797
USGT11-14    1.152    0.199    0.947
USGT11-16    1.216    0.105       0.988
USGT11-18    0.943    0.180 0.932
USGT11-20    0.980    0.256   0.903
USGT11-22    0.967    0.394   0.964
USGT11-24    1.605    0.318   0.964

The ICPMS data reported by Miami (Dr. Wu) were only for full depth stations. Two of the demi stations were determined at sea on the same day that full depth stations were run specifically GT11-17 and GT 16; GT11-23 and GT11-22. Thus, these samples were adjusted using the same slope and intercept used for the full depth station. However, the other demi station samples were determined on separate days. Thus, the samples cannot currently be adjusted with existing ICPMS data.

The following demi stations have not been adjusted and are not reported: USGT11-05, USGT11-11, USGT11-13, USGT11-18, and USGT11-21.

Shipboard dissolved Fe: Accuracy results:
There are no shipboard results for the recoveries of Certified Reference Materials (CRMs) or consensus reference materials (e.g., SAFe water) as these samples are too acid to be run using this protocol. Instead these data are corrected using the shore-based ICP MS values which are calibrated against SaFe samples.

Shipboard dissolved Mn: Details about calibration:
The Mn concentration of sample was calculated using the quadratic polynomial (x= the peak height, y= the Mn standard concentrations, y= M0+M1x+M2x^2). The coefficients (M0, M1, M2) of the polynomial were obtained based on the standard curve. This was produced by the standard addition method adding (+0µL (+0 nM), +100µL (+3.842 nM), +200 µL (+7.683 nM) of the spike standard solution added into the 200mL of seawater which had been filtered through a 0.2µm pore-size Acropack cartridge.). A typical standard curve is shown in Figure 2.

Figure 2. Mn standard curve from the standard addition method, the polynomial fit, and the coefficients of the polynomial approximation. (click on the image to view a larger version)

The shipboard standard solution was 7.683 µM and had been prepared by gravimmetric dilution from a high purity standard solution (1000µg/mL) at the shore-based laboratory of University of Hawaii. The precision of the method, the relative standard deviation (RSD, %), and detection limit (3sigma; D.L.) were calculated on each analysis day from the triplicate analysis of a spiked standard seawater solution. Results are shown in Mn Table 1, below.

Overall average precision (from Mn Table 1 below) is 2.9%, and the average detection limit is 0.05 nM.

Shipboard dissolved Mn: Blank calculation:
There is no blank calculation of the data made at sea. Shore based adjustment of the shipboard data was made by comparison between the shore-based ICP MS data measured by J Wu at RSMAS, and the shipboard calculated data, the intercept being the blank offset between the methodologies. The ICPMS data has been submitted separately. In general, the difference was very small (average difference was + 0.029nM) which is similar to the analytical uncertainty of the shipboard method. The detailed value of the blank for each station is shown in Mn Table 2.

Shipboard dissolved Mn: Accuracy results:
There are no shipboard result for the recoveries of Certified Reference Materials (CRMs) or consensus reference materials (e.g., SAFe water) as these samples are too acid to be run using this protocol. Instead these data are corrected to the shore-based ICP MS values which are calibrated against SaFe samples.

Mn Table 1. Precision, Relative Standard Deviation, and Detection Limit of the shipboard FIA method for dissolved Mn
                              Concentration of std. solution for the precision (nM)      RSD %     D.L (nM)
USGT11-01       1.243                                                                                       2.1   0.079
USGT11-02           1.027         3.4           0.103
USGT11-03           0.653         2.2           0.043
USGT11-05           0.533         3.9           0.062
USGT11-06    0.620         1.6   0.030
USGT11-08       0.443             1.6     0.021
USGT11-10       1.109             2.1   0.070
USGT11-11    0.708             3.6   0.076
USGT11-12       0.396                                                                                       3.2   0.038
USGT11-13         0.271                                                                                       2.7           0.022
USGT11-14&15    0.334         1.5    0.015
USGT11-16&17    0.183                                                                                       1.1    0.006
USGT11-18           0.160             4.2 0.020
USGT11-19    0.128             1.2           0.005
USGT11-20    1.371         1.2           0.050
USGT11-21       0.713        8.5           0.181
USGT11-22&23    0.123         4.9           0.018
USGT11-24       0.563         3.6   0.061

Mn Table 2.
Station                                             Blank (Intercept)    Memo
USGT11-01                                        0.0298
USGT11-02                                        0.0626
USGT11-03         -0.0151
USGT11-05        0.000
USGT11-06        0.0812
USGT11-08     0.0099
USGT11-10                                        0.0546
USGT11-11       0.000
USGT11-12        0.0610
USGT11-13                                        0.000
USGT11-14                                        0.0383
USGT11-15*                                      0.0383
USGT11-16 (hydrothermal plume)    -0.00910
USGT11-17                                      -0.0091
USGT11-18                                        0.0154
USGT11-19        0.000
USGT11-20                                        0.256                    Blank problem
USGT11-21        0.029                No ICP MS use average blank correction
USGT11-22       -0.0047
USGT11-23                                       -0.0047
USGT11-24                                     -0.0334

Shipboard dissolved Al data: UH Shipboard methodology:
UH Dissolved Al samples are run at sea using standards prepared from a bulk standard which has been diluted using gravimetry. We make two bottles of each working standard and take both bottles to sea but only use one of the bottles at sea (unless we think a bottle has been compromised). We bring both the unused and the partially used bottles of standard back to our laboratory for long term intercalibration. Aliquots of the shipboard standard are added to 200ml seawater samples that are obtained from a single bottle of 1L of seawater that has been acidified and microwaved and then cooled, to produce a standard curve that has +0, +20.77, +31.53 nM added Al. The concentration of Al in any sample is obtained by dividing the peak height of the sample by the slope of the peak height/nM of the standard curve. The intercept of the standard curve is not used since the "0" standard is not a true 0, but is the amount of Al already in the water used to produce the standards. We have not found that there is any procedural blank in this method (from double spiking samples with buffer and acid) and thus do not subtract any blank from the calculated values.

Comparison between UH and NIOZ Al data sets at BATS
In our opinion, the offset between the University of Hawaii and the NIOZ Al data sets in the Atlantic crossover station at Bermuda (Station S), appears to be a result of standard disagreement. This conclusion is based on the following analysis.

A plot of the two profiles shows the vertical distribution of the two data sets (24 NIOZ samples collected 12 June 2010 and 36 UH samples collected 19 November 2011). There are significant differences between the two profiles in the upper waters, particularly the surface waters, but this is hardly surprising given the 1.5 years between the two occupations and the times of year that they were conducted (June and November). Given that the upper water concentrations of dissolved Al are predominantly a function of atmospheric deposition and its incorporation into subsurface waters (Sub-Tropical Mode Waters) through subduction processes we have not attempted to derive conclusions from this part of the water column since they are likely to be highly temporally variable.

We are also excluding the data deeper than 4002 m since the shape of the two profiles which both appear to be oceanographically consistent appear to be quite different from each other, which we take to also be a difference due to variations in the strength of, or presence of, a bottom nepheloid layer between the two occupations.

Instead we focus on the region from 974m to 4002m where there is a clear and consistent offset between the two data sets.

A complicating factor in making the comparison though is that the samples from the two data sets are not at the same depths. In order to make a cross plot we have binned the two data sets to correspond to similar sampling depths (see attached table for the binning). Most of the samples are within ~50m depth, but two samples, (`3600 and 3900 are ~ 100m apart).

The plot of the 9 corresponding data points from the two, shows a very clear correlation between the data sets (R=0.993) with a slope of 1.251. Virtually all of the data fall exactly on this line. If the somewhat anomalous point at 2500m in the UH data set is eliminated R increases slightly to R=0.9965 and the slope drops slightly to 1.24.

The implication of this is that the difference between these data sets is the result of a simple standard offset. A blank offset would lead to a constant difference between the data sets, which is not what this plot implies.

We do not have access to the NIOZ standards and have been informed by R. Middag that there are no aliquots left of the standard they actually used at sea. He is trying to track down the original stock standard that their shipboard standard was prepared from and we plan to obtain an aliquot of this to check against our standard.

Meanwhile, in an attempt to identify any possible source of the offset in our own standards we have rerun both our original concentrated standard used to produce our shipboard standards as well as the actual standards that we used on board ship during both GT10 and GT11 cruises. The results are as follows:
- UH Concentrated standard (used to produce the shipboard standards) was diluted to produce a standard curve. (Standard A)
- GT10 and GT11 shipboard standards were then run and calculated against this concentrated standard curve.
- Another concentrated Al standard (Standard B) was procured from a different laboratory at UH and a dilution was made of this standard and it was run against Standard A.

Results of the standard intercomparison as a % of the expected value--all relative to Standard A:
GT10 93.0%
GT11 95.3%
Standard B 95.5%

Conclusions: Relative to Standard B, Standard A is 4.5% high, which could be the result of evaporation in that bottle. However these results which imply our standard maybe 4.5% high are preliminary and we intend to rerun all the standards again when we can get an aliquot of the NIOZ standard. We are also purchasing a new stock standard to evaluate the agreement between Standards A and B.

An evaporated stock standard will lead to an underestimate in the true value of the sample, so if anything our current understanding of our Al values is that they maybe 4.5% lower than the true value. We note an ~2.5% offset between the GT10 and GT11 standards which maybe a result of evaporation of the GT10 shipboard standard. We will compare the unused GT10 standard with the newly purchased concentrated standard when it arrives. We also hope to have an aliquot of the NIOZ standard at that point.

We also note that the NIOZ data are reported as per kg, whereas ours are volume based (we measure by volume at sea and so report in those units). Depending on the value they used to convert their data (we are assuming here that their values were originally on a per volume basis), the NIOZ data would be ~2.8% lower than the UH data.